In Vivo Analysis of Aicda Gene Regulation: A Critical Balance between Upstream Enhancers and Intronic Silencers Governs Appropriate Expression

The Aicda gene encodes activation-induced cytidine deaminase (AID). Aicda is strongly transcribed in activated B cells to diversify immunoglobulin genes, but expressed at low levels in various other cells in response to physiological or pathological stimuli. AID’s mutagenic nature has been shown to be involved in tumor development. Here, we used a transgenic strategy with bacterial artificial chromosomes (BACs) to examine the in vivo functions of Aicda regulatory elements, which cluster in two regions: in the first intron (region 2), and approximately 8-kb upstream of the transcription start site (region 4). Deleting either of these regions completely abolished the expression of Aicda-BAC reporters, demonstrating these elements’ critical roles. Furthermore, we found that selectively deleting two C/EBP-binding sites in region 4 inactivated the enhancer activity of the region despite the presence of intact NF-κB-, STAT6- and Smad-binding sites. On the other hand, selectively deleting E2F- and c-Myb-binding sites in region 2 increased the frequency of germinal-center B cells in which the Aicda promoter was active, indicating that E2F and c-Myb act as silencers in vivo. Interestingly, the silencer deletion did not cause ectopic activation of the Aicda promoter, indicating that Aicda activation requires enhancer-specific stimulation. In summary, precise regulation of the Aicda promoter appears to depend on a coordinated balance of activities between enhancer and silencer elements.


Introduction
Activation-induced cytidine deaminase (AID), which is encoded by Aicda, is central to antigen-induced immunoglobulin (Ig) diversification, that is, somatic hypermutation (SHM) and class-switch recombination (CSR) in activated B cells [1,2]. AID triggers the induction of nicking cleavage at the Ig loci [3,4]. Although the target specificity of AID is limited to Ig genes, in which SHM and CSR take place, AID can target many other genes, including proto-oncogenes, although with much lower frequencies [5][6][7][8][9]. It has been suggested that this type of nonphysiological, off-target attack by AID is involved in tumorigenesis of not only B cells, but also other cell lineages [10]. Supporting this idea, artificially overexpressing AID in transgenic mice causes tumors in non-B cells such as T lymphoma, lung tumor, and hepatoma [11,12]. Moreover, Aicda knockout drastically delays tumor development in animal models of plasmacytoma, which is associated with suppressed Ig-Myc translocation [13,14]. AID is thus suspected to be an endogenous mutagenic enzyme.
Given this self-mutating activity of AID, one would expect Aicda to be regulated very strictly to avoid any leaky expression. Indeed, several reports have shown that strong AID expression is virtually restricted to germinal-center B cells in vivo [15][16][17]. However, there is growing evidence that cells other than activated B cells can express AID when exposed to strong stimulation. In particular, AID expression is induced in gastric epithelial cells by Helicobacter pylori infection, providing insight into the mechanisms of pathogenassociated tumorigenesis-namely, that AID may act as a mutagen in inflammation-associated tumor development [18]. Other tumorigenic pathogens, including the EB, hepatitis C, and HTLV-1 viruses, are reported to induce or enhance AID expression [19][20][21][22]. The NF-kB pathway appears to be involved in AID's induction in H. pylori-infected gastric epithelial cells, suggesting that the inflammatory signals associated with infection by this pathogen may be involved in inducing AID [18].
Under physiological conditions, AID is also expressed in immature B cells and in some activated T cells at a level one to two orders of magnitude lower than that in germinal-center B cells; the role of such low amounts of AID is unclear [16,17,23,24].
Aicda is transiently expressed in a fraction of activated T cells, including a subset of T cells that produces IL-10, suggesting that Aicda-inductive stimuli correlate with the development of this Tcell subset [17]. In the case of immature B cells, AID may assist in removing B cells carrying self-reactive Igs on their surface [25,26]. These findings suggest that marginal levels of AID might have physiological roles other than Ig diversification, although the precise nature of those roles is largely unclear. Taken together, AID expression differs according to cell lineages and conditions, and these differences might account for the variety of physiological and pathological cellular responses elicited by this molecule.
Genome sequencing revealed four regions in and around Aicda that are well-conserved between humans and mice (regions 1, 2, 3, and 4) [16,27,28]. The downstream region (region 3) was shown to be important using Aicda bacterial-artificial-chromosome (BAC) constructs, although in vitro reporter assays have not clarified its regulatory function [16,28]. Several Aicda-regulation studies have shown that the Aicda promoter (region 1) is not particularly specific to B cells, and promotes transcription in various types of cells [27,28]. A HoxC4-binding site located in the promoter region is reported to induce Aicda transcription by increased HoxC4 levels [29,30]. Aicda also contains major regulatory elements in regions 2 and 4 [27,28]. Region 2, which is located in the first intron, contains binding sites for B-cell-specific transcription factors such as Pax5 and E [27,28,31], as well as E2F-and c-Myb-binding sites, which exert a strong silencing effect on the promoter [28]. Region 4, located about 8-kb upstream of the transcription-initiation site, contains stimulation-responsive elements for STAT6, NF-kB, Smad, and C/EBP. A previous in vitro reporter assay of mutations in these elements led us to propose a balanced regulation model of the Aicda promoter, in which B-cell-specific and stimulationresponsive enhancers coordinate to counteract the silencers, derepressing the Aicda promoter [28].
To understand the physiological mechanisms regulating Aicda, here we used BAC DNA to generate several lines of transgenic mice, and used these mice to verify the in vivo activity of Aicda regulatory elements identified by the previous in vitro reporter assay [28]. The present study clearly demonstrates that Aicda is regulated in vivo by the balance between the positive and negative transcription factors that were identified in our in vitro experiments.

Generation of Transgenic Mice
To generate BAC transgenic mice, we obtained the BAC clone RP24-68I7, which is 190 kb in length and harbors the entire Aicda gene, from BACPAC CHORI (http://bacpac.chori.org/) [32]. To generate the Aid-cre-cd2 construct, an expression cassette of cre IRES hCD2, the intracellular domain of which was deleted, was inserted into exon 2 of mouse Aicda on the BAC by homologous recombination in bacteria, as described previously [32] (Fig. 1A). We derived transgenic constructs with a series of deletions (Fig. 1A) from the Aid-cre-cd2 construct. The Red/ET recombination kit (Gene Bridges, Germany) was used according to the manufacturer's instructions for all the BAC modifications. In brief, we constructed vectors carrying two homology arms (either an 80-bp synthesized oligonucleotide or a PCR product of approximately 0.5 kb) adjacent to the DNA region we wished to delete or replace. The homology arms were ligated with donor DNA containing PGK-Neo cassettes floxed by Frt sequences. DNA fragments for targeting were prepared by PCR and transfected into BACbearing E. coli with the pRedET plasmid. Recombinants were selected with kanamycin, and the PGK-Neo cassette was removed by transfecting the FLPe plasmid. When introducing secondary modifications to the Aid-cre-cd2 BAC construct, the Frt orientation was reversed in the second targeting construct to avoid accidentally deleting regions between the initial Frt site and the newly introduced Frts. Clones with the reversion were excluded after transfection with the FLPe plasmid.
The following primers were used to amplify the homology arms: The dME-ac-cre transgene was generated in two steps. First, we amplified the fragment containing the c-Myb-binding motifs using the primers 59arm-ME-Fwd: 59-CTTTTCCTTCTCCTCCTT-CTCC-39 and 59arm-ME-Rv: CCTCTCCTCGGGATGTCCC-TCC. We then cloned the amplified fragment into the pGEM-T easy vector (Promega). The whole plasmid was PCR-amplified with the primers myb-mut-F 59-NNGGATCCACATCCT-GAGCCCTCAAAAAGCA-39 and myb-mut-R 59-NNGGAT-CCTTGTCTAGCATGTGTGAGGTCTTC-39. The PCR product was digested by BamHI and self-ligated. The other homology arm, the adjacent region of the E2F site, was amplified with the primers 39arm-ME-Fwd: 59-GCTCCTAGCTAGAGTTGAG-GGG-39 and 39arm-ME-Rv: 59-TCTAATCCAGCCAGTTT-TAAAC-39. Two homology arms were ligated with both ends of the PGK-Neo cassette, and were used for homologous recombination in E. coli harboring Aid-cre-cd2. The nucleotide sequences of the modified area in the final constructs were checked by subcloning and sequencing.
All mice were bred and maintained in specific pathogen-free conditions at Kyoto University and at the RIKEN CDB.
Rosa-tdRFP [33] mice were bred and maintained in specific pathogen-free conditions at the Institute of Laboratory Animals at the Graduate School of Medicine, Kyoto University. These animals were intercrossed to obtain double-mutant animals.

Immunization of Mice
Sheep red blood cells (SRBC) were washed by PBS for three times and were suspended in PBS at a concentration of 5610 9 /ml. Hundred micro-l of the SRBC suspension was given intraperitoneally.

Southern Blot and PCR Analysis
DNA for PCR and Southern blot analysis was obtained by tail biopsy. For Southern blotting, the DNA was digested with HindIII, separated on a 0.8% agarose gel, and transferred to a nylon membrane (Roche). The blots were air-dried, and the DNA crosslinked using UV light. The DIG-labeled DNA probes indicated in Fig. 1 were used for hybridization in ULTRAhyb Buffer (Ambion; Applied Biosystems). All procedures using the DIG application system (Roche) were performed according to the manufacturer's recommendations. Screening of the transgenes was done by PCR with primers CreFP33 59-CGGCATGGTG-CAAGTTGAATAAC-39 and CreRP34 59-GCTAAC-CAGCGTTTTCGTTCTGC-39, which amplify 458 bp DNA of the cre-coding region.

Construction of Aicda BAC Transgenic Mice
To examine the in vivo function of elements that regulate Aicda, we used a BAC transgenic strategy. We used the BAC clone RP24-68I7, which covers the 190-kb region of Aicda and its flanking regions, to establish a reliable system for monitoring Aicda expression [17]. We used a basic construct, Aid-cre-cd2, which harbors a bicistronic expression cassette for human CD2 and cre in the AID coding region (Fig. 1A) [32]. The Aicda BAC contains three regulatory regions (regions 2, 3, and 4) in addition to the promoter (region 1), as assessed by an in vitro culture system [16,27,28]. The locations and binding motifs of regions 2 and 4 are described above (Fig. 1A and the reference [28] for detailed information).
To examine the in vivo function of the Aicda regulatory elements identified in vitro, we added deletion mutations to the basic construct to obtain four additional constructs: dR2-ac-cre, in which region 2 was entirely deleted; dME-ac-cre, in which the region 2 c-Myb-and E2F-binding sites-the silencing elementswere deleted; dR2-ac-cre, in which region 4 was entirely deleted; and dCEBP-ac-cre, which contains a restrictive deletion of the two C/EBP sites ( Fig. 1A and Fig. S1). We generated transgenic mice carrying these BAC constructs and crossed them with the rosa-tdRFP reporter mouse [33]. In this system, Aicda promoter activation can be visualized by hCD2 and RFP for ongoing and past AID expression, respectively (Fig. 1B). It was previously reported, that the cre-loxP-mediated irreversible expression of reporters at the rosa26 locus enables more sensitive detection than the direct staining of hCD2, even allowing the visualization of transient and/or marginal activation of the Aicda promoter in immature B cells and memory T cells [17]. All mice described in this paper were crossed with the rosa26-tdRFP reporter strain unless otherwise described.

Aicda Expression is Positively Regulated by Regions 2 and 4
We chose 3 or 5 lines of transgenic mice for each construct (Table 1), each line carrying a single copy of the transgene (approximately 4 transgene copies were detected in the transgenic mouse described previously) (Fig. 1C) [17]. In all of these Aid-cre-cd2 lines, about one percent of the splenic B cells were RFP+, indicating that the basic BAC transgene was expressed similarly in these animals (Table 1). RFP and hCD2 were strongly expressed in germinal-center B cells, in which Aicda is actively transcribed; about 35% of the B220+CD38-gated population was positive for both reporters (Fig. 2). We next examined germinal-center B cells in the Peyer's patches of mice carrying manipulated BAC reporter constructs. Deletions of region 2 or 4 completely abolished the Aicda expression on the transgene (Fig. 2). Aicda was similarly inactivated in the independent founder lines dR2-ac-cre and dR4ac-cre (five each) ( Table 1). We conclude that both region 2 and region 4 are essential for normal Aicda expression in activated B cells in vivo.
Region 4 contains binding motifs for NF-kB, STAT6, and Smad3/4, which are involved in the signaling pathways of CD40, IL-4, and TGF-b, respectively [34][35][36]; an in vitro reporter assay also demonstrated the importance of two C/EBP sites in this region [28]. The dCEBP-ac-cre construct, in which the two C/ EBP sites in region 4 were deleted, along with the intervening sequence (Fig. S1), consistently failed to express the Aicda reporters, indicating that these C/EBP sites are in fact indispensable for region 4's enhancer activity in activated germinal-center B cells ( Fig. 2 and Table 1).

E2F and c-Myb Motifs Suppress Aicda Activity
In contrast to the enhancer function of the C/EBP sites in region 4, the E2F-and c-Myb binding sites act as silencers in vitro [28]. To determine whether these binding sites were responsible for silencing Aicda in germinal-center B cells, we examined transgenic lines carrying the dME-ac-cre construct. In contrast to the dR2-ac-cre transgenic mice, both reporters were expressed at elevated levels in the Peyer's patch germinal-center B cells in dME-ac-cre transgenic mice compared with Aid-cre-cd2 mice (Fig. 2). The percentage of germinal-center B cells that were positive for hCD2 or RFP in the dME-ac-cre mice was almost double that seen in Aid-cre-cd2 transgenic mice. The mean fluorescent intensity of hCD2+ B cells was similar in the Aid-cre-cd2 and dME-ac-cre mice (Fig. 2), suggesting that the efficiency of Aicda expression was similar once activated. In addition, the percentage of RFP-positive B cells in various lymphoid organs was significantly higher in the dME-ac-cre than in the Aid-cre-cd2 transgenic mice (Fig. 2 and Table 1). These results suggest that the Aicda promoter was more readily activated in dME-ac-cre than in Aid-cre-cd2 mice. We sorted germinal-center B cells, as well as B cells at other developmental stages, and used RT-qPCR to determine whether the increased percentage of cells positive for hCD2 or RFP correlated with the mRNA level. Consistent with the FACS analysis, the reporter-gene mRNA levels in the sorted cells were significantly higher in the dME-ac-cre than in the Aidcre-cd2 transgenic mice (Fig. 3). Taken together, these results indicate that the E2F and c-Myb elements negatively modulate the Aicda promoter's activity in germinal-center B cells in vivo.
To determine whether deleting these silencer elements, E2F and c-Myb, would upregulate the modest AID expression found in immature B cells, we examined the reporter expression in developing B cells in dME-ac-cre mice. These B cells did not robustly express hCD2 or RFP in immature stages (Fig. 4AB). However, the hCD2 mRNA levels in the sorted preB and transitional B cells were elevated in the dME-ac-cre (p,0.05) as compared to the Aicda-cre-cd2 mice (Fig. 3B). Notably, the RT-qPCR detection appeared to be more sensitive than BAC reporters. These results support the idea that the E2F and c-Myb elements in region 2 negatively regulate the Aicda promoter in immature B cells.
Germinal centers in Peyer's patches develop spontaneously in response to continuous stimulation from gut flora [37]. To determine whether the increased induction of RFP-positive cells in the dME mutant was dependent on chronic stimuli, we immunized naïve mice by intra-peritoneal sheep red blood cell (SRBC) injection and analyzed the germinal-center B cells on day 7, the early phase of the immune response [38]. Germinal-center B cells in the spleen were gated as B220+CD38-cells [39]. As in the Peyer's patches, a much higher percentage of dME-ac-cre spleen germinal-center B cells were positive for RFP or hCD2 than Aicda-cre-cd2 cells ( Figure 5). Therefore, the c-Myb and E2F elements' silencing activity negatively regulates Aicda promoter activation at the initiation stage of B-cell activation.

Deletion of Silencer Elements in Region 2 does not Enhance the Aicda Expression in T cells
Since region 2's silencing activity is not limited to B cells [28,40], it might be responsible for repressing Aicda in other cell lineages. Aicda is weakly expressed in chronically activated T cells that coincide with the development of the IL-10-producing T-cell subpopulation [17]. Reporter expression did not indicate a significant fraction of RFP+ cells among the T cells of the dMEac-cre mice (Fig. 6). Although we were able to detect RFP+ T cells in our previously reported transgenic mice, which carried multiple copies of the basic construct Aid-cre-cd2, this was not the case with the present single-copy transgenic mice (Table 1). Since the dMEac-cre construct was efficiently expressed in germinal-center B cells, this lack of expression is not caused by the functional inactivation of the transgene itself. Although c-Myb and E2F  All animals were crossed with rosa-tdRFP. Three mice from each line were analyzed. Percentage of RFP+ B cells in mice carrying the dME-ac-cre transgene was significantly higher than in mice carrying the Aid-cre-cd2 transgene. P = 0.0067 (Spleen), P = 0.0034 (mLN), P = 0.0021 (Peyer's patch) by two-tailed unpaired Student's t-test. mLN, mesenteric lymph node. doi:10.1371/journal.pone.0061433.t001 strongly suppress the Aicda promoter, even a marginal level of Aicda expression in T cells appears to require additional positive regulation [28].

Discussion
The AID molecule is central to the antigen-induced alteration of immunoglobulin genes in activated B cells, and its mutagenic activity can induce tumorigenesis in various types of cells [10]. AID is also thought to play a part in immature B-cell selection and T-cell activation, though it is expressed at much lower levels in these cells than in activated B cells [17,25,26]. It is thus clear that the Aicda promoter is finely regulated by cell lineage and the cellular environment. Determining how the Aicda gene is repressed in most cells, moderately expressed in response to some environmental stimuli, and fully expressed in activated B cells is critical for clarifying AID's physiological and pathological functions. However, Aicda expression under physiological conditions has not been elucidated in previous in vitro or ex vivo cellular studies [16,[27][28][29][30][31]41].

Region 2's Critical Roles in B-cell-specific AID Expression
In the present study, we demonstrated that regions 2 and 4 are important for inducing Aicda in activated B cells in vivo. Although we did not evaluate the activity of region 3 [16], all of the constructs used in this study contained region 3. Therefore, it is clear that region 3 cannot functionally compensate for the lack of region 2 or region 4. Region 2 contains binding sites for B-cellspecific transcription factors, namely, the Pax5 and E proteins. The present in vivo results are consistent with our previous in vitro report showing enhancer activity in this region [28,31]. In addition to the B-cell-specific enhancers, E2F-and c-Myb-binding sites in region 2 functioned as silencers in the in vitro reporter assay. These silencers counteracted positive regulatory elements in region 2 and region 4, in both B and non-B cells [28]. To elucidate the mechanism for lineage-specific and activation-specific AID induction, it was important to evaluate the E2F and c-Myb sites' silencing function under physiological conditions. As expected, these elements disturbed Aicda promoter activation in in vivoactivated B cells. Deletion of the E2F and c-Myb sites (dME-accre) drastically increased the percentage of hCD2-positive B cells in germinal centers, suggesting that these elements reduce the frequency of Aicda promoter activation within the gated population. On the other hand, when comparing the Aid-cre-cd2 and dME-ac-cre hCD2-positive germinal-center B cells, the mean fluorescent intensity of hCD2 staining was not significantly different. Since the hCD2 fluorescent intensity correlates with the reporter gene's transcription efficiency in individual cells, this observation indicates that the E2F and c-Myb sites do not affect the active reporter gene's transcription efficiency. These results may suggest that E2F and c-Myb negatively regulate the Aicda promoter's activation, but do not affect its transcription efficiency once activated. Aicda is usually expressed strongly only in activated B cells. If Aicda silencing were not stably maintained in other cells, even minimal exposure to various environmental stimuli could induce this dangerous internal mutagen. In this context, E2F proteins are known to associate with the pRB protein to function as a constitutive and tentative gene repressor [42]. Because the silencers in region 2 act in other cell lineages in addition to B cells [28], these silencers would also be expected to contribute to B-cell-specific Aicda suppression. We expected to observe ectopically expressed Aicda reporters in the absence of the silencers (dME-ac-cre mice). However, we did not observe ectopic or enhanced reporter expression in either immature B cells or developing T cells, in which the Aicda promoter is activated modestly under normal conditions [17]. This finding suggests that removing these silencers is not sufficient to induce ectopic Aicda expression. It is likely that the activity of the B-cell-specific enhancer proteins (Pax5 and E proteins) is not sufficient to activate the Aicda promoter in most cells, including immature B cells and activated T cells; this indicates that a coordinated balance between enhancers and silencers is critical for lineage-and activationspecific AID expression. Alternatively, there might be other unidentified silencers that are required for maintaining lineage specificity. However, this is less likely, since previous extensive deletion analysis did not find other obvious silencer elements, either in the AID gene or in its flanking regions [28].
A previous study clearly showed that a BAC transgenic system using cre recombinase coupled with floxed RFP can detect lowlevel Aicda expression [17]. However, we could not visualize Aicda expression records in immature B and activated T cells using this system in the present study. Since Aicda expression in these cells has been confirmed by other methodologies by various groups, it is unlikely that our previous observation was an artifact [16,17,23,24]. Rather, it seems that the sensitivity of the current study's system is lower than that reported by Qin et al. (2011). It is likely that this difference is due to the number of transgene copies; transgenic mice used in the previous study carried about 4 copies, while the current study used only animals with single-copy integration.

Region-4 Enhancers Positively Regulate Aicda
Region 4's role in Aicda regulation was first demonstrated in vitro by Tran et al. [28]. Our current study convincingly demonstrated in vivo the critical role of region 4's enhancer elements, which include binding sites for NF-kB, STAT6, Smad3/4, and C/EBP. NF-kB, STAT6, and Smad are major transcription factors in the CD40, IL-4, and TGF-b signaling pathways, respectively. Although dCEBP-ac-cre selectively deleted only the tandem C/ EBP-binding elements, leaving the other elements intact, it completely abolished the expression of Aicda reporters. This demonstrated C/EBP's indispensable role in regulating Aicda, as was previously shown in vitro.
While some factors are well-known to contribute to CSR induction, the involvement of C/EBP proteins has not been previously appreciated. Chromatin immunoprecipitation assays have suggested that C/EBPb binds to region 4 [28]. Since C/ EBPb can physically interact with NF-kB and modulate its activity [43][44][45], C/EBPb and NF-kB can function cooperatively and are often involved in signaling pathways mediated by toll-like receptors and cytokines, which are associated with inflammatory reactions [43,[45][46][47][48].
For example, in renal medullary interstitial cells, a mutation of the C/EBP-binding site abolishes NF-kB-dependent activation of the COX2 promoter, which contains a tandem array of NF-kBand C/EBP-binding sites [45]. These results suggest that the C/ EBP pathway dominantly regulates the tonicity-induced COX2 expression. The NF-kB-and C/EBP-binding sites in region 4 are separated by almost 100 bp, which is relatively close, although not as close as in the COX2 gene. It is well established that AID is induced by CD40 stimulation, which activates NF-kB. In this context, C/EBPb's critical role in Aicda expression in vivo is reasonable. It will be interesting to elucidate the molecular mechanisms by which C/EBP and NF-kB interact and cooperatively regulate Aicda expression. In summary, the activation of AID expression in vivo requires not only B-cell-specific enhancers in region 2, but also stimulation-dependent enhancers in region 4. Figure S1 Confirmation of deletion/replacement of regulatory elements on BAC constructs. PCR analysis confirmed the deleted of region 2 (0.5-kb) in dR2-ac-cre and that of region 4 (0.8-kb) in dR4-ac-cre (A) (B). Sequencing analyses to confirm the modifications introduced to dME-ac-cre (C) and dCEBP -ac-cre (D) were shown. The wild type sequences around the targeted elements (boxed) are presented. Sequencing chromatograms were shown below; corresponding region were indicated by dotted lines. (PDF)