Uracil-DNA Glycosylase in Base Excision Repair and Adaptive Immunity

Genomic uracil is a DNA lesion but also an essential key intermediate in adaptive immunity. In B cells, activation-induced cytidine deaminase deaminates cytosine to uracil (U:G mispairs) in Ig genes to initiate antibody maturation. Uracil-DNA glycosylases (UDGs) such as uracil N-glycosylase (UNG), single strand-selective monofunctional uracil-DNA glycosylase 1 (SMUG1), and thymine-DNA glycosylase remove uracil from DNA. Gene-targeted mouse models are extensively used to investigate the role of these enzymes in DNA repair and Ig diversification. However, possible species differences in uracil processing in humans and mice are yet not established. To address this, we analyzed UDG activities and quantities in human and mouse cell lines and in splenic B cells from Ung+/+ and Ung−/− backcrossed mice. Interestingly, human cells displayed ∼15-fold higher total uracil excision capacity due to higher levels of UNG. In contrast, SMUG1 activity was ∼8-fold higher in mouse cells, constituting ∼50% of the total U:G excision activity compared with less than 1% in human cells. In activated B cells, both UNG and SMUG1 activities were at levels comparable with those measured for mouse cell lines. Moreover, SMUG1 activity per cell was not down-regulated after activation. We therefore suggest that SMUG1 may work as a weak backup activity for UNG2 during class switch recombination in Ung−/− mice. Our results reveal significant species differences in genomic uracil processing. These findings should be taken into account when mouse models are used in studies of uracil DNA repair and adaptive immunity.

Uracil is occasionally incorporated instead of thymine during DNA synthesis and is in addition generated by spontaneous and enzymatic deamination of cytosine. In proliferating cells, misincorporation of dUMP (generating U:A pairs) is likely the most frequent route to DNA uracil, estimated to ϳ10 4 uracil residues in the human genome per cell generation (1). Spontaneous deamination has been calculated to occur at a rate of 70 -200 events per cell per day (2) but is considered more harmful because of the generation of mutagenic U:G mispairs. In activated B cells, cytosine residues within specific regions in the Ig genes are deaminated by activation-induced cytidine deaminase. This is an essential step in somatic hypermutation and class switch recombination (CSR) 2 that generates antibodies with increased antigen affinity and altered effector functions, respectively. Hence, activation-induced cytidine deaminase-generated U:G mismatches serve as key intermediates in adaptive immunity, and further processing requires the involvement of DNA repair proteins that otherwise have antimutagenic functions (2)(3)(4)(5). The mechanisms regulating processing of uracil by error-free repair or mutagenic/recombinogenic pathways are presently not well understood.
Mammalian cells express several UDGs, including mitochondrial UNG1 and nuclear UNG2, SMUG1, TDG, and MBD4 (2). These are all capable of cleaving the N-glycosidic bond between the deoxyribose and uracil, creating a substrate for the base excision repair (BER) pathway (2). UNG2 removes misincorporated uracil by postreplicative DNA repair (6) but also has a role in repair of U:G mismatches (7) and in Ig diversification (3). Substrate specificities of different UDGs and their preference for various DNA contexts partially overlap (2). However, their distinct functions in various genomic contexts in vivo remain poorly understood. The initiating uracil-DNA glycosylase influences the downstream processing of the uracil site (8). Thus, mapping the contribution of each UDG enzyme is likely the key to understanding how processing of uracil is routed into BER versus mutagenic processing (somatic hypermutation and CSR).
Gene-targeted mice are valuable models for the function of orthologs in man and are used extensively to investigate mechanisms involving genomic uracil processing. Ung Ϫ/Ϫ mice show a 20-fold increased risk of developing B cell lymphomas (9,10), which suggests that disrupted uracil processing is involved in development of B cell malignancies. In humans, the level of * This work was supported by The Research Council of Norway (National Program for Research in Functional Genomics); The Norwegian Cancer Association; the Cancer Fund at St. Olavs's Hospital in Trondheim, Norway; and The Svanhild and Arne Must Fund for Medical Research. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Figs. S1-S11 and Tables S1-S6. 1 To whom correspondence should be addressed: switched isotypes (IgG and IgA) in serum is severely reduced in hyper-IgM patients lacking functional UNG (4). In comparison, although class switching of B cells from Ung Ϫ/Ϫ mice is inhibited in vitro, Ung Ϫ/Ϫ mice display only a partial reduction (30 -50%) of IgG and IgA in serum (3,5,11). This demonstrates that class switching in mice occurs in the absence of UNG although at low levels. CSR therefore seems to be less UNG-dependent in mice than in humans, suggesting that the various UDGs may contribute differentially to processing of uracil in the two species. Indeed, significant species differences were recently demonstrated for other DNA repair/genomic maintenance pathways, which may account for the divergent phenotypes of human and mouse cells lacking the Werner syndrome helicase (12).
Here we examined species differences in genomic uracil processing between man and mouse. Activities and quantities of the major UDGs as well as total uracil BER activities were analyzed in a panel of human and mouse cell lines. Moreover, we characterized the catalytic properties of purified recombinant human and mouse UNG2, SMUG1, and TDG proteins. Finally, UNG and SMUG1 activities were monitored in CSRactivated B cells isolated from spleen from Ung ϩ/ϩ and Ung Ϫ/Ϫ backcrossed mice.
Our results show that overall uracil excision capacity was ϳ15-fold higher in human cells compared with mouse cells. Surprisingly, whereas BER of misincorporated uracil (U:A) was higher in the human cell lines, BER of deaminated cytosine (U:G) was not significantly different between human and mouse cells. A markedly higher U:G activity of SMUG1 likely explains this difference. At the protein level, SMUG1 and TDG were the most abundant uracil-DNA glycosylases in mouse cells, whereas UNG dominated in human cells. Moreover, we showed that SMUG1 (activity per cell) was not down-regulated in activated mouse B cells and may thus help compensate for UNG deficiency during CSR in Ung Ϫ/Ϫ mice. In summary, our results demonstrate that there are species differences in uracil processing in mouse and man. This is important to consider when using knock-out and transgenic mouse models in studies of genomic uracil repair and adaptive immunity.

EXPERIMENTAL PROCEDURES
Cell Culture and Preparation of Whole Cell Extracts-Cell lines and growth media are listed in Table 1. The keratinocyte cell line HaCaT was provided by Dr. Norbert E. Fusenig (German Cancer Research Centre, Heidelberg, Germany), and FUJ is a lymphoblastoid cell line derived from UNG-deficient patient number 2 (13). If not otherwise indicated, cell lines were purchased from American Type Culture Collection (ATCC). All cell media were supplemented with 10% FBS, 2 mM L-glutamine, 2.3 g/ml Fungizone, and 0.1 mg/ml gentamicin. Cell lines were cultured at 37°C at 5% CO 2 . Cells from 40 -60 plates (15 cm) were harvested at 50 -70% confluence and counted in a Bürker chamber. Whole cell extracts were prepared as described (14). Protein concentrations were measured using the Bradford method (Bio-Rad). Extracts were aliquoted, snap frozen in liquid N 2 , and stored at Ϫ80°C.
An oligonucleotide UDG assay (U:G substrate) was performed as described (7). Briefly, carboxyfluorescein-labeled uracil-containing oligonucleotide (CATAAAGTGUAAAGC-CTG) was annealed to the complementary strand containing G opposite U. Activity was measured in 10-l assay mixtures containing (final) 20 nM U:G substrate, 1ϫ UDG buffer, 10 nM APE1 (apurinic/apyrimidinic endonuclease), and various amounts of cell extract or recombinant UDGs and incubated at 37°C for 10 -60 min (supplemental Figs. S1, S2, and S3). UNG and SMUG1 were inhibited by 0.1 g of uracil-DNA glycosylase inhibitor (Ugi) and 0.1-2 g of neutralizing SMUG1 IgG (PSM1) per assay, respectively. Preimmune rabbit IgG was included in controls. Apurinic/apyrimidinic sites were cleaved by 50 l of 10% piperidine followed by heating at 90°C for 20 min. Substrate and product were separated by PAGE, laserscanned in a Typhoon Trio imager, and analyzed using ImageQuant TL software (GE Healthcare).
BER Assay-Substrates (cccDNA) were prepared as described (15). For BER incorporation assays, 250 ng of cccDNA (0.118 pmol) were incubated with 20 g of whole cell extract at 37°C for 25 min in 40-l reactions containing (final) 40 mM HEPES-KOH, pH 7.8, 70 mM KCl, 5 mM MgCl 2 , 0.5 mM DTT, 250 M NAD ϩ , 2 mM ATP, 4.4 mM phosphocreatine, 2.5 g of creatine kinase, 0.1 g/l BSA. For substrates containing U:A, the reactions were supplemented with 50 nM dTTP and 3Ci of [␣-33 P]dTTP. U:G substrate reactions were supplemented with 50 nM dCTP and 3Ci of [␣-33 P]dCTP. The concentrations of the other dNTPs were 10 M. Reactions were stopped by addition of (final) 25 mM EDTA, 0.5% SDS, 150 g/l proteinase K (37°C for 30 min). DNA was purified by phenol-chloroform extraction and ethanol precipitation with 10 g of glycogen as carrier, resuspended in the buffer recommended, and digested with restriction endonucleases XbaI and HincII. Following PAGE (15%), fixation, and drying, restriction fragments were quantified using ImageQuant software (Fujifilm).
Expression and Purification of His-tagged Proteins-Constructs for expression of His-tagged human UNG2 (pET28a-hUNG2) and human SMUG1 (pET28a-hSMUG1) were published previously (13,16). The human TDG construct pPRS202b (17) and pET28a-mTDG were a gift from Professor Primo Schär (Department of Biomedicine, Institute of Biochemistry and Genetics, University of Basel, Basel, Switzerland). Human TDG cDNA was subcloned into the BamHI and SalI sites of the pET28A vector (Novagene), generating pET28a-hTDG. The mouse UNG2 expression construct pTrcHis-mUNG2 was a gift from Dr. Javier Di Noia (Institut de Recherches Cliniques de Montréal, Montréal, Canada), and cDNA was subcloned into the EcoRI and NheI sites of pET28a, generating the pET28a-mUNG2 expression construct. The mouse SMUG1 reading frame was PCR-amplified from the cDNA clone IRAVp968A10146D using primers generating a 5Ј NdeI site and a 3Ј HindIII site. The PCR product was cloned into pET28a, generating pET28a-mSMUG1. All constructs were verified by DNA sequencing.
Generation of Polyclonal Antibody against Mouse UNG2-The antibody against mouse UNG2 was prepared by subcutaneous injection of 100 g of purified recombinant mouse UNG2 in Freund's complete adjuvant (Sigma-Aldrich) into New Zealand White rabbits. Three subsequent booster injections were given with 2-3-week intervals. Antiserum was collected 15 days after the last immunization. The IgG fractions were purified with HiTrap protein A HP columns (GE Healthcare) and further affinity-purified over a HiTrap N-hydroxysuccinimide-activated HP column (GE Healthcare) coupled to recombinant UNG2. The purified antibody against mouse UNG2 was named PUMA206.
Immunoprecipitation (IP) and Quantitative Western Blot Analysis-We used PUMA206 (this work) for IP of UNG2, polyclonal rabbit PSM1 directed against human SMUG1 (7) for IP of SMUG1, and anti-human TDG polyclonal rabbit serum (kindly provided by Professor Primo Schär) for IP of TDG. Antibodies were covalently coupled to Dynabeads protein A (Invitrogen) using bis(sulfosuccinimidyl) suberate cross-linker according to the manufacturer's instructions. 15-30 l of antibody-coated beads were incubated with 100 -2000 g of cell extract protein (supplemental Table S1) overnight at 4°C. The same antibody was used for IP of both human and mouse glycosylases. Samples were washed (3 ϫ 1 ml) with 10 mM Tris-HCl, pH 7.5, and protein was eluted from the beads in 15 l of 1ϫ NuPAGE lithium dodecyl sulfate sample buffer (Invitrogen) containing 1 mM DTT at 70°C for 10 min. To prepare standard curves, dilutions of corresponding His-tagged recombinant proteins (supplemental Table S2) were spiked into cell extract from the UNG Ϫ/Ϫ lymphoblastoid cell line FUJ (Table 1) and immunoprecipitated in parallel with the cell extract panel. To validate the efficiency of the IP protocol, residual UNG and SMUG1 activity was measured in the extracts after precipitation.
Proteins were separated on NuPAGE Novex Bis-Tris gels (Invitrogen) and electroblotted onto Immobilon PVDF (Millipore) or Hybond LFP membranes (GE Healthcare). Membranes were blocked in PBS, 0.1% Tween, 5% fat-free dry milk. For detection, we used polyclonal rabbit IgG (PU059 and K101) directed against the common catalytic domain of the human UNG proteins (UNG⌬84) (18), polyclonal mouse anti-SMUG1 (Abnova, H00023583-A01) directed against residues 2-79 of human SMUG1, and anti-mouse TDG rabbit serum (kindly provided by Professor Primo Schär). Secondary antibodies were HRP-conjugated swine anti-rabbit IgG (Dako) that was developed using SuperSignal West Femto (Pierce) and visualized on an Eastman Kodak Co. Image Station 4000R or Alexa Fluor 532 goat anti-mouse antibody that was visualized in a Typhoon Trio imager (GE Healthcare). Standard curves were generated by the plotting signal density of the His-tagged standards against the concentration and used to interpolate the amounts of target protein in the cell extracts.
Ung Ϫ/Ϫ Backcrossed Mice, Isolation of B Cells, and Preparation of Extracts from Activated B Cells-C57BL/6J-129SV Ung Ϫ/Ϫ mice (19) were backcrossed to the C57BL/6J strain for eight generations to create Ung ϩ/Ϫ mice with a near homogenous background. Heterozygous Ung ϩ/Ϫ mice were bred to generate Ung ϩ/ϩ and Ung Ϫ/Ϫ offspring in a specific pathogenfree animal facility at the Norwegian University of Science and Technology (Trondheim, Norway) according to rules and regulations of the Federation of European Laboratory Animal Science Associations. All mouse experiments were approved by the Norwegian Animal Research Authority (ID-10/2089). Mice were genotyped by PCR. Briefly, 50 ng of genomic DNA from tail biopsies were used in 20-l reactions containing 0.2 M primers (forward, GGCCACCCTGACAAATCCCC and reverse, CACGGACCTAATCAAGCTCACG), 1 unit of Platinum Taq polymerase (Invitrogen), 1ϫ Platinum Taq buffer, 1.5 mM MgCl 2 , 0.2 mM dNTP and amplified by 30 cycles (98°C for 10 s, 65°C for 30 s, and 72°C for 2 min).
Naïve resting B lymphocytes were purified from spleen taken from 8-month-old Ung ϩ/ϩ and Ung Ϫ/Ϫ backcrossed mice using a MACS mouse B cell isolation kit (Miltenyi Biotec) or EasySep mouse B cell enrichment kit (STEMCELL Technology) according to the manufacturers' instructions. The isolated B cells were cultured in RPMI 1640 medium (supplemented with 10% heat-inactivated FBS, 1 mM sodium pyruvate, 50 M ␤-mercaptoethanol, 2 mM L-glutamine, 0.1 mg/ml penicillin/ streptavidin) and stimulated with 40 g/ml LPS and 20 ng/ml IL-4. For preparation of extracts prior to enzyme activity measurements, 10 6 cells in 1 ml of medium were seeded per well in a 12-well plate. Cells from 10 wells (10 ml) were harvested at the time points indicated. Total/living cells were counted in a Bürker chamber after staining with trypan blue. Cells were washed in PBS, and protein was extracted as described previously (14).
Staining and FACS Analysis-All reagents were from BD Biosciences if not stated otherwise. Proliferation staining (CFSE) was performed using the CellTrace TM CFSE cell proliferation kit (Invitrogen) according to the recommended protocol. For CSR assay, 4 ϫ 10 5 cells/well were seeded in duplicates in flat bottomed 96-well plates and stimulated with LPS and IL-4 for 96 h as described above. GolgiPlug TM was added to the cells 4 h before harvest according to the manufacturer's recommendations. The cells were treated with EDTA (2 mM), washed twice with PBS, stained with LIVE/DEAD violet viability stain (Invitrogen), and blocked with fragment crystallizable receptor antibody (2.4G2) and normal mouse serum (Invitrogen). Cells were washed in PBS, fixed and permeabilized using CytoFix/ Cytoperm TM , and washed in PermWash TM containing saponin. Intracellular staining using fluorescently tagged antimouse antibodies (IgG1-APC and CD19-PerCP-Cy5.5) and the succeeding wash was performed in PermWash. Cells were suspended in 300 l of CellFix. Samples (240 l), unstained cells, and bead compensation controls were acquired using a FACS Aria cell sorter. Viable CD19 ϩ lymphocytes were analyzed for proliferation (CFSE) and IgG1 expression using FlowJo version 7.6 for PC software.
Statistical Analysis-t tests were performed to determine the significance level (p value) of -fold variation of mean UDG levels between human and mouse cells. The relationships between variables (UDG activities, protein levels, and/or molecules/cell) were evaluated by linear regression analysis. Best fit curves and coefficients of determination (R 2 ) were calculated. p values represent the significance level of the slope of curve different from 0, which corresponds to no correlation. Principal component analyses were performed as described (20) using in-house software written in Python and displayed as a biplot (21). The data sets were normalized to equal maximum values.

Human Cells Exhibit Higher Uracil Excision Capacity than
Mouse Cells-To investigate whether there are differences in initiation of uracil processing between man and mouse, we first analyzed total UDG activity in whole cell extracts from a panel of human and mouse cell lines. This panel included normal cells (fibroblasts), embryonic cells, and cancer cells of different origins (epithelial, sarcoma, and lymphoid) ( Table 1). We analyzed activity in two extract batches independently prepared from each cell line.
Uracil excision activities measured against long U:A substrate, mimicking uracil incorporated during replication, were higher in all human extracts compared with the mouse cell extracts. The difference was substantial with a mean UDG activity ϳ10-fold higher in human cell extracts compared with mouse extracts (Fig. 1A) and ϳ15-fold higher when comparing uracil excision activity per cell (Fig. 1B). UDG activity levels of the human cell lines are in accordance with a previous study measuring UDG activity (U:A) in a human cell line panel, including nine normal fibroblast and more than 40 human cancer cell lines (22). This long U:A substrate favors detection of UNG activity in the extracts. Consistent with its specialized role in postreplicative repair of incorporated uracil, UNG displays high catalytic turnover of uracil from U:A contexts (7) and is stimulated by proliferating cell nuclear antigen (10). All known UDGs recognize uracil in a U:G context. We therefore measured excision of uracil from a double-stranded U:G oligonucleotide substrate that mimics deaminated cytosine. In accordance with UDG activity measured on the long U:A substrate, uracil excision from U:G substrate was ϳ9-fold higher in human cell extracts compared with mouse extracts (Fig. 1C) and ϳ14-fold higher when comparing uracil excision activity per cell (Fig. 1D).
We then addressed whether the capacity of complete repair was different in human and mouse cells. To this end, we measured BER by an incorporation assay using extracts from the cell line panel and cccDNA substrates containing a single U:A or U:G lesion. In line with the uracil excision results, the mean BER activity of incorporated uracil (U:A) was severalfold higher in the human cell lines (Fig. 1E). Surprisingly, complete BER of deaminated cytosine (U:G) was not significantly different between human and mouse cells (Fig. 1F), indicating that BER of U:A and U:G may proceed by different mechanisms. Still, a markedly higher total uracil excision capacity was observed in human cells than in mouse cells. Thus, our results reveal considerable species differences in processing of uracil. UNG Activity Is Higher in Human Cells, and SMUG1 Activity Is Higher in Mouse Cells-To measure UNG and SMUG1 activities separately in the cell lines, we used Ugi and neutralizing SMUG1 antibodies that inhibit UNG and SMUG1 activity, respectively. No uracil excision from the U:G substrate was detected when both UNG and SMUG1 were inhibited, even with 10-fold more extract protein and a prolonged incubation time. This indicates that UNG and SMUG1 are the main contributors to remove deaminated cytosine in both species (supplemental Fig. S1, panel IV). All cell lines displayed UNG activity although at varying levels ( Fig. 2A). Compared with mouse, there was more than 6-fold higher UNG activity per total cell protein in the human extracts ( Fig. 2A) and 11-fold higher UNG activity when calculating activity per cell (Fig. 2B).
A surprising increase in U:G excision activity was observed in all extracts after inhibition of SMUG1. A likely explanation for this apparent paradox is that when substrate is limited UNG and SMUG1 may compete for binding to the same substrate. SMUG1, which has high affinity for U:G substrate and low catalytic turnover (7), will therefore reduce the overall U:G turnover rate by preventing the much more catalytically efficient UNG from accessing the substrate (7).
Contrary to mean UNG activity, mean SMUG1 activity was ϳ8-fold higher in the mouse cells than in the human cells (Fig.  2, C and D). On average, SMUG1 activity (U:G) constituted ϳ50% of the total UDG activity (U:G) in mouse cells (compare Figs. 1C and 2C and Figs. 1D and 2D). In human cells, however, SMUG1 activity contributed less than 1% of the total uracil excision activity from U:G substrate (compare Figs. 1C and 2C and Figs. 1D and 2D). These results suggest that excision of deaminated cytosine is predominantly performed by UNG in human cells, whereas SMUG1 and UNG contribute almost equally in mouse cells.
Mouse SMUG1 Removes Deaminated Cytosine More Efficiently than Human SMUG1-The differences in UNG and SMUG1 activity in human and mouse cells could potentially be caused both by different relative abundances of the enzymes and by species-specific differences in catalytic efficiency of the enzymes. To address these possibilities, we first expressed and purified recombinant UNG, SMUG1, and TDG from both species (supplemental Fig. S5). We also examined whether TDG was able to act on the specific U:G oligonucleotide substrate used for screening of the cell lines. Precise molecular masses were measured by mass spectrometry (MS), which verified that all six purified UDGs were non-degraded full-length proteins (supplemental Table S3). We then measured specific enzymatic activity of the recombinant UDGs using buffer conditions identical to those used for the cell lines. Recombinant human and mouse UNG2 displayed essentially equal specific activities against U:G substrate (Fig. 3A) and long U:A and long ssU substrates (Fig. 3B). Conversely, recombinant mouse SMUG1 displayed an almost 6-fold higher U:G excision activity than human SMUG1 (Fig. 3C) but was only slightly more active against long U:A and ssU substrates (Fig. 3D). The higher specific activity of mouse SMUG1 likely explains the high SMUG1 activity measured in the mouse cell extracts.
Like UNG2, catalytic activities of purified recombinant human and mouse TDG were similar (Fig. 3E). However, the catalytic activity of TDG was ϳ100and 1000-fold lower than the activity of SMUG1 and UNG2, respectively (Fig. 3F), which is consistent with the lack of detectable TDG activity in cell extracts (supplemental Fig. S1, panel IV).

UNG Is the Most Abundant UDG in Human Cells, whereas SMUG1 and TDG Are the Most Abundant UDGs in Mouse
Cells-We further quantified the level of UNG1, UNG2, SMUG1, and TDG in all the cell lines. Human UNG1 and UNG2 were quantified from Western blot analysis of whole cell extracts. All other quantifications were performed after IP of the target proteins prior to Western blot analysis. Standards were generated by spiking serial dilutions of recombinant Histagged UDGs (human and mouse UNG2, SMUG1, or TDG) into cell extracts and were included on each blot (supplemental Figs. S6 and S8).
Both UNG1 (mitochondrial) and UNG2 (nuclear) were readily detectable in 100 g of human cell extract except in the UNG Ϫ/Ϫ cell extract as expected (supplemental Fig. S8). Although the UNG antibody detected recombinant mouse UNG2 almost as well as the human protein (supplemental Fig.  S7), we were not able to detect UNG in mouse cell extracts by direct Western blot analysis. We therefore performed IP from 2 mg of cell extract protein prior to quantitative Western blot analysis. The total level of UNG proteins (UNG1 ϩ UNG2) in the cell lines is illustrated in Fig. 4A. Levels of UNG1 and UNG2 separately are shown in Fig. 4B. On average, human cell extracts contained ϳ20-fold higher levels of UNG than mouse cell extracts (Fig. 4A). Thus, the observed higher UNG activity in human cells compared with mouse cells is due to a higher level of UNG protein. A high statistical correlation between UNG activity and quantified UNG levels further validated this conclusion (supplemental Fig. S9A and Table S4).
Quantification of SMUG1 revealed a slightly higher level of SMUG1 protein in mouse cells compared with human cells (Fig. 4C). However, this was only ϳ2-fold and not statistically significant (p ϭ 0.06). A correlation plot demonstrated that SMUG1 is more active in mouse cells than in human cells (supplemental Fig. S9B and Table S4). This is in line with the higher catalytic activity of recombinant mouse SMUG1 (Fig. 3C).
Western blot analysis of TDG after IP identified bands corresponding to unmodified TDG as well as a modified form (TDG-M) likely representing sumoylated TDG (23) (supplemental Fig. S8). Contrary to the UNG proteins, TDG protein was detected at significantly higher levels (3.4-fold) in the mouse cells compared with the human cells (Fig. 4D).
We directly compared the quantity of each UDG by calculating the actual number of molecules per cell. Overall, UNG2 was the most abundant UDG per human cell, whereas SMUG1 and TDG were both more abundant than UNG in mouse cells. Human cell lines contained on average 420,000 UNG2 molecules, 71,000 SMUG1 molecules, and 52,000 TDG molecules per cell, whereas the mouse cell lines contained 15,000 UNG2, 140,000 SMUG1, and 159,000 TDG molecules per cell (Fig. 4E and supplemental Table S5).
The number of UNG2 molecules per cell was relatively low in the human lymphocyte cell lines. This was surprising considering the important role of UNG2 in affinity maturation of antibodies. However, the diverse functions of UNG2 in B cells may require a more stringent regulation of UNG2 in these cells. Alternatively, UNG2 is up-regulated in human epithelial cancer cells. In contrast, mouse lymphocytes contained the highest level of UNG2 of all mouse cells analyzed. Still, it should be noted that the number of UNG2 molecules per cell was ϳ3-fold higher in human lymphocytes than in mouse lymphocytes.
Furthermore, correlations between numbers of the various UDG molecules were different in human and mouse cell lines. In human cells, there were significant positive correlations between all three glycosylases (UNG2, SMUG1, and TDG), whereas less significant correlations were identified for the FIGURE 3. Specific activities of purified recombinant human and mouse UDGs. All values represent the mean of at least three experiments, and standard deviations are indicated as error bars. Representative PAGE gels from U:G activity assays are shown in supplemental Fig. S3. A, activity of human and mouse UNG2 against U:G substrate. B, specific activity of human and mouse UNG2 measured by a standard UDG assay on long U:A and ssU substrate. C, activity of human and mouse SMUG1 against U:G substrate. D, activity of human and mouse SMUG1 measured by a standard UDG assay on long U:A substrate and long ssU. E, activity of human and mouse TDG against U:G substrate. F, specific activity of human and mouse UDGs on U:G substrate. mouse cell lines. However, the mouse cell lines showed a significant negative correlation between total UNG level and SMUG1 ( Fig. 4F and supplemental Table S6). This suggests that expression of the various UDGs is differently regulated in the two species.
A principal component analysis (20) of all glycosylase activities and quantitative data was carried out to distinguish between variability from cell types (fibroblast, epithelial, and lymphocyte) and variability due to species. The principal component analysis revealed that 71% (component 1) of the variability distinguished between human and mouse samples, whereas only 20% (component 2) of the variability was due to cell types (supplemental Fig. S10). Taken together, we conclude that there are marked species differences in the initial step of genomic uracil processing between human and mouse cells.
UNG and SMUG1 Activities in CSR-activated Splenic Mouse B Cells Are Comparable with Activities in Mouse Cell Lines-Uracil processing plays an important role in Ig diversification. We therefore monitored UDG activities in primary mouse B cells stimulated to undergo CSR. We isolated splenic B cells from 8-month-old Ung ϩ/ϩ and Ung Ϫ/Ϫ backcrossed mice and stimulated the cells with LPS ϩ IL-4 to induce IgG1 switching. CFSE staining of the cells before stimulation revealed that wild type and Ung Ϫ/Ϫ B cells proliferated at the same rate (six generations after 96 h), having an equal distribution of cells per generation in both wild type and knock-out cells (Fig. 5A). CSR was analyzed by FACS, and a clear IgG-positive population was detected in Ung ϩ/ϩ cells (7.6% of total) but not in Ung Ϫ/Ϫ cells (0.2% of total). For Ung ϩ/ϩ mice, IgG1-positive B cells constituted as much as 30 -40% in generations 5 and 6, whereas the same generations with Ung Ϫ/Ϫ B cells contained less than 1% IgG1-positive cells (Fig. 5B). This clearly verifies inhibition of CSR in vitro in the Ung Ϫ/Ϫ backcrossed mice used in this study, which is in accordance with previous reports (3,5).
Cells were harvested at several time points after stimulation to monitor UNG and SMUG1 activity. Cell volume and total protein content of B cells increase during stimulation mainly due to increased cytoplasm. We therefore analyzed the UNG and SMUG1 activities in extracts made from the same number of cells (representing the same number of nuclei) at each time point. Stimulation with LPS and IL-4 induced UNG activity more than 20-fold in Ung ϩ/ϩ B cells after 72 h, whereas UNG activity in the Ung Ϫ/Ϫ B cells was not detectable (Fig. 5C and  supplemental Fig. S11A). The maximum level of UNG activity per cell in stimulated primary B cells was similar to the activity in the mouse lymphoid cell line YAC. Moreover, the UNG activity (related to total protein) in B cells stimulated for more than 48 h was at the same level as the mean UNG activity for all mouse cell line ( Fig. 5D and supplemental Fig. S11B). The low relative levels of UNG activity in the mouse cell lines are therefore comparable with UNG activity levels in stimulated primary mouse B cells isolated from spleen. SMUG1 was reported previously to be down-regulated in activated B cells (24). We found, however, that SMUG1 activity per cell actually increased by ϳ40% after B cell activation. Moreover, SMUG1 activity was identical in Ung ϩ/ϩ and Ung Ϫ/Ϫ B cells (Fig. 5E). All activity was inhibited by Ugi ϩ SMUG1 antibody (supplemental Fig. S2, panel III), demonstrating that the assay conditions were specific for SMUG1 with no measurable contribution from TDG and MBD4. As for UNG activity, SMUG1 activity per cell in primary activated B cells was similar to that of the YAC lymphoid mouse cell line (Fig.  5E). In accordance with previous results (24), SMUG1 activity relative to total cell protein decreased markedly during stimulation (Fig. 5F). However, this value is influenced by the strong increase in cytoplasmic cell protein after stimulation. We therefore consider activity per cell (representing one nucleus) the most relevant value when measuring change in activity during stimulation. As for UNG, the SMUG1 activity (relative to total cell protein) in activated B cells was comparable with those measured in mouse cell lines (Fig. 5F). Taken together, activated primary splenic mouse B cells express both UNG and SMUG1 at levels comparable with those found in mouse cell lines. Moreover, SMUG1 (activity per cell) was not down-regulated in cells undergoing CSR contrary to previous assumptions (24).

DISCUSSION
Gene-targeted mice generally represent valuable models for the function of orthologs in man and have been used extensively in the study of genomic uracil processing and antibody diversification (3, 5, 9 -11, 19, 24, 25). In addition, studies on the mechanism of the cytostatic drug 5-fluorouracil frequently utilize mouse models and cell lines. 5-Fluorouracil treatment may result in 5-fluorouracil in DNA and increased levels of genomic uracil, both of which are substrates for UDGs (10, 26 -28). Our results show that there are important differences in initial steps of uracil processing in mice and humans. This may have implications for the use of mouse models in studies on genomic uracil repair, adaptive immunity, and responses to cytostatic drugs.
Our results reveal a considerably higher uracil excision capacity in human cells than in mouse cells (10 -20-fold) due to significantly higher levels of UNG protein. The level of nuclear UNG2 in human cells is cell cycle-regulated at the transcriptional level (29). The mouse UNG2 promoter, however, shows limited homology to the human counterpart (30), indicating in regulation of expression of the UNG gene. Human UNG2 is  Table S5. F, linear regression analysis on the relationship between the levels of various UDGs (molecules per cell) in human and mouse cell lines. More correlation coefficients are listed in supplemental Table S6. MEF, mouse embryonic fibroblast.
further regulated by stepwise phosphorylation of serines and threonine (Ser 23 , Thr 60 , and Ser 64 ) during the cell cycle that apparently marks UNG2 for breakdown in the G 2 phase (2,29,31). These phosphorylation sites are not conserved in mouse (2), indicating that mouse UNG2 lacks this level of regulation. It is therefore tempting to speculate whether the observed lower level of UNG2 in mouse cells is related to a less fine tuned regulation during the cell cycle. This should be addressed in future studies.
Mouse SMUG1 was much more active on the U:G substrate than human SMUG1. Previously, we have identified a "wedge" motif in human SMUG1 ( 244 NPQANK 249 ) that confers U:G specificity (8). The proline (human SMUG1 Pro 245 ) in this motif is strictly conserved in SMUG1 from all vertebrates as well as bacteria. In mouse, however, the corresponding residue is an alanine. Kinetic analysis of a human SMUG1-P145A mutant, mimicking mouse SMUG1, shows a 7-fold increase in U:G turnover (k cat ). Thus, this single residue may largely account for the higher SMUG1 activity observed for mouse cells.
Generally, in the human cell line panel, the UNG2 amount was estimated to be 10 5 -10 6 molecules per cell ( Fig. 4E and supplemental Table S5). These results are in accordance with a previous report of 1.2 ϫ 10 5 UNG2 molecules per cell in normal human fibroblasts (32). UNG2 is cell cycle-regulated with ϳ3-fold higher levels in S phase (2,33). A significant fraction of UNG2 localizes to sites of replication in S phase where it has a distinct role in repair of incorporated uracil (33). The number of replication forks in human cells was recently reported to be ϳ3000 (34). Hence, there would be as many as 100 -1000 UNG2 molecules per replication fork but maybe only ϳ5-10 in mouse cells. This suggests that in human cells there is a sufficient excess of UNG2 molecules to carry out replication-independent removal of deaminated cytosine from U:G mismatches outside of replication forks. In mouse cells, however, a more active SMUG1 may be needed for efficient repair of deaminated cytosine.
We found that UNG and SMUG1 activities in splenic B cells were in the same range as for the mouse cell lines. Thus, mouse cell lines are representative for the uracil processing activities in primary activated mouse cells. In contrast to UNG activity, which is up-regulated in activated B cells, SMUG1 activity was reported to be down-regulated in B cells after LPS ϩ IL-4 stimulation (measured as SMUG1 activity per total cell protein) (24). However, we found that SMUG1 activity per cell was, in fact, slightly up-regulated (40%) during stimulation. We there- fore suggest that the higher SMUG1 activity in mouse cells may at least in part explain why Ung Ϫ/Ϫ mice display only a partial reduction of switched isotypes (IgG and IgA) in serum (3,5), whereas reduced class switching is more pronounced in human UNG-deficient patients (4).
Other DNA repair proteins also contribute to early steps of adaptive immunity. For instance, the mismatch recognition complex MSH2-MSH6 is involved in CSR and somatic hypermutation by recognizing U:G mismatches and subsequent generation of strand breaks. CSR is ablated in Ung Ϫ/Ϫ Msh2 Ϫ/Ϫ and Ung Ϫ/Ϫ Msh6 Ϫ/Ϫ double knock-out mice, which have IgG1 levels less than 2% of the wild type IgG1 serum level. In comparison, Ung Ϫ/Ϫ mice have IgG1 levels at ϳ50 -70% of wild type, whereas Msh2 Ϫ/Ϫ mice have normal IgG levels (3,5,11). Thus, the MSH2-MSH6 complex is essential for CSR when the Ung gene is inactivated. Moreover, these double knock-out mice reveal that the endogenous levels of other UDGs are unable to support CSR in a UNG2-and MSH2/6-double deficient background. However, retroviral expression of catalytically deficient UNG2 mutants, with only ϳ1% residual activity, rescued CSR in mouse Ung Ϫ/Ϫ B cells (25,35,36). This reveals that high uracil excision activity is not essential to generate substrates for CSR. Furthermore, when human SMUG1 is overexpressed in mouse Ung Ϫ/Ϫ and Ung Ϫ/Ϫ Msh2 Ϫ/Ϫ B cells, it is able to target uracil in immunoglobulin genes and initiate CSR. Possibly, SMUG1-initiated strand breaks in switch regions may occasionally escape repair and act as intermediates in CSR. Interestingly, SMUG1-dependent CSR was more efficient in Ung Ϫ/Ϫ cells than in Ung Ϫ/Ϫ Msh2 Ϫ/Ϫ double knock-outs (25), indicating that SMUG1 cooperates with the MSH2-MSH6 complex. Overexpression of TDG in either Ung Ϫ/Ϫ or Ung Ϫ/Ϫ Msh2 Ϫ/Ϫ B cells, on the other hand, did not restore switching (25). TDG is therefore not likely to play any role in either normal CSR or as a CSR-initiating backup activity in the absence of UNG.
In wild type mice, however, overexpression of human SMUG1 has been shown to reduce the extent of switching and mutations (24). Conceivably, overexpression of SMUG1 may inhibit CSR by overriding UNG processing of U:G substrates. Hence, the role of SMUG1 in processing activation-induced cytidine deaminase-generated U:G mismatches seems to be mainly initiation of normal repair and not CSR. This may be important to avoid untargeted mutations outside Ig loci. Notably, UNG2 and MSH2/6, the main U:G processing factors in somatic hypermutation and CSR, also contribute significantly to normal repair of activation-induced cytidine deaminasegenerated U:G mispairs outside Ig loci in activated B cells (37).
Our results reveal that there are considerable species differences in the initial steps of genomic uracil processing. This has to be taken into account when using mouse models in studies of genomic uracil repair, adaptive immunity, and responses to cytostatic drugs such as 5-fluorouracil. Future work should investigate whether there also are species differences in downstream steps in BER and mutagenic processing during antibody diversification. Finally, knocking out SMUG1 in Ung Ϫ/Ϫ mice will clarify whether uracil excision by SMUG1 is involved together with MSH2/6 in the residual isotype switching detected in Ung Ϫ/Ϫ mouse serum.