The Werner syndrome protein limits the error-prone 8-oxo-dG lesion bypass activity of human DNA polymerase kappa

Human DNA polymerase kappa (hpol κ) is the only Y-family member to preferentially insert dAMP opposite 7,8-dihydro-8-oxo-2′-deoxyguanosine (8-oxo-dG) during translesion DNA synthesis. We have studied the mechanism of action by which hpol κ activity is modulated by the Werner syndrome protein (WRN), a RecQ helicase known to influence repair of 8-oxo-dG. Here we show that WRN stimulates the 8-oxo-dG bypass activity of hpol κ in vitro by enhancing the correct base insertion opposite the lesion, as well as extension from dC:8-oxo-dG base pairs. Steady-state kinetic analysis reveals that WRN improves hpol κ-catalyzed dCMP insertion opposite 8-oxo-dG ∼10-fold and extension from dC:8-oxo-dG by 2.4-fold. Stimulation is primarily due to an increase in the rate constant for polymerization (kpol), as assessed by pre-steady-state kinetics, and it requires the RecQ C-terminal (RQC) domain. In support of the functional data, recombinant WRN and hpol κ were found to physically interact through the exo and RQC domains of WRN, and co-localization of WRN and hpol κ was observed in human cells treated with hydrogen peroxide. Thus, WRN limits the error-prone bypass of 8-oxo-dG by hpol κ, which could influence the sensitivity to oxidative damage that has previously been observed for Werner's syndrome cells.


INTRODUCTION
DNA damage is a barrier to faithful replication of the genome. Biological systems are continuously exposed to agents that react with DNA and alter its physicochemical properties, which can impair normal nucleic acid metabolism (1,2). DNA adducts that are not repaired can block replication fork progression (3,4). The activation of replication stress response pathways can lead to the recruit-ment of specialized DNA polymerases (pols) to perform translesion synthesis across adducts (5). The Y-family DNA polymerases are central to this process (6). Humans have four Y-family members that participate in a number of pathways involved in cellular responses to DNA damage (6). Each human Y-family pol bears unique structural characteristic and lesion bypass properties (7). Understanding the distinct mechanism of action for each of the four human Y-family pols is an ongoing area of investigation, as the diversity of form and function found within this family of enzymes is remarkable.
The Y-family member human DNA polymerase kappa (hpol ) was first identified as a homolog of the dinB gene product in Escherichia coli and is perhaps best recognized for its involvement in bypass of bulky minor-groove DNA adducts, as evidenced by its in vitro bypass properties and cellular responses to the aryl hydrocarbon receptor activation (8)(9)(10). The N-clasp of hpol has been proposed to help facilitate accurate and efficient bypass of N 2 -dG adducts (11,12). There is evidence that hpol contributes to bypass of other DNA adducts, as the enzyme appears to be recruited to sites of DNA damage following exposure to methylmethane sulfonate and participates in tolerance of O 6 -methylguanine adducts (13,14). Mis-regulation of hpol is observed in many tumors, including primary brain tumors (15)(16)(17). Indeed, experiments have illustrated that even moderate over-expression of hpol (∼2-fold) can have extremely negative consequences toward genome stability, including reduced replication fork rates, stimulation of double-strand break formation, aberrant activation of homologous recombination and increased chromosomal damage (18). Thus, studying the catalytic properties of hpol could provide insights into mechanisms that contribute to mutagenesis and genomic instability in human disease states, such as cancer.
An interesting feature of hpol catalysis is its error-prone bypass of the ubiquitous oxidative DNA lesion 7,8-dihydro-8-oxo-2 -deoxyguanosine (8-oxo-dG) (19,20). The 8-oxo-dG adduct has been widely studied because it is highly mutagenic and elevated levels of the lesion have been associ-ated with cancer and aging (21,22). Reactive species, such as 1 O 2 , · OH, one-electron oxidants and peroxynitrate, can react with guanine to form 8-oxo-dG (23,24), and increased 8-oxo-dG levels have been linked to metastasis in several cancer types, including brain and breast cancer (25)(26)(27)(28). It has been estimated that somewhere in the range of several hundred to a few thousand 8-oxo-dG lesions are formed in a cell per day (25,29). The most common mutations associated with 8-oxo-dG adducts are G → T transversions (30). This is due to the fact that many DNA pols readily insert deoxyadenosine monophosphate (dAMP) opposite 8-oxo-dG (31)(32)(33)(34)(35)(36)(37). The preferred mis-insertion of dAMP opposite 8-oxo-dG by hpol is noteworthy because it is apparently the only Y-family pol that copies the lesion in such an error-prone fashion and with essentially no reduction in catalytic efficiency (20,(38)(39)(40)(41). Structural comparisons with other Y-family pols suggest that hpol apparently lacks electrostatic contacts between the little finger domain and the templating 8-oxo-dG that are necessary for stabilization of anti-oriented 8-oxo-dG, which promotes effective pairing of dC:8-oxo-dG in other Y-family pols, such as Dpo4 from Sulfolobus solfataricus (42). Importantly, a recent study showed that hpol contributes to the mutagenic replication of 8-oxo-dG in human cells (43). Therefore, understanding hpol bypass of 8-oxo-dG could provide important insights into error-prone DNA replication.
Many studies have illustrated the importance of proteinprotein interactions in determining the accuracy and efficiency of translesion DNA synthesis (TLS) (19,39,(44)(45)(46). Interactions with the Werner syndrome protein (WRN) can stimulate hpol activity against 8-oxo-dG in vitro (47), but mechanistic features and the impact of WRN upon TLS fidelity remain unclear. The multi-functional WRN enzyme is involved in DNA replication and repair pathways, including MUTYH-mediated repair of 8-oxo-dG (48)(49)(50)(51)(52)(53)(54). Therefore, we set out to study the functional interaction between the genome caretaker WRN and Y-family member hpol in order to better understand the mechanism of action and potential ramifications upon translesion synthesis across DNA adducts, such as 8-oxo-dG.

MATERIALS AND METHODS
All unlabeled deoxynucleoside triphosphates (dNTPs) were obtained from GE Healthcare Life Sciences (Piscataway, NJ, USA). All oligonucleotides used in this work were synthesized by Integrated DNA Technologies (Coralville, IA, USA) and purified using high-performance liquid chromatography (HPLC) by the manufacturer, with analysis by matrix-assisted laser desorption timeof-flight mass spectrometry. The primer sequence used for activity assays and kinetic analysis of insertion opposite dG and 8-oxo-dG was 5 -6-carboxyfluorescein (FAM)-TTTGGGGGAAGGATTC-3 and the primer used for kinetic analysis of next-base extension was either 5 -FAM-TTTGGGGGAAGGATTCC-3 or 5 -FAM-TTTGGGGGAAGGATTCA-3 . The template sequence used in all pol assays was 5 -TCACXGAATCCTTCCCCC-3 , where X represents either dG or 8-oxo-dG.

Expression and purification of recombinant proteins
The hpol  was expressed and purified as described previously (20). WRN constructs were expressed and purified as described previously (55), except for WRN 500-1150 , WRN 500-1092 , WRN 500-949 and WRN 949-1092 . These WRN constructs possessed N-terminal 6xHis-glutathione transferase (GST) affinity tags. The His-GST-WRN fusion proteins were expressed in E. coli BL21 (DE3) Gold cells (Agilent Technologies, Santa Clara, CA, USA). Cells were grown at 37 • C and 250 rpm for 3 h (OD 600 = 0.5-0.6), followed by induction for 3 h (37 • C and 250 rpm) by addition of isopropyl ␤-D-1-thiogalactopyranoside (1 mM), and finally harvested by centrifugation. Buffer containing 50-mM Tris-HCl (pH 7.4), 0.5-M NaCl, 10% glycerol (v/v), 5-mM ␤-mercaptoethanol (␤-ME), lysozyme (1 mg/ml) and a protease inhibitor cocktail (Roche, Basel, Switzerland) was added to the harvested pellet. The suspension was sonicated and supernatant recovered from an ultracentrifugation step (35 000 g, 1 h, 4 • C). The protein was purified by two affinity steps using Ni Sepharose (GE Healthcare Life Sciences) followed by Gluthatione Seph3rose 4B beads (GE Healthcare Life Sciences). Briefly, the protein was bound to a Nichelating column in 50-mM Tris-HCl (pH 7.4) buffer containing 0.3-M NaCl, 10% glycerol (v/v) and 5-mM ␤-ME. The column was washed with 40-60-mM imidazole and the protein eluted as a single peak in 400-mM imidazole. Following dialysis to remove imidazole and lower the concentration of NaCl to 0.15 M, the protein was added to the GST column in 25-mM Tris-HCl (pH 7.4) buffer containing 0.2-M NaCl, 10% glycerol (v/v) and 5-mM ␤-ME. After washing, the protein was then cleaved from the His-GST tag by treatment with HRV 3C protease (Thermo Scientific) on the GST column, according to the methods suggested by the manufacturer. The highly pure proteins (Supplementary Figure S1) were stored at -80 • C in the HEPES buffer (pH 7.5) containing 0.2-M NaCl, 5-mM ␤-ME and 30% glycerol. The DNA binding capacity of the WRN 949-1092 construct was analyzed as a means of verifying proper folding of this short construct (Supplementary Figure S2).

Full-length extension polymerase assays
Fluorescein-labeled primer was annealed to template oligonucleotide by heating a 1:2 molar ratio of primer:template DNA (p/t-DNA) to 95 • C for 5 min and then slow cooling to room temperature (RT). The p/t-DNA was then incubated with hpol prior to extension in the presence or absence of WRN. For full-length extension experiments, hpol was added to the tube first, followed by addition of WRN (where indicated). The DNA substrate was added after WRN and the solution was incubated on ice for 30 min prior to starting the reaction. Each reaction was initiated by adding dNTP·MgCl 2 (0.25 mM of each dNTP and 5-mM MgCl 2 ) solution to a pre-incubated hpol ·DNA complex (2-nM hpol and 200-nM DNA). Where indicated WRN (100 nM) was incubated with hpol and DNA. Unless otherwise stated, all enzymatic reactions were carried out at 37 • C in 50-mM HEPES buffer (pH 7.5) containing 60-mM KCl, 5-mM dithiothreitol, 100-g ml −1 bovine serum albumin (BSA) and 10% (v/v) glycerol. At the indicated time points, 4-l aliquots were quenched with 16 l of a 95% formamide (v/v)/20-mM ethylenediaminetetraacetic acid (EDTA)/0.1% bromophenol blue (w/v) solution and were separated by electrophoresis on a 16% polyacryamide (w/v)/7-M urea gel. The products were then visualized using a Typhoon imager (GE Healthcare Life Sciences) and quantified using ImageQuant TM software (GE Healthcare Life Sciences). The total product formed over time was fit to a single-exponential equation [Equation (1)]: where A is the product formed, k 1 is the rate constant defining polymerization under the conditions used for the experiment being analyzed and t is the time.

Steady-state kinetic analysis of polymerase activity
Single-nucleotide incorporation by hpol (2 nM) on DNA substrates (200 nM) was measured over a range of dNTP concentrations (0-250 M). The reaction buffer and the general protocol was the same as that used for full-length extension assays. Time points were chosen such that only dNTP insertion by hpol  (and not exonucleolytic degradation of the primer) was observed for reactions with either full-length WRN (a.a. 1-1432) or the WRN exo domain (a.a. 1-333). Products were analyzed as described for full-length extension assays. The initial portion of the velocity curve was fit to a linear equation in the program Graph-Pad Prism (GraphPad, San Diego, CA, USA). The resulting velocity was plotted as a function of dNTP concentration and then fit to a hyperbolic equation, correcting for enzyme concentration, to obtain estimates for the turnover number (k cat ) and Michaelis constant (K M,dNTP ).

Pre-steady-state kinetic analysis of hpol polymerase activity
All pre-steady-state experiments were performed using a KinTek RQF-3 model chemical quench-flow apparatus (KinTek Corp., Austin, TX, USA). The hpol 19-526 (25 nM) enzyme was pre-incubated with FAM-labeled p/t-DNA (50 nM). The reaction was initiated by rapid mixing of the enzyme·DNA solution with a solution containing MgCl 2 (5 mM) and varying concentrations of dNTP (from 0.5 M to 1 mM). Polymerase catalysis was stopped by the addition of 50-mM EDTA (pH 9.0). Product formation was either fit to a single-exponential equation [Equation (1)] or to a singleexponential equation with a second linear phase [Equation (2)]: where A is the product formed in the burst phase, k 1 is the rate constant defining polymerization under the conditions used for the experiment being analyzed, k 2 is the observed second-order rate of product formation and t is the time. Equation (1) was only used to fit pre-steady-state results for low concentrations of dNTP (0.5-1-M dNTP) where the second phase of the reaction curve was not apparent.

Measurement of WRN 3 -5 exonuclease activity
Exonucleolytic degradation of polymerase products was measured by incubating WRN exo (a.a. 1-333; 100 nM) with the indicated 14/18-mer DNA substrates (200 nM). For experiments measuring WRN exo activity, p/t-DNA (200 nM) was added to the reaction tube followed by addition of WRN  . The solution was allowed to incubate for ∼30 min and then initiated by addition of MgCl 2 (5 mM). For experiments to test competing exo and pol activities, hpol (2 nM) was added to the reaction mixture first, followed by DNA (200 nM) and finally, WRN  . The mixture was allowed to incubate for ∼30 min before initiating the reaction by addition of dGTP (100 M) and MgCl 2 (5 mM). In both sets of experiments, a zero time point was collected by adding quench to the reaction mixture prior to the addition of dGTP/Mg 2+ . Reactions were quenched, products separated and quantified in a manner identical to that described above for the polymerase assays. WRN-catalyzed degradation of FAM-labeled 14-mer primer was quantified and the rate of degradation was estimated by linear regression analysis. The sum of all degraded products was measured to estimate the rate of WRN action. To quantify exo activity in reactions containing hpol , we quantified the substrate, pol products and exo products as total DNA and then calculated the amount of substrate degraded by WRN as the fraction of total DNA in the lane. The rate of exo product formation was estimated by fitting the initial portion of the velocity curve to a linear function.

Pull-down assays with recombinant WRN and hpol
Glutathione Sepharose 4 Fast Flow beads (GE Healthcare) were washed extensively in binding 25-mM HEPES (pH 7.5) buffer containing 0.1-M NaCl, 10% (v/v) glycerol and 5-mM ␤-ME. Two sets of beads were prepared for the pull-down experiments. The washed beads were incubated with either purified GST protein or purified GSTtagged hpol  . Both sets of beads were incubated at 4 • C overnight with constant agitation in the binding buffer. The beads were again washed thoroughly with the binding buffer to remove unbound protein and used for subsequent pull-down experiments with the different WRN constructs. GST-coated beads served as the negative control for all experiments testing specific pull-down of WRN constructs by hpol  . For each experiment, 10 l of proteincoated beads (∼1200-pmol protein) were mixed with the WRN construct (WRN 1-333 , WRN 949-1092 , WRN 500-949 and WRN 500-1092 ) in a total volume of 500-l binding buffer for a final WRN protein concentration of 1 M (∼500pmol protein). The tubes were incubated for 16 h at 4 • C with constant agitation, after which they were centrifuged at 5000 rpm for 5 min. The supernatant, which represented unbound fraction, was collected, centrifuged under vacuum to dryness and re-suspended in 40 l of dH 2 O. The beads were then washed three times with 1 ml of binding buffer each time. Washed beads were then boiled in 20 l of sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) loading buffer for 10 min to elute bound proteins and then diluted to 40 l with 1X phosphate buffered saline (PBS). Both the unbound fractions (loaded 20 l, 50% of total supernatant) and bound proteins eluted from the beads (loaded 20 l, 50% of total eluted protein) were separated on a 4-20% gradient SDS-PAGE and stained with Coomassie Brilliant Blue for visualization. The stained gels were scanned on an ImageQuant LAS 4000 scanner (GE Healthcare), and the band intensities were quantified using ImageQuant software. For each experiment, the fraction of WRN protein bound to the beads was calculated by dividing the band intensity of protein eluted from the beads by total intensity (obtained by adding intensities in unbound supernatant and the eluted lanes). Specific binding of each WRN protein to hpol  was expressed as a ratio of fraction bound to the GST-tagged hpol  beads divided by the fraction bound to the GST-coated beads.

Immunofluorescence assays for determining the co-localization of WRN and hpol
HeLa cells used for this study were kindly provided by Dr Timothy Chambers (University of Arkansas for Medical Sciences, Little Rock, AR) and maintained in Dulbecco's minimum essential medium (Life technologies, Grand Island, NY, USA) containing 10% (v/v) heat-inactivated fetal bovine serum (Life technologies) and supplemented with penicillin and streptomycin (Life technologies). For Immunofluorescence studies, HeLa cells (0.3 × 10 6 ) cultured on glass coverslips were treated with 0.5 mM H 2 O 2 (Sigma) for 4 h. Untreated cells were prepared as a control. After 4 h, cells were fixed with 3.7% (v/v) formaldehyde for 20 min at RT. The fixed cells were then permeabilized and blocked by soaking in 0.2% (v/v) Triton X-100 and PBS containing 5 mg/ml BSA for 30 min at RT. The fixed and permeabilized cells were then incubated for 1 h at RT with primary antibodies diluted in blocking solution: rabbit anti-WRN and mouse anti-hpol (Novus biological, 1:100). The slides were washed with PBS and incubated for 1 h at RT with secondary antibodies diluted in blocking solution: Dylight 488 conjugated goat anti-mouse IgG (Thermo Scientific Pierce, Rockford, IL, USA; 1:100) and Dylight 550 conjugated goat anti-rabbit IgG (Thermo Scientific Pierce, Rockford, IL, USA; 1:100). After washing with PBS, coverslips were mounted on Vectashield (Vector Laboratories) containing 4 ,6-diamidino-2-phenylindole (DAPI) and images were captured on an Olympus IX81 fluorescence microscope. Cells possessing one or more yellow foci were scored as positive for WRN/hpol co-localization. At least 50 nuclei were scored in each of three independent experiments.

The exonuclease and RecQ C-terminal domains of WRN are required for full stimulation of hpol primer extension properties
WRN is a protein with several enzymatic functions and multiple domains ( Figure 1A) (48). We began our study by testing the ability of full-length WRN and a series of WRN truncation mutants to stimulate extension by hpol on unmodified DNA templates ( Figure 1B). All WRN stocks were checked for contaminating polymerase activity by performing extension experiments out to an hour in the absence of hpol (data not shown). Extension assays were repeated at least twice to ensure the reproducibility of any stimulation of polymerization by WRN. Total product formed in each lane was plotted as a function of time and the resulting curve was fit to a single-exponential equation to obtain an estimate of the rate constant for primer extension by hpol . To compare the relative effect of different WRN constructs on hpol extension activity, we divided the rate constant for primer extension in the presence of WRN by the rate constant for primer extension when hpol is alone and multiplying by a factor of one hundred to yield percent activity. In this way, we were able to make quantitative comparisons for stimulation of pol extension activity by different WRN constructs.
Similar to our previous results and those reported by the Loeb group with hpol (47,55), we find that fulllength WRN stimulates hpol extension activity on unmodified template DNA ( Figure 1C). The rate constant for total product formation is increased ∼4-fold by inclusion of full-length WRN 1-1432 ( Figure 1D). We then went on to test the domain requirements for WRN-mediated stimulation of hpol activity by performing extension assays with truncated versions of the WRN protein. We find that the Helicase and RNaseD C-terminal (HRDC) deletion construct (a.a. 1-1092/E84A) stimulates hpol extension activity ∼4-fold, similar to full-length WRN ( Figure 1E and Supplementary Figure S3). As we observed previously with hpol (55), deletion of the RecQ C-terminal (RQC) domain (a.a. 1-949/E84A) reduces the effect upon hpol to a little less than 2-fold enhancement. A similar reduction in the level of pol stimulation was observed for a WRN construct possessing the ATPase/RQC/annealase domains (a.a. 500-1150), indicating a need for both RQC and exo domains for full stimulation of hpol extension activity. The WRN ATPase/RQC construct (a.a. 500-1092) stimulated hpol activity 2-fold but the WRN ATPase domain alone (a.a. 500-949) did not stimulate extension at all (Figure 1E). Adding either the WRN exo domain (a.a. 1-333) or the RQC domain (a.a. 949-1092) to the reaction mixture results in an ∼2-fold stimulation of hpol extension activity ( Figure 1E). These results indicate that both the WRN exo and RQC domains are important for stimulation of hpol extension activity, consistent with results obtained with hpol (55).

Physical interaction between WRN and hpol
The pol extension results showed that WRN stimulates the activity of the hpol catalytic core and that full stimulation required both exo and RQC domains of WRN. These results imply that the two proteins bind to one another but there is no previous evidence for a physical interaction between WRN and hpol . To determine if WRN and hpol physically interact with one another, we performed in vitro pull-down experiments with recombinant enzymes. A GST-tagged version of hpol (GST-hpol ) was incubated with glutathione sepharose beads and WRN. The specificity of the interaction with hpol was confirmed by incubating each WRN construct with GST-coated beads (i.e. no hpol present in the binding mixture). We used four WRN constructs to investigate the interaction with hpol : The WRN 500-1092 ATPase/RQC construct bound to GST-hpol -coated beads nearly 2-fold better than the negative control GST beads (Figure 2A). Likewise, the WRN exo domain (a.a. 1-333) was found to interact specifically with GST-hpol -coated beads ( Figure 2B). The strongest interaction was observed with the WRN RQC domain (a.a. 949-1092), which was enriched almost 3-fold over the negative control GST-coated beads ( Figure 2B). Of the four constructs tested, only the WRN 500-949 , which lacks both the RQC and exo domains, failed to bind specifically to GSThpol -coated beads ( Figure 2B). These results confirm that WRN and hpol physically interact through both the RQC and exo domains, consistent with the pol extension results and similar to our previous study with WRN and hpol (55).
We then went on to investigate whether WRN and hpol co-localize inside cells in response to DNA damage. To examine the potential cellular interaction between the two enzymes, we treated HeLa cells with H 2 O 2 (500 M) for up to 4 h and used immunofluorescence to monitor the spatial distribution of WRN and hpol . In untreated cells, there is very little overlap in the signal for hpol and WRN (Figure 2C). Treatment with H 2 O 2 results in an increased number of discrete foci for both enzymes and an increase in colocalization ( Figure 2C). The number of cells with WRN and hpol co-localized increases from less than 10% in untreated cells to almost 80% of cells showing overlap following exposure to H 2 O 2 ( Figure 2C and D). The finding that WRN and hpol are found co-localized in response to oxidative damage is the first to show that these two enzymes might encounter each other inside human cells. We then went on to examine the mechanism of WRN action on TLS by hpol .

Steady-state kinetic analysis of WRN-mediated stimulation of hpol activity on undamaged DNA
To ascertain whether WRN altered single-nucleotide insertion kinetics for hpol , we measured the turnover number (k cat ) and the Michealis constant (K M,dNTP ) for hpolcatalyzed single-nucleotide insertion reactions. From these values we calculated the specificity constant (k cat /K M,dNTP ), which is often considered an estimate of catalytic efficiency and we use the two terms (i.e. specificity constant and catalytic efficiency) interchangeably throughout the text. We first examined steady-state kinetics for dCTP insertion opposite unmodified template dG. Similar to extension assays, those WRN constructs possessing the RQC domain appear to impart the largest stimulatory effect on the k cat for hpol -catalyzed insertion of deoxycytidine monophosphate (dCMP) opposite template dG (  . 500-1150 and a.a. 949-1092) only increase the specificity constant ∼2-fold, consistent with the extension results. There is a very slight increase in the k cat for hpol insertion of dCMP opposite dG in the presence of WRN 1-333 and WRN 1-949/E84A , but there is essentially no change in the specificity constant for either of the constructs lacking the RQC ( Table 1). As such, the WRN RQC domain appears to be a stronger factor in stimulating the efficiency of singlenucleotide insertion by hpol on undamaged templates.

Steady-state kinetic analysis of WRN-mediated stimulation of hpol activity on 8-oxo-dG-containing template DNA
We were interested in examining whether WRN could modulate the DNA damage bypass properties of hpol . A previous study showed qualitative stimulation of hpol -catalysis against several DNA adducts (47), but the effect upon the efficiency and accuracy of polymerization remains unclear. While the TLS properties of hpol have been studied with a variety of adducts, we chose to investigate 8-oxo-dG for three reasons: (i) hpol is by far the most error-prone Yfamily member at bypass of 8-oxo-dG, (ii) there is evidence that hpol contributes to mutagenic bypass of 8-oxo-dG in human cells and (iii) crystal structures are available with hpol in ternary complex with 8-oxo-dG-modified DNA. In this way, the results obtained with hpol in the presence of WRN can be assessed within the context of rigorous mechanistic studies and could provide insights into cellular mechanisms of mutagenesis.
We performed steady-state analysis of hpol -catalyzed single-nucleotide insertion across from 8-oxo-dG. Singlenucleotide insertion kinetics reveal that full-length WRN 1-1432 stimulates the k cat for dCMP insertion opposite 8-oxo-dG a little over 3-fold ( Table 2). The addition of full-length WRN also decreases the K M,dNTP for dCTP ∼3-fold. Importantly, the addition of WRN increases the specificity constant for dCMP insertion opposite 8-oxo-dG ∼10-fold relative to that observed for hpol alone. The WRN HRDC-deletion construct (a.a. 1-1092/E84A) also stimulates the specificity constant for dCMP insertion opposite 8-oxo-dG over 10-fold. The ATPase/RQC (a.a. 500-1150) and RQC (a.a. 949-1092) constructs increase the efficiency of dCMP insertion opposite 8-oxo-dG over 4and 5-fold, respectively. However, kinetic analysis of hpol -catalyzed insertion of dCMP opposite 8-oxo-dG in the presence of WRN exo (a.a. 1-333) revealed no change in the specificity constant. Thus, it would appear that the RQC domain is primarily responsible for WRN-mediated enhancement of hpol insertion of dCMP opposite 8-oxo-dG.
We then examined hpol -catalyzed mis-insertion of dAMP opposite 8-oxo-dG. Normally, wild-type hpol catalyzes the mis-insertion of dAMP opposite 8-oxo-dG ∼10to 15-fold more effectively than it does dCMP insertion opposite 8-oxo-dG (20,56). Adding full-length WRN 1-1432 to the reaction mixture only increases the specificity constant for dAMP insertion opposite 8-oxo-dG 1.3-fold (Table 3), much less than the effect observed for dCMP insertion. A similar lack of an effect is observed for the other WRN constructs, the exception being the WRN exo domain. The addition of the WRN 1-333 exo domain results in a dimin-  For the determination of kinetic values, graphs of product formation versus dNTP concentration were plotted using nonlinear regression analysis (one-site hyperbolic fit) in the GraphPad prism program. The standard error of the fit is reported for each value. ished k cat for dAMP insertion opposite 8-oxo-dG, but the specificity constant increases slightly because of a reduced Michaelis constant. Based on the larger changes observed for dCMP insertion opposite 8-oxo-dG, it does not seem as though WRN has an impact on hpol -catalyzed misinsertion of dAMP opposite 8-oxo-dG. The major conclusion from these results is that the addition of WRN improves the fidelity of hpol catalysis opposite 8-oxo-dG since it shifts the balance from an ∼14-fold preference for dAMP insertion opposite 8-oxo-dG to a less than 2-fold preference for error-prone nucleotide selection.

Pre-steady-state kinetic analysis of WRN-mediated stimulation of hpol activity on 8-oxo-dG-containing template DNA
The improved fidelity of hpol opposite 8-oxo-dG observed in the steady-state kinetic assays was primarily driven by WRN-mediated stimulation of the k cat value for dCMP insertion opposite 8-oxo-dG. We next sought to understand whether the altered k cat was due to an increased rate constant for polymerization (k pol ), as we observed previously for hpol on unmodified DNA (55). To investigate this possibility, we performed pre-steady-state kinetic analysis of hpol -catalyzed insertion of either dCMP or dAMP opposite 8-oxo-dG in either the presence or absence of full-length WRN.
Pre-steady-state experiments were performed under enzyme-limiting concentrations with varying concentrations of either dCTP or dATP (Figure 3). The observed rate constant (k obs ) was plotted as a function of dNTP concentration in order to determine k pol and K d,dNTP for both error-prone and error-free insertion across from 8oxo-dG. The addition of WRN stimulates k pol for hpolcatalyzed insertion of dCMP opposite 8-oxo-dG 6-fold but does not greatly alter the affinity of the incoming nucleotide ( Figure 3A and Table 4). The efficiency of the reaction (k pol /K d,dNTP ) is increased 4-fold for dCMP insertion opposite 8-oxo-dG. The addition of WRN does not alter the efficiency of dAMP insertion by hpol ( Figure 3B). The overall efficiency of the mis-insertion reaction is essentially the same whether WRN is present or not (Table 4). From these results, we conclude that WRN stimulates the rate constant k pol value for hpol -catalyzed insertion of dCMP opposite 8-oxo-dG.

The effect of WRN on hpol -catalyzed next-base extension past 8-oxo-dG
Previous studies indicate that hpol extends dA:8-oxo-dG base pairs (bp) about 2-fold more efficiently than it does dC:8-oxo-dG (56). The preference for extending from a mispaired dA:8-oxo-dG was largely due to an increase in the Michaelis constant for extension from dC:8-oxo-dG bp relative to that for dA:8-oxo-dG. We wanted to determine if WRN could modulate extension by hpol in a way that might alter the fidelity of TLS across 8-oxo-dG.
We measured the steady-state kinetic parameters for hpol -catalyzed extension from four different bp: (i) dC:dG, (ii) dA:dG, (iii) dC:8-oxo-dG and (iv) dA:8-oxo-dG. The template base for these p/t-DNA substrates was dC. Accordingly, insertion of dGMP was measured with the four different primer-template substrates in the presence and absence of full-length WRN. As with the other steady-state experiments, we did not observe WRN 1-1432 exonuclease activity under the conditions used to measure single-nucleotide insertion by hpol (Supplementary Figure S4). For unmodified DNA substrates, we find that WRN stimulates hpol -catalyzed extension from dC:dG bp but not from dA:dG mis-pairs substrates (Table 5). We observe a slightly different effect for 8-oxo-dG containing DNA substrates. Interestingly, the inclusion of full-length WRN leads to an increase in the k cat for extension from dC:8-oxo-dG bp, with little or no effect on the K M,dNTP (Table 5). In this way, WRN increases the efficiency for extension from the dC:8oxo-dG bp about 2.4-fold. The k cat for extension from dA:8oxo-dG is decreased in the presence of WRN, while the K M,dNTP is increased, which results in a little over 2-fold reduction in the specificity constant (k cat /K M,dNTP ) for extension from the mis-pair relative to extension by hpol alone ( Table 5). The net effect of WRN on pol fidelity is an approximate 5-fold increase in the accuracy of hpol extension from 8-oxo-dG containing substrates since accurate extension is improved 2.4-fold while extension of dA:8-oxo-dG is decreased 2-fold.

Proofreading activity of the WRN exonuclease domain
Next, we tested whether the isolated WRN exo domain could preferentially degrade 8-oxo-dG containing bp (Figure 4A). In the absence of hpol , the WRN 1-333 exo domain exhibits a clear preference for the degradation of dA-containing mis-pairs regardless of whether 8-oxo-dG is present or unmodified dG. The rate of degradation was ∼0.9 nM min −1 for the two dA-containing mis-pairs, whereas the rate of degradation for dC:dG and dC:8-oxo-dG bp was 0.33 and 0.40 nM min −1 , respectively ( Figure  4B). The rate of WRN-catalyzed dA:8-oxo-dG degradation is ∼2-fold faster than dC:8-oxo-dG.
We also tested whether WRN 1-333 exo activity could degrade primers when competing with hpol activity ( Figure  4C). In the presence of hpol -catalyzed nucleotide insertion, the rate of exonucleolytic degradation of dA:dG bp by WRN 1-333 is 2.7 nM min −1 , while dA:8-oxo-dG bp are degraded at a rate of 2.0 nM min −1 ( Figure 4D). Conversely, WRN 1-333 exo activity degrades dC:dG and dC:8-oxo-dG bp at a rate of 0.07 and 0.30 nM min −1 , respectively. Thus, WRN 1-333 is 39-fold more active on dA:dG mis-pairs than dC:dG bp, and WRN 1-333 degrades dA:8-oxo-dG bp 6.7fold faster than dC:8-oxo-dG bp when hpol extension activity competes directly with the exonuclease. When considering WRN exo activity on 8-oxo-dG-containing bp within the context of polymerization by hpol , it is important to note that, on its own, hpol exhibits a 5-fold preference for extension from dA:8-oxo-dG mis-pairs compared to dC:8oxo-dG bp (Table 5). This preferential extension may be due in part to the fact that dA:8-oxo-dG mis-pairs are geometrically similar to dA:dT bp since 8-oxo-dG is rotated about the glycosyl bond (χ ) into a syn orientation to form a Hoogsteen-type bp with dA (57). Importantly, hpol extension from dC:8-oxo-dG bp is as efficient as extension from dA:8-oxo-dG when WRN is present in the reaction mixture (Table 5). Thus, WRN exo activity on dA:8-oxo-dG substrates could potentially impart additional accuracy by Nucleic Acids Research, 2014, Vol. 42, No. 19 12035  13.2 ± 0.5 1.5 ± 0.4 8.8 Kinetic parameters were obtained as indicated in Table 1.  Table 4. The observed rate constant for dNTP insertion (k obs ) was plotted as a function of dNTP concentration and fit to a quadratic equation. The standard error of the fit is reported for each value.  removing mis-pairs inserted by hpol during error-prone bypass of 8-oxo-dG and allowing extension from error-free dC:8-oxo-dG bp.

DISCUSSION
TLS is an important mechanism for tolerating DNA damage and Y-family pols are central to the bypass of many DNA adducts (5-7). Y-family pol activity is regulated through a variety of mechanisms mostly involving interactions with the sliding clamp, but additional protein-protein interactions that modify Y-family pol TLS properties have been identified (39,44,(58)(59)(60)(61)(62)(63)(64)(65)(66)(67)(68). Interactions with WRN have been shown to stimulate the in vitro activity of Y-family pols and this stimulation occurs on templates that contain DNA adducts (47,55). We were interested in studying the mechanism by which WRN altered the kinetics of polymerization by hpol in order to better understand potential ramifications upon TLS. We find that the rate constant for full-length extension by hpol on unmodified DNA is increased in the presence of WRN and that this stimulation can be mostly attributed to the RQC domain. We observe some stimulation of hpol extension by WRN constructs that contain the exo domain, indicative of a twopart (bipartite) functional interaction between the two proteins which depends on both the RQC and exo domains of WRN. At the level of single-nucleotide insertions, the RQC domain of WRN appears to be most important for enhancing hpol activity on undamaged DNA templates. The bipartite functional interaction between WRN and hpol involves a physical interaction based on pull-down assays. Also, co-localization of WRN and hpol in HeLa cells exposed to H 2 O 2 is consistent with the idea that the physical/functional interaction observed with recombinant enzymes may be carried over into the cellular context.
As noted before, hpol is the most error-prone of the four human Y-family pols at bypass of the ubiquitous 8-oxo-dG lesion (20,(38)(39)41,56,(69)(70)(71). The preferential insertion of dAMP opposite 8-oxo-dG by hpol can be attributed in part to interactions between Met-135 and the damaged template strand which stabilize 8-oxo-dG in the syn conformation, which promotes Hoogsteen-type bonding with an incoming dATP (20). Compared to other Y-family members, such as Dpo4 from S. solfataricus and pol , hpol appears to lack side chains in the little finger domain which facilitate anti-oriented 8-oxo-dG through hydrogen bonding with the O 8 oxygen of 8-oxo-dG (42). Mutating Leu-508 in the little finger domain of hpol to lysine partially restores the electrostatic contact with the lesion as it produces an enzyme that is much less error-prone at the insertion step opposite 8-oxo-dG, only preferring dATP ∼3.2-fold over dCTP (20). By way of contrast, Rev1 uses a protein-template-directed mechanism to act as a deoxycytidyl transferase during 8oxo-dG bypass (71), and human pol iota (hpol ) relies upon a narrowed active-site cleft to prevent base pairing between syn-oriented 8-oxo-dG and dATP (69), which results in dC:8-oxo-dG insertions being preferred 30-fold over dA:8-oxo-dG mis-pairs (70). Strangely enough, hpol inserts dCMP opposite 8-oxo-dG only ∼2-fold more effectively than dGMP (70), but it is still more accurate than hpol at bypass of 8-oxo-dG. This is perhaps due to the similarity noted between syn-oriented 8-oxo-dG and dT, as hpol is known to preferentially insert dGMP opposite template dT (72).
Our kinetic analyses indicate that WRN attenuates the error-prone nature of hpol -catalyzed bypass of 8-oxo-dG through a functional interaction involving the RQC and exo domains of WRN. WRN constructs possessing the RQC domain stimulate accurate insertion of dCMP opposite 8-oxo-dG, primarily by increasing k cat and k pol (Tables 2 and 4). The exo domain of WRN preferentially degrades dA:8-oxo-dG mis-pairs but allows unperturbed extension of dC:8-oxo-dG bp (Figure 4). To calculate the total effect of WRN on error-free bypass by hpol we multiplied 5.2 min −1 M −1 (the specificity constant for dCMP insertion opposite 8-oxo-dG in the presence of fulllength WRN) by 0.14 min −1 M −1 (the specificity constant for extension from dC:8-oxo-dG bp in the presence of WRN), which yields a value of 0.73 min −1 M −1 . This value estimates the efficiency of accurate bypass (i.e. both insertion and extension steps) by hpol in the presence of WRN. For error-prone bypass we multiplied 8.9 min −1 M −1 (the specificity constant for dAMP insertion opposite 8-oxo-dG in the presence of full-length WRN) by 0.13 min −1 M −1 (the specificity constant for extension from dA:8-oxo-dG bp in the presence of WRN), which yields a value of 1.16 min −1 M −1 . This value estimates the efficiency of error-prone bypass (i.e. insertion and extension) by hpol in the presence of WRN. We then divided the value for dAMP insertion/extension (1.16 min −1 M −1 ) by 2, since WRN exhibits a 2-fold preference for degrading dA:8-oxo-dG mis-pairs over dC:8-oxo-dG bp (Figure 4), which decreases the overall efficiency of error-prone bypass to 0.58 min −1 M −1 . Thus, the efficiency of errorfree insertion/extension (0.73 min −1 M −1 ) by hpol in the presence of WRN is actually 1.3-fold higher than the efficiency of error-prone insertion/extension/degradation (0.58 min −1 M −1 ). WRN-mediated stimulation of hpol -catalyzed error-free bypass of 8-oxo-dG is even greater if we consider that degradation of dA:8-oxo-dG bp is 6.7fold more rapid than dC:8-oxo-dG bp when hpol and WRN 1-333 exo activities compete directly ( Figure 4C and D). The cumulative effect of adding WRN to the reaction mixture is a 70-fold change in the kinetics of hpol -catalysis at the insertion and extension steps of bypassing 8-oxo-dG ( Figure 5). While hpol remains relatively error-prone at bypass of 8-oxo-dG even in the presence of WRN, it is possible that additional interacting proteins, such as PCNA and RPA, could modulate the fidelity of lesion bypass further, as was observed for human pols and (39).
The stimulation of hpol by WRN reported here is similar in many respects to previous studies with hpol (47,55). In both cases, stimulation of TLS pol action appears to be driven through functional interactions primarily involving the WRN RQC. Pull-down experiments with recombinant enzymes indicate that both the WRN exo and RQC domains can physically interact with hpol ( Figure 2A and B), again similar to results with hpol (55). More-over, we show that hpol and WRN co-localize in HeLa cells following exposure to oxidative damage ( Figure 2C and D). A previous report from the Loeb group showed that WRN and hpol co-localize following exposure to ultraviolet radiation sub-type C (UV-C) irradiation (47). This colocalization with hpol was found to occur in the absence of either WRN helicase activity (i.e. with the WRN K577M mutant) or the WRN exo domain (47). Only WRN constructs possessing the RQC domain co-localized with hpol in response to UV irradiation (47), consistent with our results showing that the RQC is the most important domain for stimulation of TLS pol activity.
The cellular role for hpol in the replication of 8-oxo-dG is not entirely clear, but it is interesting to consider how the functional interaction between WRN and hpol might influence 8-oxo-dG mutagenesis. A recent study used the supF mutagenesis assay to show that knock-down of hpol by siRNA reduced the frequency of G → T tranversions associated with mutagenic replication of 8-oxo-dG (43), suggesting that hpol copies unrepaired 8-oxo-dG in an errorprone fashion in human cells. Whether WRN modulates hpol bypass of 8-oxo-dG in cells remains to be seen, but WRN-deficient cells are known to accumulate mutations much faster than WRN-proficient cells (50,73). Moreover, WRN has been shown to modulate the efficiency of 8-oxo-dG bypass by hpol during MUTYH-mediated repair of dA:8-oxo-dG mis-pairs through a direct physical interaction between the core polymerase domain of hpol and either the helicase domain (a.a. 500-946) or C-terminal region (a.a. 949-1432) of WRN (51). However, neither the helicase nor the exo catalytic activity of WRN was required for enhancement of hpol activity. The stimulation of hpol by WRN was specific for 8-oxo-dG containing substrates and increased product formation ∼3-fold, similar to what we observe with hpols and . One difference between our results and the study investigating WRN modulation of hpol is that WRN apparently stimulated both dCMP and dAMP insertion opposite 8-oxo-dG and did not appear to modify hpol fidelity one way or another. However, kinetic analysis of nucleotide selection by hpol was not performed and could yet reveal subtle changes in WRN-mediated stimulation of nucleotide insertion rates and/or affinities.
In summary, we have shown that WRN-mediated stimulation of hpol occurs through a physical and functional interaction that appears to involve the WRN exo and RQC domains. The modulation of hpol activity by WRN improves the fidelity of the TLS pol on both undamaged and 8oxo-dG substrates by increasing the rate constant for polymerization (k pol ). Co-localization of these two enzymes occurs in cells treated with H 2 O 2 . The cellular role for TLS pols in replication of 8-oxo-dG lesions leads us to conclude that WRN may act to suppress error-prone DNA synthesis by hpol following exposure to conditions that promote oxidative DNA damage. Experiments to examine the effects of WRN on hpol -mediated bypass of 8-oxo-dG in cells are ongoing.

SUPPLEMENTARY DATA
Supplementary Data are available at NAR Online.