Genomic Uracil Homeostasis during Normal B Cell Maturation and Loss of This Balance during B Cell Cancer Development

Abstract

Activation-induced deaminase (AID) converts DNA cytosines to uracils in immunoglobulin genes, creating antibody diversification. It also causes mutations and translocations that promote cancer. We examined the interplay between uracil creation by AID and its removal by UNG2 glycosylase in splenocytes undergoing maturation and in B cell cancers. The genomic uracil levels remain unchanged in normal stimulated B cells, demonstrating a balance between uracil generation and removal. In stimulated UNG^−/− cells, uracil levels increase by 11- to 60-fold during the first 3 days. In wild-type B cells, UNG2 gene expression and enzymatic activity rise and fall with AID levels, suggesting that UNG2 expression is coordinated with uracil creation by AID. Remarkably, a murine lymphoma cell line, several human B cell cancer lines, and human B cell tumors expressing AID at high levels have genomic uracils comparable to those seen with stimulated UNG^−/−splenocytes. However, cancer cells express UNG2 gene at levels similar to or higher than those seen with peripheral B cells and have nuclear uracil excision activity comparable to that seen with stimulated wild-type B cells. We propose that more uracils are created during B cell cancer development than are removed from the genome but that the uracil creation/excision balance is restored during establishment of cell lines, fixing the genomic uracil load at high levels.

INTRODUCTION

When B lymphocytes are activated through antigen presentation, they acquire hypermutations in the immunoglobulin (Ig) genes, facilitating affinity maturation of antibodies. An enzyme, activation-induced deaminase (AID), initiates these events by converting cytosines in DNA to uracil (1–4). The introduction of this rare base into DNA leads to a very high frequency of base substitution mutations in the Ig variable domain (known as somatic hypermutations [SHMs]; reviewed in references 5 and 6). Generation of uracils is also the first step in the creation of strand breaks in the switch regions of Ig genes, leading to the replacement of the μ constant domain with other domains such as γ, in a process called class-switch recombination (CSR; reviewed in reference 7). AID also binds near the transcription start sites of a large number of actively transcribed genes (8) and mutates a number of additional genes, including those encoding BCL-6, MYC, PAX-5, and PIM1 (9–12). The uracils generated by AID are thought to be removed by the nuclear form of the uracil-DNA glycosylase, UNG2, creating abasic sites that are repaired by error-prone copying mechanisms causing hypermutations (13, 14). Another uracil-DNA glycosylase, SMUG1, is normally considered the backup enzyme for UNG2 (15), but overproduction of SMUG1 is required for it to complement an UNG^−/− mutant during CSR (16). Additionally, DNA mismatch repair (MMR) also plays a role in shaping the mutation spectrum in SHM (17).

There is a strong connection between expression of AID and cancers in animals. Constitutive expression of AID in mice causes T cell cancers (18). Many human B cell lymphomas and some leukemias that arise during the maturation of B lymphocytes in germinal centers (GC) constitutively express AID (19, 20). This is probably because AID is required for generating both of the double-strand breaks involved in the c-myc/IgH translocations that are a hallmark of B cell cancers (21, 22). Additionally, UNG^−/− mice develop B cell hyperplasia and lymphomas at higher frequency than normal mice, suggesting that B cell maturation in the absence of UNG promotes oncogenic transformation (23). Based on such observations, it has been suggested that uracils generated by AID cause mutations and/or strand breaks that drive cellular transformation in some of the B cells undergoing maturation in germinal centers and resulting in hematopoietic cancers (24).

Despite the wide acceptance of the idea that AID converts cytosines in DNA to uracil, no study has yet identified or quantified uracils in B cell tumor genomes. The only reports of uracils created by AID in normal B cells have been in mouse models of antibody maturation that have focused only on the Ig genes (25, 26). Furthermore, the balance between uracil creation in the B cell genome by AID and its removal by UNG2 or other repair enzymes has not been examined. For example, it is not known whether the targeting of a large number of genes by AID (8, 27, 28) results in uracil accumulation in the genomes of B cells undergoing normal maturation in germinal centers. It is also not known whether the B cell cancers that constitutively express AID at high levels have enhanced repair capabilities that eliminate the excess uracils generated through cytosine deamination. To begin to address such issues, we initiated a study of genomic uracils in normal and malignant B cells from both mice and humans. We quantified the uracils and also determined the ability of these cells to eliminate uracils through DNA repair. We find that while normal B cells undergoing maturation efficiently eliminate most uracils created by AID, most B cell lymphomas and leukemias are unable to eliminate many of the AID-generated uracils despite the presence of uracil excision activity in their nuclei.

MATERIALS AND METHODS

Mammalian tissue harvesting and cell lines.

The experiments involving animals either were performed at the University of Toronto and approved by that university's Animal Care Committee or were done at Wayne State University and approved by its Institutional Animal Care and Use Committee. AID^−/− mice used for some of the experiments were kindly provided by Patricia Gearhart (National Institute on Aging). The CH12F3 cell line was kindly provided by Kefei Yu (Michigan State University). The human cell lines used in this study are described in Table S1 in the supplemental material. Ramos 7 and Ramos 1 were previously referred to as Ramos clone 7-3 and Ramos clone 1-12, respectively (29).

Spleen and liver tissues were harvested from 12-to-18-week-old wild-type (WT), UNG^−/−, AID^−/−, and MSH2^−/− mice on the C57BL/6 background. Single-cell suspensions of splenocytes were cultured in RPMI 1640 media (HyClone) supplemented with 10% fetal bovine serum (HyClone) and 50 μM β-mercaptoethanol (Sigma-Aldrich). Following stimulation of primary splenocytes with 25 μg/ml of lipopolysaccharides (LPS) (Escherichia coli serotype 055:B5; Sigma-Aldrich), the switching of antibodies to the IgG3 isotype made by B cells was confirmed using fluorescence-activated cell sorting (FACS).

All human cell lines were cultured in suspension in RPMI 1640 (HyClone) media supplemented with 10% fetal bovine serum (HyClone) and 1% penicillin-streptomycin. CH12F3 cells were cultured in suspension in RPMI 1640 media supplemented with 10% fetal bovine serum, 50 μM β-mercaptoethanol, and 1% penicillin-streptomycin. CH12F3 cells were stimulated with 1 ng/ml of recombinant human transforming growth factor β1 (TGF-β1) (R&D Systems), 10 ng/ml of recombinant mouse interleukin-4 (IL-4) (R&D Systems), and 1 μg/ml of purified anti-mouse CD40 (eBiosciences) to induce switching to the IgA isotype, and the expected result was confirmed using FACS.

Cancer patient samples, human peripheral B cells, and naive mature tonsillar B cells.

The patient samples were collected according to a protocol approved by the Human Investigation Committee of Wayne State University, and the patient backgrounds are summarized in Table S2 in the supplemental material. Following informed consent, peripheral blood was collected from lymphoma or leukemia patients at the St. John Hospital Van Elslander Cancer Center Lymphoma Clinic. Peripheral blood mononuclear cells (PBMCs) were isolated from patient whole blood using LymphoPrep density-gradient media (ProGen Biotechnick GmbH). Monocytes were depleted from the PBMC fraction by allowing the cells to adhere to a sterile plastic surface for 2 h at 37°C and recovering the nonadherent lymphocyte population. T lymphocytes were removed using magnetic bead separation with Dynabeads pan CD2 (Dynal). FACS analysis confirmed the recovery of >90% B lymphocytes in the remaining cell populations used for gene expression and uracil quantification analysis. Normal blood was collected from donors in apheresis cones or filters and was provided by Martin Bluth (Detroit Medical Center Transfusion Services). Mononuclear cells were isolated from the normal blood samples using Ficoll-Paque Premium density gradient media (GE Healthcare). B cells were separated from other mononuclear cells using human B cell isolation kit II (MACS).

Tonsils were obtained as surgical waste specimens from anonymous patients undergoing resection of hypertrophic tonsils at Children's Hospital of Michigan, with the approval of the Institutional Review Board of Wayne State University. Tonsillar mononuclear cells were isolated using Ficoll-Paque Premium density gradient media (GE Healthcare). These cells were sorted by flow cytometry using IgD-fluorescein isothiocyanate (FITC) (Southern Biotec catalog no. 2032-02), CD19-phycoerythrin (PE)-Cy7 (eBioscience catalog no. 25-0199-42), CD38-allophycocyanin (APC) (Biolegend catalog no. 303510), and CD27-PE (BD catalog no. 555441) antibodies to isolate IgD⁺ CD19⁺ CD38⁻ CD27⁻ B cells. 7AAD (BD 559925) was used to exclude dead cells. Purified naive mature B cells were cultured in RPMI 1640 (HyClone) media supplemented with 10% fetal bovine serum (HyClone) and 1% penicillin-streptomycin and stimulated with 500 ng/ml of CD40-L (Peprotech) and 50 ng/ml of IL-4 (Peprotech). Switching of antibodies to IgG1 was confirmed with FACS at 7 days following stimulation.

Generation of AID^−/− UNG^−/− mice.

AID^−/− female mice were crossed with UNG^−/− male mice to generate AID^+/− UNG^+/− mice (F1). F1 males were bred with F1 females to generate offspring (F2). Genotyping of F2 progeny was carried out using PCR. The presence of the WT AID allele was confirmed using primers AID-811 (5′-CTGAGATGGAACCCTAACCTCAGCC) and AID-G4 (5′-CACGATTTTCTACAAATGTATTCCAGC), and the presence of the disrupted AID allele was confirmed using primers AID-G3 (5′-GGGCCAGCTCATTCCTCCACTC) and AID-G4. The presence of the WT UNG allele was confirmed using primers Up-310 (5′-GCCCATCTTGGAAACTCAAA) and Lower-18 (5′-CCAGTCTGGCTTGGTTACCTTG), and the presence of the disrupted allele was confirmed using primers Up-310 and Neo-1524R (5′-CGTCAAGAAGGCGATAGAA). Six of 150 F2 mice were found to be AID^−/− UNG^−/−.

Uracil quantification assay.

The assay is described schematically in Fig. S1 in the supplemental material. Genomic DNA was extracted using a Qiagen blood and cell culture kit and was digested with HaeIII restriction endonuclease. The endogenous abasic sites in DNA were blocked by incubating with methoxyamine (Mx; Fisher Scientific), and the DNA was ethanol precipitated to remove Mx and treated with E. coli uracil DNA-glycosylase (New England BioLabs) and an aldehyde-reactive probe (ARP; Dojindo Laboratories). It was extracted with phenol-chloroform to remove the enzyme, ethanol precipitated, and passed over G-50 columns (GE Healthcare) to remove residual reagents. An oligonucleotide duplex, 5′-T37UT37/5′-A37GA37, served as a uracil standard, and several dilutions of this DNA were processed in parallel with genomic DNA. The DNA samples were spotted onto a positively charged nylon membrane (Immobilon-Ny+; Millipore) using a vacuum-filtration apparatus (Bio-Rad Bio-Dot vacuum apparatus). The membrane was preequilibrated with StartingBlock buffer (Fisher Scientific) followed by incubation in a solution of streptavidin-Cy5 (GE Healthcare). The membrane was washed several times in TBS-T (25 mM Tris, 3 mM KCl, 140 mM NaCl, 1% Tween 20) and scanned on a Typhoon 9210 Phosphorimager for fluorescence. The DNA from each sample was quantified using SYBR gold dye (Invitrogen) and a Tecan Genios microplate reader. Cy5 fluorescence was analyzed using ImageJ software, and the intensities of the 75-mer duplexes were used to generate a calibration plot of Cy5 fluorescence versus uracil amounts. The raw fluorescence numbers were adjusted for background fluorescence signal, and the numbers of uracils in the genomic DNAs were determined by interpolating their fluorescence intensities in the calibration plot. These amounts were normalized for the amount of DNA loaded for each sample to calculate the number of uracils per 10⁶ bp (1 million bp = ∼1.05 × 10⁻¹⁵ g). The value corresponding to the uracil amount calculated in this way was divided by the “control” DNA value, and this ratio is reported here. In the case of the murine samples, the control was DNA of unstimulated splenocytes from WT mice or DNA from unstimulated CH12F3 cells. For human lymphoma cell lines or tissue samples, the control sample consisted of peripheral B lymphocytes from the blood of volunteers.

RNA extraction and RT-PCR.

Total RNA was extracted from cells using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. Reverse transcription (RT) of RNA was carried out using an oligo(dT) primer and a ProtoScript Moloney murine leukemia virus (M-MuLV) First Strand cDNA synthesis kit (New England BioLabs). To determine relative gene expression levels, the cDNA was amplified utilizing HotStart Taq polymerase and an RT2 SYBR green/ROX kit (SA Biosciences). The primers used for RT-PCR are described in Table S3 in the supplemental material, and the amplification was performed using an Applied Biosystems 7500 fast real-time PCR system.

Construction of AID knockdown cell lines and transient transfection with the AID mutant.

Transfection-grade short hairpin RNA (shRNA) plasmids (KH12741N; SA Biosciences) were prepared using a HiSpeed Plasmid Maxi kit (Qiagen). One million log-phase FSCCL, Raji, and Ramos cells were electroporated with 5 to 15 μg of shRNA-AID or shRNA-scrambled plasmids in 4-mm-long cuvettes (Gene Pulser Xcell). Three million Daudi, DLCL2, and Toledo cells were transfected with shRNA-AID and shRNA-scrambled using FuGeneHD reagent (Promega). Stable clones of shRNA-AID and shRNA-scrambled cell lines were selected with 500 to 700 μg/ml of G418 (HyClone). Two million Ramos 1 cells were transfected with transfection-grade pMSCV-EV (empty vector), pMSCV-E58A, or pMSCV-AID using FuGeneHD reagent (Promega). Cells were cotransfected with pMAX-GFP (Lonza), and the transfection efficiency was ∼60% at 48 h posttransfection.

Uracil excision activity assays.

The whole-cell extracts were prepared by harvesting cells by centrifugation, washing them with buffer I (10 mM Tris-HCl [pH 8.0]), and then resuspending them in buffer II (10 mM Tris-HCl [pH 8.0], 200 mM KCl, 2 mM EDTA, 40% [vol/vol] glycerol, 0.5% NP-40, 2 mM dithiothreitol [DTT]) with protease inhibitor (Halt protease inhibitor cocktail; Thermo Scientific). The mixture was incubated at 4°C for 3 h with gentle rocking, and cellular debris was removed by centrifugation at 25,000 × g at 4°C. The supernatant was used as cell-free whole-cell extract. Nuclear extracts of cells were prepared using a NucBuster protein extraction kit (Novagen, Millipore). Whole-cell and nuclear protein concentrations were measured using a Bio-Rad protein assay kit, and extracts were snap-frozen in liquid nitrogen and stored at −80°C.

To assay uracil excision, 24 pmol of a 6-carboxyfluorescein (FAM)-labeled oligonucleotide (5′-ATTATTAUCCATTTATT) was incubated with extracts for 10 min at 37°C in a 20-μl total reaction mixture with 1 mM EDTA, 1 mM DTT, and 20 mM Tris-HCl (pH 8.0) with or without excess uracil DNA-glycosylase inhibitor (New England BioLabs). A single-stranded substrate was used in these assays to ensure that double-strand-specific glycosylases, including SMUG1, would not act on the uracil. Reactions were stopped by heating to 95°C for 5 min followed by incubation with proteinase K (Qiagen) for 60 min at 50°C. Abasic sites were incised by adding NaOH to 0.1 M and heating to 95°C for 5 min. Reaction products were separated by electrophoresis (15% PAGE, 7 M urea, Tris-borate-EDTA buffer), visualized using a Typhoon 9210 scanner, and quantified using ImageJ gel analysis software. The initial part of each set of reaction mixtures (≤50% product formation) was fitted to a linear regression equation, and the slope of the line was used to calculate the activity as pmol of product/min and was converted to specific activity (pmol of product/min/μg of protein in the extract) based on protein quantification of the extracts.

Nuclear cytosine deamination activity assay.

Nuclear extracts of cells were prepared using a NucBuster protein extraction kit (Novagen, Millipore) according to the manufacturer's instructions. A total of 8 pmol of a FAM-labeled oligonucleotide containing the sequence 5′-ATTATTACCCATTTATT was incubated with nuclear cell extracts. The target cytosine for AID in this oligomer is in a WRC context (W = A/T; R = purine) and is underlined in the sequence. The reaction buffer contained 10 mM Tris-HCl (pH 7.5), 1 mM EDTA, 1 mM DTT, 150 mM NaCl, and RNase A (5 ng/μl). The reaction mixtures were incubated at 37°C with a total reaction volume of 20 μl and terminated by the addition of 1,10-phenanthroline (Sigma-Aldrich) to reach a 5 mM concentration. E. coli UNG (New England BioLabs) (0.5 U) was added to the reactions to excise any uracils introduced through cytosine deamination, and the incubation was continued at 37°C for 60 min. Any abasic sites resulting from uracil excision were cleaved by the addition of NaOH to reach a concentration of 0.1 M and incubation at 95°C for 5 min. Reaction products were analyzed by electrophoresis, and specific activity for cytosine deamination was calculated in the same manner in which uracil excision activity was calculated.

RESULTS

Genomic uracil levels following murine B lymphocyte stimulation.

Total cellular DNA was isolated from lipopolysaccharide-stimulated (LPS-stimulated) splenocytes from 12- to-18-week-old C57BL/6 mice with different genetic backgrounds, and the uracil sites were labeled and quantified using a biochemical method. This method uses Escherichia coli UNG enzyme to excise uracils from DNA and labels the resulting abasic sites with biotin using a chemical called aldehyde-reactive probe that reacts specifically with the aldehydic form of deoxyribose sugar (see Fig. S1A in the supplemental material) (30, 31). The biotins are bound with Cy5-labeled streptavidin and quantified (see Fig. S1A and B). This assay has been previously used to quantify uracils in genomic DNA (32–34). A fluorescence scan of a nylon membrane containing splenocyte DNA after 3 days of stimulation is shown in Fig. 1A. In this and other studies, different amounts of a duplex containing a U·G mispair were also processed in parallel and applied to the membrane to generate a calibration plot (Fig. 1A; see also Fig. S1C). Fluorescence signals from genomic DNA of mice with different genetic backgrounds were converted to uracil amounts using the calibration plot, and then uracil amounts in wild-type (WT) splenocytes prior to stimulation were used to calculate the fold increases in uracil levels. The results from a typical experiment are presented in Fig. 1B.

FIG 1.

Genomic uracils in murine splenocytes. (A) Image of a fluorescence scan of a nylon membrane with equal amounts of DNA from triplicate samples of unstimulated splenocytes (−LPS) and splenocytes stimulated with LPS (+LPS) for 3 days. The splenocytes were obtained from mice with different genetic backgrounds, and the DNAs were labeled with Cy5 at sites of uracils as described in Fig. S1A in the supplemental material. Liver DNAs from the same mice served as negative controls. As shown at the top, different amounts of an oligonucleotide duplex containing uracils were spotted onto the membrane to create a calibration plot (see Fig. S1C). (B) Quantification of uracils in splenocyte DNA following stimulation with LPS. The fluorescence intensities from a membrane similar to the one shown in panel A were used to quantify the uracils and were normalized with respect to the uracil amount in WT splenocytes prior to stimulation. In these and all other experiments reported here, means and standard deviations of the results from triplicate samples are reported. (C) Time course of genomic uracil levels in stimulated WT and AID^−/− splenocytes. Uracils were quantified and the data analyzed as described for panel B. (D) Fluorescence scan of a nylon membrane on which equal amounts of DNAs labeled at uracil sites from unstimulated WT and UNG^−/− mouse splenocytes and unstimulated CH12F3 cells were spotted. DNAs from splenocytes stimulated with LPS or CH12F3 cells stimulated with CIT (anti-CD40 antibody, IL-4, and TGF-β) for 3 days were labeled for uracil quantification in parallel and also spotted onto the membrane. Different amounts of the oligonucleotide duplex standard with uracil were similarly processed and spotted at the bottom of the membrane. (E) Relative genomic uracil levels in LPS-stimulated splenocytes from UNG^−/− and AID^−/− UNG^−/− mice normalized to uracil levels in unstimulated WT splenocytes.

In WT or AID^−/− mice, no change in uracil levels was observed for at least 5 days in repeated experiments (Fig. 1B and C; see also Fig. S2A and B in the supplemental material). This lack of difference between WT and AID^−/− splenocytes throughout the 5-day stimulation period was unexpected since AID mutates many non-Ig genes and causes genome-wide translocations (24). Similarly, no increases in genomic uracils were seen in stimulated splenocytes from MSH2^−/− mice in repeated experiments (Fig. 1A and B; see also Fig. S2A and B). This shows that, despite the role played by mismatch repair in shaping SHM spectra and strand bias and in CSR (10, 17, 35), MMR does not play the primary role in the removal of uracils generated by AID across the genome. This result is consistent with the well-established role of MMR in correcting replication errors and with a previous study that compared relative effects of Ung-dependent repair and MMR on SHMs in some AID target genes (10).

In contrast, the genomic uracil levels increased more than 11-fold in UNG^−/− splenocytes within the first 3 days following stimulation and then decreased back to the levels seen in unstimulated cells by the fifth day (Fig. 1B). The high uracil level in genomes of UNG^−/− splenocytes at day 3 was reproducible (Fig. 1B, D, and E; see also Fig. S2A and B in the supplemental material); in fact, in most experiments, increases of greater than 11-fold were observed (see Fig. S2B and below). The factors likely to contribute to the decrease in genomic uracil levels beyond day 3 include a decrease in AID gene expression, repair of uracils by backup DNA glycosylases, and replicative dilution of incorporated uracils. These data are qualitatively consistent with the increased uracil levels in the immunoglobulin genes of stimulated UNG^−/− B cells (25), but the magnitude of the increase we detected is much greater than that reported previously. To demonstrate that the presence of uracils was caused by AID, we quantified genomic uracils in stimulated splenocytes from AID^−/− UNG^−/− mice. There was no detectable increase in uracil levels in AID^−/− UNG^−/− cells following stimulation, while stimulated UNG^−/− cells acquired ∼15-fold-higher levels of uracils within 72 h (Fig. 1E). Together these data suggest that splenocyte stimulation with LPS causes a rapid and large increase in genomic uracil levels due to AID but that an UNG2-mediated repair pathway eliminates the uracils in WT cells.

AID and UNG2 gene expression following B cell stimulation.

To study the balance between uracil creation by AID and uracil elimination by base excision repair, mRNA levels of AID, UNG2, and SMUG1 genes were determined in stimulated mouse splenocytes over several days. In all genetic backgrounds except AID^−/−, AID gene expression levels increased about 20-fold (∼2% to ∼40% of GAPDH [glyceraldehyde-3-phosphate dehydrogenase]) during the first 3 days following stimulation and then dropped off (Fig. 2A through D). We tested nuclear extracts from WT and UNG^−/− cells for cytosine deamination activity using a fluorescently labeled single-stranded DNA substrate. Deamination activities were similar in the two types of cells and peaked at day 3 (Fig. 2E). Thus, both AID mRNA levels and DNA-cytosine deamination activities follow the same general pattern of increase and decline in the two genetic backgrounds.

FIG 2.

Gene expression and enzyme activities in stimulated splenocytes. The mRNA levels of AID, UNG2, and SMUG1 genes relative to the GAPDH expression level (set at 100) are shown. The gene expression levels were determined prior to and subsequent to LPS stimulation of splenocytes from WT (A), AID^−/− (B), UNG^−/− (C), and MSH2^−/− (D) mice. (E) A time course of DNA-cytosine deamination activity in nuclear extracts prepared from LPS-stimulated splenocytes from WT and UNG^−/− mice. Specific activity was calculated as picomoles of DNA substrate converted to product per microgram of protein in the extract per minute. (F and G) A fluorescently labeled DNA oligomer containing uracil was incubated with nuclear extract from WT or AID^−/− LPS-stimulated splenocytes, and the specific activity of uracil excision was determined. The activity was determined in the presence (+Ugi) or absence (−Ugi) of the UNG enzyme inhibitor UGI.

In WT mice, UNG2 gene expression paralleled AID gene expression, increasing 19-fold compared to its unstimulated level during the first 3 days and decreased thereafter (Fig. 2A). The increases in UNG2 mRNA levels following stimulation of WT splenocytes were correlated with increases in single-strand-specific uracil excision activities in both nuclear extracts (Fig. 2F) and whole-cell extracts (S. Shalhout and A. S. Bhagwat, unpublished results). UGI, a specific inhibitor of the UNG enzymes, almost completely eliminated the uracil excision activity (Fig. 2F). Such upregulation of UNG2 following B cell stimulation has been noted before (36).

Although UNG2 gene expression increased in AID^−/− splenocytes following their stimulation, the magnitude of the increase was smaller than in WT mice (Fig. 2A and B). In contrast to this result, a previous study using transcriptome sequencing (RNA-Seq) found UNG expression to be about the same in stimulated B cells from WT and AID^−/− mice (37). To resolve this issue, we determined uracil excision activity in nuclear extracts from AID^−/− cells and found that it also peaked at day 3 poststimulation but that its magnitude was smaller than the peak activity seen in WT cells (compare Fig. 2F and G). This activity was eliminated when UGI was added to the reaction mixture, showing that it was UNG specific (Fig. 2G). No expression of UNG2 gene was detected in UNG^−/− splenocytes (Fig. 2C), and in MSH2^−/− splenocytes, UNG2 gene expression followed the same pattern as in WT cells (Fig. 2D).

In contrast to UNG2 gene expression, there was no increase in the expression of SMUG1 gene in any of the genetic backgrounds tested (Fig. 2A through D). Similarly, splenocyte stimulation did not lead to large increases in expression of other uracil-specific DNA glycosylases (TDG and MBD4) following stimulation of WT splenocytes (see Fig. S3 in the supplemental material). Thus, despite recent reports that SMUG1 contributes more to uracil excision than UNG2 in mice (36) and that SMUG1 contributes to CSR and SHM in mice (38), we found that UNG2 is the only uracil-specific DNA glycosylase whose expression is upregulated in response to B cell stimulation. Thus, it must be responsible for the repair of an overwhelming majority of uracils created by AID in stimulated WT B lymphocytes. In other words, B cells undergoing maturation ex vivo display homeostasis in genomic uracil levels.

Uracil accumulation in a murine B cell line despite high UNG2 activity.

To determine whether the lack of uracil accumulation in the genome of WT B cells was also true in germinal center-derived cancer cell lines, we studied the CH12F3 cell line. This murine B cell lymphoma-derived line expresses AID upon stimulation with a cocktail of anti-CD40 antibody, IL-4, and transforming growth factor β (CIT) and switches from the IgM isotype to IgA (39). We determined the levels of genomic uracils and expression of AID, UNG2, and SMUG1 genes in this cell line following its stimulation and compared this pattern with the pattern seen in WT and UNG^−/− murine splenocytes.

Three days after stimulation, CH12F3 cells accumulated uracils in their genome to a level about 14 times the unstimulated level (Fig. 3A). As was observed in LPS-stimulated UNG^−/− splenocytes (Fig. 1B), the uracil content decreased after day 3, returning to the unstimulated level by day 10. To directly compare the uracil amounts in the genomes of these two types of murine cells, equal amounts of DNA from primary WT and UNG^−/− splenocytes and from CH12F3 cells were labeled at uracil sites and the samples were spotted on nylon membranes for quantification. A fluorescence scan of the membrane showed roughly equal intensities of spots for DNA from UNG^−/− splenocytes and CH12F3 cells after 3 days of stimulation (Fig. 1D). Quantification of these spots confirmed that the two sets of DNAs had similar levels of uracil content and that these amounts were much higher than those seen with the unstimulated CH12F3 cells and both unstimulated and stimulated WT primary splenocytes (see Fig. S4A in the supplemental material). Therefore, stimulated CH12F3 cells behave like stimulated UNG^−/− cells in terms of uracil accumulation.

FIG 3.

Genomic uracils, gene expression, and enzyme activities in CH12F3 cells. (A) Time course of change in genomic uracil levels in CH12F3 DNA with or without CIT stimulation. (B) Time course of change in AID gene expression and cytosine deamination activity in nuclear extracts prepared from CIT-stimulated CH12F3 cells. (C) UNG2 expression in CH12F3 cells. In panels B and C, AID and UNG2 gene expression is reported relative to GAPDH gene expression (set at 100). (D) Time course of change in uracil excision activity in nuclear extracts prepared from stimulated CH12F3 cells in the presence or absence of UGI. (E) Uracil excision activity in nuclear extracts from stimulated WT splenocytes (+LPS) and CH12F3 cells (+CIT). DNA containing uracil was treated with the indicated amount of extract, and the products were electrophoresed on a denaturing gel following strand cleavage. “Day 0” indicates unstimulated cells, and “Day 3” indicates cells stimulated for 3 days.

Both the AID and UNG2 expression profiles of CH12F3 cells were similar to those of stimulated WT splenocytes (compare Fig. 3B and C with Fig. 2A). The kinetics of AID gene expression in CH12F3 cells was paralleled by nuclear DNA-cytosine deamination activity (Fig. 3B), and the same pattern was seen with UNG2 gene expression and nuclear uracil excision activity (Fig. 3C and D). Furthermore, the peak specific activities of cytosine deamination and uracil excision in the nuclear extracts of CH12F3 cells were comparable to those in WT splenocytes (compare Fig. 2E with Fig. 3B and Fig. 2F with Fig. 3D). SMUG1 gene expression did not change detectably following CH12F3 stimulation (Shalhout and Bhagwat, unpublished). Thus, CH12F3 cells accumulated uracils in their genome following their stimulation despite the fact that their AID and UNG2 gene expression and nuclear enzymatic activity profiles were similar qualitatively and quantitatively to those in WT splenocytes.

To directly compare gene expression in CH12F3 cells with gene expression in murine splenocytes, real-time PCR was performed on equal amounts of cDNA from the two cell types using the same set of primers and internal controls. The results showed that in both of the cell types, AID and UNG2 gene expression peaked at day 3 and the peak mRNA levels of both the genes were about 20% to 30% lower in CH12F3 cells than in the splenocytes (see Fig. S4B and C in the supplemental material). To compare the UNG2 excision activities of the two cell types, we performed parallel uracil excision assays with nuclear extracts from WT splenocytes and CH12F3 cells prior to stimulation and at 3 days following stimulation and electrophoresed the products on the same gel. The gel showed that CH12F3 cells had uracil excision activity comparable to that of WT splenocytes (Fig. 3E). Thus, stimulated CH12F3 cells show AID gene expression, deamination activity, UNG2 gene expression, and uracil excision activity similar to those seen with stimulated WT splenocytes but accumulate uracils like stimulated UNG^−/− splenocytes.

Genomic uracil levels of human tonsillar B cells and AID-expressing B cell lines.

To extend these observations to human B cells, we obtained naive B cells from tonsils of two human donors, stimulated them with anti-CD40 antibody and IL-4, and monitored AID expression and genomic uracils over 7 days. AID expression increased substantially in both sets of samples after 4 days of stimulation, but the magnitudes of the increase were different in the two samples (Fig. 4A). Regardless, the genomic uracil level changed little over the 7 days in either sample (Fig. 4B). This shows that, like the WT murine splenocytes, normal human B cells maintain their genomic uracil levels following stimulation despite an increase in AID expression.

FIG 4.

AID gene expression and genomic uracil levels in stimulated human tonsillar B cells. (A) Relative AID expression levels in naive mature B cells from two independent donor tonsils. Isolated B cells were stimulated with anti-CD40 antibody and IL-4 for 7 days. AID expression is normalized to GAPDH expression (with the day 0 level set to 1.0). (B) Quantification of the genomic uracil levels in the ex vivo-stimulated tonsillar B cells from panel A relative to uracil levels in DNA from unstimulated cells (with the day 0 level set to 1.0).

To assess whether the unexpected correlation between high AID levels, high UNG2 activity, and high genomic uracil levels in CH12F3 cells extends also to human B cell cancers, we studied 13 cell lines created from human hematopoietic malignancies. They included seven lines from cancers of GC origin—two Burkitt lymphoma (BL) lines, two follicular lymphoma (FL) lines, and three diffuse large B cell lymphoma (DLBCL) lines. Additionally, three hematopoietic cell lines not derived from GC—a multiple myeloma (MM) B cell line and two T cell lines—were included. Three classical Hodgkin's lymphoma (cHL) cell lines were also included in this study. Although many cHL tumors show some evidence of GC development (40), they usually do not express AID at high levels (41). These cell lines are described in Table S1 in the supplemental material. These experiments used peripheral B lymphocytes from healthy human volunteers instead of tonsillar B cells as negative controls because of the availability of the former cell type in larger quantities.

There were clear differences between the group containing BL, DLBCL, and FL cell lines and the other lines in terms of both AID gene expression levels and uracil levels. While the AID gene expression levels in the group containing BL, DLBCL, and FL ranged from about 4% to 8% relative to GAPDH gene expression (“high-AID group”) (Fig. 5A, top panel, closed bars), the non-GC-derived lines, cHL lines, and normal cells had AID gene expression levels below 1% (“low-AID group”) (Fig. 5A, top panel, open bars). When nuclear DNA-cytosine deamination activities were determined from several high-AID and low-AID cell lines, the results were consistent with the AID mRNA levels in the cell lines (compare the top panel of Fig. 5A with Fig. 5B).

FIG 5.

AID gene expression, cytosine deamination activity, and genomic uracil levels in human cancer cell lines. Abbreviations: cHL, classical Hodgkin's lymphoma; T-Cell, T-cell lymphoma/leukemia; MM, multiple myeloma; FL, follicular lymphoma; DLBCL, diffuse large B cell lymphoma. Raji and Daudi are Burkitt lymphoma-derived cell lines. The cell lines are numbered and are described in Table S1 in the supplemental material. (A) AID gene expression and genomic uracil levels in 13 human lymphoma-derived lines and circulating human B cells. (Top panel) The AID expression levels are shown relative to the GAPDH gene expression level (set at 100). The asterisk indicates no detectable expression. (Bottom panel) Genomic uracil levels relative to those in circulating human B lymphocytes (set at 1.0). (B) Nuclear cytosine deamination activity for some of the cell lines. The asterisks indicate that no deamination activity was detected under these conditions. (C) Uracil levels in the Raji cell genome over several passages. The uracil level in Raji cells at passage 1 (P1) was set at 100.

The genomic uracil levels were also quite different between the two groups—the high-AID cells contained uracils at ∼80- to 120-fold the level found in normal peripheral B cells (Fig. 5A, bottom panel, closed bars). In contrast, some of the low-AID B cell lines contained genomic uracils at levels that were only two to five times those seen with normal peripheral B cells. The T cell lines had uracil levels similar to those of peripheral B cells (Fig. 5A, bottom panel, open bars). When the genomic uracil levels in these cell lines were plotted against their AID gene expression levels, the high-AID and low-AID lines clearly separated into distinct groups (see Fig. S5A in the supplemental material). Stimulated tonsillar cells had somewhat higher uracil levels than peripheral B cells, and the low-AID lines had uracil levels comparable to those found in the tonsillar cells (see Fig. S6).

Genomic instability is a hallmark of cancer (42), and it seemed possible that the B cell cancer cell lines are not stable with regard to genomic uracil content. In particular, it is possible that continual cytosine deamination by AID increases the uracil content of high-AID cell lines with time. However, when the Raji cell line was followed over many passages, no consistent increase or decrease in the uracil content was observed (Fig. 5C). Thus, the B cell lymphoma-derived high-AID cell lines displayed stable uracil content, albeit at much higher levels than normal B cells.

Uracil excision activity in human B cell cancers overexpressing AID.

One possible cause of elevated uracil levels in high-AID cell lines and tumor samples is a reduction in uracil excision activity. To test this possibility, we performed real-time PCR measurements of UNG2 and SMUG1 mRNA on all human cell lines as well as uracil excision activity assays on several cell lines. The mRNA quantification studies showed that high-AID cell lines also expressed UNG2 and SMUG1 genes at moderately higher levels than peripheral B cells (Fig. 6A). While the higher UNG2 expression seen in these cells was similar to the increased UNG2 expression seen in stimulated WT murine splenocytes and CH12F3 cells, the increase in SMUG1 expression was not observed in murine cells (Fig. 2; see also Fig. S3 in the supplemental material) (Shalhout and Bhagwat, unpublished).

FIG 6.

UNG2 expression and uracil excision activity in human lymphoma cell lines. (A) The UNG2 and SMUG1 expression levels in the same human cell lines as described for Fig. 5A relative to the GAPDH expression level. (B) Uracil excision activity in nuclear extracts prepared from a subset of the human lymphoma cell lines in the absence of UNG inhibitor UGI (left) or in its presence (right).

The uracil excision activity assay was performed for two low-AID lines and seven high-AID lines in the presence or absence of UGI. The high-AID nuclear extracts showed 50% to 100% more uracil excision activity than the low-AID extracts tested (Fig. 6B, left panel). This activity was abolished when UGI was included in the reactions (Fig. 6B, right panel), confirming that it was due to UNG2. Furthermore, the specific activities of the nuclear extracts from high-AID cell lines were comparable to the peak activity seen in nuclear extracts from murine splenocytes (compare Fig. 2F and 6B). These results show that the UNG2 gene expression is not defective in high-AID human B cell cancer cell lines and that they contain nuclear uracil excision activity comparable to that seen with stimulated WT murine splenocytes. Despite this activity, they contain very high levels of uracils in their genome.

Genomic uracil levels change with AID level.

We wanted to determine whether increasing or decreasing AID gene expression in human cell lines caused commensurate changes in genomic uracil levels. To investigate this, we chose previously generated variants of the Ramos cell line that express AID at different levels (29). These included cell lines Ramos 7 (high AID gene expression) and Ramos 1 (low AID gene expression) (29) and lines created through transfection of a plasmid expressing AID into Ramos 1 (Ramos A.1, A.2, and A.5) or the empty vector (Ramos C.1) (43).

There was good correlation between AID gene expression and genomic uracil levels in these cells. For example, AID gene expression was much higher in Ramos 7 than in Ramos 1 cells and was higher in Ramos A.2 and A.5 cells than in Ramos C.1 cells (Fig. 7A, top panels). Correspondingly, the genomic uracil level in Ramos 7 cells was higher than in Ramos 1 cells, and there was a similar relationship between genomic uracil levels in the Ramos A.2 and A.5 lines and in Ramos C.1 cells (Fig. 7A, bottom panels). Thus, Ramos-derived cell lines with higher AID expression show a pattern of higher genomic uracil levels as seen in the high-AID group of cell lines (Fig. 5A). To test whether the catalytic activity of AID was required for its ability to increase genomic uracil levels, a catalytically inactive version of the AID (E58A) was transfected into Ramos 1 cells. Both WT AID and E58A AID gene transfection increased AID mRNA levels (Fig. 7B, left panel), but only the catalytically active version increased genomic uracil levels (Fig. 7B, right panel).

FIG 7.

AID gene expression and genomic uracil levels in Ramos-derived lines. (A) (Top panel) AID expression in Ramos 1 and Ramos 7 lines and in the Ramos 1-derived lines transfected with empty vector (Ramos C.1) or with an AID expression plasmid (Ramos A.1, A.2, and A.5). (Bottom panel) Genomic uracil levels in the same cell lines. In both cases, the numbers were normalized relative to uracils in normal circulating human B cells (set to 1.0). The asterisks indicate undetectable expression. (B) (Left panel) Relative expression levels of WT AID and the AID E58A mutant following transient transfection of Ramos 1 cells. The AID level in Ramos 1 transfected with an empty vector was set to 1.0. (Right panel) Genomic uracil levels in the same transient-transfection experiment relative to uracils in the empty vector control (set to 1.0). (C) Effects of AID knockdown on human lymphoma cell lines. The AID mRNA levels and genomic uracil levels are presented as the percentages of the mean values for the line expressing scrambled shRNA. (Top panel) AID expression in cell line expressing AID-specific shRNA. (Bottom panel) Genomic uracil levels in the same cell lines.

Interestingly, UNG2 gene expression was highest in the three lines expressing AID at very high levels, Ramos 7, Ramos A.2, and Ramos A.5 (see Fig. S7A in the supplemental material). This suggests that when Ramos cells were transfected with AID-expressing plasmids, endogenous UNG2 gene expression levels may have lagged behind increasing AID expression levels, allowing the genomic uracil levels to rise, but that the uracil excision ability eventually caught up with uracil generation by AID stabilizing the genomic uracil content at a higher level than before.

To study the effects of reducing AID gene expression on genomic uracil levels, human cell lines expressing AID at high levels were transformed with plasmids expressing AID-specific shRNA. The levels of success of the knockdown (KD) differed among the cell lines, but AID gene expression in all the lines was reduced by ∼75% or greater compared to the expression seen with a control line expressing scrambled shRNA (Fig. 7C, top panel). All the AID KD lines had substantially reduced genomic uracil levels compared to their scrambled shRNA counterparts, but the degrees of reduction were different for different cell lines (Fig. 7C, bottom panel). The variability in the reduction in genomic uracil levels among different KD lines may have been due to differences in genetic backgrounds between the cell lines and the fact that they represent three different B cell lymphomas. For example, SMUG1 levels or dUTP pools may be different in different cell lines and this may have secondary effects on genomic uracil levels. It should be noted that UNG2 gene expression changed in the same direction as the change in AID gene expression in these cell lines (see Fig. S7B in the supplemental material).

Genomic uracil content in AID-expressing human B cell tumors.

We obtained tumor samples from six pretreatment lymphoma patients (called patients P5 through P10) (see Table S2 in the supplemental material) and determined the levels of AID gene expression and genomic uracil in their cells. AID gene expression was higher in FL and DLBCL patient tumors than in the three marginal-zone lymphoma (MZL) patient samples (Fig. 8A, open bars). MZL tumors are thought to originate from post-GC marginal-zone B cells (44); hence, the low AID gene expression in these tumors is consistent with their origin. The patients from whom the tumor samples were obtained were unrelated individuals, and their cancers were at different stages and had different cytogenetic markers (see Table S2). Consequently, it is not surprising that the patients within each group (patients with DLBCL or FL and patients with MZL) showed considerable variation in AID gene expression levels (Fig. 8A). Regardless, the genomic uracil levels were much higher in FL and DLBCL tumors than in normal B cells and the MZL tumors. The uracil levels in FL and DLBCL tumors were comparable to those seen in most FL and DLBCL cell lines (compare Fig. 5A and 8A, closed bars). In contrast, the MZL tumors had amounts of uracils in their DNA that were only slightly higher than those seen with normal B cells (Fig. 8A, closed bars).

FIG 8.

Gene expression and genomic uracil levels in patient tumor samples. Abbreviations: MZL, marginal-zone lymphoma; FL, follicular lymphoma; DLBCL, diffuse large B cell lymphoma; CLL, chronic lymphocytic leukemia. The patients and the numbering are described in Table S2 in the supplemental material. (A) AID expression levels and uracil levels in lymphoma patient samples. Genomic uracil levels are indicated relative to those in normal circulating human B lymphocytes (set at 1.0). (B) AID expression levels and uracil levels in CLL patient samples. (C) UNG2 and SMUG1 gene expression in the same CLL patient B cells. Expression data are normalized to GAPDH expression set at 100.

We also obtained tumor samples from eight patients with chronic lymphocytic leukemia (CLL; patients P1 through P4 and P11 through P14) (see Table S2 in the supplemental material). CLL tumors are known to be heterogeneous with respect to AID gene expression (45, 46), and we also found a range of AID gene expression in the CLL patient samples. In contrast, there was little tumor-to-tumor variation in UNG2 and SMUG1 levels (Fig. 8C). However, there was a strong correlation between AID gene expression and genomic uracil levels (Fig. 8B). When the data for all 14 patients were combined, there was a monotonic relationship between the AID gene expression level and the genomic uracil level (Spearman's nonparametric rank correlation coefficient ρ = +0.95) (see Fig. S5C in the supplemental material). Thus, the correlation observed between elevated AID gene expression and accumulation of uracils in the genomes of human cell lines holds also for the patient tumor samples (see Fig. S5A and C).

DISCUSSION

AID participates in the only known biological process in which an organism damages its genomic DNA bases for programmed alterations in its genetic information. Previously, genetic evidence and in vitro biochemistry of AID had established the hypothesis that AID converts cytosines in DNA to uracils. Cellular attempts to repair uracils create mutations and strand breaks that are the molecular cause of SHM and CSR. However, the interplay between uracil creation by AID and removal by DNA glycosylases or mismatch repair machinery had not been studied in either normal B cell maturation or B cell cancer development. We examined this interplay here and arrived at some novel conclusions.

First, we showed that UNG^−/− splenocytes acquired substantial amount of uracils upon stimulation and that this increase was abolished when both AID alleles were inactivated (Fig. 1E). Although this was not unexpected given the DNA-cytosine deamination activity of AID in vitro (1–4), it is the first direct demonstration that AID causes uracil accumulation in B cell genomes. Using transient-transfection assays, we also showed that the catalytic ability of AID was required for its ability to increase genomic uracil levels (Fig. 7B). We cannot eliminate the possibility that, like one report in yeast (47), high transcription in stimulated B cells causes increased utilization of dUTP during replication. However, our results clearly show that AID in stimulated cells deaminates genomic cytosines and that this is the principal cause of increase in genomic uracils.

Second, most of the uracils introduced by AID must lie outside the Ig genes and their total number may be large. In UNG2-defective splenocytes, the genomic uracil levels increase following LPS stimulation by a factor of about 11 to 60 (Fig. 1B and E; see also Fig. S2B and S4A in the supplemental material). To ensure that the uracil quantification comparisons between different cells were valid, we assayed every DNA sample in triplicate, processed all the samples in an experiment in parallel, applied them to the same nylon membrane, and included the complete range of uracil standards on every membrane (Fig. 1A and D; see also Fig. S2A). However, we have refrained from stating a specific number for the genomic uracils partly because our assay uses convenient but nonphysiological DNA as the uracil standard. Also, a range of values from ∼200 (48) to ∼15,000 (49) has been reported for the amount of uracil in the unstimulated lymphocyte genome. Thus, the genomic uracil load in stimulated UNG^−/− B lymphocytes is at least several thousand per haploid genome and may be much higher. As Maul et al. (25) found only ∼1 extra uracil per 1 kbp of V(D)J and switch regions of Ig genes following B cell stimulation, most of the AID-generated uracils must lie outside the Ig genes.

Third, when WT murine or human B lymphocytes were stimulated ex vivo, they kept a perfect balance between uracil creation and elimination. In fact, the genomic uracil content of LPS-stimulated murine WT cells was similar to that of LPS-stimulated AID^−/− cells within the sensitivities of our assay (Fig. 1C). The lack of increase in uracil levels was seen with three different stimulation regimens, LPS alone (Fig. 1), LPS plus IL-4 (see Fig. S8 in the supplemental material), and IL-4 plus anti-CD40 antibody (Fig. 4). The absence of a global increase in uracils in stimulated WT splenocytes suggests that UNG2-dependent repair pathways quickly eliminate nearly all the uracils introduced by AID in the murine genome, establishing genomic uracil homeostasis during their germinal-center maturation (Fig. 9, upper panel).

FIG 9.

Homeostasis in genomic uracil creation and excision. (Top panel) Homeostasis during normal B cell development in germinal centers. Uracil creation by AID and uracil excision by UNG2 are kept in balance during normal B cell-development, resulting in low genomic uracil levels. (Bottom panel) The balance between uracil creation and removal is lost during the development of precancer B cells in favor of uracil creation. This is a “DNA repair crisis” for the precancerous cells leading to accumulation of mutations and strand breaks in addition to uracils and resulting in genomic instability. This may lead to cellular transformation. When uracil removal equalizes with uracil creation, stable lymphoma or leukemia cell lines are established. These cell lines maintain the high uracil levels introduced during the repair crisis.

Fourth, there is a functional disconnect between UNG2-specific uracil excision activity and genomic uracil levels in both human and murine B cell cancers expressing AID at high levels. While genomic uracil levels increased in primary B cells following their stimulation only when the UNG2 gene was inactivated, the B cell cancers had high uracil levels despite UNG2 gene expression and robust nuclear uracil excision activity. In fact, the nuclear excision activities in stimulated WT splenocytes and high-AID cell lines were comparable (Fig. 2F and 6B). This was particularly striking in stimulated CH12F3 cells. At day 3 following their stimulation, CH12F3 cells and WT splenocytes had comparable levels of AID gene expression (see Fig. S4B in the supplemental material), UNG2 gene expression (see Fig. S4C), and UNG2-specific nuclear uracil excision activity (Fig. 2F and 3D and E). Despite this, only CH12F3 cells accumulated genomic uracils. Similarly, when the Ramos 1 cell line was transformed with an AID expression vector to produce cell lines with higher AID expression, both UNG2 gene expression and genomic uracil levels in these cells increased (Fig. 7A; see also Fig. S6A).

There are several possible reasons for the disconnect between UNG2-specific uracil excision activity and genomic uracil levels in cancer cells. The principal function of UNG2 in cells is to excise uracils introduced into DNA during replication through dUTP utilization by replicative polymerases (15, 36). Consequently, UNG2 protein in B cells may have to be recruited to loci where AID acts. The recruitment may occur through a combination of chromatin modifications, protein modifications, and the help of molecular chaperones. It is known that when the Ig locus undergoes SHM and CSR, the local chromatin undergoes epigenetic modifications (50), and a protein factor may recruit UNG2 to these chromatin tags. Alternatively, since the action of AID requires transcription of target genes (51), UNG may be recruited to the transcript elongation complexes. One candidate for such a chaperon is translesion synthesis polymerase Rev1, which was shown to have a structural role in bringing UNG2 to switch regions in an AID-dependent manner during CSR (52). Additionally, UNG2 or some accessory protein required for its action may have to be covalently modified for it to be effective in the removal of uracils generated by AID. The apparent ineffectiveness of UNG2 enzyme in CH12F3 in eliminating uracils suggests that one or more of these factors may be downregulated in these cells.

Fifth, the uracil levels in the Raji cell line did not significantly increase over many passages during an 18-month period. This means that genomic uracil levels in this high-AID cancer cell line are fairly stable. It is likely that if uracil levels were to continue to increase in high-AID cancer lines, they would become unstable through acquisition of mutations, strand breaks, and other genetic abnormalities. Therefore, our results suggest that when a cell line is established from a B cell tumor expressing AID at high levels, a new steady-state level of genomic uracil is reached at which uracil creation and excision are balanced. However, this steady-state uracil level is much higher than that found in pre-B or peripheral B cells and does not represent homeostasis in genomic uracils. Increasing or decreasing AID levels in these cells changes the genomic uracil levels in the same direction (Fig. 7).

To explain these observations, we propose that some B cells lose the coordination between uracil creation and elimination during their germinal-center maturation. This could happen because of poor targeting of UNG2 protein to AID-generated uracils. This would lead to a diminished ability to excise uracils created by AID, creating a “DNA repair crisis” (Fig. 9, lower panel). If this situation continues for a number of cell generations, it would probably lead to high genetic instability, accumulation of mutations, and cell death. However, the excision ability in some cells eventually catches up with uracil creation due to a combination of falling AID levels, increasing UNG2 levels, UNG2 modifications, and synthesis of a molecular chaperone, resulting in the establishment of a stable cell line. This new steady-state condition occurs at a uracil level that is much higher than the level in unstimulated cells (Fig. 9, lower panel). In contrast, B cells from lymphoma and leukemia patients are in the midst of the DNA repair crisis mentioned above and have not reached steady-state conditions. Their AID, UNG2, and genomic uracil levels may change with time and are different for different patients. This is the likely source of the variability of these parameters within patient groups with nominally similar diseases (Fig. 8A and B) and implies that patient tumor samples with high AID levels may not have uniformly high UNG2 expression or genomic uracil levels.

An AID-UNG2 imbalance in favor of the former enzyme is likely to increase transition mutations. In UNG^−/− murine cell lines that did not express AID, the lacI and HPRT genes, respectively, showed mutations rates 1.5-fold and 5-fold higher than those seen with WT cells (15, 53). Due to the high AID levels in B cells undergoing maturation, a deficiency in UNG expression could transiently result in much higher mutation frequencies and this could activate oncogenes and inactivate tumor suppressors. It is known that UNG^−/− mice show an incidence of B cell lymphomas that is more than 20-fold higher than that seen with WT mice and have a shorter life span (54). However, the genetic changes in these tumors have not been characterized (54). A recent study of follicular lymphomas in human patients found that both the V(D)J and non-Ig targets of AID continued to acquire mutations during tumor evolution (55), but more-extensive deep-sequencing studies are needed to confirm that high-AID tumors evolve by acquiring mutations biased toward C-to-T transitions.

AID is not the only member of the AID/APOBEC (apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like) family of DNA-cytosine deaminases that is thought to act on cancer genomes and cause mutations. Analysis of cancer genome sequences has revealed that a number of cancers carry base substitution mutations with AID/APOBEC “signatures.” This conclusion is based mostly on the sequence context and types of base substitutions found and includes breast, bladder, and cervix cancers in addition to those of hematopoietic origin (56–58). This analysis is strengthened by the observation that many breast cancer cell lines and tumor samples express APOBEC3B at high levels (59). In at least two breast cancer cell lines, genomic uracil levels were also elevated (59). At present, it is not known whether occasional overexpression of APOBEC3B or any other APOBEC gene whose product is transported into the nucleus causes a compensating increase in uracil excision activity and maintenance of genomic uracil homeostasis. If the development of nonhematopoietic cancers overexpressing APOBECs also involves the type of DNA repair crisis outlined above (Fig. 9), then understanding how normal B cells undergoing germinal-center maturation strike a balance between uracil creation by AID and uracil elimination by UNG2 may be critical to understanding how many cancers develop.