Unfilled gaps by polβ lead to aberrant ligation by LIG1 at the downstream steps of base excision repair pathway

Abstract Base excision repair (BER) involves the tightly coordinated function of DNA polymerase β (polβ) and DNA ligase I (LIG1) at the downstream steps. Our previous studies emphasize that defective substrate-product channeling, from gap filling by polβ to nick sealing by LIG1, can lead to interruptions in repair pathway coordination. Yet, the molecular determinants that dictate accurate BER remains largely unknown. Here, we demonstrate that a lack of gap filling by polβ leads to faulty repair events and the formation of deleterious DNA intermediates. We dissect how ribonucleotide challenge and cancer-associated mutations could adversely impact the ability of polβ to efficiently fill the one nucleotide gap repair intermediate which subsequently results in gap ligation by LIG1, leading to the formation of single-nucleotide deletion products. Moreover, we demonstrate that LIG1 is not capable of discriminating against nick DNA containing a 3′-ribonucleotide, regardless of base-pairing potential or damage. Finally, AP-Endonuclease 1 (APE1) shows distinct substrate specificity for the exonuclease removal of 3′-mismatched bases and ribonucleotides from nick repair intermediate. Overall, our results reveal that unfilled gaps result in impaired coordination between polβ and LIG1, defining a possible type of mutagenic event at the downstream steps where APE1 could provide a proofreading role to maintain BER efficiency.


Introduction
Base excision repair (BER) is a critical process for repairing base lesions and abasic sites, thereby preventing the mutagenic and lethal consequences of DNA damage (1)(2)(3).The BER pathway involves a series of sequential enzymatic steps which requires direct transfer of repair intermediates from one enzyme to the next in the pathway, in a process referred to as substrate-product channeling, to prevent the accumulation of toxic strand break intermediates in cells (3)(4)(5).In short-patch (SP)-BER, a damaged base is first removed by a lesion-specific DNA glycosylase, resulting in an apurinic / apyrimidinic (AP)site in double-stranded DNA, which is then recognized and cleaved by AP-Endonuclease 1 (APE1), generating a singlestrand break with 3 -OH and 5 -deoxyribose phosphate (5 -dRP) ends ( 1 ,2 ) Next, DNA polymerase (pol) β removes the 5 -dRP group via its lyase activity and subsequently catalyzes template-directed gap filling DNA synthesis through its nucleotidyl transferase activity ( 1 ,2 ).BER is completed with a final DNA ligation step by DNA ligase (LIG) 1 or LIG3 α that catalyzes phosphodiester bond formation between 3 -OH and 5 -P ends of the nick ( 1 ,2 ).Studies indicate that LIG1 and LIG3 α can be interchangeable in BER, apparently being selected, at least in part, by the choice between either the SPor long-patch (LP)-BER subpathway ( 3 ).After initial damage recognition and its processing by DNA glycosylase and APE1 activities, respectively, the nick repair product generated by pol β following correct nucleotide incorporation into gap DNA is channeled to LIG1 or LIG3 α for its subsequent nick sealing during the downstream steps of the BER pathway ( 3 ).Our biochemical and structural studies have revealed that, in certain situations, BER responses can lead to mutagenic repair where the gap filling coupled to the DNA ligation steps of the repair pathway can be compromised (6)(7)(8)(9).
Pol β is an error-prone polymerase that incorporates mismatches approximately one out of every 5000 nucleotide insertion events during BER ( 10 ).Ribonucleotide triphosphates (rNTPs) are much more abundant than deoxyribonucleotide triphosphates (dNTPs) in human cells, and rNTP misincorporation by DNA polymerases constitutes a major source of ribonucleotides embedded into genomic DNA, which could make the phosphodiester backbone more labile to DNA damage, influence DNA conformation, increase the mutation rate and hinder replication fork progress (11)(12)(13)(14)(15).However, the impact of ribonucleotides on BER pathway coordination at the downstream steps involving rNTP incorporation by errorprone pol β and subsequent nick sealing by DNA ligase remains undefined.
Moreover, 30% of a variety of human tumors such as lung, gastric, colorectal and prostate cancer have been found to express pol β variants carrying single amino acid substitutions in distinct regions of the protein (16)(17)(18).These pol β cancer-associated variants possess aberrant activity in vitro such as diminished fidelity, stemming from reduced discrimination against incorrect nucleotides, leading to mutator phenotypes (Y265C and E288K), slower rate of dRP lyase (L22P) or polymerase (S229L) activity and a complete lack of gap filling activity (E295K) (17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30).Previous studies have also shown that the expression of these pol β variants in mouse or human cells leads to cancer phenotypes such as an increased mutation frequency, permanent cellular transformation, aberrant DNA repair, accumulation of toxic BER intermediates and ultimately genomic instability, suggesting a pivotal role for pol β-mediated high-fidelity DNA synthesis in carcinogenesis.( 17 , 20 , 22 , 24 , 27 , 28 ).Yet, how the efficiency of substrateproduct channeling at the downstream steps of BER could be adversely affected in case of unfilled gap mediated by pol β cancer-associated variants remains entirely unknown.
In the present study, we comprehensively investigated the impact of canonical, mismatched, damaged and ribonucleotide incorporation by pol β on BER pathway coordination between gap filling by pol β and DNA ligation by LIG1.Our findings revealed that while pol β is inefficient at inserting dNTP and rNTP mismatches, LIG1 seals the one nucleotide gap repair intermediate itself, referred to as gap ligation, resulting in the formation of deletion mutagenesis products.Upon mutation of the steric gate residue of pol β (Y271A), we observed efficient 8oxorGTP incorporation and its subsequent mutagenic ligation, while LIG1 fails following pol β 8oxodGTP insertion, resulting in the formation of abortive ligation product.Moreover, our findings demonstrated that pol β cancer-associated variants Y265C, E288K and E295K leave the gap repair intermediate unfilled even in the presence of correct nucleotides, which leads to gap ligation by LIG1 and the formation of deletion mutagenesis products.
In the current work, we also extensively characterized the substrate specificity of LIG1 and APE1 to elucidate the efficiency of nick sealing versus proofreading at the final steps of the BER pathway for a variety of nick repair intermediates containing canonical, mismatched, damaged or ribonucleotides at the 3 -end.Our results demonstrated that LIG1 cannot discriminate against a 'wrong' sugar and is able to ligate almost all possible 12 ribonucleotide mismatches, while it shows distinct efficiency for the ligation of deoxyribonucleotide mismatches.Finally, we showed that APE1 can remove a 3 -base through its proofreading exonuclease activity from the nick DNA substrates containing ribonucleotides or mismatches depending on the base-pairing architecture.Furthermore, LIG1 can seal the nick repair intermediate with 3 -8oxodG or 3 -8oxorG opposite A from which APE1 is able to remove the damaged base efficiently as well.
Overall, our findings revealed that ribonucleotide discrimination by pol β is the key determinate which prevents ribonucleotides from interfering with BER pathway coordination re-sulting in faulty repair events or deleterious DNA intermediates, and APE1 / LIG1 interplay plays a critical role for ensuring accurate repair at the downstream steps.Additionally, the present study demonstrates that unfilled gaps left by pol β, wild-type or cancer-associated mutants, are erroneously ligated by LIG1, resulting in a deviation from canonical BER pathway coordination and the formation of aberrant repair intermediates with one nucleotide deletions.

Preparation of DNA substrates
Oligodeoxyribonucleotides with and without a 6carboxyfluorescein (FAM) label were obtained from Integrated DNA Technologies.One nucleotide gap DNA substrates with template base A or C and a FAM label at the 5 -end were used in pol β nucleotide insertion assays ( Supplementary Table S1 ).One nucleotide gap DNA substrates with template base A or C and FAM labels at both 3and 5 -ends were used in pol β nucleotide insertion coupled to DNA ligation assays ( Supplementary Table S2 ).Nick DNA substrates with preinserted 3 -deoxyribo-or ribo-nucleotide mismatches and FAM label at the 3 -end were used in DNA ligation assays ( Supplementary Tables S3 and S4 ).Nick DNA substrates with preinserted 3 -deoxyribo-or ribo-nucleotide mismatches and FAM label at the 5 -end were used in APE1 exonuclease assays ( Supplementary Tables S5 and S6 ).

Protein purifications
Human wild-type his-tag LIG1 full-length (1-919) was overexpressed in Rosetta (DE3) pLysS Esc heric hia coli ( E. coli ) cells and grown in Terrific Broth (TB) media with kanamycin (50 μg ml −1 ) and chloramphenicol (34 μg ml −1 ) at 37 • C. Once the OD 600 reached 1.0, the cells were induced with 0.5 mM isopropyl β-D-thiogalactoside (IPTG), and overexpression continued overnight at 20 • C.After centrifugation, the cells were lysed in lysis buffer containing 50 mM Tris-HCl (pH 7.0), 500 mM NaCl, 20 mM imidazole, 2 mM βmercaptoethanol, 5% glycerol and 1 mM Phenylmethylsulfonyl fluoride (PMSF) by sonication at 4 • C. The cell lysate was pelleted at 31 000 × g for 90 min at 4 • C. The cell lysis solution was filter clarified and then loaded onto a His-Trap HP column that was previously equilibrated with binding buffer containing 50 mM Tris-HCl (pH 7.0), 500 mM NaCl, 20 mM imidazole, 2 mM β-mercaptoethanol and 5% glycerol.The column was washed with binding buffer and then eluted with an imidazole gradient (0-500 mM) at 4 • C. The collected fractions were then subsequently loaded onto a HiTrap Heparin column that was equilibrated with binding buffer containing 20 mM Tris-HCl (pH 7.0), 50 mM NaCl, 2 mM β-mercaptoethanol and 5% glycerol, and the protein was eluted with a linear gradient of NaCl up to 1 M. LIG1 protein was further purified by a Superdex 200 gel filtration column in buffer containing 50 mM Tris-HCl (pH 7.0), 200 mM NaCl, 1 mM DTT and 5% glycerol.Human DNA ligase (LIG) 3 α (pET-24b) full-length (1-922) protein was overexpressed in BL21(DE3) E. coli cells in LB media at 37 • C for 8 h and induced with 0.5 mM IPTG.The cells were harvested, lysed at 4 • C, and then clarified as described above.The supernatant was loaded onto HisTrap HP column and purified with an increasing imidazole gradient (0-300 mM) elution at 4 • C. The collected fractions were then further purified by Superdex 200 Increase 10 / 300 chromatography in the buffer containing 50 mM Tris-HCl, pH 7.0, 500 mM NaCl, glycerol 5%, 1 mM DTT.
Human wild-type full-length (1-335) pol β with a GST-tag (pGEX-6p-1) was overexpressed in BL21(DE3) pLysS E. coli cells in TB media at 37 • C. When the OD 600 reached 1.0, the cells were induced with 0.5 mM IPTG, and the overexpression continued overnight at 20 • C.After cell lysis at 4 • C by sonication in lysis buffer containing 1 × PBS (pH 7.3), 200 mM NaCl, 1 mM Dithiothreitol (DTT) and 1 mM PMSF, the cell lysate was pelleted at 31 000 × g for 90 min and then filter clarified.The clarified supernatant was loaded onto a GSTrap HP column, washed with lysis buffer, and eluted with elution buffer containing 50 mM Tris-HCl (pH 8.0), and 10 mM reduced glutathione.To cleave the GST-tag, the recombinant pol β protein was incubated with PreScission Protease for 16 h at 4 • C in buffer containing 1 × PBS (pH 7.3), 200 mM NaCl and 1 mM DTT. Pol β protein was then subsequently passed through a preequilibrated GSTrap HP column to remove the cleaved tag, and the protein without the GST-tag was then further purified by loading onto a Heparin HP column that was equilibrated with binding buffer containing 20 mM Tris-HCl (pH 7.0), 50 mM NaCl, 2 mM β-mercaptoethanol and 5% glycerol.The protein was then eluted using a linear gradient up to 1 M NaCl.Finally, pol β was purified by a Superdex 200 gel filtration column in buffer containing 50 mM Tris-HCl (pH 7.0), 200 mM NaCl, 1 mM DTT and 5% glycerol.Plasmid DNA coding Y271A (pGEX-6p-1), as well as Y265C, E288K and E295K mutations (pGEX-4T-3) were generated by site directed mutagenesis and the coding sequences of the mutants were confirmed by DNA sequencing prior to purification.The steric gate mutant protein, pol β Y271A, was overexpressed and purified as described above.Pol β cancerassociated mutants Y265C, E288K and E295K were purified as described above except the GST-tag cleavage step.
Human his-tag wild-type full-length (1-335) APE1 was overexpressed in BL21(DE3) E. coli cells in Lysogeny broth (LB) media at 37 • C. When the OD 600 reached 1.0, the cells were induced with 0.5 mM IPTG, and the overexpression continued overnight at 28 • C.After centrifugation, the cells were lysed in lysis buffer containing 50 mM Tris-HCl (pH 7.0), 500 mM NaCl, 20 mM imidazole, 2 mM β-mercaptoethanol, 5% glycerol and 1 mM PMSF by sonication at 4 • C. The lysate was pelleted at 31 000 × g for 90 min at 4 • C and then the supernatant was filter clarified.The clarified supernatant was loaded onto a HisTrap HP column and eluted with an increasing imidazole gradient (0-300 mM) at 4 • C. The collected fractions were then subsequently loaded onto a HiTrap Heparin column and eluted with a linear gradient of NaCl up to 1 M.The recombinant APE1 protein was then further purified by a Superdex 200 gel filtration column in buffer containing 20 mM Tris-HCl (pH 7.0), 200 mM NaCl and 1 mM DTT.
All proteins purified in this study were stored in aliquots at -80 • C. Protein quality was evaluated on a 10% SDS-PAGE gel, and protein concentrations were measured using the absorbance at 280 nm.

Nucleotide insertion assays
Pol β nucleotide incorporation assays were performed to examine nucleotide insertion efficiency of pol β into gap DNA containing template base A or C. The reaction mixture contains 50 mM Tris-HCl (pH 7.5), 100 mM KCl, 10 mM MgCl 2 , 1 mM ATP, 1 mM DTT, 100 μg ml −1 BSA, 10% glycerol, DNA substrate (500 nM), and dNTP , rNTP , 8-oxodGTP or 8-oxorGTP (10 μM) in a final volume of 10 μl.The reaction was initiated by the addition of pol β (wild-type, Y271A, Y265C, E288K or E295K) at a final concentration of 100 nM, and incubated at 37 • C for the times as indicated in the figure legends.The reaction products were then quenched by mixing with an equal amount of gel loading buffer (95% formamide, 20 mM EDTA, 0.02% bromophenol blue and 0.02% xylene cyanol) and then separated by electrophoresis on an 18% polyacrylamide gel.The gels were scanned with a Typhoon PhosphorImager (Amersham Typhoon RGB), and the data were analyzed using ImageQuant software as described previously.

Nucleotide insertion coupled to DNA ligation assays
Coupled assays were used to measure pol β nucleotide insertion and DNA ligation in the same reaction simultaneously using one nucleotide gap DNA substrates containing template base A or C. The reaction mixture contains 50 mM Tris-HCl (pH 7.5), 100 mM KCl, 10 mM MgCl 2 , 1 mM ATP, 1 mM DTT, 100 μg ml −1 BSA, 10% glycerol, DNA substrate (500 nM) and dNTP , rNTP , 8-oxodGTP or 8-oxorGTP (10 μM) in a final volume of 10 μl.The reaction was initiated by the addition of pre-incubated enzyme mixture of LIG1 and pol β (wild-type, Y271A, Y265C, E288K or E295K) at a final concentration of 100 nM each, and incubated at 37 • C for the times as indicated in the figure legends.The reaction samples were then quenched by mixing with an equal volume of the loading dye.The products were separated, and the data were analyzed as described above.The coupled assays were performed similarly in the presence of pol β, rNTP and Lig3 α.

DNA ligation assays
DNA ligation assays were performed to evaluate the substrate specificity of LIG1 using nick DNA substrates containing 3preinserted deoxyribo-or ribonucleotide mismatches and 3 -8oxodG or 3 -8oxorG.The ligation reaction mixture contains 50 mM Tris-HCl (pH 7.5), 100 mM KCl, 10 mM MgCl 2 , 1 mM ATP, 1 mM DTT, 100 μgml −1 BSA, 10% glycerol and DNA substrate (500 nM) in a final volume of 10 μl.The reaction was initiated by the addition of LIG1 (100 nM), incubated at 37 • C and quenched at the time points as indicated in the figure legends.The products were separated, and the data were analyzed as described above.

APE1 exonuclease assays
Exonuclease assays were performed to evaluate the substrate specificity of APE1 using nick DNA substrates containing 3preinserted deoxyribo-or ribonucleotide mismatches and 3 -8oxodG or 3 -8oxorG.The reaction mixture contains 50 mM Tris-HCl (pH 7.5), 100 mM KCl, 10 mM MgCl 2 , 1 mM ATP, 1 mM DTT, 100 μg ml −1 BSA, 10% glycerol and DNA substrate (500 nM) in a final volume of 10 μl.The reaction was initiated by the addition of APE1 (1 μM) and incubated at 37 • C for the time points as indicated in the figure legends.The products were separated, and the data were analyzed as described above.

DNA-binding measurements by BioLayer Interferometry assay
DNA-binding kinetics of pol β, LIG1 and LIG3 α was measured by BioLayer Interferometry (BLI) assays in real time using the Octet QKe (Fortebio) using one nucleotide gap DNA substrate with 3 -biotin label.Streptavidin (SA) biosensors were used to attach the biotin labeled DNA.BLI experiments were performed at 20 • C in 96-well microplates with agitation set to 1000 rpm.The SA biosensors were hydrated at 20ºC for 20 min in the buffer containing 50 mM Tris-HCl pH 7.5, 100 mM KCl, and 1 mM DTT.The sensors were then immersed in DNA (40 nM) in the buffer for 300 s.After recording an initial baseline for 60 s, the sensors with DNA were exposed to the concentration range of the proteins as indicated in the figure legends.DNA binding was performed for 240 s association, and then for 240 s dissociation.In all measurements, the affinity constants ( K D ), the association ( k on ) and dissociation ( k off ) rates were calculated using the ForteBio Data Analysis software with 1:1 binding model.The association rate = k on [ligand][analyte] and the dissociation rate = k off [ligand-analyte].At equilibrium, forward and reverse rates are equal.All images were drawn using GraphPad Prism 7.

Impact of pol β ribomismatch incorporation on DNA ligation
We first investigated the impact of pol β ribonucleotide insertion on BER pathway coordination at the downstream steps by performing coupled assays containing pol β wild-type, LIG1, rNTP and one nucleotide gap DNA substrate (Figure 1 A).For ribonucleotide mismatches rATP, rGTP and rCTP, we mainly observed ligation products of the one nucleotide gap DNA substrate, referred to as gap ligation hereafter (Figure 1 B,C).These gap ligation products were more abundant in reactions including pol β and gap DNA substrate with template base C (Figure 1 C) over template base A (Figure 1 B).Gap versus nick ligation was revealed by the difference in the size of the products between the ligation of gap DNA itself versus nick sealing after pol β correct nucleotide insertions dTTP:A and dGTP:C by LIG1 in the control reactions (Figure 1 B and C, lane 1).Consistent with previous reports (31)(32)(33), we also showed inefficient incorporation of rNTPs by pol β ( Supplementary Figure S1 ).
Ribonucleotides are susceptible to chemical modification, through a variety of mechanisms, which causes damage to the structure and function of the ribonucleotide ( 34 ,35 ).A common form of ribonucleotide damage is oxidative damage, caused by endogenous or exogenous reactive oxygen species, which leads to oxidized ribonucleotides such as 8-oxoguanosine-5 -triphosphate (8oxorGTP) ( 36 ,37 ).As demonstrated in the X-ray crystal structure of pol β, the active site can accommodate 8oxorGTP opposite templating C and mutagenic A in anti -and syn -conformation, respectively ( 31 ).In the present study, we also investigated the ligation efficiency following pol β oxidized ribonucleotide incorporation in coupled reactions including pol β wild-type, 8oxorGTP and LIG1.Our results demonstrated that LIG1 attempts to ligate the one nucleotide gap DNA substrate, creating gap ligation products (Figure 1 D).We additionally observed more abundant gap ligation products in the presence of the gap DNA substrate with template C over template A, similar to ribonucleotide mismatch reactions (Figure 1 D, lanes 3-6 versus 8-11).In the insertion assays containing pol β alone, our results are consistent with our coupled assays which show that pol β wild-type is inefficient at inserting 8oxorGTP ( Supplementary Figure S1 ).We performed the same experiments with LIG3 α and obtained similar results showing the gap ligation products in the presence of pol β and rNTP and the preference for template base C over A ( Supplementary Figure S2 ).

Impact of pol β steric-gate mutant on the mutagenic nick sealing of ribonucleotide incorporation products
The discrimination against ribonucleotide by repair and replication DNA polymerases is essential for the maintenance of genome integrity (11)(12)(13)(14).Previous structure / function studies have reported the steric gate role of the Tyr(Y)271 backbone carbonyl of pol β, which clashes with the ribose 2 -OH of the incoming rNTP and discourages pol β from incorporating ribonucleotides (31)(32)(33).Furthermore, it has been shown that the Y271A mutation leads to ∼12-fold increase in ribonucleotide insertion by pol β as demonstrated by a > 10-fold loss in sugar discrimination and an increase in the insertion of 8oxorGTP:A, having a similar efficiency to a non-damaged matched ribonucleotide (31)(32)(33) In order to compare the impact of increasing the ribonucleotide tolerance of pol β on the nick sealing efficiency at the final steps of the BER pathway, we next investigated the ligation of rNTP and 8oxorGTP incorporations by pol β Y271A in coupled assays as described above (Figure 2 A).Our results showed the formation of gap ligation products (Figure 2 B-C), which was similar to the ligation of gap DNA itself we observed with pol β wild-type (Figure 1 ).We obtained slightly more nick ligation products in the coupled reaction including canonical rNTP:templating base pair rGTP:C (Figure 2 C, lanes 8-11).Similarly, we observed both gap and nick ligation products for 8oxorGTP:A (Figure 2 D, lanes 2-6).Even with increased ribonucleotide tolerance, there was no nick ligation products for 8oxorGTP:C (Figure 2 D, lanes 8-11).The insertion results with pol β Y271A alone demonstrated that upon increased ribonucleotide tolerance, pol β is capable of inserting rGTP opposite C as well as 8oxorGTP opposite template A more efficiently ( Supplementary Figure S3 ).Furthermore, we showed similar nick sealing efficiency by LIG1 after correct dGTP:C insertions by pol β wild-type versus the Y271A mutant in control coupled assays, demonstrating that the Y271A mutation does not effect the canonical gap filling activity of pol β and LIG1 can seal resulting nick repair product efficiently ( Supplementary Figure S4 ).
Overall, our findings demonstrate the formation of single deletion mutagenesis intermediates that could be formed due to unfilled gaps by pol β in the presence of rNTPs or 8ox-orGTP when LIG1 attempts ligating the gap repair intermediate at the downstream steps of the BER pathway.These findings also highlight the importance of ribonucleotide discrimination by pol β for accurate BER pathway coordination.

Comparison of ligation efficiency after pol β mismatch incorporation
To compare the impact of deoxyribo-versus ribo-nucleotide mismatch insertions and subsequent DNA ligation at the  R epresentativ e gel images of three independent repeats.downstream steps of the BER pathway, we also analyzed the nick sealing efficiency by LIG1 after pol β dNTP mismatch insertions in coupled assays ( Supplementary Figure S5 A).
As we previously reported ( 7 ), our results demonstrated that LIG1 cannot ligate the nick repair products of pol β mismatch insertions opposite template base A or C ( Supplementary Figure S5 B and C ).Similar to the coupled assays including rNTP mismatches, in both cases, the prod-ucts of pol β dNTP mismatch insertion coupled to ligation were mainly the ligation products of the one nucleotide gap DNA itself, as revealed by the difference with the nick ligation products after pol β correct dTTP:A or dGTP:C insertion.We previously demonstrated that oxidized nucleotide incorporation by pol β confounds DNA ligase resulting in the accumulation of cytotoxic ligation failure products ( 6 ).Consistent with this report, our results showed a failure in the ligation along with mutagenic nick sealing after pol β 8-oxodGTP insertion opposite template base A ( Supplementary Figure S5 D).Furthermore, pol β insertion assays showed an inefficiency of inserting dNTP mismatches opposite A or C, consistent with what we observed in our coupled assays ( Supplementary Figure S6 ).
Inef ficient inser tion of cor rect nucleotides by pol β cancer-associated mutants leads to gap ligation of unfilled gaps by LIG1 It has been extensively reported that germline and tumorassociated variants of pol β catalyze aberrant BER that leads to genomic instability (16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30).One of these pol β variants, E295K, which has been identified in both gastric and colon carcinoma, appears to lack DNA polymerase activity while showing the same level of dRP lyase function and gap DNA binding affinity as the wild-type enzyme ( 19 , 21 , 22 ).The pol β E295K variant interferes with BER and cannot fill singlenucleotide gaps in an efficient manner, leading to the accumulation of BER intermediates, formation of double-strand breaks, induced cellular transformation and an increase in sister chromatid exchanges ( 19 , 21 , 22 ).Furthermore, pol β cancer-associated variants that exhibit mutator phenotypes, Y265C and E288K, have been previously shown to affect the enzyme's ability to discriminate between correct and incorrect dNTPs, resulting in filling gaps in an error-prone manner with increased mutagenesis in vitro and also inducing cellular transformation when expressed in mouse cells, leading to genome instability ( 18 ,23-25 ).In light of our observations above demonstrating gap ligation in the case of inefficient incorporation of dNTP or rNTP mismatches by pol β wild-type, in the present study, we examined the impact of correct nucleotide insertions by pol β cancer-associated variants Y265C, E288K and E295K on nick sealing by LIG1 in coupled assays.Under our reaction conditions, pol β dGTP:C nucleotide insertion assays showed moderate gap filling for Y265C and E288K, and no detectable gap filling activity for E295K (Figure 3 A,B).
In the coupled assay including pol β and LIG1 (Figure 3 C), we demonstrated that the E295K mutant is unable to insert a correct nucleotide, and that LIG1 seals the one nucleotide gap DNA itself (Figure 3 D, lanes [13][14][15][16].Interestingly, we observed both gap ligation and nick ligation products for pol β Y265C and E288K variants (Figure 3 D, lanes 3-6 and 8-11, respectively).More gap ligation product was formed in the coupled reaction including E288K, which could be due to less efficient dGTP:C insertion by this pol β variant (Figure 3 B, lines [8][9][10][11].This is consistent with our coupled results containing pol β wild-type with rNTP or dNTP mismatch, which show no insertion and gap ligation products only (Figures 1 -2 ).Taken together, these results could mimic cellular conditions in which pol β leaves gaps unfilled during the DNA synthesis step of the BER pathway, such as in gastric carcinoma cancer patients carrying the somatic E295K mutation in the POLB gene, where canonical repair pathway coordination could be confounded at the downstream steps.Additionally, in cellular contexts where the gap filling activity of pol β is slower, such as in the colorectal cancer patients with E288K mutation or as shown in Lupus disease with pol β Y265C somatic mutation, LIG1 can attempt to ligate the gap leading to the formation of aberrant repair intermediate.

Gap ligation of one nucleotide gap DNA by LIG1
To understand whether gap ligation of one nucleotide gap DNA (Figures 1 -3 ) is dependent on pol β or free nucleotide (rNTP or dNTP), we performed ligation assays containing only LIG1 and the one nucleotide gap DNA substrate ( Supplementary Figure S7 A).Our results demonstrated that LIG1 exhibits different efficiency to seal the gap depending on the identity of the template base ( Supplementary Figure S7 B).Interestingly, we observed much more efficient gap ligation when the template base was C compared to A, T or G ( Supplementary Figure S7 C).This suggests that gap ligation by LIG1 demonstrates substrate specificity.We observed similar efficiency of gap ligation in the ligation versus coupled reactions with the pol β E295K variant that is deficient in gap filling activity (Figure 3 ) as well as coupled reactions with pol β wild-type in the presence of either rNTPs or dNTP mismatches (Figure 1 ).Furthermore, we demonstrated that LIG3 α can also seal the gap itself as efficient as LIG1 ( Supplementary Figure S7 D).Taken together, this suggests that gap ligation is pol β and free nucleotide independent and both BER ligases are capable of this aberrant ligation.Lastly, we measured the real-time DNA-binding kinetics of pol β, LIG1 and LIG3 α for one nucleotide gap DNA using BLI assay.Our results demonstrated very similar binding affinity with the equilibrium binding constants ( K D ) in the range of 3-6 nM for all BER enzymes finalizing the repair pathway at the downstream steps ( Supplementary Figure S8 ).

Ligation of the nick repair intermediates with 3 -ribonucleotides by LIG1
In the present study, in addition to the coupled assays including pol β, rNTP mismatch, one nucleotide gap DNA and LIG1, we comprehensively investigated the ligation efficiency of LIG1 in the presence of a single ribonucleotide at the 3end of nick DNA for all possible 12 ribo-mismatches, i.e. 3 -preinserted rA, rG or rC opposite templates A, T, G or C ( Supplementary Figure S9 A).These nick DNA substrates mimic the ribonucleotide insertion products of DNA polymerases before LIG1 seals the resulting nick intermediate during Okazaki fragment maturation of DNA replication or at the last ligation step of almost all DNA repair pathways.
The ligation of 3 -preinserted ribonucleotide mismatches by LIG1 yielded efficient nick sealing for almost all 12 ribo-mismatches (Figure 4 We observed a robust accumulation of ligation products for the Watson-Crick base paired nick DNA substrates containing 3 -rA:T (Figure 4 B and Supplementary Figure S9 C   (Figure 4 A and Supplementary Figure S9 B, lanes 10-16).Overall, our results revealed a time-dependent increase in the amount of nick sealing products with > 60% ligation for every 3 -ribo-mismatch except 3 -rG:A (Figure 4 ).
In control experiments, we confirmed the ligation of the repair intermediates with 3 -preinserted Watson-Crick base pairs containing 3 -dA:T, 3 -dT:A, 3 -dC:G and 3 -dG:C by LIG1 ( Supplementary Figure S10 ).When compared to these control reactions, the results of ribo-mismatch ligations reveal a lack of efficient sugar discrimination by LIG1 for the repair intermediates containing a single ribonucleotide at the 3 -end.

Ligation of the nick repair intermediates with 3 -mismatches by LIG1
To compare the substrate specificity of LIG1 for nick DNA containing a single deoxyribo-versus ribonucleotide mismatch at the 3 -end, we also performed ligation assays in the presence of all possible 12 non-canonical mismatches, i.e .3 -preinserted dA, dT, dG or dC opposite templates A, T, G or C.

Ligation of the nick repair intermediates with 3 -8o x odG versu s 3 -8o x orG
We next investigated the ligation efficiency of nick DNA substrates containing 3 -preinserted 8-oxodG versus 8-oxorG ( Supplementary Figure S12 A).Consistent with our previous reports ( 6 , 7 , 9 ), we observed mutagenic nick sealing of 3 -8oxodG:A, while the ligation of nick DNA substrate containing 3 -8oxodG:C was less efficient resulting in the formation of DNA-AMP intermediate products ( Supplementary Figure S12 B) and ∼4to 8-fold difference in the amount of ligation products (Figure 4 E).On the other hand, our results demonstrated mutagenic end joining of oxidized ribonucleotide-containing nick DNA substrates 3 -8oxorG:A and 3 -8oxorG:C ( Supplementary Figure S12 C).We observed a faster accumulation of ligation products for both 3 -8oxorG:A and 3 -8oxorG:C (Figure 4 F).Overall, LIG1 does not effectively discriminate against 3 -8oxorG just as it fails to discriminate against all possible 12 ribonucleotide mismatches.

Removal of 3 -mismatches from the nick repair intermediates by APE1
In addition to its endonuclease function, using the same rigid active site, APE1 has been shown to exhibit 3 -5 -exonuclease activity to remove mismatches and various forms of oxidative damage incorporated by pol β during BER ( 8 , 38 , 39 ).Yet, it is unclear whether APE1 could serve as a proofreader of ribonucleotide-containing repair intermediates that could be formed due to aberrant BER pathway coordination at the final steps.In the present study, we comprehensively investigated the substrate specificity of APE1 exonuclease activity for nick DNA substrates containing all 12 possible mismatches, ribonucleotides, and oxidized bases at the 3 -end (Figures 5 -6 ).
For the nick DNA substrates containing ribonucleotide mismatches, our results demonstrated that APE1 shows distinct efficiency depending on the hydrogen bonding characteristics of the terminal base pair and the identity of the 3 -ribonucleotide (Figure 5 and Supplementary Figure S13 ).We observed poor efficiency in removing ribonucleotides that were Watson-Crick base paired to their complementary DNA base: 3 -rA:T, 3 -rC:G and 3 -rG:C (Figure 5 B,D).The highest amount of ribonucleotide removal products were obtained for the nick DNA substrates containing 3 -rA:A, 3 -rA:G, 3 -rA:C, 3 -rC:A, 3 -rC:T and 3 -rC:C mismatches (Figure 5 A,D).Interestingly, APE1 shows poor efficiency for the removal of 3 -rG opposite template A, T and G.
We then extended the analysis of APE1 exonuclease activity for the repair intermediates containing all possible 3 -preinserted 12 non-canonical mismatches (Figure 6 and Supplementary Figure S14 ).Our results demonstrated that APE1 exhibits no significant differences for the purine:pyrimidine, pyrimidine:purine or pyrimidine:pyrimidine mispairs.There was a variability with purine:purine mismatches such as 3 -dA:A, which shows the most efficient mismatch removal out of all tested nick DNA substrates (Figure 6 A).Furthermore, we observed the lowest mismatch excision for purine:purine mispair substrates of differing purines 3 -dG:A and 3 -dA:G (Figure 6 A,C).As expected, APE1 was relatively less efficient for the removal of 3 -bases from the nick DNA substrates containing Watson-Crick canonical ends ( Supplementary Figure S15 ).Lastly, we compared the efficiency of 3 -8oxodG versus 3 -8oxorG removal by APE1 (Figures 5 E and 6 E).Our results showed that APE1 can remove 3 -8oxodG or 3 -8oxorG damaged bases from the nick DNA substrates containing template base A more efficiently than template C ( Supplementary Figure S16 ).
Our overall results revealed that APE1 shows distinct efficiencies for the removal of mismatched bases depending on the 3 -end:template base pair architecture of the nick repair intermediate.It can also proofread nicks containing a damaged base depending on the strength of base paring interaction with the template.However, for almost almost all nick DNA substrates with a single ribonucleotide at 3 -end, excluding those with 3 -rG, our results demonstrate that APE1 is capable of efficiently proofreading ribonucleotide mismatches.

Discussion
BER involves the coordinated channeling of repair intermediates from one enzyme to the next in the repair pathway, which prevents the accumulation of potentially toxic strandbreak intermediates in cells.While it is generally assumed that DNA repair operates to preserve genome integrity, this mechanism is not always precise.Evidence, particularly from our studies, is emerging that BER can contribute to genome instability if normal coordination breaks down (6)(7)(8)(9).Inaccurate BER, stemming from deviations in the proper repair pathway coordination from pol β gap filling to DNA ligase nick sealing at the downstream steps, may result in an accumulation of aberrant repair intermediates which could lead to cellular toxicity and genome instability.The one nucleotide gap formed after APE1-mediated strand incision, which requires 5 -end processing, gap filling synthesis and subsequent nick sealing, could be a particularly deleterious intermediate that interferes with the completion of accurate repair.Failures in the DNA synthesis step by pol β, could lead to persistent one nucleotide gaps if left unfilled, which is expected to be a deleterious repair intermediate ( 17 ,40 ).In the present study, we investigated how LIG1 responds at the last nick sealing step if pol β is unable to fill the one nucleotide gap during the prior DNA synthesis step of the BER pathway.Our results demonstrate that aberrant gap filling by pol β results in ligation of the one nucleotide gap itself, which is especially mutagenic as it leads to a single-nucleotide deletion product.Our first observation was that unfilled gaps by pol β are ligated by LIG1 in the presence of ribonucleotides, which exist several orders of magnitude higher than deoxyribonucleotides within cells, especially within terminally differentiated cells, where the discrepancy can be even higher (11)(12)(13)(14).Pol β is an error prone polymerase belonging to the X-family of DNA polymerases which utilize a backbone carbonyl to clash with the 2 -OH of rNTPs ( 31-33 ,41-43 ).Previous work has highlighted that this strategy results in a polymerase with a steric 'fence' rather than the steric 'gates' of A-and B-family of polymerases ( 32 ,33 ) Due to the lower sugar discrimination of pol β, as compared to replicative polymerases, as well as the abundance of rNTPs, it has previously been estimated that pol β inserts a rNTP every 81 insertion events under physiological conditions ( 32 ).However, under our assay conditions, we did not detect rNTP incorporation nor ligation of ribonucleotide incorporation products.Upon mutation of the steric gate residue, we observed ribonucleotide incorporation products and nick ligation products for rGTP:C and 8oxorGTP:A insertion and coupled reactions, respectively, indicating that the Y271A mutation allows for insertion of Watson-Crick or Hoogsteen base pairs.We also observed defective gap filling by the pol β cancerassociated variants Y265C, E288K and E295K.We found that reduction or ablation of gap filling by these pol β mutants results in the accumulation of gap ligation products by LIG1.These results indicate that other pol β cancer-associated variants with known aberrant BER activity, including R152C, S229L, G231D and I260M, may also result in defective gap filling and accumulation of single-nucleotide deletion mutagenesis products ( 17 , 30 , 44-50 ).We also note that since more gap ligation products were observed for gap DNA containing template base C over all other template bases, it is possible that cancer patients with these mutations might preferentially accumulate dG deletions from the genome in response to defective BER stemming from unfilled gaps by pol β.Future structural studies of LIG1 in complex with gap DNA will be needed to explain the template C preference of LIG1.We hypothesize that in order to bring the 3 -OH and 5 -P of the gap close enough to react, LIG1 could either be looping out the templating base or locally melting the upstream strand.It is known that LIG1 enforces an underwound conformation upstream of the nick, so perhaps this flexibility allows close positioning of the reactive groups ( 51 ).If a specific interaction is made between the looped out templating nucleotide and the LIG1 active site, this mechanism could potentially explain the apparent substrate specificity of gap ligation.
Additionally, we demonstrated that gap ligation of the one nucleotide gap is pol β and free nucleotide independent so it is possible that in other scenarios where one nucleotide gaps are found, if the gap is not filled efficiently by repair or replication polymerase, LIG1 may attempt to ligate it.Furthermore, gap ligation is not specific to LIG1, as other non-human ligases, such as T4 DNA ligase, have been previously shown to ligate one nucleotide gaps as well ( 52 ).
In the case of ribonucleotide incorporation by repair or replicative DNA polymerases, the resulting nick product can serve as a substrate for LIG1 to seal at the last ligation step.In the present study, we observed that LIG1 is capable of efficient ligation of almost all ribonucleotide-containing mismatches at the 3 -end of the nick, which is in stark contrast to the efficiency of deoxyribonucleotide-containing mismatches (Figure 7 A,B).This demonstrates a complete lack of sugar discrimination by LIG1 and highlights the importance of pol β for excluding ribonucleotides during BER, as any incorporated ribonucleotides would be readily ligated by LIG1.Mechanistically, it is known that ligases enforce a 2 -endo to 3endo sugar pucker conformational change in the 3 -nucleotide which is critical for catalysis; therefore, it is possible that the reason 3 -ribonucleotides are efficiently ligated is because ribonucleotides naturally prefer the 3 -endo conformation, which allows close-to-optimal positioning of reactive groups, even in the absence of proper base pairing between the 3nick terminus ( 8 , 51 , 53 ).However, future structure / function studies of LIG1 in complex with nick DNA containing 3ribonucleotides will be critical to confirm this hypothesis.
APE1 has previously been shown to serve as a compensatory proofreading enzyme for pol β-mediated errors introduced during BER ( 8 , 38 , 39 ).Our results revealed for the first time that APE1 is capable of the removal of a single 3 -ribonucleotide from the nick repair intermediates, particularly 3 -rA and 3 -rC while it was less efficient on 3 -rG containing nicks.This suggests that inserted 3 -rG might escape proofreading by APE1 and be incorporated into the genome more frequently, especially due to the highly efficient ligation of 3 -rG:C and 3 -rG:T.Furthermore, we observed that APE1 was capable of removing 3 -8oxorG with moderate efficiency opposite template A and C, suggesting that APE1 serves as a potential line of defense against the incorporation of this highly mutagenic lesion into the genome.Additionally, APE1 shows different efficiency for the removal of 3 -mismatched base depending on the architecture of the 3 -end; however, all 12 mismatches were capable of removal by APE1, demonstrating the broad range of suitable substrates for APE1 exonuclease activity (Figure 7 C,D).From these results, as well as our previous study showing physical and function interaction of APE1 / LIG1, we suggest that APE1 plays an unappreciated role as a vital proofreader of pol β to prevent mutagenic repair, further supporting the notion that the multi-protein BER complex, including APE1, pol β and LIG1, is required for faithful BER ( 8 ).Without the exonuclease activity of APE1, errors introduced by pol β would be more likely to be incorporated into the genome, leading to greater BER-induced genome instability.
In light of the insights gained from this study, we propose the following working model for how unfilled gaps left by pol β could affect substrate-product channeling at the down-stream steps of BER (Figure 7 E).As the main BER DNA polymerase, pol β discriminates against rNTPs and does not insert them often, but when pol β fails to insert any nucleotide, whether due to ribonucleotide challenge or disease-associated mutation, LIG1 may attempt to ligate the one nucleotide gap which results in a single-nucleotide deletion.This singlenucleotide deletion may be even more deleterious then an embedded ribonucleotide, since at least the ribonucleotide still retains base coding potential.Therefore, pol β has a limited amount of time to incorporate nucleotides from the nucleotide pool before LIG1 ligates the one nucleotide gap.When pol β inserts a rNTP, it is swiftly channeled to LIG1 which ligates the nick, finalizing incorporation of the inserted ribonucleotide into genomic DNA.Therefore, the ribonucleotide discrimination of pol β is the only line of defense for preventing BERmediated ribonucleotide incorporation.Instances where pol β has greater ribonucleotide tolerance, or when the gap filling activity of pol β is substituted by a polymerase with greater ribonucleotide tolerance, will likely result in increased BERmediated ribonucleotide incorporation.However, as APE1 can efficiently remove the majority of 3 -ribonucleotides from a nick DNA substrate, it is possible that in the multiprotein BER complex, channeling between pol β ribonucleotide insertion and LIG1 ligation is interrupted by the proofreading activity of APE1, thereby preventing mutagenic incorporation and subsequent ligation at the downstream steps.Additionally, in disease states where the gap filling activity of pol β is compromised, any other X-family repair DNA polymerase, such as pol λ or pol μ, may take over to prevent the mutagenic consequences of unfilled gaps ( 54 ,55 ).Future experiments will be directed at understanding how the multi-protein BER complex consisting of APE1, pol β and LIG1 function together to ensure accurate repair by excluding ribonucleotides and preventing unfilled gaps.Defining the molecular determinants that dictate BER accuracy, particularly in the context of pathway coordination, is critical to fully understand disease mechanisms and how defects in this system contribute to disease risk.

Figure 1 .
Figure 1.Ligation efficiency of pol β ribonucleotide insertion by LIG1.( A ) Scheme shows the substrate and products observed in the coupled assay.( B and C ) Lane 1 is the positive control showing the ligation of pol β wild-type dTTP:A (B) and dGTP:C (C) insertion products by LIG1.Lanes 2, 7 and 12 are the negative enzyme controls of gap DNA substrates containing template base A or C. Lanes 3-6, 8-11 and 1 3-1 6 are the reaction products in the presence of pol β wild-type, LIG1 and rNTP as indicated in the figure, and correspond to time points of 0.5, 1, 3 and 5 min.R epresentativ e gel images of three independent repeats.( D ) Lane 1 is the positive control showing the ligation of pol β wild-type dTTP:A insertion product by LIG1.Lanes 2 and 7 are the negative enzyme controls of gap DNA substrates containing template base A or C. Lanes 3-6 and 8-11 are the reaction products in the presence of pol β wild-type, LIG1 and 8-o x orGTP as indicated in the figure, and correspond to time points of 0.5, 1, 3 and 5 min.R epresentativ e gel images of three independent repeats.

Figure 2 .
Figure 2. Ligation efficiency of pol β steric gate mutant Y271A ribonucleotide insertions by LIG1.( A ) Scheme showing the substrate and products observed in the coupled assay.( B-C ) Lane 1 is the positive control showing the ligation of pol β Y271A dTTP:A (B) and dGTP:C (C) insertion products by LIG1.Lanes 2, 7 and 12 are the negative enzyme controls of gap DNA substrates containing template base A or C. Lanes 3-6, 8-11 and 13-16 are the reaction products in the presence of pol β Y271A, LIG1 and rNTP as indicated in the figure, and correspond to time points of 0.5, 1, 3 and 5 min.R epresentativ e gel images of three independent repeats.( D ) Lane 1 is the positive control showing the ligation of pol β Y271A dTTP:A insertion product by LIG1.Lanes 2 and 7 are the negative enzyme controls of gap DNA substrates containing template base A or C. Lanes 3-6 and 8-11 are the reaction products in the presence of pol β Y271A, LIG1, and 8-o x orGTP as indicated in the figure, and correspond to time points of 0.5, 1, 3 and 5 min.

Figure 3 .
Figure 3. Impact of pol β cancer-associated mutants on the ligation of correct nucleotide insertion products by LIG1.( A ) Scheme shows the substrate and product observed in the insertion assay.( B ) Lane 1 is the positive control showing dGTP:C insertion product by pol β.Lanes 2, 7 and 12 are the negative enzyme controls of gap DNA substrates containing template base C. Lanes 3-6, 8-11 and 1 3-1 6 are the insertion products by pol β mutants Y265C, E288K, E295K, respectively, and correspond to time points of 0.5, 1, 3 and 5 min.Representative gel images of three independent repeats.( C ) Scheme shows the substrate and products observed in the coupled assay.( D ) Lane 1 is the positive control showing the ligation of pol β dGTP:C insertion products by LIG1.Lanes 2, 7 and 12 are the negative enzyme controls of gap DNA substrates containing template base C. Lanes 3-6, 8-11 and 1 3-1 6 are the reaction products in the presence of pol β mutants Y265C, E288K, E295K, respectively, and correspond to time points of 0.5, 1, 3 and 5 min.R epresentativ e gel images of three independent repeats.

Figure 4 .
Figure 4. Ligation efficiency of the repair intermediates with 3 -preinserted ribonucleotide mismatches and 8-o x odG v ersus 8-o x orG.( A-D ) Graphs sho w the time-dependent changes in the amount of ligation products for nick DNA substrates containing 3 -preinserted ribonucleotide mismatches opposite template base A, T, G and C. ( E-F ) Graphs show the time-dependent changes in the ligation products for nick DNA substrates containing 3 -preinserted 8-o x odG v ersus 8-o x orG.T he data are presented as the a v erages from three independent e xperiments ± SD.R epresentativ e gel images sho wing time-dependent product formation are presented in Supplementary Figures S9 -S12.

Figure 5 .
Figure 5. R emo v al of 3 -ribonucleotide mismatc hes from the nic k repair intermediate by APE1.( A -E ) Graphs show the time-dependent c hanges in the amount of remo v al products of APE1 for 3 -ribonucleotide mismatches from the nick DNA substrates containing template base A, T, G, C and 3 -8-o x orG.The data are presented as the averages from three independent experiments ± SD.The gel images showing the time-dependent substrate and product formation are presented in Supplementary Figure S13 .

Figure 6 .
Figure 6.R emo v al of 3 -mismatc hes from the nic k repair intermediate b y APE1.( A -E ) Graphs sho w the time-dependent changes in the amount of remo v al products of APE1 for a mismatched base 3 -dA, 3 -dG, 3 -dC, 3 -dT and 3 -8-o x odG from the nick DNA substrates containing template base A, T, G and C. The data are presented as the averages from three independent experiments ± SD.The gel images showing the time-dependent substrate and product formation are presented in Supplementary Figure S14 .

Figure 7 .
Figure 7. Efficiencies of APE1 and LIG1 for proofreading versus nick sealing of nick repair intermediates with 3 -deoxyribonucleotide and ribonucleotide mismatc hes.( A -B ) Dot plots sho w LIG1 efficiency, represented b y the colors indicated in the figure, f or the ligation of the nick DNA substrates with 3 -deoxyribonucleotide (A) and 3 -ribonucleotide (B) mismatches for 3 min.( C-D ) Dot plots show the APE1 efficiency, represented by the colors indicated in the figure, for the exonuclease removal of a base from the nick DNA substrates with 3 -deoxyribonucleotide for 4 min (C) and 3 -ribonucleotide for 3 min (D) mismatches.( E ) The working model showing how the deviations in the substrate-product channeling process in the presence of ribonucleotides impact the efficiency of BER at the downstream steps of the repair pathway.