Phosphorylation Directly Regulates the Intrinsic DNA Cytidine Deaminase Activity of Activation-induced Deaminase and APOBEC3G Protein*

The beneficial effects of DNA cytidine deamination by activation-induced deaminase (AID; antibody gene diversification) and APOBEC3G (retrovirus restriction) are tempered by probable contributions to carcinogenesis. Multiple regulatory mechanisms serve to minimize this detrimental outcome. Here, we show that phosphorylation of a conserved threonine attenuates the intrinsic activity of activation-induced deaminase (Thr-27) and APOBEC3G (Thr-218). Phospho-null alanine mutants maintain intrinsic DNA deaminase activity, whereas phospho-mimetic glutamate mutants are inactive. The phospho-mimetic variants fail to mediate isotype switching in activated mouse splenic B lymphocytes or suppress HIV-1 replication in human T cells. Our data combine to suggest a model in which this critical threonine acts as a phospho-switch that fine-tunes the adaptive and innate immune responses and helps protect mammalian genomic DNA from procarcinogenic lesions.

Activation-induced deaminase (AID) 2 and APOBEC3G (A3G) are the archetypal members of a larger family of polynucleotide cytidine-to-uridine deaminases with critical functions in adaptive and innate immunity (1,2). All mammals have AID, apolipoprotein B mRNA-editing catalytic subunit 1 (APOBEC1), APOBEC2 and variable numbers of APOBEC3s (A3) ranging from one in rodents to seven in most primates, including humans (A3A/B/C/D/F/G/H) (3). AID, A1, A3A, A3C, and A3H are single domain proteins with one zinc-coordinating active site, whereas several A3s, including rodent A3 and human A3B, A3D, A3F, and A3G, are double domain proteins with two zinc-coordinating motifs (both are conserved, but typically only one is active). Atomic structures for the catalytic domain of human A3G(4 -7) and APOBEC2 (8) are available, and these enable structure-function studies and homology models.
Considerable effort has been devoted to understanding the multiple mechanisms that combine to regulate AID and A3G activity. First, the transcription of each of these genes is tissuespecific, with AID being expressed predominantly in B lymphocytes and A3G in most cell types (9 -11). Second, AID and A3G transcription levels are up-regulated by distinct signal transduction pathways (STAT/NFB for AID and NFAT/IRF for A3G) (12,13). Third, AID expression is regulated by at least one microRNA, miR-155 (14,15). Fourth, both proteins are predominantly cytoplasmic, with AID having additional nuclear import and export capabilities (16 -20). Finally, both proteins are subject to proteasome-dependent degradation, AID in the nuclear compartment (21) and A3G in the cytoplasmic compartment (22)(23)(24).
E. coli Mutation Assays-The rifampicin resistance assay has been published (41,42). BW310 E. coli cells were transformed with pTrc99a-AID or pTrc99a-A3G expression constructs and plated on ampicillin-containing media. Four individual colonies were picked and seeded into media containing 1 mM isopropyl 1-thio-␤-D-galactopyranoside and 100 g/ml ampicillin. After shaking overnight at 37°C, the cultures were plated on ampicillin media to obtain viable cell counts and to 100 g/ml rifampicin-containing media for mutational frequency. AID and A3G expression levels were determined by Western blot analysis using antibodies against AID (EK25G9, Cell Signaling Technology, Inc.) or A3G (#10201 rabbit anti-A3G polyclonal serum provided by J. Lingappa through the AIDS Research and Reference Reagent Program).
DNA Binding Assays-Protein samples prepared from E. coli as above were used for AID DNA binding reactions. 6 pmol of purified protein was diluted serially 1:2 and mixed with 0.5 pmol 32 P-labeled oligo (5Ј-ATT ATT ATT ATT CCA ATG GAT TTA TTT ATT TWR CTA TTT ATT T) in binding buffer (10 mM HEPES (pH 7.6), 10% glycerol, 100 mM KCl, 10 mM MgCl 2 , 100 M EDTA, 500 M DTT). Reactions were incubated for 30 min at 37°C before separation on a 7% Tris borate-EDTA acrylamide gel. The gel was dried and imaged by phosphoimager (Storm, Molecular Dynamics).
Protein samples purified from human cells from the above were used for A3G DNA binding reactions. 25 pmol of purified protein was diluted serially 1:2, and each dilution was mixed with 1 pmol fluorescein-labeled oligo (5Ј-ATT ATT ATT ATT CCA ATG GAT TTA TTT ATT TAT TTA TTT ATT T-fluorescein) in binding buffer. The reactions were incubated for 30 min at 37°C before separation on a 7% Tris borate-EDTA acrylamide gel. The free and bound oligos were then detected by fluorescence imaging (FLA-5000, Fuji).
Oligo-based Deaminase Assays-Protein samples purified from human cells as above were used for A3G deaminase activity reactions. Starting with 1.2 pmol purified protein, 2-fold serial dilutions were made and mixed with 1 pmol of substrate oligo (5Ј-ATT ATT ATT ATT CCA ATG GAT TTA TTT ATT TAT TTA TTT ATT T-fluorescein), 0.1 g/l RNase A (Qiagen), and 0.001 units/l uracil DNA glycosylase (NEB). The reactions were incubated at 37°C for 2 h, and then NaOH was added to 100 M before incubating at 90°C for another 30 min. The reactions were separated on a 16% Tris/urea-acrylamide gel and visualized by fluorescence imaging (FLA-5000, Fuji).
HIV Restriction Assays-CEM-SS and CEM-GFP (courtesy of M. Malim) were maintained in RPMI 1640 (Invitrogen) supplemented with 10% FBS (Hyclone). Stable cell lines expressing pEGFP-N3, pEGFP-N3-A3G, or mutants were generated in the permissive cell line CEM-SS by electroporating linearized DNA and selecting for stable integrants with 1 mg/ml G418 (Mediatech) as described (38). Clones were confirmed to have similar expression levels by Western blot analysis using an antibody against A3G (rabbit polyclonal raised against a C-terminal peptide). Virus was produced by transfecting HIV-1 provirus using TransIT-LT1 (Mirus) HEK293T cells maintained in DMEM (Invitrogen) supplemented with 10% FBS (Hyclone). 48 h after transfection, virus-containing supernatants were filtered with a 0.45-m filter. Viruses were then titered using the CEM-GFP reporter cell line as described (38). Spreading infections were initiated by adding virus to CEM-SS stable cell lines at a multiplicity of infection of 0.05. Supernatants from infected cultures were collected at 2-to 4-day intervals and added to CEM-GFP. After 48-hours, the cells were fixed in 4% paraformaldehyde and analyzed for GFP expression by flow cytometry (Quanta SC MPL, Beckman Coulter). Procedures for the detection of A3G in producer cells and viral particles have been described (44).
Class-switch Recombination Assays-All experiments were conducted in accordance with the University of Minnesota Animal Care and Use Committee guidelines. The C57BL/6 AID Ϫ/Ϫ mice have been described (45). Ex vivo CSR assays were conducted by purifying resting B-cells from spleen by magnetic sorting (130-090-862, Miltenyi Biotec). Isolated B-cells were then cultured in RPMI supplemented with 10% FBS, 50 ng/ml IL-4, and 50 g/ml LPS. After 48 h, the media were replaced with transducing viral supernatant supplemented with 20 mM HEPES and 16 g/ml polybrene and centrifuged (600 ϫ g, 2 h, 30°C). The cells were then resuspended into fresh media containing IL-4 and LPS and cultured for 4 days. Efficiency of switching to IgG1 was determined by staining with anti-IgG1-PE (BD Biosciences) and analyzed by flow cytometry (FACSCanto II, BD Biosciences).

RESULTS
AID-Thr-27, A3G-Thr-32, and A3G-Thr-218 Are Homologous and Located within a Region of High Sequence and Structural Conservation-Prior studies demonstrated phosphorylation of AID-Thr-27 in vivo and in vitro by mass spectrometry and radiolabeling(31, 33, 35) and A3G-Thr-32 by immunoblot-ting (34). We noted that these two threonines are homologous to A3G-Thr-218, whose high-resolution structures have shown to be located within a solvent-accessible loop (4 -7) (Fig.  1A). This threonine anchors a conserved motif that matches a consensus PKA phosphorylation site (R-H/R-X-T) (46) (Fig.  1B). Notably, nearly all AID/A3 family members have homologous threonine or serine residues at this exact position (Fig.  1C). Rare exceptions are only apparent in specific mammalian lineages (carnivores and rodents) or in redundant or inactive domains (most alleles of human A3H are unstable) (47). In the catalytic domain of A3G, the first arginine in this motif (Arg-215) is located adjacent to the catalytic glutamate, and it has been implicated in binding substrate single-stranded DNA (5)(6)(7). Taken together, these observations, and particularly the high level of conservation and the structural positioning next to the active site, suggest that phosphorylation and dephosphorylation may serve as a posttranslational switch that helps control the DNA deaminase activity of these mutagenic enzymes.
PKA and CaMKII Phosphorylate A3G-Thr-218 in Vitro-AID-Thr-27 and A3G-Thr-32 can be phosphorylated by PKA (31,(33)(34)(35). To determine whether these observations extend to A3G-Thr-218, we asked whether recombinant PKA could phosphorylate a peptide representing the soluble loop in which this residue resides, VRGRHET 218 YLCYE. We found that PKA could readily phosphorylate this peptide but not a T218A mutant derivative that is otherwise identical (Fig. 1D). Similarly, CaMKII, which also phosphorylates R-X-X-T motifs (46), was able to phosphorylate the A3G-Thr-218 peptide but not the alanine mutant derivative. Both enzymes were also able to phosphorylate a serine in a control peptide (Kemptide). These data demonstrate that A3G-Thr-218 is a suitable substrate for at least two kinases, PKA and CaMKII.
Phospho-mimetic Mutations Inhibit DNA Cytidine Deaminase Activity-To address whether phosphorylation is capable of attenuating the DNA cytidine deaminase activity of AID and A3G, phospho-null and phospho-mimetic derivatives of these proteins were tested in an E. coli-based activity assay. The rifampicin-resistance (Rif R ) mutation assay has been used extensively to assess intrinsic DNA cytidine deaminase activity (41,42). Consistent with prior reports, AID and A3G triggered 3-and 4-fold increases in the median Rif R mutation frequency compared with catalytically inactive controls, AID-E58Q and A3G-E259Q (Fig. 2, A and B). In comparison, phospho-mimetic AID-T27E and A3G-T218E proteins also showed greatly reduced activity approaching background levels. Phospho-null alanine mutants showed slightly higher levels of mutator activity. Mutation of another predicted surface threonine in AID (Thr-140) or the homologous threonine in the non-catalytic N-terminal domain of A3G (Thr-32) had little effect. All proteins expressed similarly in E. coli, indicating that these data are not due to variable protein expression levels (lower panels in Fig. 2, A and B).
To ask whether these observations extended to A3G purified from human cells, we used a DNA oligonucleotide deamination assay optimized to measure catalytic activity. As expected, wild-type A3G catalyzes dose-dependent cytidine-to-uridine deamination of labeled deoxy-oligonucleotide substrates, which following uracil excision and NaOH-mediated phosphodiester backbone cleavage, is detected as a shorter DNA fragment (Fig. 3, A, C, and E). As anticipated from the E. coli mutation experiments, the A3G phospho-mimetic variant T218E showed considerably lower levels of catalytic activity. Interestingly, the A3G phospho-null variant T218A showed significantly elevated levels of catalytic activity consistent with a proportion of the wild-type protein being already phosphorylated (and thereby inactivated) in HEK293T cells. Taken together, the E. coli and the purified protein activity data indicate that phosphorylation of the conserved threonine, AID-Thr-27 or A3G-Thr-218, may serve to attenuate the intrinsic DNA deaminase activity of these proteins (supported further by HEK293T cell extract data in supplemental Fig. S1).
DNA Binding Is Unaffected by Phospho-mimetic Substitutions-To ask whether the diminished catalytic activity of the phospho-mimetic substitution mutants is due to defective ssDNA binding, we tested the ssDNA binding ability of AID and A3G in electrophoretic mobility shift assays. A3G-myc-his used in the deaminase reactions above was used for ssDNA binding experiments. Purified protein was diluted serially, incubated with a fluorescently labeled oligo, and fractionated on a native polyacrylamide gel. As expected, A3G and the catalytic mutant E259Q bound ssDNA in a dose-dependent manner (Fig. 3, B, D, and E) (48). Likewise, A3G-T218A and A3G-T218E had nearly identical ssDNA binding abilities, which were indistinguishable from the wild-type enzyme (Fig. 3, B, D, and E).
Similar EMSA experiments were done with wild-type AID and mutant derivatives, but the sensitivity of the assay had to be increased by using a radiolabeled ssDNA oligo substrate. Again, the wild-type and the phospho-null and phospho-mimetic variants produced near identical mobility shifts (Fig. 3, F, G, and H). As a control to demonstrate the specificity of AID for ssDNA, an AID-R24E mutant was analyzed in parallel and shown to be defective in DNA binding. This arginine is conserved and homologous to A3G-Arg-215, which NMR chemical shift perturbation and mutagenesis experiments have implicated strongly in DNA binding (5)(6)(7). Additional EMSA data can be found in supplemental Fig. S2. Overall, these EMSA results clearly show that phospho-mimetic substitutions in A3G and AID do not cause visible decreases in the ability of each protein to bind ssDNA.
Mutants of AID and A3G Localize Normally within Living Cells-The subcellular localization of AID/APOBEC family members has been well studied (16 -20). A3G is predominantly cytoplasmic. AID is also mostly cytoplasmic, but it is imported into the nuclear compartment by an importin-␣ pathway and exported back to the cytoplasm by a CRM1 pathway. To ask whether our phospho-null or phospho-mimetic mutants retain normal, steady-state subcellular distributions, we performed a series of AID/A3G-GFP localization studies in living HeLa cells. No detectable alteration in the steady-state cytoplasmic distribution of A3G-EGFP, AID-EGFP, or their mutant derivatives was detected (Fig. 4, A and B). Moreover, experiments done in the presence and absence of the CRM1 inhibitor leptomycin B indicated that the nuclear import and export activities were also intact for all AID-EGFP constructs. These data therefore indicate that A3G, AID, and their mutant derivatives are capable of interacting with the cellular factors responsible for localization and, furthermore, that AID is able to enter the nucleus, where it will have the opportunity to access the immunoglobin locus, its physiologic DNA deamination target.
AID-T27E Is Defective for Class-switch Recombination-One of the physiological functions of AID is catalyzing cytidine-to-FIGURE 2. Intrinsic DNA cytidine deaminase activity of AID and A3G constructs. A and B, results from E. coli-based Rif R mutation assays, with each X representing data from an independent culture. Median mutation frequencies are indicated by horizontal bars and numbers. Also shown are Western blot analyses of AID or A3G constructs from representative cultures with a nonspecific band (NSB) as a loading control.
uridine deamination events in immunoglobulin heavy chain gene switch region DNA and thereby triggering additional DNA repair processes that ultimately manifest as antibody isotype switch recombination (1,45). Therefore, as a functional test of AID activity, we assayed the phospho-null and phosphomimetic mutants in an ex vivo B-cell CSR system (30 -33). Naïve splenic B-lymphocytes were isolated from AID-deficient mice; cultured in the presence of IL-4 and LPS to induce cell division and isotype switching from IgM to IgG1; transduced with retroviruses encoding AID-IRES-EGFP, mutants of AID, or EGFP alone; and 4 days later subjected to flow cytometry for IgG1-positve cells. Mock-(not shown) or EGFP-virus transduced cells remained AID-defective and showed no class switching to IgG1 (Fig. 5). Also, as expected, wild-type AID expression complemented the endogenous defect and enabled class switching to IgG1 in a significant proportion of cells (rep- resentative plots in Fig. 5A and average of four experiments in B). Conversely and surprisingly, neither T27A nor T27E was capable of promoting the switch to IgG1 despite similar protein expression levels (Fig. 5, A, B, and C).
The T27E result was anticipated on the basis of the lower level of catalytic activity, but not DNA binding or localization activities, elicited by this mutant. However, the aforementioned data on AID-T27A showing normal deaminase, ssDNA binding, and cellular localization/trafficking activities strongly suggested that this variant would be capable of normal or even elevated CSR levels, in stark contrast to the defect in CSR shown here. This result makes the CSR data set more difficult to interpret. One possibility, noted previously (31), is that phosphorylation of Ser-38 may depend first upon phosphorylation of Thr-27. An alternative may be that each of these residues has a distinct mechanistic contribution to CSR, with our studies favoring a role for Thr-27 in regulating catalysis.
A3G-T218E Lacks HIV-1 Restriction Activity-A3G potently inhibits HIV-1 replication by blocking reverse transcription and deaminating viral cDNA cytosines to uracils (2). This antiviral activity is most evident in HIV-1 lacking viral infectivity factor (Vif), a small basic protein that triggers A3G degradation. Thus, a rigorous test of the functional activity of A3G is whether it suppresses the spreading infection of Vif-deficient HIV-1 (38,39,49). We therefore created a panel of CEM-SS T cell lines stably expressing wild-type A3G-EGFP, an E259Q catalytic mutant control, a phospho-null T218A construct, or a phospho-mimetic T218E protein. As anticipated from prior studies, wild-type A3G completely suppressed the replication of Vifdeficient HIV-1, and its strong antiviral effect was largely dependent upon the integrity of the catalytic glutamate Glu-259 (50, 51) (Fig. 6A). A3G-T218A showed wild-type levels of restriction consistent with full or elevated levels of enzymatic activity. In contrast, A3G-T218E failed to prevent the replication of Vif-deficient HIV-1. However, this mutant protein did cause reproducible delays in peak viral replication consistent with severely attenuated but not fully defective catalytic activity. As additional controls, N-terminal A3G-T32A or T32E substitutions had no discernable effect, and all cell lines supported similar levels of Vif-proficient HIV-1 spreading infection ( Fig.  6A and data not shown). It is notable that, although we were able to confirm A3G-Thr-32 phosphorylation by mass spectrometry, we found no differences in the subcellular localization, HIV restriction capacity, or Vif susceptibility in alanine-or glutamate-substituted derivatives ( Fig. 4 and data not shown).
An additional possibility is that A3G-T218E may not restrict HIV-1 because it is not efficiently packaged into budding viri-  ons. To test and eliminate such a possibility, we harvested virus produced from HEK293T cells expressing A3G and mutants thereof and blotted for the presence of A3G in these viral particles. We found no significant difference in the ability of any of the mutants to get into virions as compared with wild-type A3G (Fig. 6C).

DISCUSSION
The AID/APOBEC family of cytidine deaminases is an important facet of the adaptive and innate immune responses in humans. However, their mutagenic activity must be tightly regulated to prevent potentially detrimental off-target effects. Regulation of these proteins has been described at multiple lev-els, including transcription, microRNAs, cytoplasmic localization, proteasomal degradation, and phosphorylation (see introduction). Here we describe a novel phosphorylation regulatory mechanism capable of attenuating the intrinsic deaminase activity of AID and A3G. In this study, we demonstrate that phospho-mimetic substitution of a highly conserved threonine renders these proteins inactive in several independent assays. We show that ssDNA binding ability and steady-state subcellular localization (and for AID, also trafficking) are unaffected, indicating that these proteins are structurally intact. In functional assays, this modification prevents AID from facilitating CSR and A3G from restricting HIV-1⌬Vif replication. It is intriguing that two neighboring phosphorylation sites can have such contrasting effects on the function of AID, with Ser-38 phosphorylation enabling interaction with replication protein A and allowing CSR and somatic hypermutation, and Thr-27 phosphorylation rendering the protein inactive. This begs the question of how PKA is regulated to distinguish between these neighboring residues. Further studies are warranted to better understand these posttranslational regulatory events and investigate the possible involvement of other Ser/Thr kinases that can also recognize PKA consensus motifs, such as CaMKII described here.
The obvious utility of posttranslational regulation by phosphorylation is 2-fold (illustrated by the model in supplemental Fig. S3). First, a threonine-or serine-phosphorylated DNA deaminase would possess a low level of DNA deaminase activity and pose less of a threat to genomic DNA. Genomic DNA integrity is further ensured by the fact that AID, A3G, and many other A3 proteins are predominantly cytoplasmic. Second, signal transduction pathways, which are critical for both adaptive and innate immune responses, could readily switch on DNA deaminase activity by triggering the removal of the phosphate group (phosphatase or phosphotransferase activity). This would ensure an expedited immune response that could be further bolstered by up-regulating AID or A3 expression at the transcriptional and/or translational levels.
We propose that the posttranslational modification of AID and the A3 proteins by phosphorylation provides a means of directly controlling the intrinsic DNA cytidine deaminase activity of these proteins (supplemental Fig. S3). It is likely that this mechanism will be conserved in vertebrates because residues homologous to AID-Thr-27, A3G-Thr-32, or Thr-218 are apparent in almost all other known polynucleotide cytidine deaminase family members (Fig. 1). It is further possible that defects in these signal transduction pathways may manifest as immunodeficiency syndromes (overphosphorylated protein), autoimmune diseases (underphosphorylated protein), and/or carcinogenesis (underphosphorylated protein), especially in combination with other regulatory defects.