Functional requirements of AID’s higher order structures and their interaction with RNA-binding proteins

Significance

This paper demonstrates that activation-induced cytidine deaminase (AID), an essential enzyme in antigen-induced antibody diversification, forms distinct ribonucleoprotein complexes depending on its structural states: namely monomers or dimers. The identified RNA-binding proteins are required for the function of AID: namely DNA cleavage or recombination. In addition, the complex formation between AID and heterogeneous nuclear ribonucleoproteins (hnRNPs) is RNA-dependent.

Abstract

Activation-induced cytidine deaminase (AID) is essential for the somatic hypermutation (SHM) and class-switch recombination (CSR) of Ig genes. Although both the N and C termini of AID have unique functions in DNA cleavage and recombination, respectively, during SHM and CSR, their molecular mechanisms are poorly understood. Using a bimolecular fluorescence complementation (BiFC) assay combined with glycerol gradient fractionation, we revealed that the AID C terminus is required for a stable dimer formation. Furthermore, AID monomers and dimers form complexes with distinct heterogeneous nuclear ribonucleoproteins (hnRNPs). AID monomers associate with DNA cleavage cofactor hnRNP K whereas AID dimers associate with recombination cofactors hnRNP L, hnRNP U, and Serpine mRNA-binding protein 1. All of these AID/ribonucleoprotein associations are RNA-dependent. We propose that AID’s structure-specific cofactor complex formations differentially contribute to its DNA-cleavage and recombination functions.

Activation-induced cytidine deaminase (AID), which is expressed in antigen-stimulated mature B cells, is essential for Ig somatic hypermutation (SHM) and class-switch recombination (CSR) (1, 2). AID induces DNA breaks at the variable (V) and switch (S) regions during SHM and CSR, respectively (3, 4). Although both processes are initiated by AID-induced DNA cleavage, point mutations at the V region are executed mostly by error-prone DNA repair whereas CSR is accomplished by recombination of cleaved ends at donor and acceptor S regions (5, 6). However, the detailed mechanisms by which AID carries out the two mechanistically distinct functions for SHM and CSR have yet to be uncovered (7). Studies on AID mutants revealed that AID’s N- and C-terminal domains are distinctly required for its DNA-cleavage and recombination functions, respectively (8–10). Mutations at the N terminus of AID impair SHM as well as CSR whereas those at the C terminus abrogate CSR only and show increased SHM activity. Recent studies demonstrated that the CSR process after DNA cleavage, including the synapsis formation between cleaved ends, is impaired with the C-terminally defective AID, indicating that AID’s C terminus confers a CSR-specific recombination function, independent of AID’s DNA cleavage function, by an unknown mechanism (11, 12).

AID belongs to the APOBEC (apolipoprotein B mRNA-editing enzyme catalytic polypeptide) family of cytidine deaminases (CDDs) and shows high sequence homology with APOBEC1 (A1) (1, 13, 14), which edits apolipoprotein B (APOB) mRNA. The APOB mRNA editing ability of A1 is highly dependent on its cofactors, A1CF/ACF (15, 16) and RBM47 (17), both of which belong to the heterogeneous nuclear ribonucleoprotein (hnRNP) family. Recently, two A1CF-like hnRNPs, hnRNP K and hnRNP L, were identified as the cofactors of AID and found to be involved in the cleavage and recombination of DNA, respectively (18). Because the N and C termini of AID differentially regulate two functions of AID—cleavage and recombination, respectively—we speculated that the AID termini would be critical for function-coupled cofactor association. For instance, the N or C terminus of AID may function as a molecular switch that induces an AID–AID interaction, enabling AID to exert distinct physiological functions through its association with cofactors. Regrettably, however, there is little structural information available that can explain any of AID’s regulatory modes of action, including its cofactor association mechanisms, in the context of its physiological functions.

Although a significant amount of structural information is available for a number of APOBEC family members, the 3D structures of A1 and AID are yet to be resolved (19, 20). The CDD family of enzymes exists in nature in a variety of structural forms, including monomeric, dimeric, and tetrameric forms, and comparative structural modeling using the yeast CDD structure predicts a dimeric structure for both A1 and AID (21, 22). On the other hand, homology modeling with the APOBEC2 (A2) crystal structure, which seems to be a tetramer composed of two head-to-head interacting dimers, predicts that AID forms a tetramer (23). Notably, A2 was later reported to exist as a monomer in solution (24). Similarly, an atomic force microscopic (AFM) study found that AID exists in the cell predominantly as a monomer associated with a single-strand DNA substrate (25). However, the same AFM dataset was interpreted differently by another group of investigators, who concluded that AID probably forms an A2-like tetramer in solution (26). The modeling of AID’s catalytic pocket in reference to eight APOBEC family members suggested that most of the AID–DNA complex remains in an inactive state due to occlusion by the substrate DNA, which may explain its weak catalytic activity for cleaving DNA in vitro (27).

One of the limitations of the computational modeling of AID’s structure is that AID’s N-and C-terminal sequences are substantially different from those of other APOBEC members and thus reside outside the modeling template. Although the structural outcome of a protein can differ by a variety of reasons, including the methods applied (28), none of the AID studies mentioned above explain why the C-terminal deletion of AID leads to the loss of CSR function only. Therefore, model-based computational simulation may not explain the physiological structure–function relationship of AID in B cells.

Here, we explored AID’s structure–function relationship using a bimolecular fluorescence complementation (BiFC) assay, which detects homo- or heteromeric protein–protein interactions in live cells (29, 30). For the homomeric interaction assay, the target protein is fused to two nonfluorescent halves of a green or red fluorescent protein. An interaction between two of the target proteins brings the two nonfluorescent halves of the fluorescent protein into close proximity, reconstituting the fluorescence. The BiFC assay thus allows a rapid analysis of the dimerization of a protein of interest in live cells.

By combining this assay with other biochemical approaches, such as coimmunoprecipitation (co-IP) and glycerol gradient sedimentation, we revealed the presence of both monomeric and dimeric forms of AID in analyzed cells. Intriguingly, C-terminal AID mutants that lost CSR function showed a severe dimerization defect, suggesting that AID’s C terminus is required to stabilize the dimeric structure that is required for CSR. We also showed that the AID monomer and dimer associate with different RNA-binding proteins (RBPs) to form ribonucleoprotein (RNP) complexes. Based on these findings, we propose that the monomeric AID–RNP complex includes hnRNP K (18) and contributes to the DNA cleavage function of AID whereas the dimeric AID–RNP complexes include hnRNP L (18), hnRNP U (31), or Serpine mRNA-binding protein 1 (SERBP1) (32) and contribute to the recombination step of CSR.

Results

Detection of APOBEC Homodimer Formation by the Monomeric Kusabira Green-BiFC Assay.

The monomeric Kusabira Green-BiFC (mKG-BiFC) assay detects the association of two nonfluorescent mKG fragments (mKG-N and mKG-C), which occurs through the interaction of fusion partners (33). To validate the mKG-BiFC–based protein–protein interaction system for the APOBEC family, we first tested the four members of this family that are known to undergo homomeric interactions. The tetrameric structure of APOBEC2 (A2), composed of two dimeric units, was revealed from its crystallographic structure (23). Biochemical studies and atomic force microscopy observations showed that APOBEC3G (A3G) formed a dimer (34, 35) or an oligomer (36). APOBEC3A (A3A) was very recently crystalized as a dimer (37). Also, the dimeric structure of APOBEC1 (A1) was suggested by biochemical studies (13, 38, 39).

We examined the dimer formation of A2, A3G, A3A, and A1 using the mKG-BiFC assay. We generated four BiFC constructs (A, B, C, and D) for each APOBEC family member, as depicted in Fig. S1A. The split fragments of mKG, designated mKG.N (N-terminal half) and mKG.C (C-terminal half), were fused to the N or C terminus of the full-length APOBEC construct: A, mKG.N-APOBEC; B, mKG.C-APOBEC; C, APOBEC-mKG.N; and D, APOBEC-mKG.C (Fig. S1A). As expected, pairwise (AB, AD, CB, or CD) transfection of the fusion constructs in human embryonic kidney (HEK) 293T cells generated BiFC-positive cells for all of the ABOBECs (Fig. S1B). All of the A2–A2 interaction combinations produced BiFC with a high mean fluorescence intensity (MFI), which may reflect A2’s high-affinity homomeric (di/tera) interaction property in live cells. The A3G–A3G interaction also produced intense BiFC signals, in agreement with a previous A3G-interaction study using an mCherry-BiFC assay (40). A3A and A1 showed relatively weaker but substantial BiFC signals. Interestingly, the maximum BiFC signal was obtained for A1 by the AB combination whereas the CD combination was the most efficient for the rest. For each APOBEC, none of the individual mKG fusion constructs (A, B, C, or D) alone produced an appreciable BiFC signal, confirming that the mKG-BiFC assay specifically detected APOBEC–APOBEC dimerization.

Fig. S1.

Detection of APOBEC–APOBEC interaction by the mKG-BiFC assay. (A, Left) The monomeric kusabira green (mKG) fluorescent protein can be dissected into two nonfluorescent fragments: mKG.N (cyan) and mKG.C (yellow). mKG.N and mKG.C were fused to APOBEC members (light orange) via a linker sequence. Fluorescence reconstitution (green) occurred when the two fragments of the fluorescent protein were assembled by the interaction of APOBEC proteins. (Right) Structure of the mKG-BiFC constructs mKG.N-APOBEC (construct A), mKG.C-APOBEC (construct B), APOBEC-mKG.N (construct C), and APOBEC-mKG.C (construct D). The mKG N-terminal (mKG.N) and C-terminal (mKG.C) fragments were fused to either the C-terminal (constructs C and D) or the N-terminal (constructs A and B) end of full-length APOBEC (constructs A and B) and were cotransfected into HEK293T cells in four possible pairs of combinations (AB, AD, CB, and CD). (B, Left) Frequency histograms from the FACS analysis of HEK293T cells transfected with individual or pairwise mKG-BiFC constructs of APOBEC2, APOBEC3G, APOBEC3A, and APOBEC1. Mean fluorescence intensity values represent the interaction strength in each histogram. (Right) Mean fluorescence intensity of BiFC-positive cells shown as bar graphs. (C) Fluorescence microscopy images of HEK293T cells expressing the mKG-BiFC constructs of APOBEC2, APOBEC3G, APOBEC3A, APOBEC1, and AID. The pairs of mKG-BiFC constructs transfected are shown above the images. Only the pairwise combinations that produced the maximum MFI for each protein are shown. (D) FACS analysis of HEK293T cells transfected with individual and pairwise combinations of mKG.N, mKG.C vectors. As a reference, BiFC data of AID (CB combination) is shown (taken from Fig. 1A). MFI (%) values are plotted next to the FACS profile. Glycerol gradient fractionation was performed using total cell extracts of HEK293T cells cotransfected with mKG.N and mKG.C vectors. Fluorescence values of the fractions are plotted; the profiles of the mock and WT AID are included as reference from Fig. 4A.

Evidence for AID Dimer Formation by the mKG-BiFC Assay.

Using the mKG vector system, we next examined the AID–AID interaction in HEK293T cells. BiFC signals generated upon pairwise cotransfection of the AID–mKG fusion constructs were measured by FACS (Fig. 1 A and B). Remarkably, the CB pair produced the highest MFI (Fig. 1 B and C) although all of the four interacting combinations generated BiFC-positive cells with nearly the same expression amount of all of the four constructs (Fig. 1D). We observed that none of the fusion constructs alone produced a fluorescence signal above background. We also verified that the coexpression of mKG.N and mKG.C did not produce any BiFC signal (Fig. S1D). Furthermore, the efficiency of the AID–AID interaction was assessed by a direct co-IP analysis (Fig. 1E), which mostly agreed with the BiFC MFI profile (Fig. 1 B and C). Fluorescence microscopic observation showed that the BiFC signal of AID was localized to the cytoplasm (Fig. S1C), which was previously shown to be the major localization site of AID (Fig. S1C) (41). The expected subcellular localizations were observed for the other APOBECs, confirming that the BiFC signals represent native characteristics of the APOBEC proteins.

Fig. 1.

Evidence of AID dimer formation in living cells by the mKG-BiFC assay. (A) Schematic of the AID–AID dimerization examined by the mKG-BiFC assay. Fluorescence reconstitution (green) occurs when the nonfluorescent N- and C-terminal mKG fragments (mKG.N and mKG.C) are assembled by AID homodimerization. The two mKG fragments fused to full-length AID (light orange) are shown in cyan and yellow, respectively. The diagram below shows the four types (types A–D) of fusion constructs used in the mKG-BiFC assay for AID. In these constructs, AID was placed either at the C or N terminus of the mKG fragments, and the constructs were transfected into HEK293T cells alone or in combination as indicated above each FACS profile (B). Values inside the histogram plots represent the mean fluorescence intensity (MFI) of the BiFC-positive cells. (C) Mean fluorescence data of the BiFC-positive cell population from three independent experiments are shown as bar graphs. (D) Western blot analysis of the mKG–AID fusion constructs. (E) AID dimer formation detection by the anti-FLAG IP of A and C coexpressed with B and D, which had a FLAG epitope. Coprecipitated proteins were detected by immunoblotting with an anti-AID antibody.

AID C-Terminal Mutants That Are Defective in CSR Are Unable to Form a Dimer.

To evaluate the functional relevance of AID dimer formation, we examined the BiFC signals using AID mutants that showed clear functional defects. AID C-terminal mutants are of particular importance because they lose CSR function but retain strong SHM activity (Fig. 2A). Such well-characterized C-terminal AID mutants include JP8Bdel (R183X), P20 (34-aa insertion at residue 182), JP41 (R190X), and JP8B (26-aa frameshift replacement at residue 183) (8). Remarkably, each mutant showed severely defective BiFC signal generation in all four pairwise combinations even though they were all expressed well (Fig. 2 B and C). These results clearly demonstrated that the C-terminal mutations severely impaired the homodimer-formation ability of AID in live cells. In addition, all of the C-terminal mutants were also defective in their heteromeric interaction with WT AID (Fig. 2E). To confirm the loss of the homo- and heterodimerization abilities, we performed co-IP analyses with the C-terminal mutants (Fig. 2 D and F). Although the WT AID could efficiently pull down its WT AID counterpart, none of the C-terminal mutants could pull down either their respective counterparts or the WT AID. Thus, both the BiFC and co-IP analyses confirmed that the homo- and heterodimerization abilities of AID were lost in the C-terminal mutants.

Fig. 2.

AID C terminus mutants are defective in dimerization. (A) Structures of AID C-terminal mutants and their CSR and SHM efficiencies in reference to WT. The CSR and SHM efficiencies of the mutants are from published reports. (B) FACS analysis of HEK293T cells transfected with pairwise combinations of the indicated mKG fusion constructs of WT AID or C-terminal AID mutants. MFI values of BiFC are shown in the respective histogram plots. The expression level of each mutant is shown next to its FACS profile. Tubulin was used as a loading control. (C) Percentage of MFI (mean ± SD; n = 3) is shown only for the CB combination. (D) Immunoprecipitation to analyze the homodimerization of the AID C-terminal mutants was performed by coexpressing the CB combination of BiFC constructs in HEK293T cells. The FLAG epitope was fused to the C terminus of the AID mutants in the B constructs. Cell lysates were IPed with anti-FLAG M2 agarose, and the coprecipitated proteins were analyzed by immunoblotting with an anti-AID antibody. (E) BiFC assay to determine the heterodimer formation between WT AID and the C-terminal mutants. Frequency histograms of HEK293T cells cotransfected with the CB combination of BiFC constructs, as indicated above each plot. The data from three independent experiments are shown next to the FACS profile. (F) Immunoprecipitation to analyze the heterodimerization of the AID C-terminal mutants with WT AID. The cotransfection and IP procedure were as described above (D).

The AID N Terminus Is also Required for Dimer Formation.

Although the BiFC signal intensity did not necessarily reflect the proximity between the two fusion termini of AID, the maximum signal obtained by the combination of C (AID-mKG.N) and B (mKG.C-AID) might suggest the involvement of AID’s N terminus in the AID–AID dimer formation. Thus, we generated a CB pair of mKG-AID constructs, with serial deletions at the N terminus of AID (∆N5, ∆N10, and ∆N26), and examined their homodimer formation by the BiFC assay (Fig. 3 A and B). The serial N-terminal truncations caused a progressive loss of the BiFC signal (Fig. 3 B and C) although all of the mutants were expressed well.

Fig. 3.

The N terminus of AID is also required for AID dimerization. (A) Schematics of serially truncated N-terminal mutants of AID, in which the first 5, 10, and 26 amino acids were deleted. (B) Frequency histograms of the FACS analyses of HEK293T cells transfected with WT AID and N-terminally truncated mutants. BiFC constructs were cotransfected in pairwise combinations as indicated above each FACS profile. The Western blot analysis is shown for each AID expression profile, using tubulin as a loading control. (C) Percent MFI plot summarizing the results of the CB combination in the BiFC assay (mean ± SD; n = 3). The data from the other combinations are not shown because the CB combination produced the highest dimerization signal for WT AID. (D, Left) Representative FACS data showing the IgG1 switching efficiency of the N-terminally truncated AID mutants in AID^−/− splenic B cells. Numbers indicate the percentage of IgG1 positive (+) cells among the GFP-gated population in each FACS profile. (Right) Relative frequencies of CSR and SHM of WT AID and its N-terminal mutants. (E, Left) BiFC analysis examining the heterodimeric interaction between WT AID and its N-terminal mutants. Frequency histograms of the FACS analysis of HEK293T cells transfected with the indicated pairwise combinations of BiFC constructs. (Right) Bar plots show the % MFI (mean ± SD; n = 3). (F and G) Immunoprecipitation analysis of the homo- and heterodimerization of N-terminal mutants by coexpressing CB combinations of BiFC constructs in HEK293T cells. AID was FLAG-tagged at the C terminus in the B construct. Cell lysates were IPed with anti-FLAG M2 agarose, and the coprecipitated proteins were analyzed by immunoblotting with an anti-AID antibody.

Similar to the C-terminal mutants, the N-terminally truncated AID mutants (∆N5, ∆N10, and ∆N26) barely formed heterodimers with the WT AID (Fig. 3E). Co-IP analysis using homomeric and heteromeric combinations confirmed that the deletion of as few as 10 aa at the AID N terminus was sufficient to completely disrupt the homo- and heterodimeric interactions (Fig. 3 F and G). Examination of the CSR efficiency of the N-terminal mutants in AID^−/− spleen B cells revealed a similar progressive loss of this activity (Fig. 3D). Taken together, these findings show that AID’s intact N terminus is important for the AID–AID interaction and for CSR activity.

Separation of AID Monomer and Dimer on a Glycerol Gradient.

Although the BiFC signal in live cells indicated a homomeric interaction of AID, it could not provide further information about the AID tertiary structure that might incorporate other proteins and RNA. Thus, we performed glycerol gradient centrifugation of the cell extracts from HEK293T cells expressing the CB combination of mKG-AID constructs. In reference to the background fluorescence signal from HEK293T cell extracts, the CB combination of mKG-AID revealed two major fluorescence peaks at around fractions 3–12 and 19–23, which we called the low molecular weight (LMW) and high molecular weight (HMW) regions, respectively (Fig. 4A).

Fig. 4.

Analysis of AID monomer and dimer by glycerol gradient sedimentation. (A) BiFC signal profile of the glycerol gradient fractions of the total cell extracts of HEK293T cells transfected with the CB combination of BiFC constructs of WT or mutant AID, as indicated. A total of 23 fractions collected from the top to bottom of the gradient were subjected to fluorescence intensity measurement. Plot shows a representative dataset, and the positions of protein molecular weight standards are indicated by open triangles. A mock fluorescence profile was obtained from HEK293T cells that were not transfected with any AID construct. (B) Anti-AID immunoblots of the glycerol gradient fractions. (C) Anti-FLAG and anti-HA IP of AID from the gradient fractions. Cell lysate was prepared from HEK293T cells transfected with pairwise mKG-CB combinations of WT AID, where the C- and B-constructs harbored AID-Flag and AID-HA, respectively. Both input and IPed fractions were analyzed by immunoblotting (IB) with an anti-AID antibody. Fractions enriched with monomers (mono) and dimers/multimers are indicated below the blots. (D) Anti-FLAG IP analysis of the gradient fractions derived from WT and mutant (P20 and JP8B) AID-expressing cells. HEK293T cells were transfected with the CB combination of BiFC constructs in which the B-construct harbored Flag-tagged AID. Both input and IPed fractions were analyzed by immunoblotting with an anti-AID antibody.

Western blot analyses of the gradient fractions showed that the distribution of AID, fused with mKG-N and mKG-C fragments, in general corresponded well with the two fluorescence peaks (Fig. 4B), indicating that the reconstituted BiFC signal was stable after cell lysis. To distinguish the distribution of AID monomer and dimer, we performed IP using gradient fractions corresponding to the two major BiFC peak areas (Fig. 4C). Because the two mKG-AID constructs were tagged with different epitopes (FLAG and HA), pulling down HA-fused AID by anti-FLAG IP, and vice versa, provided further evidence of dimerized AID in the individual fractions. The failure of reciprocal AID co-IP suggested the exclusive presence of the monomer. These results showed that the monomeric AID was exclusively distributed in fractions 3–4 whereas the dimeric form was predominantly in fractions 5–12 and fractions 19–23.

In the gradient analyses of AID mutants defective at either the C terminus (P20 and JP8B) or N terminus (ΔN10), the BiFC signal was dramatically decreased at both the LMW and HMW regions compared with the WT AID (Fig. 4A), in agreement with the requirement of AID’s N and C termini for dimer formation (Fig. 4B), and the presence of dimer at the HMW region. The anti-FLAG IP of the P20 and JP8B mutants could not pull down their untagged counterparts (Fig. 4D), indicating that they did not form a homodimer, although they were broadly distributed between fractions 3 and 12. Because the AID C-terminal mutants were all shown to be monomeric, their wide distribution along the gradient suggested that some of the monomers associated with other proteins and RNAs. Collectively, these results showed that AID distributed in the LMW region could be subdivided into a narrow region of exclusive monomer (fractions 3–4) and a broader region of monomer and dimer (fractions 5–12) and that the HMW fractions of AID seemed to contain dimers.

AID Forms Multiple RNP Complexes.

To examine the multimeric complexes containing AID, extracts of HEK293T cells expressing the CB combination of mKG-AID constructs were first run on the glycerol gradient after RNase or high salt (500 mM) treatments. Strikingly, either RNase or salt treatment caused the BiFC peak in the HMW region to disappear completely, suggesting that the HMW region contains higher order AID–RNP complexes (Fig. 5A). In contrast, these treatments did not abolish the BiFC peak in the LMW region. Although the peak was slightly broadened, it was possibly due to contamination of AID species dissociated from the HMW complex.

Fig. 5.

RNA-dependent association of AID with RBPs. (A) Glycerol gradient (10–60% wt/vol) sedimentation analysis of the total cell extract of HEK293T cells transfected with the CB combination of BiFC constructs of WT AID. Cell extracts were treated with either 150 μg/mL RNase A or 500 mM NaCl and loaded onto the gradient. A total of 23 fractions were collected from the top to the bottom of the gradient, followed by fluorescence intensity measurement of the fractions, which was plotted. The profiles of the mock and the WT AID from Fig. 4A are included as reference. An anti-AID Western blot below the plot shows the distribution of AID in the fractions after RNase or salt treatment. (B) Effect of AID expression and/or RNase A treatment on the glycerol gradient distribution of seven RNA-binding proteins (RBPs), analyzed by immunoblotting (IB) with their corresponding antibodies. Indicated is the presence (+) or absence (−) of AID and RNase A. (C) Co-IP analysis of the association of RBPs with AID using the total cell extract of HEK293T cells transfected with the CB pair of WT AID BiFC constructs, in which the B-construct harbored FLAG-tagged AID. Before the co-IP analysis, the cell lysates were either untreated or treated with RNase A or 500 mM NaCl. FLAG-tagged AID was pulled down by anti-FLAG, and co-IPed proteins were analyzed by immunoblotting with RBP-specific antibodies, as indicated.

We next examined the distribution profiles of AID-interacting RBPs (hnRNP I/PTBP1, SERBP1, PABP, hnRNP U, and hnRNP C) that we discovered during AID co-IP (18, 42) in addition to hnRNP K and hnRNP L, which are known to interact with AID through RNA. We generated the migration profiles of the RBPs under three different conditions: with or without AID and after RNase treatment (Fig. 5B). HnRNP K, hnRNP I, and SERBP1, but not hnRNP L, PABP, hnRNP U, or hnRNP C, showed a clear change in their distribution patterns in the presence of AID. The distribution profiles of all of the RBPs were strongly affected by RNase treatment. Particularly, hnRNP U and hnRNP C, which were normally found in the HMW region, shifted dramatically to the LMW region.

The direct co-IP analysis of whole-cell extracts confirmed the association of these RBPs with AID, and their interactions were highly sensitive to both RNase and salt treatment (Fig. 5C). Notably, these treatments did not affect AID dimer formation, given that the Flag IP of tagged AID (B construct) could pull down its untagged counterpart (C construct), suggesting that the BiFC-mediated AID dimer was quite stable once formed. All of the RBPs examined were almost completely dissociated from AID after RNase or salt treatment, except hnRNP M, which formed a complex with AID at the HMW region (Fig. S2 A and B). No association of AID with hnRNP Q or hnRNP E1/PCBP1 was detected (Fig. S2 A and B).

Fig. S2.

Interaction of selected RBPs with AID and its C-terminal mutants. (A) Analysis of the association of hnRNP M, hnRNP Q, and hnRNP E1 with AID using the total cell extract of HEK293T cells transfected with the CB combination mKG-BiFC constructs of WT AID, in which AID in the B-construct was FLAG-tagged. Cell lysates were either untreated (−) or treated (+) with RNase A or 500 mM NaCl and immunoprecipitated (IPed) with anti-FLAG, and the coprecipitated proteins were analyzed by immunoblotting with the indicated antibodies. Mock and CB above the blots indicate the cells expressing no AID and WT AID, respectively. (B) Analysis of the association of hnRNP M, hnRNP Q, and hnRNP E1 with AID in the fractions obtained after glycerol gradient (10–60% wt/vol) sedimentation with a shorter duration (5 h). The total cell extract was prepared from HEK293T cells transfected with the mKG-CB combination of WT AID, in which AID in the B construct was FLAG-tagged. All of the gradient fractions (1–23) were individually subjected to fraction-specific immunoprecipitation (IP) with an anti-FLAG antibody. Both the input and IPed fractions were subjected to immunoblotting with the antibodies shown. (C) Comparison of the interaction of various RBPs with WT and AID C-terminal mutants. The cell extract preparation and FLAG IP were as described in B. Antibodies used for immunoblotting are shown on the right. (D) Distribution of RBPs in the gradient fractions obtained after glycerol gradient (10–60% wt/vol) sedimentation of the total cell extract of HEK293T cells (Mock) with a longer duration (17 h). Each gradient fraction (1–23) was individually subjected to IP with anti-FLAG, and the coprecipitated proteins were analyzed by immunoblotting with the specific antibodies, as indicated.

We also tested whether the monomeric C-terminal mutants could interact with the RBPs, by performing co-IP experiments of various C-terminal mutants using a whole-cell extract. The results showed that hnRNP K, hnRNP L, PABP, and hnRNP U were co-IPed with all of the C-terminal mutants as efficiently as with WT AID. HnRNP I and SERBP1 were also co-IPed with the C-terminal mutants but to a lesser degree. In particular, SERBP1 showed much less association with the C-terminal mutant JP8B (Fig. S2C). These findings support our assumption that the broad mobility of AID was due to its complex formation with various RBPs. A portion of the AID dimer can be found in the HMW region probably because of their association with multiple RBPs.

We also examined whether AID and the above RBPs had the same distribution profile in B cells. Intriguingly, endogenous or overexpressed AID in CH12F3-2A B cells appeared in the LMW and HMW regions with a similar distribution as in HEK293T cells (Fig. S3). Moreover, the migration profiles of the RBPs also matched well with those obtained using HEK293T cells (Fig. 5).

Fig. S3.

Higher order complex formation of AID in B cells. Western blot analysis of the glycerol gradient (10–60% wt/vol) fractions obtained from the total extract of CIT-activated CH12F3-2A B cells harboring AID-GFP-HA. The cells expressed endogenous AID upon CIT activation, and AID-GFP-HA was constitutively expressed. The distributions of AID and the RBPs indicated on the left were analyzed by immunoblotting with their corresponding antibodies.

Higher Order AID Complexes Contain Distinct RNPs.

To examine whether the IPed RBPs were associated with the monomeric or dimeric AID, we performed co-IP experiments with AID in individual fractions of the glycerol gradient (Fig. 6). By applying the reciprocal co-IP approach, we first determined the relative distributions of AID monomers and dimers in the gradient, followed by detection of hnRNP K and hnRNP L in the first 12 fractions, where they mostly migrated (Figs. 5B and 6A). Monomeric AID, found exclusively in fractions 3–4, was co-IPed with hnRNP K whereas hnRNP K alone was quite heavily distributed through fractions 3–8. In contrast, hnRNP L began to be strongly co-IPed with the AID from fraction 5 and tailed to fraction 12 in the LMW region. Thus, hnRNP K showed a preferential association with the AID monomer whereas hnRNP L seemed to be associated with the AID dimer (Fig. 6A). Because direct co-IP analysis also confirmed that hnRNP K and hnRNP L interacted with the C-terminal AID mutants that existed only as monomers (Fig. S2C), the absence of hnRNP L at the monomeric AID fraction is likely due to the fact that the monomeric AID–hnRNP L complex contains additional proteins.

Fig. 6.

Monomeric and dimeric AID form RNP complexes. (A) Glycerol gradient (10–60% wt/vol) sedimentation analysis of the total cell extract of HEK293T cells transfected with the CB combination of BiFC constructs of WT AID, in which HA and FLAG were fused to the C and B construct, respectively. The gradient fractions (1–12) were subjected to IP by anti-FLAG or anti-HA, followed by immunoblot detection of AID, hnRNP K, and hnRNP L in the input and IPed fractions, as indicated Top and Right. (B and C) BiFC fluorescence intensity profiles of AID–AID interaction in the glycerol gradient fractions obtained from the 17-h and 5-h runs. (D and E) Analysis of the association of RBPs with AID in the fractions obtained after glycerol gradient (10–60% wt/vol) sedimentation of the total cell extract of HEK293T cells transfected with the CB BiFC constructs of WT AID, in which AID in the B-construct was fused to FLAG. Glycerol gradient sedimentation was performed for 17 h (D) and 5 h (E), and each fraction (1–23) was subjected to IP by anti-FLAG. Both the input and IPed fractions were subjected to immunoblotting with the indicated antibodies. Fractions containing different forms of AID are indicated by bars below the blots.

We also examined whether the resolution of differential complex formation of AID could be affected in a glycerol gradient run for a shorter duration (5 h). Interestingly, the shorter centrifugation revealed three BiFC peaks (I, II, and III) (Fig. 6C). Presumably, the latter two peaks merged into the HMW region in the longer run (Fig. 6B).

The peak I at fractions 1–5 seems to contain the monomer only at fraction 1 and the monomer/dimer at fractions 2–5. The peak II at fractions 6–10 seems to contain mostly dimers complexed with RBPs whereas peak III may consist of AID multimers/oligomers (Fig. 6 C and E). In general, RBP distributions after AID co-IP were much narrower than those detected in the inputs, indicating that only small fractions of RBP form complexes with AID. In particular, SERBP1 was found to be at fractions 1–10, indicating that SERBP1 associates with the monomer as well as dimer of AID. hnRNP I, PABP, hnRNP U, and hnRNPC were largely found at peak II, indicating that these RBPs form complexes with the AID dimer. We therefore generated a comprehensive overview of the other newly identified AID-associated RBPs (hnRNP I, SERBP1, PABP, hnRNP U, and hnRNP C) by running two types of glycerol gradients (17 h and 5 h), followed by the co-IP analysis of each fraction (Fig. 6 D and E). Because hnRNP Q and hnRNP E1 did not show any interaction with AID, they could not be detected in any of the fractions subjected to AID IP (Fig. S4 A and B).

Fig. S4.

Glycerol gradient sedimentation analysis of mKG fused APOBECs. Glycerol gradient (10–60% wt/vol) sedimentation analysis of the total cell extract of HEK293T cells transfected with the CD combination of BiFC constructs of APOBEC2, APOBEC3G, APOBEC3A, and the AB combination of BiFC constructs of APOBEC1. A total of 23 fractions were collected from the top to the bottom of the gradient, and the fluorescence intensity of the fractions was plotted as arbitrary units (a.u.). The Western blot of the glycerol gradient fractions of each APOBEC is shown below its respective fluorescence profile.

In parallel, we performed a control IP experiment using gradient fractions from HEK293T cells expressing no AID, which did not show any specific IPed products (Fig. S2D). Taken together, our findings indicate that the AID monomer and dimer associate with distinct sets of RBPs and RNAs, resulting in the broad distribution of AID in the LMW and HMW regions in glycerol gradient sedimentation analysis.

Requirement of RNA-Binding Proteins for AID’s Functions.

We previously characterized hnRNP K and hnRNP L as AID cofactors for DNA cleavage and postcleavage recombination, respectively (18). We therefore examined whether CSR was affected by other newly identified AID-associated RBPs. Knockdown (KD) of SERBP1 and hnRNP U in CH12F3-2A cells reduced the CSR in accordance with their KD efficiencies (Fig. 7 A and B). Less, but significant, reduction of CSR was observed by KD of hnRNP I. We then examined the effect of the KD of these RBPs on AID-induced DNA cleavage. A ligation-mediated PCR (LM-PCR)-based double-stranded break (DSB) assay showed a significant decrease in the DNA-cleavage signal, especially in the case of SERBP1 and hnRNP U KD (Fig. 7 C and D). However, the DNA-break signal became comparable with the control in the presence of T4 polymerase, which blunts the DNA ends with high efficiency, suggesting that SERBP1 or hnRNP U deficiency causes a defect in the DNA end processing that is required for DNA repair at the recombination phase of CSR. Given that DNA cleavage itself was not affected by the KD of hnRNP L, SERBP1, hnRNP U, or hnRNP I, these RNPs are likely to be involved at the postcleavage recombination phase of CSR. For example, hnRNP L was found to be involved in S–S synapsis regulation (18). Taken together, it can be envisaged that multiple RBP factors are required to execute different steps of DNA recombination. Most importantly, CSR-specific RBPs seem to associate with the AID dimer whereas hnRNP K associates with the AID monomer as well as C-terminal mutants that are proficient in DNA cleavage and SHM. These results further strengthen our view that the RNP complex formation specific to the higher order AID structures is distinctly involved in regulating DNA cleavage and recombination (Fig. 7E and Table S1).

Fig. 7.

CSR requires SERBP1 and hnRNP U at the DNA repair phase of CSR. (A) Representative FACS profiles of IgA switching in CH12F3-2A cells treated with the indicated siRNAs. Cont., control. Cells stimulated for IgA switching by CIT or left untreated are indicated by CIT(+) and CIT(−), respectively. The percentage of switched cells after 24 h of CIT stimulation is indicated in each FACS profile. (B) IgA switching and KD efficiency plots were generated from independent experiments. Immunoblots confirmed the specific KD of SERBP1and hnRNP U in CH12F3-2A cells. (C) AID-induced double-strand DNA break detection by the LM-PCR assay in CH12F3-2A cells treated with the indicated siRNAs. T4 polymerase-treated (T4+) and -untreated (T4−) DNA samples were subjected to Sμ-specific LM-PCR, followed by Southern blot hybridization with an Sμ-specific probe. Fourfold titrations of the input DNA were performed, and the semiquantitative PCR of the GAPDH gene served as an internal control for the DNA samples used. (D) Quantitative representation of the LM-PCR signals generated, obtained by the densitometric analysis of all four lanes per sample. (E) A proposed model of AID’s structure–function relationship. AID monomers and dimers associate with a distinct set of RBPs and can form various higher order RNP structures containing yet unidentified RNAs. Such AID–RNP complexes are predicted to contribute uniquely in DNA cleavage and recombination.

Table S1.

AID-interacting RBPs that contribute to DNA cleavage and recombination

RBPs	Defects of RBP KD					Interaction with AID
	CSR	DSB	Repair	Synapse	WT	AID mutants			In presence of		Complex with AID
						ΔC	ΔN	G133P/V	High Salt	RNaseA	Monomer	Dimer
hnRNP K	✚	✚	nd	−	✚	✚	−	✚	−	−	✚	−
hnRNP L	✚	−	−	−	✚	✚	−	✚	−	−	✚	−
SERBP1	✚	−	✚	nd	✚	✚	nd	±	−	−	✚	✚
hnRNP U	✚	−	✚	_	✚	✚	nd	−	−	−	−	✚
hnRNP I	±	−	✚	nd	✚	✚	nd	±	−	−	−	✚
PABP	±	nd	nd	nd	✚	✚	−	±	−	−	−	✚
hnRNP C	−	nd	nd	nd	✚	nd	nd	±	−	−	−	✚

ΔC, C-terminal deletion; ΔN, N-terminal deletion; ✚, positive; −, negative; ±, uncertain; nd, not determined.

We also examined the gradient profiles, run under identical conditions, of the four APOBECs with their best interacting mKG combinations (Fig. S4). Unlike AID, all of the APOBEC family members showed single BiFC peaks that corresponded to their dimer positions. Although A3G showed a slightly broader distribution, the majority of the protein was located under the single BiFC peak (Fig. S4B). Strikingly, A2 produced a very sharp fluorescence peak that matched perfectly with its narrow protein distribution profile (Fig. S4A), which may indicate a higher self-interaction and a low association with other cellular proteins and RNAs in HEK293T cells.

Discussion

Based on the structural templates of A2 and A3G, several homology-modeling studies have postulated monomeric, dimeric, and tetrameric structures for AID (23, 25, 27, 43–46). However, the amino acid sequences of AID’s N and C termini do not match those of the APOBEC templates whose structures have been completely or partially resolved (27). Thus, the computational modeling studies are not informative for elucidating the molecular basis of AID’s N- and C-terminal–specific functions in DNA cleavage and recombination during SHM and CSR. Our BiFC analysis of AID revealed that both of its termini are essential for homodimer formation. Intriguingly, a head-to-tail oriented association by BiFC showed the strongest fluorescence signal, which may give some insights into the AID dimeric structure, which has not yet been crystallized. Notably, the BiFC signals of AID and A1 were much weaker than those of A2, A3G, and A3A, which showed no construct (combination) dependency. It is thus possible that AID and A1 have a unique dimerization property. The weaker intrinsic dimerization property of AID may be particularly beneficial for its ability to form multiple higher order structures through its interchangeable association with different functional cofactors. The BiFC technique itself may have stabilized the dimer and helped us to detect the complexes with higher sensitivity. Consistent with this possibility, the BiFC-positive AID dimer was found to be stable during glycerol gradient centrifugation and did not dissociate upon RNase or high-salt treatment.

The AID IP products also contained a number of RBPs, including hnRNPs that are known to form multiprotein complexes with other proteins and RNAs. These binding partners partly explain why AID forms higher order structures with broad mobilities on the glycerol gradient sedimentation. Although we previously identified hnRNP K and hnRNP L by the direct IP of AID, we could not reveal any relationship between AID’s structure and cofactor association (18). The detection of AID’s dimerization by BiFC and the separation of monomers and dimers by glycerol gradient centrifugation enabled us to demonstrate functionally important RBP interactions with AID’s various structural states. We confirmed that AID’s association with RBPs was sensitive to RNase and to high-salt treatment, suggesting involvement of RNA–protein interaction in complex formation. AID seems to form multimers that do not seem to interact with the functional RBPs identified.

We showed that hnRNP K, previously identified as a putative DNA-cleavage cofactor, preferentially associated with the monomeric AID. On the other hand, hnRNP L, which is involved in S–S synapse during CSR, associated with the AID dimer as well as the monomer. The dimer fraction of AID was associated with several other critical RBP cofactors, which seem to be involved in CSR. Among those, hnRNP U and SERBP1, whose KD significantly affected CSR, were confirmed to be involved in the postbreak repair step required for recombination. Although hnRNP C and hnRNP M were detected in larger AID–RNP complexes, they did not seem to contribute to CSR function (18).

Based on these findings, we favor the idea that the monomeric AID–hnRNP K complex is involved in DNA cleavage for SHM and CSR whereas the dimeric AID–RNP complex is involved in the CSR-associated recombination function, which absolutely requires the C terminus of AID. This scenario is consistent with the observation that C-terminal AID mutants are CSR-defective. Moreover, the requirement of both the N and C termini for stable dimer formation and the requirement of the N terminus for the associations with hnRNP K and hnRNP L indicate why both the N and C termini are critical for AID’s function. In fact, many AID mutants with N-terminal deletions in humans and mice are known to lose both CSR and SHM. In contrast, all of the C-terminal mutants that failed to form dimers retained their SHM activity, suggesting that monomeric AID (without the C terminus), which interacts with hnRNP K, is sufficient for the DNA-cleavage function. Furthermore, CSR seems to require the dimeric structure of AID, which forms larger complexes with hnRNP L, hnRNP U, hnRNP I, SERBP1, or other factors to fully execute recombination.

Consistent with this view, the AID mutant G23S (47), which has compromised SHM but intact CSR activity, and the mutant S3A (48, 49), which has a higher CSR activity than WT, both had intact dimerization ability and also migrated to the HMW region upon glycerol gradient centrifugation (Fig. S5). Curiously, mutations at the putative RNA-binding motif (G133V and G133P) (50) resulted in loss of the BiFC signal and an altered association with the RBPs (Fig. S6) that are important for DNA cleavage and recombination for CSR. Despite their lesser expression level compared with WT AID, both G133V and G133P mutants showed a stronger association with hnRNP K and hnRNP L than with hnRNP U, hnRNP I, SERBP1, or PABP. In particular, these AID mutants showed a nearly complete loss of interaction with hnRNP U (Fig. S6C). These results suggested that these mutants are unable to form dimers or to interact with CSR-associated AID–RNP complexes. Because AID’s dimerization and RNP complex formation are critical for CSR, the CSR defect of these mutants can be explained by their inability to associate with the RBPs and RNA essential for CSR.

Fig. S5.

S3A and G23S mutants have intact dimerization ability. (A, Top) Schematic representation of the S3A and G23S AID mutants. (Bottom) Calculated relative frequencies of CSR and SHM of the S3A and G23S mutants. (B, Left) Frequency histograms from the FACS analysis of HEK293T cells transfected with the CB combination of mKG-BiFC constructs indicated above each plot. Mean fluorescence intensity values of BiFC-positive cells obtained from WT AID, S3A, and G23S transfectants are indicated inside the histogram profiles. (Right) Western blot analysis of the WT AID and mutants in the cell lysates prepared from HEK293T cells transfected with the indicated mKG-BiFC constructs. (C) Glycerol gradient (10–60% wt/vol) sedimentation analysis of the total cell extract of HEK293T cells transfected with the mKG-BiFC constructs of S3A and G23S mutants. The fluorescence intensity of the fractions was measured and plotted, and the AID distribution was analyzed by immunoblotting with an anti-AID antibody. The fluorescence intensity profiles of Mock and WT AID (150 mM NaCl) from Fig. 4 are included as a reference.

Fig. S6.

Loss of BiFC signal and RBP association in the G133P and G133V AID mutants. (A) Schematic representation of the G133P and G133V AID mutants. (B) FACS analysis of HEK293T cells transfected with the pairwise combinations of BiFC constructs indicated above each FACS profile. (C) Analysis of the association of RBPs with AID using the total cell extract of HEK293T cells transfected with the mKG-CB combination of WT AID and the G133P and G133V mutants. In each case, the FLAG epitope was fused to AID in the B-construct for IP. Cell lysates were IPed with anti-FLAG, and the coprecipitated proteins were analyzed by immunoblotting with the antibodies for each RBP, shown on the right.

It is critical to understand the functional properties of the AID C terminus because its loss does not affect DNA cleavage but abolishes AID’s dimerization and CSR recombination function. The extreme C terminus of A1 is also important for dimerization; the A1 dimer associates with the cofactor A1CF and forms the functional APOB–mRNA editing complex (38, 39). A1 not only associates with A1CF or RBM47, but also interacts with other RBPs and hnRNPs (AB, Q, C, and K) (17, 51–54). These cofactors regulate APOB mRNA editing, either positively or negatively, and IL8 mRNA stability (54). The functional relevance of all of the A1-associated RBPs is not yet fully understood, but they may participate in a variety of RNA-processing events by A1. In fact, large-scale sequence analyses by two independent groups recently revealed novel A1-specific C-to-U editing signatures in the UTR of a dozen mRNAs (55, 56). Therefore, the dimeric form of AID, like that of A1, may form an as-yet unidentified RNA-editing complex(es) in association with its CSR-specific RBP cofactors, such as hnRNP L, hnRNP U, SERBP1, and others. Indeed, we found that dimeric AID migrated to the HMW region with distinct BiFC peaks and interacted with different RBPs. Moreover, the identification of A1CF-like hnRNP cofactors for AID suggests that AID can form functional RNP complexes similar to A1.

Interestingly, A3A, which is known to mutate the genome heavily and is implicated in breast cancer development, was recently found to edit specific RNAs in a cell type-specific manner (57, 58). Although the crystal structure study of A3A suggested a cooperative dimerization model for its action on DNA, the structural basis of A3A’s RNA-editing and RNP complex-forming abilities is still unknown (37, 59).

Because A1 and other APOBEC members have been suggested to exist also as monomers in cells, it is likely that AID exists in an equilibrium between monomer and dimer. The low affinity of AID’s dimerization probably helps it transform dynamically from one state to another and thus to form distinct function-specific AID–RNP complexes. In the present study, we demonstrated that specific AID-associated RBPs have the potential to direct AID’s function toward either DNA cleavage or recombination. Because CSR recombination can be regulated at either the S–S synapse or the DNA end-processing repair phase (6, 11, 12), we speculate that at least two independent mRNAs are involved (7, 11, 60). Thus, it is reasonable to propose that distinct RBPs are involved in AID–RNP complex formation and that their KD affects different stages of CSR. Currently, it is not clear whether these RBPs are part of a single megacomplex or of isolated complexes with AID because the loss of a potential factor by KD might affect the stability of the other components in an AID–RNP complex. Nevertheless, we think that specific RBPs in distinct AID–RNP complexes assist AID in recombination-relevant RNA editing, which may lead to generation of novel proteins involved in the S–S synapse and end-joining phases. Further study will be required to fully explore the DNA-cleavage and recombination-specific AID–RNP complexes and their target RNAs. It is also possible that AID possesses an as-yet unknown novel mode of translational regulation, by which a protein is up- or down-modulated when its coding mRNA becomes bound to a specific AID–RNP complex.

Materials and Methods

Bimolecular Fluorescence Complementation Assay.

The plasmids expressing the AID–mKG fusion constructs were transferred either individually or in pairwise combinations (Fig. S1) into HEK293T cells in 12-well plates. FACS analysis was performed 48 h after transfection in a BD Biosciences FACSCalibur cytometer. The mKG-BiFC–positive cells were detected in the FITC channel, and the data were analyzed in the live gate using CellQuest software (BD Biosciences). The mKG-BiFC assay was described previously (29, 33), and the stepwise details are also available in the manufacturer’s instruction manual (Code No. AM-1100; MBL).

Glycerol Gradient Sedimentation Assay.

HEK293T cells were plated on 10-cm dishes and transfected with a pair of AID–mKG fusion constructs. In parallel, mock transfection was performed to prepare the control cell extract. Forty-eight hours after transfection, the cells were lysed in a lysis buffer [30 mM Tris⋅HCl (pH 7.4), 150 mM NaCl, 10% (vol/vol) glycerol, 1% Triton X-100, 0.05% Na-deoxycholate, and 5 mM EDTA] supplemented with protease inhibitors (Roche). To analyze the effect of RNase A treatment, HEK293T cell extracts were preincubated at room temperature for 15 min with RNase A (130 μg/mL). To analyze the effect of high-salt treatment, the cells were lysed with the lysis buffer containing 500 mM NaCl. Next, 1 mL of clarified cell extract was layered on top of an 11-mL 10–60% (wt/vol) glycerol gradient prepared in the same lysis buffer using a Gradient Master. The gradient was centrifuged (40,000 rpm; 17 h or 5 h at 4 °C) in a P40ST rotor (Hitachi Ultracentrifuge CP 70MX), and 0.5-mL fractions were collected in a BioComp gradient fractionator. The fluorescence intensity of each fraction was measured with an EnVision Multilabel Plate Reader (Perkin-Elmer). The fractions were also subjected to various analyses, such as immunodetection and co-IP. To analyze the endogenous AID and overexpressed AID in B cells, a CH12F3-2A clone was used that constitutively expresses an AID–GFP fusion protein. Cells were stimulated by CD40L, IL4, and TGFβ (CIT) for 24 h to induce endogenous AID expression (61). The cells were then lysed and subjected to gradient centrifugation and fraction collection, as described above.

Immunoprecipitation (co-IP) from Total Cell Extract.

HEK293T cells were plated on six-well plates and transfected with the desired sets of AID–mKG fusion constructs. Cells were harvested 36 h after transfection and lysed as described above. Clarified cell lysates with an equal amount of total protein were incubated with either anti-FLAG M2 agarose (Sigma) or EZview Red anti-HA affinity gel (Sigma), using standard IP protocols. The IP complex was washed three to four times in the same buffer containing 150 mM NaCl. The bound proteins were eluted into SDS-sample buffer by boiling the beads. The eluted proteins were subjected to SDS/PAGE analysis and to immunoblotting with the specified antibodies. RNase- and high salt-treated cell extracts were prepared as described above, and the IP complex in each case was washed with buffer containing either RNase or 500 mM NaCl.

Construction of mKG fused AID, CSR assay, gene knockdown in CH12F3-2A cells (62, 63), and DNA break analysis (64, 65) are described in SI Materials and Methods. Antibodies used are listed in Table S2.

Table S2.

List of antibodies used in this study

Antibody	Company	Catalog no.
mAID	eBioscience	K10211
Tubulin	Calbiochem	CP06
hnRNP K(d-6)	Santa Cruz	sc-28380
hnRNP L(d-5)	Santa Cruz	sc-48391
hnRNP U	Bethyl	A300-689A
hnRNP C1/C2	Santa Cruz	sc-32308
PTBP1	MBL	RN001P
SERBP1	MBL	RN056PW
PABP	Santa Cruz	sc-28834
hnRNP M1-4(1D8)	Santa Cruz	sc-20002
PCBP1	Abcam	ab74793
hnRNP Q[I8E4]	Abcam	ab10687
APOBEC1	Santa Cruz	sc-11739
APOBEC2	Santa Cruz	39985
APOBEC3G	Abcam	38604
FLAG	Sigma	F3165
HA-Tag (6E2)	Cell Signaling	2999S

SI Materials and Methods

Cells and Culture Conditions.

Human embryonic kidney (HEK) 293T cells were cultured in DMEM (Invitrogen) supplemented with 10% (vol/vol) FBS. These cells were used for the BiFC assay. For the CSR assay, primary B cells were isolated from the spleens of 6- to 8-wk-old AID-deficient (AID^−/−) mice using the BD IMag B Lymphocyte Enrichment Set-DM (557792) and cultured at 1.0 × 10⁶ cells per mL in complete RPMI medium (1). The mouse B-cell lymphoma cell line CH12F3-2A expressing Bcl2 (62, 63) was used for the CSR assay and gene knockdown experiments.

Monomeric Kusabira Green Fusion Constructs.

The mKG fusion constructs of AID and APOBECs were designed as depicted in Fig. 1 and Fig. S1. We generated four constructs (A, B, C, and D) encoding the protein of interest (AID/APOBEC) fused to either the N- or C-terminal fragments of mKG protein, mKG.N (168 aa), and mKG.C (51 aa). There was a 24-aa glycine-rich linker sequence between the mKG fragment and the protein of interest to provide greater mobility to the fusion protein. Primer sequences were designed to fuse AID/APOBEC in-frame with the mKG fragments at either the N or C terminus. The coding sequences for AID/APOBEC were mostly cloned into the KpnI and EcoRI sites of the respective mKG vectors (Code No. AM-1100; MBL). For co-IP experiments, AID was tagged at its C terminus with a FLAG or HA epitope.

CSR Assay in Primary B Cells.

To analyze the CSR efficiency of WT and mutant AID, primary B lymphocytes were isolated from AID-deficient mice as described above, and cultured at 1.0 × 10⁶ cells per mL in complete RPMI medium containing 25 μg/mL LPS and 7.5 ng/mL IL-4, to undergo class switching to IgG1. For the retroviral transduction, the cells were preactivated before infection by culturing in the presence of LPS and IL4 for 48 h. The retroviral supernatants were prepared, and spleen cell infection was performed using standard protocols. The IgG1 expression was examined by flow cytometry by staining the cells with biotinylated anti-IgG1 (Pharmingen) and APC-conjugated streptavidin (eBioscience) on day 3. The IgG1 switch efficiency was calculated from the infected GFP-positive cells in the live gate.

Gene KD in CH12F3-2A Cells and DNA Break Assay.

To knock down the expression of specific genes of interest, chemically modified Stealth siRNA oligonucleotides (Invitrogen) were introduced into CH12F3-2A cells using the Nucleofector 96-well electroporation system (Lonza) (62, 63). After electroporation, the cells were cultured for 24 h and then stimulated by CIT for another 24 h to induce IgA switching. Cells were stained with FITC-conjugated anti-IgM (eBioscience) and PE-conjugated anti-IgA (eBioscience) and subjected to FACS analysis using a FACSCalibur (BD Biosciences). The IgM-to-IgA switching efficiency was examined in the live cell population.

For the LM-PCR–based DSB assay, the cells were stimulated for CSR as described above, and the live cells were embedded in low-melt agarose plugs and processed for linker ligation as described previously (64, 65). The samples were treated with T4 polymerase (Takara) before linker ligation, and the ligated DNA was subjected to GAPDH DNA PCR analysis to adjust the DNA input before LM-PCR. Threefold dilutions of the input DNA were amplified by KOD-FX-Neo polymerase (Toyobo). The PCR products were separated by electrophoresis on 1% agarose gels and validated by Southern blotting using a 5′ Sμ probe; the primers and probe sequences were the same as described previously (64).