|Home | About | Journals | Submit | Contact Us | Français|
Rearranged Ig V regions undergo AID-initiated diversification in sequence to produce either nontemplated or templated mutations, in the related pathways of somatic hypermutation and gene conversion. In chicken DT40 B cells, gene conversion normally predominates, producing mutations templated by adjacent pseudo-V regions; but impairment of gene conversion switches mutagenesis to a nontemplated pathway. We recently showed that the activator, E2A, functions in cis to promote diversification, and that G1 phase of cell cycle is the critical window for E2A action. By single-cell imaging of stable AID-YFP transfectants, we now demonstrate that AID-YFP can stably localize to the nucleus in G1 phase, but undergoes ubiquitin-dependent proteolysis later in cell cycle. By imaging of DT40 PolyLacO-λR cells, in which polymerized lactose operator tags the rearranged λR gene, we show that both the repair polymerase Polη and the multifunctional factor MRE11/RAD50/NBS1 localize to λR, and that λR/Polη colocalizations occur predominately in G1 phase, when they reflect repair of AID-initiated damage. We find no evidence of induction of γ-H2AX, the phosphorylated variant histone that is a marker of double-strand breaks, and Ig gene conversion may therefore proceed by a pathway involving templated repair at DNA nicks rather than double-strand breaks. These results lead to a model in which Ig gene conversion initiates and is completed or nearly completed in G1 phase. AID deaminates single-stranded DNA, and restriction of mutagenesis to G1 phase would contribute to protecting the genome from off-target attack by AID when DNA replication occurs in S phase.
The V regions of actively transcribed Ig genes undergo physiologically induced and regulated sequence diversification, which expands and modulates the repertoire for antigen recognition and provides a dynamic response to infection by pathogenic microorganisms (1-5). V region diversification produces two distinct mutagenic signatures: somatic hypermutation introduces nontemplated single base changes, while gene conversion uses upstream pseudo-V (ψV) gene segments as templates for mutagenesis (2, 6-11). Gene conversion is the primary source of diversity in the pre-immune antibody repertoire of chicken and other fowl. Somatic hypermutation occurs in antigen-activated human and murine B cells and also diversifies the pre-immune repertoire in other species, such as sheep (12).
The mechanisms of somatic hypermutation and gene conversion are closely related, as first proposed over a decade ago (13). Both processes are initiated by attack of the B cell-specific enzyme, activation-induced deaminase (AID), on actively transcribed Ig genes (14-17). AID deaminates cytosine to uracil in DNA (18-21). Mutagenic repair then depends either on uracil-DNA glycosylase (UNG), which removes uracil to create an abasic (AP) site; or MutSα (MSH2/MSH6), which recognizes U·G mismatches (22). The multifunctional MRE11/RAD50/NBS1 (MRN) complex (23-25) associates with and is enriched at diversifying Ig genes, where it may promote V region gene conversion by employing its AP lyase activity to cleave at the AID-initiated AP site (26, 27); by tethering DNA molecules for recombination (24); and/or by carrying out DNA resection necessary for homology-dependent repair (28). The repair polymerase, Polη, participates in both gene conversion (29, 30) and somatic hypermutation (31-35), and can generate the templated mutations characteristic of normally proliferating chicken B cells. Point mutations accumulate if gene conversion is impaired by a variety of strategies, including ablation of critical factors, such as the RAD51 paralogs (RAD51B, RAD51C, RAD51D, XRCC2, and XRCC3) (36-38), or BRCA1 and BRCA2 (39, 40); or deletion of (41) or repressive chromatin modifications at (42) ψV donors, probably as a result of recruitment of this and/or other low fidelity polymerases to sites of AID-initiated DNA damage.
We recently showed that activation of Ig gene diversification depends upon cis-interactions of the Igλ allele with the regulatory factor E2A that occur in G1 phase of cell cycle (43). Those experiments took advantage of the ability to directly image diversifying λR genes in single DT40 PolyLacO-λR B cells. These cells are derivatives of the chicken bursal lymphoma line DT40, in which polymerized lactose operator (PolyLacO) has been inserted just upstream of the rearranged and expressed λ light chain gene (42, 43). By single-cell imaging, we now demonstrate that AID-YFP can stably localize in the nuclei of cells in G1 phase, but not in S phase or later stages. Colocalizations of λR with the polymerase Polη occur predominately in G1 phase. We find no evidence for induction of γ-H2AX, the phosphorylated variant histone that is a hallmark for DNA double-strand breaks (DSBs), at any stage of cell cycle. These results suggest that Ig gene conversion involves a DNA intermediate that carries a nick, rather than a DSB, and that gene conversion may initiate and be nearly completed in G1 phase of cell cycle.
DT40 and its derivative cell lines were cultured and transfected as previously described (26, 43). Leptomycin B (LMB; Sigma-Aldrich) and MG132 (Z-Leu-Leu-Leu-aldehyde; Sigma-Aldrich) were added at 50 ng/ml and 50 μM, respectively, for indicated times. 5-bromo-2′-deoxyuridine (BrdU; Sigma-Aldrich) was added at 10 μM for 30 min to pulse label cells. Irradiation was carried out at 8 Gy using a 137Cs irradiator. The sIgM-loss assay was conducted as described (26, 38, 43), and results were compared with the Mann-Whitney U test.
An AID-YFP expression construct was generated by cloning an AID cDNA of DT40 into the BglII site of pEYFP-N1 (Clontech). AIDΔC, which lacks the C-terminal 16 amino acids, was generated by amplification with primers 5′-CCAGAGATCTGAGAGAGAACCCAGCTGACATG-3′ and 5′-TTAAAGATCTGGCAGAAGGATCCGTCGGAGTTTCCTGGA-3′ and similarly cloned into the BglII site of pEYFP-N1. Three phosphorylation-site mutants, AIDT27A, AIDS38A and AIDY184A, were generated by QuikChange site-directed mutagenesis (Stratagene). A Polη-GFP expression construct (44) was provided by Dr. James Cleaver (University of California, San Francisco, CA). A Ugi expression construct (45) was provided by Dr. Michael Neuberger (Medical Research Council Laboratory of Molecular Biology, Cambridge, UK).
Immunofluorescence staining was performed as previously described (43). Primary antibodies used were: anti-GFP (3E6, 1:100; Molecular Probes); anti-NBS1 (1:25; provided by Dr. Pat Concannon, University of Virginia, Charlottesville, VA; or 1:1000; affinity purified rabbit antibody raised against a peptide corresponding to the C-terminal 20 amino acids of murine NBS1); anti-γ-H2AX (JBW3C1, 1:200; Upstate; or BL178, 1:500; Bethyl Laboratories); anti-FG-containing nuclear pore proteins (Mab414, 1:5000; Covance); anti-Golgi 58K protein (58K-9, 1:1000; Sigma); anti-ubiquitin (ab8134, 1:200; Abcam); and anti-SUMO1 (21C7, 1:200; Zymed). The BrdU labeling and detection kit I (Roche) was used to detect incorporated BrdU. Alexa Fluor 488- or 594-conjugated anti-IgG (Molecular Probes) were used as secondary antibodies. DNA was stained with 4,6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich).
For imaging of AID-YFP transfectants, we screened multiple independent clones and selected for imaging the two transfectant clones in which the AID-YFP signal was brightest. Results with these two clones were comparable, and detailed analysis of one clone is presented. In that clone, AID-YFP was more than 10-fold overexpressed relative to endogenous AID, as estimated by RT-PCR. Nuclear localization of AID-YFP was examined using a line profile tool of the softWoRx software (Applied Precision), and cells scored as positive if a YFP signal above background was detected throughout the nucleus. Cell cycle stage was determined by BrdU staining or estimated based on the nuclear size: G1, r < 4 μm; G2, r ≥ 5.2 μm (43). For experiments in which BrdU-staining was used to establish cell cycle stage, DNase treatment (Roche) was used to render DNA single-stranded and expose the newly incorporated BrdU to antibodies. This interferes minimally with the YFP signal but does not create as intense a BrdU signal as denaturation with alkali. Only cells containing five or more distinct BrdU foci were scored as S-phase cells, in order to exclude cells in other stages of cell cycle in which repair DNA synthesis was ongoing. The fraction of S phase cells as determined by BrdU staining was therefore lower than the fraction determined by propidium iodide staining and flow cytometry (36% compared to 56%).
To quantitate colocalizations in DT40 PolyLacO-λR and DT40 PolyLacO-λU derivatives, fluorescent images were acquired using the DeltaVision deconvolution microscopy system (Applied Precision) and processed and analyzed with softWoRx and Imaris (Bitplane) softwares. Fluorescence signals were sometimes partially rather than completely overlapping, which may reflect the considerable distance (~17 kb) between the PolyLacO-tag and the Vλ region, and were scored as colocalizing if the distance of fluorescent peaks was < 0.2 μm. Significance of colocalization and cell cycle distribution were analyzed with the Pearson’s χ2 test.
We assayed subcellular localization of AID in the chicken B cell line DT40, which derives from a bursal lymphoma and carries out constitutive Ig gene diversification (10, 11, 46). AID carries a strong nuclear export signal (NES) which specifies export via the nuclear pore, using the CRM1-mediated pathway, and AID has been shown to localize predominantly to the cytoplasm in mammalian cells (47-51). To examine AID localization by single-cell imaging, we constructed a vector expressing chicken AID carrying a C-terminal yellow fluorescent protein tag (AID-YFP). Tagging AID at this position has been shown not to interfere with protein function (47, 52). By analysis of sIgM loss rate (43) of 24 independent DT40 AID-YFP transfectants, we determined that stable expression of AID-YFP accelerated the clonal rate of Ig gene diversification 2.5-fold (p = 0.0014, Mann-Whitney U test), comparable to results obtained by others (e.g. (41, 47)). We imaged AID-YFP localization by fluorescence microscopy. AID-YFP localized to the cytoplasm, forming dots or flares just outside the nucleus rather than a diffuse signal (e.g. Fig. 1A, left). Others have similarly observed that, in mammalian B cells, AID cytoplasmic signals may be threadlike (47) or concentrated in pockets on the nuclear surface (51). Using quantitation with a line profile tool of the softWoRx imaging software as the criterion for distinguishing nuclear/cytoplasmic distribution of AID, we confirmed the nuclear level of AID-YFP to be very low, indistinguishable from background (Fig. 1A, right). We further verified that cytoplasmic localization of AID is determined by the conserved C-terminal NES, as shown for AID in mammals, by demonstrating that deletion of the NES resulted in nuclear accumulation (AIDΔC-YFP; e.g. Fig. 1B, left). This was also confirmed by analysis with the line profile tool (Fig. 1B, right).
AID function is regulated in part by phosphorylation, and mutation of Ser38, a protein kinase A phosphorylation site, has been shown to affect AID function in Ig gene diversification (53-57). To test the possible role of regulatory phosphorylation in nuclear localization, we asked if localization was affected by mutation at Ser38 as well as two other phosphorylation sites, Thr27 and Tyr184. We generated stable AIDT27A-YFP, AIDS38A-YFP and AIDY184A-YFP transfectants, and assayed localization of each. All localized to the cytoplasm, but AIDS38A-YFP also formed a few nuclear foci (Figure 1C), suggesting that regulatory phosphorylation at this site may be important for AID localization.
To test possible cell cycle-dependence of AID-YFP localization, we first attempted to identify the AID-YFP signal in fixed cells sorted following propidium iodide staining, which measures DNA content. The cell cycle distribution was similar in the AID-YFP transfectants compared to the parental DT40 cell line: 23% G1 phase, 56% S phase, 21% G2/M phase in DT40 AID-YFP transfectants; and 21% G1 phase, 55% S phase, 24% G2/M phase in the DT40 parent. However, the fixed and sorted cells had undergone significant loss of the AID-YFP signal. We therefore adopted alternative approaches to determine cell cycle phase. We have previously shown that nuclear radius corresponds closely to cell cycle stage in DT40 and its derivatives, and that nuclear radius < 4 μm provides a conservative criterion for scoring G1 phase cells (43). Cells were cultured with LMB for 30 min, fixed, imaged, and assigned to cell cycle stage based on nuclear radius. A nuclear AID-YFP signal was observed in 21% of cells (n = 123); and more than 80% of cells exhibiting a nuclear AID-YFP signal had nuclear radius < 4 μ, and were therefore in G1 phase (Fig. 1D, right). G1 phase cells correspond to less than 25% of the cell population, based on either cell sorting or measurements of nuclear radius. Thus there appeared to be a considerable overrepresentation of nuclear AID-YFP in G1 phase of cell cycle, and underrepresentation in later stages. In particular, these results raised the possibility that AID-YFP might be excluded from the nucleus in S phase.
To specifically assay the presence of AID-YFP in the nuclei of S phase cells, we stained with BrdU to identify cells carrying out DNA replication. Cells were cultured with LMB for 30 min, then BrdU added for an additional 30 min of culture, after which cells were fixed, DNase treated to expose incorporated BrdU to antibodies, stained with anti-BrdU antibodies and imaged (e.g. Fig. 1D, left). We scored as BrdU+ only cells containing five or more nuclear BrdU foci, to exclude cells carrying out repair DNA synthesis, and tabulated BrdU+ cells and cells exhibiting a nuclear AID-YFP signal (Table 1). Following culture with LMB followed by addition of BrdU, 19% of cells contained nuclear AID-YFP, but only 2.4% of cells containing nuclear AID-YFP were also BrdU+ (n = 288). When LMB was included after initiation of BrdU labeling, 12% of cells contained nuclear AID-YFP, and only 1.1% of cells contained both nuclear AID-YFP and BrdU+ (n = 353). Thus there is a considerable underrepresentation of AID-YFP in BrdU+ cells, suggesting that AID-YFP is depleted from the nucleus when DNA replication occurs.
Levels of AID in mammalian cells are regulated by ubiquitin-dependent proteolysis, and the half-life of nuclear AID has been reported to be shorter than that of cytoplasmic AID (2.5 hr vs. 18-20 hr (52)). This suggested that proteolysis of AID may account for underrepresentation of nuclear AID-YFP in S phase cells; and predicted that culture with MG132, an inhibitor of ubiquitin-dependent proteolysis, would stabilize nuclear AID-YFP and increase the fraction of S phase cells in which it is evident. We tested this by culturing cells with MG132 and LMB for 4 hr, then analyzing AID-YFP localization as a function of cell cycle, as determined by measuring nuclear radii of individual cells (43). Nuclear AID-YFP was evident in 18% of these cells (n = 298; e.g. Fig. 1E, left), comparable to the fraction of such cells observed following culture with LMB alone (19%). However, the cell cycle distribution of cells in which a nuclear AID-YFP signal was evident was clearly shifted, so only 13% of such cells were in G1 phase, while 69% were in S phase and 18% were in G2 phase (Fig. 1E, right; n = 55). This contrasts with the distribution of cells containing nuclear AID-YFP and treated with LMB alone (Fig. 1D, right; p < 0.001, χ2 test). Thus, culture with MG132 appears to stabilize nuclear AID-YFP in S phase cells, without increasing the fraction of G1 phase cells containing nuclear AID-YFP.
Culture with MG132 and LMB also promoted formation of a distinctive cytoplasmic spot of AID-YFP, which was shown to be juxtaposed to the Golgi apparatus by staining with anti-Golgi 58K antibodies (red, Fig. 1E, left). This spot is similar to the aggregates of ubiquitinated proteins, or “aggresomes” that form as consequence of inhibition of the proteasome, as documented for other proteins including the close relative of AID, APOBEC3G (58, 59). To ask if these bright cytoplasmic AID-YFP spots might reflect aggregates of modified protein, we asked if they colocalized with regions stained by antibodies to ubiquitin or SUMO. There was some colocalization of the bright AID-YFP spot with regions of ubiquitin but not SUMO staining (Fig. 1F).
The results above show that while AID-YFP localizes predominately to the cytoplasm, it can be found in the nucleus in G1 phase cells. AID-YFP appears to be able to enter the nucleus throughout cell cycle, but to be destabilized by ubiquitin-dependent proteolysis outside of G1 phase.
Many repair factors and transcription factors function in factories readily identifiable by fluorescence microscopy as distinctive foci within the nucleus, containing 50 or more protein molecules concentrated in a small nuclear volume (60). Single-cell imaging of colocalizations of foci of such factors with specific genes can provide valuable information on temporal regulation of transcription or repair. To image Ig genes in the nucleus, we have used a derivative of DT40, DT40 PolyLacO-λR, in which PolyLacO has been integrated by homologous gene targeting to the ψVλ array just upstream of the rearranged and expressed λR gene (42, 43). In DT40 PolyLacO-λR cells, the PolyLacO tag enables the diversifying λR gene to be imaged directly in cells expressing green or red fluorescent proteins fused to lactose repressor (GFP-LacI or RFP-LacI), but does not alter proliferation, cell cycle distribution, the clonal rate of Ig gene conversion or local chromatin structure (42, 43). Colocalizations of factors with the tagged gene can readily be quantitated in these cells by fluorescence microscopy. However, the confined nuclear space means that there will be inevitable background contributed by apparent overlap of the tagged gene and foci of transcriptional or regulatory factors. We have previously shown that the regulatory factor E2A colocalizes with the rearranged λR gene, and that this colocalization correlates with the cis-interaction essential to promote Ig gene diversification (43). In addition, by chromatin immunoprecipitation (ChIP) analysis we have shown that E2A specifically associates with the rearranged λR allele in DT40 cells, exhibiting an 18-fold enrichment at that allele relative to the unrearranged λU allele (43). Those results enabled us to ask if E2A preferentially colocalizes with the rearranged λR allele, as predicted by ChIP, and at the same time determine the background level of colocalizations evident as analyzed by fluorescence microscopy. To do this, we assayed colocalizations of E2A with the unrearranged λU light chain allele, in the DT40 PolyLacO-λU cell line. In that cell line, the PolyLacO tag is integrated in the ψVλ array upstream of the unrearranged and unexpressed λU allele (Fig. 2A). In DT40 PolyLacO-λU cells, proliferation (Fig. 2B), cell cycle distribution (Fig. 2C) and the clonal rates of Ig gene diversification (Fig. 2D) were comparable to the DT40 parental line, and to the DT40 PolyLacO-λR cells. However, while λR/E2A colocalizations were evident in 26% of DT40 PolyLacO-λR cells (n = 227) in asynchronous culture (43), λU/E2A colocalizations were evident in only 6.3% (n = 191) of DT40 PolyLacO-λU cells (Fig. 2E; p < 0.0001, χ2 test). Thus, quantitation of colocalization can provide very useful information on function of a factor at a gene.
The repair DNA polymerase, Polη, participates not only in Ig gene diversification but also in repair of UV damage and has been shown to form repair foci after UV irradiation (44). To establish when in cell cycle repair of AID-initiated DNA damage occurs, we therefore assayed temporal regulation of localization of the repair factor Polη to λR in DT40 PolyLacO-λR cells, in which PolyLacO tags the rearranged and expressed λR gene (Fig. 3A). To ensure specificity in imaging Polη, we stably transfected DT40 PolyLacO-λR RFP-LacI cells with a Polη-GFP expression construct and stained with antibodies to the GFP tag. λR/Polη-GFP colocalizations were imaged by fluorescence microscopy, and cells containing colocalized foci were scored. Cell cycle profiling showed that in these transfectants, 27% of cells were in G1 phase, 56% in S phase, and 17% in G2 phase. λR/Polη-GFP colocalizations were observed in 11% of DT40 PolyLacO-λR RFP-LacI Polη-GFP cells in asynchronous cultures (n = 1,013; e.g. Fig. 3B, left). Analysis of cell cycle distribution of λR/Polη-GFP colocalizations showed that 39% occurred in G1 phase, 56% in S phase, and 5% in G2 phase (Fig. 3B, right). The excess of λR/Polη-GFP colocalizations in G1 phase and the dearth of colocalizations in G2 phase were statistically significant (p = 0.0010, χ2 test). Similarly, we previously documented a significant excess of λR/E2A colocalizations in G1 phase (43).
To ask whether MRN localizations at λR exhibit cell cycle dependence, we imaged λR/NBS1 colocalizations, by staining DT40 PolyLacO-λR GFP-LacI cells with anti-NBS1 antibodies. λR/NBS1 colocalizations were evident in 17% of cells in asynchronous culture (n = 232; e.g. Fig. 3C, left). Analysis of cell cycle distribution of λR/NBS1 colocalizations showed that 28% occurred in G1 phase, 48% in S phase, and 24% in G2 phase (Fig. 3C, right). This did not different significantly from the cell cycle distribution of these cells (G1, 27%; S, 56%; G2/M, 17%; p = 0.13, χ2 test).
Foci of the phosphorylated variant histone, γ-H2AX, are hallmarks of DSBs (61-63). We therefore stained DT40 cells with antibodies to γ-H2AX to determine whether DSBs are an intermediate in Ig gene conversion, but this produced only very faint signals in DT40 PolyLacO-λR GFP-LacI cells, and λR/γ-H2AX colocalizations were not evident (not shown). We then quantitated γ-H2AX foci in the DT40 parental cell line, to ensure that the strong signal from the tagged λR gene would not interfere with the weak γ-H2AX signal. We found that most cells in asynchronous DT40 cultures contained no γ-H2AX foci (88%, n = 410; e.g. Fig. 3D, above left), but a small fraction of cells contained 1-3 small, faint, and punctate foci (12%; e.g. Fig. 3D, above left), similar to those observed in other cell lines proliferating under normal conditions (62, 64). Analysis of cells sorted by cell cycle stage prior to staining showed that faint γ-H2AX foci were present in 2.3% of G1 phase, 9.0% of S phase, and 7.7% of G2 phase cells (n > 300 in each case; Fig. 3D above right). In contrast, NBS1 foci were present in 37% of G1 phase, 54% of S phase, and 43% of G2 phase cells (Fig. 3D, above right), but there was no evidence of colocalization of γ-H2AX and NBS1. Following treatment with ionizing radiation (IR), which induces DSBs, brightly staining γ-H2AX foci were induced and colocalized with NBS1, as has been reported for many other cell types: 3 or more γ-H2AX/NBS1 colocalized foci were evident in 38% of cells (n = 312; e.g. Fig. 3D, lower). Thus, bright γ-H2AX foci can form in DT40 cells, but γ-H2AX foci do not characterize cells carrying out gene conversion. This suggests that if a DSB is formed in the course of gene conversion, it is present only very transiently, or does not associate with γ-H2AX in the course of repair.
G1 phase is the window in which AID can stably localize to the nucleus (Fig. 1) and in which E2A activates diversification at the Ig genes (43). This suggested that λR/Polη-GFP colocalizations in G1 phase may reflect repair synthesis at the diversifying V region. To test this, we examined the effect of uracil-DNA glycosylase inhibitor (Ugi) expression on λR/Polη-GFP colocalizations. Ugi inhibits UNG, and expression of Ugi in DT40 B cells inhibits Ig gene conversion (65). We generated stable DT40 PolyLacO-λR RFP-LacI Polη-GFP Ugi transfectants, and verified inhibition of UNG activity upon Ugi expression by assaying deglycosylation of a uracil-containing, 5′-end-labeled synthetic oligonucleotide by nuclear extracts from the parental cell line and the Ugi-expressing derivative (Fig. 4A). Single-cell imaging showed that λR/Polη-GFP colocalizations were evident in 5.4% of asynchronous DT40 PolyLacO-λR RFP-LacI Polη-GFP Ugi cells (n = 298), compared with 11% of parental DT40 PolyLacO-λR RFP-LacI Polη-GFP cells (p = 0.0035, χ2 test). In DT40 PolyLacO-λR RFP-LacI Polη-GFP Ugi cells, 19% of colocalizations occurred in G1 phase (Fig. 4B; p < 0.001 vs. 39% in the parental line, χ2 test), 75% in S phase, and 6.3% in G2 phase.
To better gauge the effect of Ugi, λR/Polη-GFP colocalizations in each phase of cell cycle were graphed as a fraction of the entire cell population. This showed that, in Ugi-expressing cells, the percentage of G1 phase cells containing λR/Polη-GFP colocalizations was reduced from 4.5% to 1.0%, or 4.4-fold (Fig. 4C; p = 0.011, χ2 test). In contrast, the percentage of S phase cells containing λR/Polη-GFP colocalizations was reduced from 6.3% to 4.0% (p = 0.20, χ2 test); and the percentage of G2 phase cells containing λR/Polη-GFP colocalizations was reduced from 0.6% to 0.3% (p = 0.63, χ2 test). Thus, only in G1 phase did Ugi expression have a significant effect on the fraction of cells containing λR/Polη-GFP colocalizations. While we cannot eliminate the possibility that Ugi expression has subtle effects on DNA metabolism throughout cell cycle, the simplest interpretation of these results is that the G1 phase-specific colocalizations reflect repair initiated by AID and UNG. This suggests that G1 phase is the window in which much of the repair synthesis that is induced in response to AID-initiated DNA damage occurs.
Using single-cell imaging to analyze temporal regulation of Ig gene conversion, we have found that two key factors act primarily in G1 phase of cell cycle. AID-YFP stably localizes to the nucleus in G1 phase, but not other phases of cell cycle; and Polη localizes to the Igλ locus predominately in G1 phase. AID initiates gene conversion, and repair synthesis by Polη is a late step in the mutagenic pathway. These results complement earlier evidence that G1 phase is key to activation of Ig gene diversification by E2A (43), and raise the possibility that gene conversion may initiate and be nearly completed in G1 phase, prior to DNA replication in S phase.
AID preferentially deaminates single-stranded DNA (19-21, 66). The transient DNA denaturation essential for replication would therefore be predicted to create a target for AID. Restricting AID attack to G1 phase would therefore protect the replicating genome from AID-initiated mutagenesis or other forms of genomic instability, such as translocation. In addition, mutations created in G1 phase could be fixed in the subsequent S phase (Fig. 5) and thereby protected from post-replicative repair which would have an adverse effect on mutagenesis essential to the immune response.
To analyze regulation of AID, we generated a snapshot of AID-YFP subcellular localization by treating cells briefly with LMB, which inhibits CRM1-dependent nuclear export, and then distinguishing S phase cells by nuclear size and/or BrdU staining. This analysis showed that AID-YFP was present in the nucleus in early stages of cell cycle, but absent from the nucleus in S and later stages, when DNA replication occurs. Inhibition of ubiquitin-dependent proteolysis altered cell cycle dependence of AID-YFP localization, increasing the fraction of S phase cells in which AID-YFP was observed. Like its close relative, APOBEC3G (59), AID-YFP accumulated in cytoplasmic aggresomes upon inhibition of proteasomal degradation. These experiments analyzed localization of tagged fluorescent protein in stable AID-YFP transfectants, so we cannot preclude the possibility that some of the observed effects were due to overexpression of AID. Nonetheless, this evidence for nuclear instability of AID-YFP in chicken B cells is consistent with analysis of mammalian cells, which demonstrated that the half-life of nuclear AID is eight-fold below that of cytoplasmic AID (52). These results therefore suggest that regulatory proteolysis prevents accumulation of AID-YFP in the nucleus at stages of cell cycle later than G1 phase.
Single-cell imaging showed that both the repair polymerase Polη and the MRN complex localize to the diversifying λR allele in DT40 B cells. The visual evidence provided by analysis of colocalizations complements genetic and biochemical evidence for function of Polη and MRN in Ig gene conversion (26-30). Imaging further suggests that localizations of Polη and MRN to the Ig genes are under distinct temporal regulation. Colocalizations of λR with Polη-GFP occurred predominately in G1 phase; and expression of Ugi, which inhibits UNG activity and thereby mutagenic repair downstream of deamination of cytosine to uracil, diminished colocalizations in G1 phase over 4-fold, but had relatively modest effect on colocalizations in other stages of cell cycle.
In contrast, colocalizations of λR with NBS1 occurred throughout cell cycle. This may reflect the multiple roles played by the multifunctional MRN complex. Gene conversion depends on factors that promote homology-dependent repair, which occurs primarily in S and G2 phases. Gene conversion may be prolonged into S phase to take advantage of elevated levels of those factors in that phase of cell cycle. If so, then MRN may function differently at the V region in each stage of cell cycle, using its AP lyase to initiate diversification in G1 phase and then resecting the DNA or tethering recipient and donor later in cell cycle. This is distinct from MRN function in switching B cells, where nonhomologous end-joining (NHEJ) promotes DSB repair, and NBS1 colocalizes with the IgH locus predominately in G1 phase (67), when the NHEJ pathway is most active (63, 68-71).
In principle, gene conversion can involve a DNA intermediate carrying either a DNA DSB or nick. In one example studied in considerable mechanistic detail, mating type switching in the yeast, S. cerevisiae, is initiated by a DNA DSB (72). However, DNA nicks have also been shown to be effective in initiating recombinational repair (73, 74). We assayed for DSBs by staining DT40 cells for foci of γ-H2AX, diagnostic of DNA DSBs (61-63). We found no evidence of such foci in normally proliferating DT40 cells, although bright γ-H2AX foci were inducible by IR. This suggests that gene conversion at the Ig loci does not involve DSBs; or if DSBs do form, they are present only very transiently.
At the diversifying Ig loci, AID, UNG and MRN act sequentially to generate a nick, in a series of steps common to gene conversion and somatic hypermutation (Fig. 5). If a nick rather than a DSB initiates recombinational repair, then this suggests a model in which a nick generated in G1 phase after successive action of AID, UNG and MRN would be repaired in G1 phase by Polη, and by other recombination factors in later cell cycle stages (Fig. 5). This model does not postulate involvement of DSBs or γ-H2AX.
Two groups have shown that deficiencies in NHEJ increase the rates of Ig gene conversion (75, 76). This has been interpreted as evidence that gene conversion depends on DSBs, based on the view that DSBs not repaired by NHEJ would then be intermediates for gene conversion. However, products of V region gene conversion rarely carry the short deletions typical of NHEJ, making it unlikely that a significant fraction enter this pathway. The absence for evidence of γ-H2AX induction in DT40 cells suggests that impaired NHEJ may not simply increase available substrates, but have other effects on Ig gene conversion. For example, impaired NHEJ may result in activation of factors that promote homology-dependent repair, and thus increase the rates of gene conversion.
γ-H2AX is not necessary for somatic hypermutation, as mice deficient in γ-H2AX exhibit normal levels of somatic hypermutation (77). In contrast, γ-H2AX is required for Ig class switch recombination, a process of regulated DNA deletion which depends upon the NHEJ pathway and involves the formation and rejoining of DNA DSBs; and γ-H2AX is evident in distinct foci in B cells carrying out class switch recombination (67). The apparent lack of participation of γ-H2AX in Ig gene conversion (see above) and somatic hypermutation (77) reinforces the mechanistic relationship between these two processes. Both can result from templated or nontemplated mutagenesis at a DNA nick (Fig. 5). In gene conversion, strand transfer occurs to permit templated mutagenesis; while the error-prone polymerases that promote somatic hypermutation can extend directly from a DNA nick. The RAD51 paralogs promote strand transfer, and deficiencies in these factors alters the outcome of repair at the chicken Ig loci, so that hypermutation occurs rather than gene conversion (36-40), providing further support for the model that gene conversion and somatic hypermutation proceed via the same DNA intermediate.
We thank Dr. Erik Larson for help with UNG assays; Monica Fujii for assistance in preliminary experiments; Greg Martin (Keck Imaging Center, University of Washington) for help with fluorescence microscopy; and Drs. Yansong Gu, Brian Kennedy, David Morris, Dennis Willerford and all members of the Maizels laboratory for discussions.
1This work was supported by National Institutes of Health Grants R01 GM39799 and R01 GM41712 (to N.M.) and National Institutes of Health Predoctoral Training Programs T32 GM07223 and T32 AG00057 (to E.C.O.).