Germ Line Variants of Human N-Methylpurine DNA Glycosylase Show Impaired DNA Repair Activity and Facilitate 1,N6-Ethenoadenine-induced Mutations*

Background: hMPG initiates repair of mutagenic and tumorigenic modified purine bases. Results: Two germ line hMPG variants showed reduced activity due to low affinity for DNA and facilitated lesion-induced mutations. Conclusion: Changes in amino acid sequence alter the function of hMPG, leading to genomic instability. Significance: Individuals possessing these hMPG variants may be at higher risk for genomic instability-related diseases. Human N-methylpurine DNA glycosylase (hMPG) initiates base excision repair of a number of structurally diverse purine bases including 1,N6-ethenoadenine, hypoxanthine, and alkylation adducts in DNA. Genetic studies discovered at least eight validated non-synonymous single nucleotide polymorphisms (nsSNPs) of the hMPG gene in human populations that result in specific single amino acid substitutions. In this study, we tested the functional consequences of these nsSNPs of hMPG. Our results showed that two specific arginine residues, Arg-141 and Arg-120, are important for the activity of hMPG as the germ line variants R120C and R141Q had reduced enzymatic activity in vitro as well as in mammalian cells. Expression of these two variants in mammalian cells lacking endogenous MPG also showed an increase in mutations and sensitivity to an alkylating agent compared with the WT hMPG. Real time binding experiments by surface plasmon resonance spectroscopy suggested that these variants have substantial reduction in the equilibrium dissociation constant of binding (KD) of hMPG toward 1,N6-ethenoadenine-containing oligonucleotide (ϵA-DNA). Pre-steady-state kinetic studies showed that the substitutions at arginine residues affected the turnover of the enzyme significantly under multiple turnover condition. Surface plasmon resonance spectroscopy further showed that both variants had significantly decreased nonspecific (undamaged) DNA binding. Molecular modeling suggested that R141Q substitution may have resulted in a direct loss of the salt bridge between ϵA-DNA and hMPG, whereas R120C substitution redistributed, at a distance, the interactions among residues in the catalytic pocket. Together our results suggest that individuals carrying R120C and R141Q MPG variants may be at risk for genomic instability and associated diseases as a consequence.

Human N-methylpurine DNA glycosylase (hMPG) initiates base excision repair of a number of structurally diverse purine bases including 1,N 6 -ethenoadenine, hypoxanthine, and alkylation adducts in DNA. Genetic studies discovered at least eight validated non-synonymous single nucleotide polymorphisms (nsSNPs) of the hMPG gene in human populations that result in specific single amino acid substitutions. In this study, we tested the functional consequences of these nsSNPs of hMPG. Our results showed that two specific arginine residues, Arg-141 and Arg-120, are important for the activity of hMPG as the germ line variants R120C and R141Q had reduced enzymatic activity in vitro as well as in mammalian cells. Expression of these two variants in mammalian cells lacking endogenous MPG also showed an increase in mutations and sensitivity to an alkylating agent compared with the WT hMPG. Real time binding experiments by surface plasmon resonance spectroscopy suggested that these variants have substantial reduction in the equilibrium dissociation constant of binding (K D ) of hMPG toward 1,N 6 -ethenoadenine-containing oligonucleotide (⑀A-DNA). Pre-steadystate kinetic studies showed that the substitutions at arginine residues affected the turnover of the enzyme significantly under multiple turnover condition. Surface plasmon resonance spectroscopy further showed that both variants had significantly decreased nonspecific (undamaged) DNA binding. Molecular modeling suggested that R141Q substitution may have resulted in a direct loss of the salt bridge between ⑀A-DNA and hMPG, whereas R120C substitution redistributed, at a distance, the interactions among residues in the catalytic pocket. Together our results suggest that individuals carrying R120C and R141Q MPG variants may be at risk for genomic instability and associated diseases as a consequence.
Cellular DNA is under constant attack from various endogenous and exogenous chemical agents that can damage the DNA through several chemical reactions, including alkylation, oxidation, and deamination (1). Consequently, cells have developed multiple DNA repair pathways to repair damaged bases and thwart cell death and base substitutions responsible for genomic instability, which is a major cause of cancer, neurodegeneration, aging, and other diseases (2,3). Among various DNA repair pathways, the base excision repair (BER) 3 pathway repairs small DNA adducts and is responsible for the removal of at least 20,000 DNA lesions per cell per day. This process begins with the removal of the damaged DNA base by the appropriate DNA glycosylase, which catalyzes the cleavage of an N-glycosidic bond, creating an abasic site (AP site) (4). The backbone of the DNA is then cleaved by an AP endonuclease, which creates a nick 5Ј to the AP site, leaving a 3Ј-hydroxyl and a 5Ј-phosphodeoxyribose moiety (4). A DNA polymerase is then recruited to the AP site and fills in the correct nucleotide, and subsequently the repair is completed by a DNA ligase (4). Human N-methylpurine-DNA glycosylase (hMPG), a monofunctional DNA glycosylase, is known to excise at least 17 structurally diverse modified purine bases, including toxic and mutagenic alkylated, deaminated, and etheno adducts, from both the major and minor grooves of duplex DNA (5-9). 1,N 6 -ethenoadenine (⑀A) is one of the major substrates of MPG and was detected in biological samples from animals and humans, including plastic factory workers and cigarette smokers (10 -13). This adduct is induced by vinyl chloride and tissue inflammation, which have been implicated in tumorigenesis of several organs, including liver, colon, and lung (12,14,15). MPG activity has been found to be highly variable  units/g of protein) even in disease-free human population and may have an adverse effect on the onset or progression of diseases such as cancer (12,16). In addition to variations in expression and post-translational modifications of the MPG protein, changes in MPG coding gene sequence may attribute to the variation in MPG activity. Such short genetic alterations in coding sequences are called non-synonymous single nucleotide polymorphisms (nsSNPs) and result in amino acid substitutions. Many DNA repair gene polymorphisms have been identified; some have been shown to affect protein function (17)(18)(19)(20) and have been associated with an increased incidence of cancer (21,22). In fact, an SNP variant of hNTH1, another BER DNA glycosylase, showed loss of activity and induced genomic instability in human and mouse cells (23). A number of hMPG nsSNPs were identified in human populations by genetic analysis and validated, but their clinical and functional significance are not known. Thus, the objective of this study was to examine the functional effect of hMPG nsSNP variants. Our approach included analysis of mutation resulting from hMPG variants in human cells and repair analysis of hMPG variants in vitro and in mammalian cells followed by comprehensive mechanism analysis in vitro.

EXPERIMENTAL PROCEDURES
Cell Lines and Culture-HEK293 and HCT116 cells were obtained from American Type Culture Collection (Manassas, VA). Immortalized Mpg Ϫ/Ϫ mouse embryonic fibroblast (MEF) cells were kindly provided by Dr. Leona Samson (Center for Environmental Sciences, Massachusetts Institute of Technology). All cell lines were cultured in DMEM supplemented with 10% FBS, 1% penicillin-streptomycin.
Retrovirus-packaging (SD3443) cells were transfected with either pRS-shMPG or pRS-scrambled vector using FuGENE (Promega, Madison, WI) transfection reagent following the manufacturer's protocol (OriGene, Rockville, MD). HCT116 cells were transduced with retrovirus particles and selected with puromycin. Resistant colonies were pooled and cultured in DMEM supplemented with 10% FBS, 1% penicillin-streptomy-cin with continuous puromycin selection. To allow expression of hMPG, constructs carrying WT and variant hMPG coding sequences were introduced with silent mutations at the shRNA target sequence, thereby becoming resistant to shRNA. Mutation was introduced using the QuikChange II XL site-directed mutagenesis kit (Agilent, Santa Clara, CA) following the manufacturer's protocol and validated by sequencing (Genewiz, Germantown, MD). The primer set for each substitution is given in Table 1.
Non-synonymous Polymorphisms in the hMPG Gene-We screened SNPs in the hMPG gene that appeared in the NCBI Short Genetic Variations (dbSNP) database. The validated entries from nsSNPs in the hMPG coding region were selected for this study and were tabulated along with their respective SNP ID and validation method ( Table 2).
Site-directed Mutagenesis to Generate Polymorphic Equivalent hMPG Variants-To generate the hMPG polymorphic variants, site-directed mutagenesis was carried out using codon-optimized WT hMPG cDNA in the expression vector pRSETB (24) using the QuikChange mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer's protocol. The primer set for each substitution is given in Table 1. The desired point substitutions were verified by sequencing reaction.
Purification and Western Blot Analysis of WT and Polymorphic Variants of hMPG-Human WT MPG and polymorphic variants of hMPG were purified as described previously for fulllength codon-optimized WT hMPG (24), and then the purified proteins were analyzed by Western blot as described previously (25).
Preparation of Oligonucleotide Substrate-⑀A-or hypoxanthine (Hx)-containing 50-mer oligonucleotide with the sequence 5Ј-TCGAGGATCCTGAGCTCGAGTCGACGXTC-GCGAATTCTGCGGATCCAAGC-3Ј (where X represents ⑀A or Hx) was purchased from Operon Technologies (Alameda, CA) and Gene Link (Hawthorne, NY), respectively. The complementary oligonucleotide containing T opposite ⑀A or Hx was synthesized by the Recombinant DNA Laboratory Core  GC CGA CGG ATG GGG CAA AAG CAG CAG CGA CCA GCT AGA G  Q22K reverse  C TCT AGC TGG TCG CTG CTG CTT TTG CCC CAT CCG TCG GC  P64L forward  C AGG GAG CGC TGC TTG GGA CTG CCC ACC ACT CCG GGC CCA TAC  P64L reverse  GTA TGG GCC CGG AGT GGT GGG CAG TCC CAA GCA GCG CTC CCT G  Y71H forward  CG CCC ACC ACT CCG GGC CCA CAC CGC AGC ATC TAT TTC TC  Y71H reverse  GA GAA ATA GAT GCT GCG GTG TGG GCC CGG AGT GGT GGG CG  Q93R forward  CGA CTG GGG TTG GAG TTC TTC GAC CGG  The ⑀Aor Hx-containing oligonucleotide was labeled at the 5Ј-end using T4 polynucleotide kinase and [␥-32 P]ATP and then annealed to complementary oligonucleotide to prepare 5Ј-32 P-labeled duplex oligonucleotide as described previously (26). hMPG-mediated Excision Activity Assay-Purified WT and all the hMPG variant proteins (2 nM each) were individually incubated with 5Ј-32 P-labeled ⑀A-DNA or Hx-DNA (1 nM) for 15 min at 37°C in assay buffer (25 mM HEPES-KOH, pH 7.9, 0.5 mM DTT, 10 g/ml nuclease-free bovine serum albumin, 150 mM NaCl, 10% glycerol) in a final volume of 20 l. The reaction was stopped by inactivating the enzyme at 75°C for 5 min. The products containing the AP sites were analyzed as described previously (26). Percent product formation was determined by quantifying the substrate and product band intensities using an imager (Chemigenius Bioimaging System, Frederick, MD) and software (GeneTool, Syngene Inc., San Diego, CA).
hMPG-mediated Excision Activity Assay in Cell Extracts-Mpg Ϫ/Ϫ MEF cells were transfected with WT, R120C, and R141C hMPG constructs. Whole cell extracts (1-3 g) containing an equal amount of expressed hMPG were individually incubated with 5Ј-32 P-labeled ⑀A-DNA (1 nM) for 15 min at 37°C in assay buffer in a total volume of 20 l. The reaction was stopped by inactivating the enzyme at 75°C for 5 min. The reaction termination and subsequent analysis of reaction products were performed as described for the excision activity assay.
Construct Preparation and Single Strand DNA (ssDNA) Isolation-We constructed a replicating phagemid as described previously (27). Essentially, pBluescript-SV40 (pBS-SV40) was prepared by cloning a DNA fragment (700 bp) containing the SV40 replication origin sequence, which was amplified from pSP189 using primers containing AflII and SapI restriction sites. The phage suspension was prepared following a standard method (28). Single-stranded pBS-SV40 DNA was isolated from the phage suspension using the QIAprep Spin M13 kit (Qiagen, Gaithersburg, MD).
⑀A or 8-Oxo-dG pBS-SV40 in Vitro Construct Preparation-The modification (⑀A or 8-oxo-dG) was placed at the EcoRI site of the pBS-SV40 construct. The preparation included three main steps: phosphorylation of the damage-containing primer oligonucleotide, annealing of the oligonucleotide to the ssDNA, and the primer extension reaction. Phosphorylation of the primers was performed by incubating 2 g of adduct-contain-ing (5Ј-CCGAGCTCGXATTCGTAATC-3Ј where X is ⑀A or 8-oxo-dG) oligonucleotide with 1ϫ T4 polynucleotide kinase (PNK) buffer, 400 nM ATP, 50 mM DTT, and 10 units of T4 polynucleotide kinase (New England Biolabs, Ipswich, MA) at 37°C for 45 min. The phosphorylated oligonucleotide was purified through a G-25 column (GE Healthcare) according to the manufacturer's protocol. Six microliters of purified oligonucleotide was incubated with 2 g of pBS-SV40 ssDNA in annealing buffer (10 mM Tris-HCl, pH 7.5, 50 mM NaCl) at 80°C for 5 min and then gradually cooled to room temperature with brief centrifugation at 50°C. The reaction was then incubated with an extension reaction mixture (1ϫ T7 DNA polymerase buffer, 1.5 mM ATP, 1.5 mM dNTP, 10 mM DTT, 160 g/ml BSA, 10 units of T7 DNA polymerase, 400 units of T4 DNA ligase) for 5 min on ice followed by 5 min at room temperature. The reaction was subsequently incubated at 37°C for 1 h after which 50 nmol of ATP and 200 units of T4 DNA ligase were added and incubated at 14°C overnight for efficient ligation. Finally, the reaction was incubated with 1ϫ Supercoil-It buffer (Bayou Biolabs, Metairie, LA) and Supercoil-It enzyme at 37°C for 3 h. Plasmid DNA was recovered using the QIAquick PCR Purification kit (Qiagen).
Site-specific Mutagenesis Assay-The damaged (⑀A and 8-oxo-dG) constructs were transfected into cells with Lipofectamine 2000 (Invitrogen) following the manufacturer's protocol. Forty-eight hours post-transfection, the cells were harvested, and the plasmids were recovered using the Qiagen Miniprep kit (Qiagen). Retrieved plasmid DNA was digested with DpnI for 1 h at 37°C. The restricted DNA was aliquoted and treated with and without EcoRI for 2 h at 37°C. Digested plasmid DNA was then transformed by electroporation into XLI-Blue cells. Colonies were manually counted, and mutation frequency was calculated by dividing the mutant colonies (EcoRI-digested) by total colonies (undigested). The colonies formed without EcoRI digestion provided the measure of total number of plasmid DNA copies, and the colonies that were formed after EcoRI digestion afforded the measure of copy numbers of mutant plasmid DNA as they are only able to form colonies because of resistance to EcoRI digestion. To validate this resistance, 30 -40 of these colonies were tested by a second EcoRI digestion followed by agarose gel electrophoresis. However, we observed only 2-3% of colonies containing WT DNA that had escaped the initial EcoRI digestion. Therefore, those colonies did not significantly affect the overall mutation frequency reported.
DNA Binding Studies Using Surface Plasmon Resonance-A 50-mer duplex containing an ⑀A (⑀A-DNA) or A (control-DNA) at the 26th position from the 5Ј-end of one strand was used for measuring enzyme-DNA interactions. Oligonucleotides were biotinylated and immobilized on streptavidincoated Biacore chips as described previously (33). Then we measured the binding parameters of WT, R120C, and R141Q hMPG (0 -30 nM) for ⑀A-DNA and control-DNA using binding buffer (10 mM HEPES-KOH, pH 7.6, 150 mM NaCl, 0.05% surfactant) at 7°C in a Biacore T100 (GE Healthcare). Various concentrations of MPG wild type and variant proteins were injected, and the surface plasmon resonance units were measured with three 60-s injections. Following each injection, the chip was regenerated with 1 M NaCl. The binding kinetics for ⑀A-DNA or control-DNA was established with a series of hMPG concentrations. Kinetic parameters for binding were calculated using a 1:1 binding model and Langmuir isotherms based on the on/off rates and protein concentrations for each protein using BIAevaluation software (Biacore, Uppsala, Sweden).
Single Turnover Kinetic Assay-The kinetic studies were carried out and analyzed as described previously (29). Typically, 54 and 68 nM WT and 1080 and 1340 nM each of R120C and R141Q variant hMPG proteins were individually incubated with 1 nM 5Ј-32 P-labeled ⑀A-DNA substrates at 37°C in hMPG excision activity assay buffer in a total volume of 100 l. Aliquots of 5 l were heat-inactivated at 80°C in preheated microcentrifuge tubes at various time points from 0 to 10 min. AP site cleavage and sample denaturation and resolution by gel electrophoresis were performed as described for the excision activity assay. Data were analyzed using the equation where A 0 represents the amplitude of the exponential phase and k obs is the observed rate constant of the reaction, which is approximately equivalent to the catalytic constant at the chemistry step (k chem ), the rate of glycosidic bond cleavage.
Burst Analysis-Reactions were carried out as described previously (29). Briefly, 6.5 nM hMPG was incubated with 70 nM 5Ј-32 P-labeled ⑀A-DNA for different lengths of time (0 -10 min) at 37°C under conditions similar to those described for the single turnover kinetic study. The data were analyzed according to the equation where k ss is the slope of the linear phase, which is the turnover rate, and A 0 represents the amplitude of the burst. The rate of product dissociation (k pd ) is estimated by the ratio of k ss over A 0 .
Prediction of hMPG Structure-Structural models of WT, R120C, and R141Q hMPG were predicted based primarily on the x-ray structure of human alkyladenine DNA glycosylase (also known as hMPG; Protein Data Bank code 1EWN). Because the hMPG three-dimensional structure lacks the first 1-78 residues, the structure of those residues were deduced using two available x-ray crystal structure templates, namely tRNA synthetase (Protein Data Bank code 1UE0) with 34% and human immunodeficiency virus type 1 trans-activator of transcription (TAT) (Protein Data Bank code 1JFW) with 31% identity at amino acid sequences to those 1-78 residues of hMPG. Predicted structural models were then energy-minimized using the consistent valence force field (CFF91) with AMBER 9.0 (30). The cutoff for nonbonded interaction energies was set to ∞ (no cutoff).
Molecular Modeling-Using the energy-minimized structure of WT full-length hMPG as the initial model, the structure predictions of R120C and R141Q were conducted using 1-ns molecular dynamics simulations with a distant-dependent dielectric constant in the SANDER module of AMBER 9.0 with the PARM98 force field parameter. The SHAKE algorithm (31) was used to keep all bonds involving hydrogen atoms rigid. Weak coupling temperature and pressure coupling algorithms were used to maintain constant temperature and pressure, respectively. Molecular dynamics simulations were performed at 0.001-ps time steps with temperature set at 300 K. Electrostatic interactions were calculated using the Ewald particle mesh method (32) with a dielectric constant at 1R ij . A nonbonded cutoff of 12 Å was used for the real part of electrostatic interactions and for van der Waals interactions. Structural analyses were performed using the SYBYL 8.2 (Tripos International, St. Louis, MO) molecular modeling program.
Statistics and Quantifications-Data are presented as the mean Ϯ S.E. of at least three independent experiments and were analyzed by Student's t test. All statistical analyses were performed and all graphs were generated using GraphPad Prism 5.0 software (GraphPad).

RESULTS
Reduced Activity of hMPG Variants Compared with WT hMPG-All proteins were 90% or more pure (Fig. 1A), and their identities were confirmed by Western blotting (Fig. 1B) using a previously well characterized anti-hMPG monoclonal antibody (24). The activity of each purified protein was measured by excision activity assay using ⑀A-DNA and Hx-DNA as substrates. Of the eight polymorphic variants, two (R120C and R141Q) showed significantly reduced excision activity (15-40%) compared with WT hMPG for ⑀A-DNA (Fig. 1, C and E) and Hx-DNA (Fig. 1, D and E). However, R141Q showed significantly less (ϳ40%; p Ͻ 0.0005) activity compared with WT (Fig. 1E), whereas R120C had only a borderline significant reduction (ϳ20%, p Ͻ 0.05) in activity compared with WT. Thus, we selected R141Q to evaluate biological consequences such as its effects on in-cell repair and mutations and for detailed mechanism studies to understand the loss of its activity. We also included R120C for biological and biochemical studies with the expectation of moderate biological effects because it had a moderate reduction in activity. Although the frequency of the R120C variant was 0.1% in a study of 840 sam-hMPG nsSNPs Increase Genomic Instability FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 ples, other cohorts demonstrated frequencies from 0 to 2.3% in a given population. R141Q showed low occurrence at 0.1% in all tested populations so far, and at this point, R141Q may be defined as a mutant rather than a polymorphism ( Table 2). The aforementioned reports were obtained from the NCBI Short Genetic Variations (dbSNP) database.
The R141Q and R120C Variants Show Reduced in-cell Repair Activity Compared with WT hMPG-To further confirm the results of the activity assay with purified recombinant R141Q and R120C hMPG, we performed the activity assay in extracts from cells expressing these proteins. R141Q, R120C, and WT hMPG expression constructs were transfected into Mpg Ϫ/Ϫ MEF cells, and efficient expression was confirmed by Western blot analysis (Fig. 2A). Whole cell extracts were assayed for MPG activity using ⑀A-DNA as a substrate. Indeed, extracts expressing R141Q and R120C hMPG demonstrated reduced activity compared with WT hMPG (Fig. 2B). We also observed that extracts from cells expressing R141Q showed less activity compared with R120C, thereby confirming our in vitro results with purified proteins (Fig. 1, C-E). We further tested several concentrations of WT and variant hMPG in the Mpg Ϫ/Ϫ MEF background by supplementing Mpg Ϫ/Ϫ MEF nuclear extracts with increasing concentrations of purified enzymes. The excision activity assay was performed with supplemented extracts, and as expected, the variant-supplemented extracts demonstrated reduced activity compared with WT-supplemented extracts at all concentrations tested (Fig. 2, C and D).
The R141Q and R120C Variants Show Reduced Repair under Heterozygous Condition-In the human population, the occurrence of SNPs is either homo-or heterozygous. Our previous experiments modeled the homozygous genotype because the activity assays were not performed in the presence of both variant and WT hMPG. Therefore, we chose to test the activity of variants when present with WT hMPG to model a heterozygous genotype that might be observed in humans. It is possible that a dominant negative effect could be observed in the heterozygous genotype condition. If variants bind substrate but are not able to catalyze excision this could lead to substrate sequestration from WT hMPG and an overall decrease in product formation even in the presence of WT hMPG. Therefore, we tested the activity of purified WT hMPG and variants in equal concentrations against ⑀A-DNA. As expected, we observed a 2-fold increase in product formation with 4.6 nM WT hMPG (representing homozygous WT) compared with 2.3 nM WT hMPG (Fig. 3A). However, when a 2.3 nM concentration of either variant was tested with an equal concentration of WT hMPG (representing the heterozygous genotype), only a 1.3-1.5-fold increase in product formation was observed compared   (Fig. 3A). Because an increase in product formation was observed with the mixing experiment, we concluded that there was no substrate hijacking with the hMPG variants. We chose to further test this result in cells. Therefore, in HCT116 cells, WT hMPG and variants were expressed to obtain similar amounts of endogenous and exogenous protein levels (Fig. 3B). Western blot analysis showed that cells expressing the empty vector represent heterozygous with one functional allele and one null allele, and expression of WT hMPG represents homozygous, whereas expression of variants represents heterozygous with one allele WT hMPG and the other variant (Fig. 3B, first panel). To confirm that the upper band was in fact exogenous hMPG, Western blot for the FLAG tag was performed that showed that the upper band was indeed exogenous hMPG (Fig. 3B, second panel). Using extracts from these combinations of endogenous and exogenous hMPG, an excision assay was performed using ⑀A and Hx as substrates. Under heterozygous conditions of variants, the total activity of the extracts was reduced compared with homozygous WT MPG extracts (Fig. 3, C and D). Although R120C under heterozygous conditions showed reduction in activity compared with homozygous WT MPG, R140Q showed the most reduction in activity for both substrates (Fig. 3, C and D). The results of R120C variant expression with endogenous WT   (Fig. 3, C and D) were consistent with those of the mixing experiment with purified proteins (Fig. 3A). Interestingly, no increase in activity was observed with expression of the R141Q variant along with endogenous WT hMPG compared with endogenous WT hMPG alone, indicating a slight dominant negative effect of the variant in cell extracts. Although there are known interactions between WT hMPG and proteins that stimulate its activity such as hHR23 (33), it is unclear how the variant may bind to this and other proteins that are part of the hMPG interactome. Furthermore, the R141Q variant itself may have a different interactome due to the amino acid change that leads to an inhibitory effect on MPG activity. In fact, a recent example of this phenomenon is described in a report of a polymorphic variant of the DNA glycosylase NEIL2 that was shown to have reduced affinity for downstream BER proteins compared with WT NEIL2 (34).

Expression of R141Q and R120C hMPG Variants Increases Mutation Frequency and MMS Sensitivity Compared with WT
hMPG-Previous reports suggest that reduced activity of DNA glycosylases may promote genomic instability (23). Given that R141Q and R120C showed reduced activity, it is possible that these variants may facilitate an increase in ⑀A-induced mutations compared with WT hMPG. We hypothesized that the mutation frequency would be inversely proportional to MPG activity so that expression of R141Q would lead to higher mutation frequency than expression of R120C. To test this hypothesis, we developed an in-cell site-specific mutagenesis assay (Fig. 4A).
The mutagenesis assay is designed to report mutation frequency of a single base adduct in the context of living human cells (Fig. 4A). The method utilizes the EcoRI restriction site to probe for adduct-induced mutagenesis. An adduct is placed within the EcoRI site of a replicating plasmid. After transfection of the plasmid into human cells, the replicated plasmid DNA is harvested at various time points post-transfection. Digestion of any recovered WT plasmid DNA with EcoRI results in linearization of the DNA and inefficient transformation into Escherichia coli, which yield little to no colony-forming units (cfu) on antibiotic-containing LB agar. Recovered plasmid DNA with mutations in the EcoRI site is resistant to EcoRI digestion and can be efficiently transformed into E. coli, and colony formation on antibiotic-containing LB agar occurs. We validated this approach by converting the first A to T, C, and G within the EcoRI restriction site, and then the mutagenesis assay was per-  FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 formed (Fig. 4B). We found that the colony forming frequency of the mutated plasmid DNA (T, C, and G) was significantly elevated compared with the wild type (A) in HEK293 cells (Fig.  4B). However, these observations were based on artificial muta-tions and not mutations because of improper repair of DNA adducts. Thus, additional validation was carried out by using two different types of DNA adducts (⑀A and 8-oxo-dG) in HEK293 cells. Our data showed a statistically significant increase (p Ͻ 0.005) in mutation frequency for both adducts compared with undamaged control (Fig. 4C). We observed a 5-fold higher mutation frequency for 8-oxo-dG (ϳ50 mutations/1000) compared with ⑀A (ϳ10 mutations/1000), which is consistent with a previous report (35). The mutant progenies in colonies from an EcoRI-restricted plate were confirmed by their resistance to further EcoRI digestion (Fig. 4D).

hMPG nsSNPs Increase Genomic Instability
Colon carcinoma HCT116 cells were used for the mutagenesis assay as Mpg knock-out mice showed elevated colon cancer incidence when the mice exhibited colitis or were subjected to Helicobacter pylori infection (36). Of note, HCT116 cells are mismatch repair-deficient. However, there has been no evidence of the mismatch repair pathway repairing BER substrates, especially ⑀A, nor does mismatch repair affect BER of ⑀A. Therefore, the presence or absence of efficient mismatch repair activity in a given cell line is irrelevant to the results of this assay. Importantly, HCT116 cells compared with other cell lines demonstrate appreciable sensitivity for detection of ⑀A-induced mutation (11,35). However, HCT116 cells express endogenous WT hMPG; hence, any results generated after expression of our variants will be obscured by the endogenous WT hMPG. To circumvent this problem, we knocked down endogenous MPG by generating stable cell lines with shRNA specific to WT hMPG (Fig. 4E). Then we complemented the hMPG knockdown HCT116 cells with transient expression of WT, R120C, and R141Q hMPG proteins with silent mutations at shRNA target sites (Fig. 4F). Under these conditions, we performed a site-specific mutagenesis assay for ⑀A and found a significant increase (p Ͻ 0.005) in mutation frequency for R141Q (ϳ36/1000) and R120C (ϳ22/1000) variants compared with hMPG WT (ϳ10/1000) (Fig. 4G). Subsequent sequencing of mutants from each variant showed ϳ92% insertion/deletion due to ⑀A in HCT116 cells. These results were consistent with a previously published report, which found ϳ89% insertion/deletion due to ⑀A in HCT116 cells (11). This consistency with other studies also validates our in vivo site-specific mutagenesis method we developed for this study.
In addition to ⑀A and Hx, MPG repairs the replication blocking toxic DNA adduct 3-methyladenine in genomic DNA that is induced by various alkylating agents, including MMS (7). Consequently, loss of hMPG renders mammalian cells sensitive to MMS (37,38). We intended to determine whether variants are responsive to MMS treatment. Wild type hMPG and variants were expressed in Mpg Ϫ/Ϫ MEF cells, and an MMS sensitivity assay was performed. We found that expression of WT MPG protected cells from MMS-induced killing as expected, whereas variants showed no protection compared with the vector control (Fig. 4H).
Reduced Affinity of hMPG Variants for ⑀A-DNA-In search of a mechanism of modulation of R141Q and R120C hMPG activity, we tested the affinity of WT hMPG and variants (Fig. 5,  B-D) for ⑀A-DNA using surface plasmon resonance spectroscopy. Our results showed that the equilibrium dissociation constant of binding (K D ) is 2.25 nM for the WT hMPG and 1340 and 197 nM for R141Q and R120C hMPG, respectively (Fig. 5E). Thus, R141Q and R120C exhibited significantly decreased affinity toward ⑀A-DNA compared with WT hMPG, thereby suggesting that the reduced enzymatic activity observed might be due to a reduction in substrate binding. Furthermore, among the variants, R141Q had a significantly higher K D and showed extremely weak ⑀A-DNA binding, consequently explaining the activity difference between the variants.
Reduced Affinity of hMPG Variants to Control-DNA-In view of the observation that WT hMPG compared with R141Q and R120C variants had significantly less affinity for ⑀A-DNA as indicated by the low K D , we further tested whether the nonspecific DNA binding was also altered. We directly measured  FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 the binding kinetics of WT and variant hMPGs to control-DNA using surface plasmon resonance spectroscopy (Fig. 6, A-C). Our results showed an apparent K D of 34.2 nM for WT hMPG and 819 and 174 nM for R141Q and R120C variants, respectively (Fig. 6D). Overall, the results from surface plasmon resonance spectroscopy indicate that hMPG variants not only lost affinity for substrate DNA, but also nonspecific MPG-DNA interaction was severely disrupted, and thereby the variants are unable to distinguish between specific and nonspecific DNA. It should be noted that the overall reduced affinity for DNA of the variants is consistent with the results of the experiments modeling the heterozygous genotype with purified proteins (Fig. 3A). We did not observe a dominant negative effect with an equal concentration of either variant and WT hMPG, which indicated that the variants were not inordinately bound to substrate and preventing substrate binding and catalysis by WT hMPG.

hMPG nsSNPs Increase Genomic Instability
WT and hMPG Variants Have Similar Chemistry, but the Variants Exhibited Slower Turnover-Prompted by the observation that the R141Q and R120C substitutions could affect product formation in ⑀A-DNA/hMPG reactions (Figs. 1 and 2), we tested whether those variations affect any of the catalytic intermediate steps other than the binding step. We conducted single turnover kinetics with hMPG proteins to measure the k chem (Fig. 5A and see "Experimental Procedures" for definition) (26,39). Two different enzyme/substrate ratios for both WT and variant proteins provided similar k chem values, ensuring that the enzymatic reactions were following single turnover conditions. However, the WT, R141Q, and R120C hMPG have relatively similar k chem values:1.4 Ϯ 0.02, 1.1 Ϯ 0.2, and 1.3 Ϯ 0.2, respectively (Fig. 7, A and C), indicating minimal effect of those polymorphic variations on the chemistry step of MPG/ ⑀A-DNA reaction.
Next, we attempted to measure the k pd (26,39). For the reaction of WT hMPG with ⑀A-DNA, we previously found a k pd value of 0.00435 Ϯ 0.00057 s Ϫ1 (39). Here, we found a similar k pd value for WT hMPG, but notably at the same hMPG and ⑀A-DNA concentrations both of the variant proteins had much lower turnover (Fig. 7B). In fact, the turnover was extremely slow, and it was neither practical nor possible to perform curve fitting to assign an accurate value of k pd for both of the variant proteins. Therefore, it is evident that R120C and R141Q affected the turnover of hMPG for ⑀A-DNA reactions in a significant fashion in addition to affecting the substrate binding step.
Prediction of Full-length hMPG Structure and Molecular Dynamics Study-So far, no three-dimensional structures are available for full-length hMPG; however, in this study, we used full-length WT and variant hMPG proteins. Therefore, in the current study, we predicted the three-dimensional structure of hMPG by molecular modeling through energy minimization and the available x-ray structure of truncated hMPG. The structure of residues 1-78 was predicted to be unfolded and looped (data not shown). We then analyzed the effect of R120C and R141Q variations on this structure by molecular modeling (Fig. 8, A-F). The Arg-120 forms salt bridges with Glu-185 and Glu-253 in its structure, and the R120C substitution abolished the charge, which is necessary for the salt bridge interactions (Fig. 8, B and C). Upon changing Arg-120 to Cys-120, a 1-ns molecular dynamic simulation trajectory revealed that Glu-253 moved toward Cys-120 to form a hydrogen bond with Cys-120 and filled the empty space created by the substitution. Furthermore, Glu-185 moved away from Cys-120 and formed a new hydrogen bond with Ser-252 in the R120C structure. These changes upon R120C substitution altered the structural folding in that region, which is naturally compressive in the WT structure. As a result the abrogation of those WT interactions in R120C may prevent hMPG from properly folding or maintaining its proper conformation once folded. The Arg-141 variation in fact had a very distinct effect: Arg-141 formed a salt bridge with DNA (Fig. 8, D and E), but the R141Q substitution completely disrupted this salt bridge interaction (Fig. 8F); the latter is required for the product formation. Thus, these variants had significant effect in the structure of the catalytic pocket of MPG and in turn affected its activity.

DISCUSSION
This study aimed to learn the effect of epidemiologically relevant substitutions/polymorphic variants of hMPG on struc-   FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 ture, function, and biological outcome, specifically in the context of genomic instability and sensitivity to a DNA-damaging agent. Two of these eight germ line variants, R141Q and R120C, had reduced enzymatic activity in vitro as well as in mammalian and mouse cell extracts (Figs. 1 and 2). Expression of these two variants in human colon cells also showed an increase in mutation frequency compared with the WT MPG (Fig. 4G). Furthermore, unlike WT MPG, the variants failed to protect MPG-null MEFs against MMS treatment (Fig. 4H). Mechanism studies showed that these two variants had reduced activity due to low affinity for DNA and loss of substrate specificity as a consequence of significant disruption in the structure of the catalytic pocket of hMPG (Figs. 5, 6, and 8). Recapitulating the heterozygous genotype for both variants, they showed reduced activity compared with homozygous WT hMPG (Fig. 3), thereby underscoring the notion that hMPG variants are not only less capable of repairing damaged DNA but may also facilitate mutation and perhaps even genomic instability (Fig. 4G). Furthermore, we may expect individuals that are homozygous for variants to show even further reduced repair and consequently genomic instability. Interestingly, in cell extracts, we did observe a slight dominant negative effect with R141Q expressed at equal levels with WT hMPG (Fig. 3, C and D), indicating that individuals heterozygous for the R141Q variant may not be protected from mutation by expression of a WT hMPG allele.

hMPG nsSNPs Increase Genomic Instability
It has been established that chronic inflammation increases cancer risk, generally by the production of reactive oxygen and nitrogen species-induced DNA damage. In this case, the expression and activity of hMPG will repair such damage, thereby indicating that hMPG may have an association with risk of disease such as cancer (40). In support of this notion, studies have shown in an MPG knock-out mouse model of episodic inflammatory bowel disease by repeated administration of dextran sulfate sodium that MPG-mediated DNA repair prevents colonic epithelial damage and reduces the severity of dextran sulfate sodium-induced colon cancer (36). Furthermore, that study proposed that MPG could be a strong candidate for genetic association studies of human colorectal cancer or gastric cancer risk (36). In fact, in our study, we found that R141Q and R120C variants enhanced mutation frequency compared with WT hMPG, and more interestingly, there was a significant increase in ⑀A-induced insertion/deletions (Fig. 4). Thus, individuals that are heterozygous or homozygous for these variants may be predisposed to colon or gastric cancer and more so if they are affected by inflammatory bowel or related diseases.
In both R120C and R141Q, an arginine residue was replaced. Apparently, the observed decrease in enzymatic activity may be explained by simple electrostatic interaction dynamics at the point of DNA contact because a critical basic residue within or around the catalytic pocket was replaced with a neutral residue (41). Moreover, for another DNA glycosylase, T4 endonuclease V, basic residues play a critical role in non-target DNA binding (42), leading to substrate binding and catalysis. Thus, the distortion due to the reduced basic property in variant proteins may have an overall adverse effect at the DNA contact point and affect the DNA binding. From the crystal structure of truncated hMPG, it was found that Arg-141 directly interacts with DNA. Therefore, in our study, it was not surprising that R141Q has ϳ595-fold less binding than the wild type (Fig. 5, B-E). In fact, our molecular dynamics study showed that due to substitution the salt bridge between DNA and the R141Q protein is completely disrupted, conferring serious consequences to the binding of hMPG protein to DNA. Notably, Arg-120 is far from the catalytic pocket. However, the R120C variant still afforded enough changes in residues near the catalytic site to affect the DNA-protein interaction (Fig. 8). In general, being a broad substrate enzyme, MPG must be flexible for DNA binding to recognize DNA lesions of varied structures (43). However, this molecular dynamics study showed that even a slight change in distance could affect the catalytic pocket of hMPG, indicating some rigidity in the structure of the catalytic pocket to ensure product formation.
It is true that the substrate binding step is generally not ratelimiting for the WT MPG reaction (39). However, our experiments demonstrated that the turnover (k ss ) of the variants is slow compared with that of the wild type protein, which could be explained by the reduced affinity of the variants toward ⑀A-DNA (Figs. 5 and 7). Notably, k pd is composed of k off values for both free base and the AP site-containing DNA. Moreover, the k ss consists of k pd and the effective rate constant (kЈ) for the process MPG ϩ ⑀A-DNA 3 MPG⅐AP site DNA⅐⑀A (26). Generally, for obtaining the definite value for product dissociation from burst kinetics, it is assumed that k pd is much slower than the kЈ, and thus, the contribution of the latter is ignored, and k ss becomes identical to k pd . However, relatively lower binding affinity toward substrate in the case of variant MPG violates that assumption; therefore, an accurate k pd value for the variant proteins could not be assigned (Fig. 7). Hence, the overall effect on product formation apparently arises from the alterations in the binding step that in turn affect the turnover step (Figs. 5-7). In this way, substrate binding is in fact the rate-limiting step of the hMPG variant-mediated excision reaction.
In conclusion, our study supports the notion that WT hMPG is critical to the maintenance of genomic integrity. The variants characterized in this work not only exhibit dramatic effects on adduct-induced mutagenesis in vivo but are also epidemiologically relevant given their prevalence in the population. It remains to be seen whether or not these hMPG variants confer increased risk for cancer and susceptibility to other conditions such as neurodegenerative diseases and aging.