New Insights in the Removal of the Hydantoins, Oxidation Product of Pyrimidines, via the Base Excision and Nucleotide Incision Repair Pathways

Background Oxidative damage to DNA, if not repaired, can be both miscoding and blocking. These genetic alterations can lead to mutations and/or cell death, which in turn cause cancer and aging. Oxidized DNA bases are substrates for two overlapping repair pathways: base excision (BER) and nucleotide incision repair (NIR). Hydantoin derivatives such as 5-hydroxyhydantoin (5OH-Hyd) and 5-methyl-5-hydroxyhydantoin (5OH-5Me-Hyd), major products of cytosine and thymine oxidative degradation pathways, respectively, have been detected in cancer cells and ancient DNA. Hydantoins are blocking lesions for DNA polymerases and excised by bacterial and yeast DNA glycosylases in the BER pathway. However little is known about repair of pyrimidine-derived hydantoins in human cells. Methodology/Principal Findings Here, using both denaturing PAGE and MALDI-TOF MS analyses we report that the bacterial, yeast and human AP endonucleases can incise duplex DNA 5′ next to 5OH-Hyd and 5OH-5Me-Hyd thus initiating the NIR pathway. We have fully reconstituted the NIR pathway for these lesions in vitro using purified human proteins. Depletion of Nfo in E. coli and APE1 in HeLa cells abolishes the NIR activity in cell-free extracts. Importantly, a number of redundant DNA glycosylase activities can excise hydantoin residues, including human NTH1, NEIL1 and NEIL2 and the former protein being a major DNA glycosylase activity in HeLa cells extracts. Conclusions/Significance This study demonstrates that both BER and NIR pathways can compete and/or back-up each other to remove hydantoin DNA lesions in vivo.


Introduction
Endogenous aerobic metabolism and variety of exogenous factors generate reactive oxygen species (ROS), which can damage macromolecules including lipids, proteins and nucleic acids. DNA has a limited chemical stability and it is one of the most biologically critical targets for ROS, with more than 80 base modifications identified so far [1]. Oxidative damage to DNA can induce mutations that cause cancer, cell death or senescence. In Archaea, Bacteria and Eukarya, as well as in some large DNA viruses, oxidatively damaged DNA bases are removed in two repair pathways: base excision repair (BER) and nucleotide incision repair (NIR) [2,3,4,5,6]. In the classical BER pathway, a DNA glycosylase excises the base, leaving as an end product either an apurinic/apyrimidinic (AP) site or a single-stranded DNA break with 39-sugar phosphate groups which must be removed prior the gap-filling synthesis step [7,8]. Alternatively, in the NIR pathway, an AP endonuclease makes an incision 59 next to a damaged base in a DNA glycosylase-independent manner, providing a proper 39-OH group for DNA polymerization and a 59-dangling damaged nucleotide [3]. Although the majority of oxidized DNA bases are removed in the BER pathway initiated by multiple DNA glycosylases [9,10] certain types of oxidative DNA damage such as the alpha-anomeric 29-deoxynucleosides (adA, adT and adC) cannot be repaired by DNA glycosylases/AP lyases but rather by the AP endonucleases in the alternative NIR pathway [11,12,13]. Furthermore, oxidatively damaged pyrimidines including 5,6dihydrothymine (DHT), 5,6-dihydrouracil (DHU), 5-hydroxyuracil (5OHU) and 5-hydroxycytosine (5OHC) are substrates for the both BER and NIR pathways suggesting that latter pathway can serve as back-up system to counteract oxidative stress [3,13,14].
In human cells the major AP endonuclease 1, APE1/Ref-1/ HAP-1, initiates NIR pathway by cleaving duplex DNA 59 next to oxidatively damaged bases [13]. In the past, APE1 was independently discovered as an abasic site-specific endonuclease homologous to the E. coli Xth protein [15] and as a redoxregulator of the DNA binding domain of Fos-Jun, Jun-Jun, AP-1 proteins and several other transcription factors [16]. In addition to AP endonuclease and NIR activities, APE1 exhibits other DNA repair activities: 39R59 exonuclease, 39-phosphodiesterase, 39phosphatase and RNase H [15]. Although pH, ionic strength and divalent cation requirements of the APE1-catalyzed NIR versus AP endonuclease are dramatically different, we have demonstrated that the intracellular environment of human cells can support NIR function [17]. The APE1 repair activities are divalent metal ion dependent, indeed, structural studies have shown that APE1 contain metal-binding site(s) that can held several metal ions including Sm +3 , Pb +2 , Ca +2 , Mn +2 and Mg +2 [18,19]. Previously, we have demonstrated that Zn +2 cations can support APE1catalyzed NIR activity in human cell-free extracts [13]. Interestingly, zinc is the second most abundant transition metal in the body after iron [20] therefore it could play biological role in stimulating APE1-NIR activity in vivo.
Previously, it has been shown that E. coli Nth, Nei and Fpg as well as S. cerevisiae Ntg1 and 2 can initiate the BER pathway for 5OH-Hyd and 5OH-5Me-Hyd residues in DNA [34,35]. Here, we demonstrate that the AP endonucleases of E. coli Nfo, yeast Apn1 and human APE1 initiate the NIR pathway by incising the duplex DNA containing 5OH-Hyd and 5OH-5Me-Hyd residues. Using Matrix Assisted Laser Desorption Ionisation Time-Of-Flight (MALDI-TOF) Mass Spectrometry (MS) analysis we show that the AP endonucleases cleave 59 next to the pyrimidine-derived hydantoin lesions when present in the duplex DNA substrate. The removal of 5OH-Hyd and 5OH-5Me-Hyd residues in the both BER and NIR pathways in bacteria, yeast and human cells has been characterized. A number of redundant DNA glycosylase activities excising the pyrimidine-derived hydantoins have been identified, including human NTH1, NEIL1 and NEIL2 DNA glycosylases. The roles of BER and NIR, as back-up complementary pathways for oxidized DNA bases in vivo, are discussed.

Results
Pyrimidine hydantoins are substrates for the NIR pathway E. coli and S. cerevisiae bifunctional DNA glycosylases initiate BER pathway by excising 5OH-Hyd and 5OH-5Me-Hyd residues when present in duplex DNA [34]. To examine whether these oxidized bases can be also substrates of the AP endonucleases involved in the NIR pathway we incubated the 39-[ 32 P]-labelled 5OH-HydNG and 5OH-5Me-HydNA substrates with the E. coli Nfo, S. cerevisiae Apn1 and human APE1 proteins. As shown in Figure 1B, all three AP endonucleases tested cleave the sugar phosphate backbone 59 next to 5OH-Hyd nucleotide, generating ,14-mer (n+1) fragment that migrates slower (lanes 4,5 and 7) than 13-mer (n) DNA glycosylases cleavage product (lanes 2 and 3). In agreement with our previous observations, APE1 efficiently incise 5OH-HydNG under NIR conditions, whereas no activity was detected under BER conditions (lanes 5 versus 6) [13]. Unexpectedly, the (n+1) AP endonuclease-cleavage fragments migrate faster than the 39-[ 32 P]-labelled 14-mer size marker oligonucleotides containing either cytosine or thymine at 59 end (lanes 4,5,7 versus 9,10), suggesting that either cleavage fragments with 59-terminal nucleotide migrate faster as compared to 14-mer size marker or that 5OH-Hyd may undergo some transformation after enzymatic cleavage.
Interestingly, all AP endonucleases tested also incised 5OH-5Me-HydNA, albeit with much lower efficiency as compared to 5OH-HydNG duplex (lanes 14,15,17 versus 4,5,7). Again, APE1 cleaved 5OH-5Me-HydNA only under NIR conditions generating the cleavage product with the same size as with 5OH-HydNG oligonucleotide (lanes 15 and 5). In contrast, Nfo and Apn1catalyzed incision resulted in two closely migrating (n+1) fragments (lanes 14 and 17), upper-band co-migrates with 14-mer size marker (lane 18) and lower-band co-migrates with 5OH-HydNG NIR-cleavage fragment (lanes 4,5,7). Taken together, these results suggest that the Apn1 and Nfo-generated upper-band corresponds to the fragment containing a 59-dangling 5OH-5Me-Hyd nucleotide whereas the lower-band is similar to that observed with 5OH-HydNG substrate and may correspond to fragment containing a decomposed form of 5OH-5Me-Hyd. We may propose that bigger molecular weight difference between 5OH-5Me-Hyd and its putative degradation product resulted in a better electrophoretic resolution on the gel as compared to 5OH-Hyd and its corresponding putative degradation product. Importantly, Nfo-catalyzed cleavage of 5OH-5Me-HydNA under NIR condition gives essentially the same proportion of two closely migrated bands on the gel as compared to the Nfos' standard reaction condition. Furthermore, longer incubations and/or simultaneous treatment of 5OH-5Me-HydNA with Nfo and APE1 do not change the migration pattern of cleavage products (Supporting Information Figure S1). The difference between APE1 and other AP endonucleases observed on 5OH-5Me-HydNA substrate may imply difference in the mechanism of action of these enzymes. Since, oxidative stress can generate 5OH-5Me-Hyd in DNA from both thymine and 5-methylcytosine residues, we investigated base-pair specificity of DNA repair enzymes using two duplexes 5OH-5Me-HydNA or 5OH-5Me-HydNG as substrates. APE1, Nfo, Apn1 and Nth incise 5OH-5Me-HydNA and 5OH-5Me-HydNG with the same efficiency (data not shown).
Study of the mechanism of action of DNA repair enzymes on the pyrimidine-derived hydantoins by MALDI-TOF mass spectrometry Previously, the mechanism of action of the AP endonucleases on oxidatively damaged bases was studied by analysing the migration pattern of 39-end labelled cleavage DNA fragments in denaturing PAGE [3]. Here, for the first time we performed MALDI-TOF MS analysis of the reaction products of the AP endonucleases and DNA glycosylases when acting on 5OH-Hyd and 5OH-5Me-Hyd lesions in duplex DNA. Analysis of the mass spectrum of the reaction products resulting from the incision of 5OH-HydNG by  22 , respectively). Importantly, the respective amounts of cleavage fragments containing ureido and 5OH-Hyd varied among the experiments. As shown in Figure 2B, the spectrum of products after incubation of 5OH-5Me-Hyd with Nfo contains a peak corresponding to the 13-mer 59-p-   Interestingly, MALDI-TOF MS analysis showed that APE1catalyzed incision of 5OH-HydNG and 5OH-5Me-HydNA gave rise only to ureido-containing fragment and no hydantoincontaining cleavage products were observed ( Figure 2C and D). In addition, APE1 extend the nick to the gap by 39R59 exonuclease activity generating shorter 8 mer downstream cleavage fragment 59-CACTTCGG ([M-H] 2 with calculated mass 2383 Da) ( Figure 2C). It should be stressed that the control MALDI-TOF MS analysis of the newly synthesized hydantoincontaining oligonucleotides prior to incubation with DNA repair enzymes confirmed the integrity of 5OH-Hyd and 5OH-5Me-Hyd nucleotides and did not reveal any trace of degradation products such as ureido adducts [30,36]. Therefore, this data strongly suggest that ureido adducts occur during and/or after the AP endonuclease-catalyzed incision. Conversion of pyrimidinederived hydantoins to ureido during NIR raises the question whether the AP endonucleases could affect stability of a damaged nucleotide during incision step. For this we examined reaction products of the APE1-catalyzed incision of 30-mer adANT oligonucleotide duplex by MALDI-TOF MS. As expected, we found two mono-charged peaks one at  Figure 2E). This result indicates that APE1 cleaves duplex DNA 59 next to adA further confirming the mechanism of action of the AP endonuclease. Importantly, analysis of mass spectrum did not reveal any chemical modifications of adA residue in the cleavage products suggesting that the AP endonucleases do not degrade 59-dangling adA residues.
Next we employed MALDI-TOF MS analysis to investigate the mechanism of action of NEIL1, a human DNA glycosylase that excises pyrimidine hydantoins residues in the BER pathway. Similar to E. coli Fpg and Nei proteins, NEIL1 is a bi-functional DNA glycosylase endowed with an AP lyase activity that incises DNA at abasic sites by a b,d-elimination mechanism and leaves single-strand DNA break carrying a phosphate residue at the 39  Figure 2F). This result corroborates with previous data obtained using denaturing PAGE separation technique ( Figure 1B).
Activity of the AP endonucleases on oligonucleotide duplexes containing thymine glycol and urea residues Thymine glycol is a major nucleobase lesion that may be formed within DNA by several oxidative processes. This oxidized form of pyrimidine residue, as hydantoins residues, exhibit a ringchain tautomerism at C6-N1 or C5-N1 bond that may lead to degradation towards fragmented products (such as ureido or formamido residues), spontaneously and/or upon oxidative and alkali conditions [37,38,39]. Therefore, we examined whether Tg is a substrate for the NIR pathway. For this we used Tg-containing oligonucleotide 34-mer, 30-mer and 19-mer duplexes Tg-34NA, Tg-30NA and 19Tg-IWNA, respectively as substrates for wild-type (WT) APE1 and NIR-deficient APE1K98E mutant. Denaturing PAGE analysis revealed that the both APE1 proteins incise 59-[ 32 P]-labelled Tg-34NA and Tg-30NA oligonucleotide duplexes but not 19Tg-IWNA duplex (Supporting Information Figure S2A). Interestingly, Tg-34 and Tg-30 oligonucleotides migrated as a double band suggesting the presence of degradation product of Tg. Indeed, the slow migrating ''upper-band'' cannot be incised by wild-type APE1 and NIR-deficient APE1-K98E mutant, whereas the fast migrating ''lower-band'' band was incised by both APE1. The MS analysis of the ''lower-band'' fragment purified form the gel showed that it contains an ureido nucleotide (data not shown) indicating that the degradation product of Tg but not intact Tg residue is a substrate for APE1. The urea-containing oligonucleotides purified from Tg-34NA and Tg-30NA were incised by the E. coli exonuclease III (Xth) protein [40] and also by the NIRdeficient APE1 (D308 and K98A/R185A) mutants (Supporting Information Figure S2A,B) whereas 5OH-HydNG and 5OH-5Me-HydNA duplexes were not suggesting that the later duplexes do not contain urea (Supporting Information Figures S1). Furthermore, APE1 can incise urea-containing DNA substrate under both NIR and BER conditions with similar efficiency suggesting that urea residue closely mimics an abasic site and can be recognized by APE1 under both reaction conditions (Supporting Information Figure S2A). Importantly, APE1-catalyzed incision of 5OH-HydNG and 5OH-5Me-HydNA oligonucleotide duplexes is strongly inhibited under BER condition thus ruling out a possible presence of ureido degradation products in our oligonucleotide preparations ( Figure 1B). Taken together these data suggest that although Tg is not a substrate for APE1, its degradation products can be repaired in the NIR pathway.

Activity of various DNA glycosylases on oligonucleotide duplexes containing 5OH-Hyd and 5OH-5Me-Hyd lesions
Previous studies have demonstrated that pyrimidine-derived hydantoins are substrates for bi-functional DNA glycosylases in E. coli and yeast. Here, we investigated whether 5OH-Hyd and 5OH-5Me-Hyd residues are also substrates for the previously characterized bacterial and human DNA glycosylases. As shown in Figure 3, the 59-[ 32 P]-labeled 5OH-HydNG and 5OH-5Me-HydNA oligonucleotide duplexes were challenged with a variety of highly purified DNA glycosylases. Since not all DNA glycosylases possess AP site-nicking activity, the samples, after incubation with the mono-functional DNA glycosylases, were treated with light piperidine treatment [10% (v/v) piperidine at 37uC for 45 min]. in order to cleave DNA at the potential abasic sites generated by the excision of the modified base. Control light piperidine treatment of 5OH-HydNG duplex resulted in slight defragmentation of the oligonucleotide to short DNA fragments indicating the presence of alkaline labile sites most likely due to spontaneous depurination ( Figure 3A, lane 2). Nevertheless, 5OH-Hyd and 5OH-5Me-Hyd residues were resistant to piperidine treatment and no cleavage at the lesion site was observed in the absence of DNA glycosylase treatment ( Figure 3A,B lane 2). In agreement with our previous observations, E. coli Fpg, Nth and Nei excise with good efficiency both hydantoins ( Figure 3A,B lanes 15-17). As expected NTH1 and NEIL1, human homologues of Nth and Nei, respectively, excise with comparable efficiency both 5OH-Hyd and 5OH-5Me-Hyd residues (lanes 7 and 9). While, NEIL2 a human paralogue of NEIL1 shows only weak activity towards hydantoins (lane 10). Despite being used in 10-fold molar excess, none of the mono-functional DNA glycosylases used: UNG, SMUG1, ANPG70, TDG, UDG, TagI, AlkA were able to excise the hydantoins. Except E. coli MUG, Mismatch Uracil DNA Glycosylase, which showed weak activity on both 5OH-HydNG and 5OH-5Me-HydNA substrates (lane 14). Interestingly, TDG, a human homologue of MUG, does not show any detectable activity on the hydantoins (lane 6). Overall, the E. coli Nei and Nth proteins show a slight preference for 5OH-5Me-Hyd than 5OH-Hyd ( Figure 3B lanes 15-16 vs 3A, lanes 15-16), whereas human NEIL1 and to more extent NEIL2, NTH1 and E. coli Fpg excise preferentially 5OH-Hyd than 5OH-5Me-Hyd residues ( Figure 3A lanes 7, 9, 10 and 17 vs 3B, lanes 7, 9, 10 and 17).

Kinetic parameters of the incision of oligonucleotide duplexes with a single hydantoin residue by various AP endonucleases and DNA glycosylases
To examine the relative efficiency of NIR and BER pathways for the removal of 5OH-Hyd and 5OH-5Me-Hyd residues, we measured the amount of cleaved oligonucleotide as a function of the Nfo, Apn1, APE1, NTH1 and NEIL1 protein concentrations (Supporting Information Figure S3). 5OH-HydNG was the preferred substrate as compared to 5OH-5Me-HydNA for both AP endonucleases and DNA glycosylases. 5OH-5Me-HydNA oligonucleotide duplex was the less preferred substrate for all AP endonucleases tested, and human APE1 was the less efficient among other AP endonucleases, quickly reaching the plateau at 1-2 nM protein and incising at most 50% of 5OH-HydNG and 5% of 5OH-5Me-HydNA (Supporting Information Figure S3B). In contrast to AP endonucleases, DNA glycosylases NEIL1 and NTH1 can efficiently excise both hydantoin residues, although they also showed slight preference to 5OH-HydNG (Supporting Information Figure S3D). Based on these assessment studies we have adjusted the DNA repair enzyme concentration and length of incubation to cleave no more than 30% of the substrate, thereby maintaining initial velocity conditions. Then we further characterized substrate specificity of the AP endonucleases and DNA glycosylases by measuring K M , k cat and k cat /K M values for incision activity using 5OH-HydNG and 5OH-5Me-HydNA substrates. As shown in Table 1, the k cat /K M values for 5OH-HydNG incision of all three AP endonucleases tested are the same order of magnitude as compared to that of E. coli Nei DNA glycosylase and 20 or 2.5 fold higher as compared to that of human DNA glycosylases NEIL1 and NTH1, respectively. The higher kinetic efficiency of the AP endonucleases towards 5OH-HydNG as compared to human DNA glycosylases suggests that the NIR pathway could be a major pathway for the removal of 5OH-Hyd residues in vivo. In contrast to Nfo, the eukaryotic AP endonucleases incise 5OH-5Me-HydNA substrate with very low efficiency as compared to human DNA glycosylases suggesting that 5OH-5Me-Hyd is removed mainly in the BER pathway in eukaryotic cells. Interestingly, among all DNA repair enzymes tested E. coli Nei and human NTH1 DNA glycosylases have the highest k cat /K M values for cleavage of 5OH-5Me-HydNA.

In vitro reconstitution of the human NIR pathway for 5OH-Hyd residues
Recently, we have reconstituted in vitro the human NIR pathway for adANT oligonucleotide duplex using four purified proteins APE1, flap endonuclease 1 (FEN1), DNA polymerase b (POLb) and DNA ligase I (LIG1) [41]. As described previously, for the NIR reconstitution assay we used 2 mM ATP in addition to 3 mM MgCl 2 which allowed us to bring the concentration of free Mg 2+ down to 1 mM, due to binding of magnesium cations by ATP. Here, we examined whether 5OH-Hyd residues can be efficiently removed in the APE1-initiated NIR pathway in vitro in DNA glycosylase-independent manner ( Figure 4). Under the reaction condition used (3 mM MgCl 2 and 2 mM ATP) which allows both NIR endonuclease activity and DNA repair synthesis, APE1 incised about 60% of 39-[ 32 P]-end labelled 5OH-HydNG substrate resulting in a 14-mer cleavage product (lane 2). Addition of FEN1 initiates strong 59R39 exonuclease degradation of 14-mer fragment (lane 4). Previously, we observed futile repair of adANT  in the presence of APE1 and LIG1 proteins [41]. However, when APE1 cleaved 5OH-HydNG oligonucleotide duplex, addition of LIG1 did not restore the 23-mer full-sized fragment yet it efficiently blocked FEN1 exonuclease (lane 7). These results indicate that APE1-induced single-strand DNA breaks next to hydantoin residues would persist if the 59-terminal dangling base is not removed. In the absence of a DNA ligase, addition of POLb further stimulates FEN1 exonuclease activity (lane 6). Finally, in the presence of all four proteins APE1, POLb, FEN1 and LIG1, we observed the completion of repair reaction resulting in nearly full restoration of the 23-mer fragment (lane 9). In order to verify the removal of 5OH-Hyd residues following incubation of 5OH-HydNG with the purified human proteins, we treated the repaired 23-mer DNA products with Nei DNA glycosylase which can incise with the high efficiency 5OH-Hyd-containing DNA ( Figure 3) [34]. The appearance of a 13-mer cleavage DNA fragment after the NIR reconstitution assay will indicate the presence of 5OH-Hyd in the repaired 23-mer fragment. As expected, in the absence of APE1, no repair reactions took place (lanes 10 and 20). Importantly, Nei-treatment of the 23-mer fragment from lane 9 which was previously incubated with all four human DNA repair proteins revealed very little amount of a 13-mer cleavage product (lane 19) indicating that the absolute majority of 5OH-Hyd residues are efficiently eliminated from the 23-mer duplex during NIR reconstitution assay. These results indicate that APE1 incises 59 next to 5OH-Hyd and allows POLb to initiate DNA stranddisplacement synthesis generating a flap-structure which is in turn cleaved by FEN1 to remove 59-dangling hydantoin nucleotide and generate single-strand break which is then sealed by LIG1. These data strongly suggest that, under the reaction condition which enables nucleotide incision activity, DNA polymerase synthesis and ligation, 5OH-Hyd residues can be efficiently removed in the APE1-initiated NIR pathway resulting in the restoration of the DNA sequence integrity.
DNA repair activities on 5OH-Hyd and 5OH-5Me-Hyd containing DNA in E. coli and human cell-free extracts Data obtained with the purified DNA repair proteins show the redundancy of the BER and NIR pathways for the removal of pyrimidine-derived hydantoin residues in DNA. Therefore, to ascertain the respective role of these two pathways in vivo, we examined AP endonuclease and DNA glycosylase activities in cellfree extracts from E. coli and human cells. To distinguish NIR and BER activities in the extracts we used 39-[ 32 P]-labelled 5OH-HydNG and 5OH-5Me-HydNA oligonucleotide duplexes as substrates. When using these substrates the DNA glycosylases/AP lyases generate a 13-mer cleavage fragment ( Figure 5A, lanes 10-12 and 22-24), whereas AP endonucleases generate a 14-mer fragment (lanes 9 and 21).
As shown in Figure 5, we detected mainly DNA glycosylase activities and very little Nfo activity on both hydantoins in the extracts from WT E. coli strain (lanes 2 and 14). As expected from the known induction of the Nfo protein by paraquat, addition of this oxidizing agent to growing cultures of E. coli increased the amount of the 14-mer fragment (lanes 3 and 15). Also, extracts from E. coli nfo mutant completely lack NIR activity on both DNA substrates (lanes 5 and 17) indicating that the NIR pathway in E. coli is absolutely dependent on the Nfo gene product. Interestingly, extracts from the E. coli nth fpg and nth nei double mutants exhibited dramatic decrease in incision activities on both 5OH-HydNG and 5OH-5Me-HydNA substrates (lanes 7, 8 and 19, 20) as compared to the extracts from E. coli WT, single nth, nfo and double fpg nei mutants (lanes 2-6 and 14-18). Importantly, in the extract from E. coli nth nei double mutant we detected only NIR activity and no DNA glycosylase (lanes 8 and 20) suggesting that Nth and Nei are major DNA glycosylases that remove 5OH-Hyd and 5OH-5Me-Hyd residues in the BER pathway. Taken together, these results suggest that in E. coli pyrimidine-derived hydantoins are mainly removed in DNA glycosylase-initiated BER pathway. However, in the absence of DNA glycosylases, the NIR pathway can serve as back-up system.
Interestingly, some reduction in the percentage of NIR cleavage products can be seen in DNA glycosylase-deficient strains: decrease from 8.6% in WT strain to 3.6% in nth mutant on 5OH-HydNG ( Figure 5B) and from 2.8% in WT strain to 1% in nth nei double mutant on 5OH-5Me-HydNA ( Figure 5C) suggesting that the NIR activity in E. coli cell-free extracts may require the presence of DNA glycosylases. To examine whether the NIR pathway functions independently of DNA glycosylase-catalyzed BER we measured the NIR and BER activities in DNA glycosylase-deficient mutants treated with paraquat. As expected, in all extracts from DNA glycosylase-deficient mutants the Nfocatalyzed NIR activity towards 5OH-HydNG and 5OH-5Me-HydNA was induced by paraquat up to the level observed in WT strain (Supporting Information Figure S4). Based on these results we suggest that the NIR pathway functions independently of BER and does not require the presence of DNA glycosylases to remove pyrimidine-derived hydantoins.
Previously, we have established that, in human cell-free extracts, the APE1-catalyzed NIR is the major activity on 5,6-dihydrouracil (DHU) and 5-hydroxycytosine (5OHC) containing DNA duplexes under NIR condition (in the presence of 0.1 mM ZnCl 2 ) [13,14]. Therefore, to assess the role of NIR pathway in the removal of 5OH-Hyd and 5OH-5Me-Hyd residues in DNA, we examined incision activities in the extracts from HeLa cells under both NIR+Zn 2+ and BER+EDTA (in the presence of 1 mM EDTA and absence of Zn 2+ cations that may inhibit the DNA glycosylases) conditions ( Figure 6A-C). Interestingly, under the NIR+Zn 2+ condition, we observed mainly NIR activity on the 39-[ 32 P]-  labelled 5OH-HydNG and 5OH-5Me-HydNA substrates and very little DNA glycosylase-dependent cleavage ( Figure 6A, lanes 2 and  9). Whereas, under the BER+EDTA condition we observed predominantly DNA glycosylase incision and no or very little NIR activity was detected (lanes 4 and 11). These results imply that the AP endonuclease and DNA glycosylase activities detected in human cell-free extracts vary dramatically upon the reaction conditions used. As expected, the APE1 silencing strongly reduces the NIR activity on 5OH-HydNG and 5OH-5Me-HydNA duplexes (lanes 3 and 10), indicating that APE1 is required for the NIR activities on hydantoins in HeLa cell-free extracts. Interestingly, down-regulation of the APE1 protein results in a dramatic increase of the DNA glycosylase activities on both DNA substrates (lanes 5 and 13). This may suggest that either the APE1 protein inhibits DNA glycosylase activities or that the transcription silencing of APE1 gene induces expression of DNA glycosylases in HeLa cells.
Biochemical data demonstrate that both purified human DNA glycosylases NTH1 and NEIL1 can efficiently repair pyrimidinederived hydantoins in vitro however neither NEIL1, nor NEIL2 specific activities were detected after incubation of 59-[ 32 P]labelled 5OH-HydNG and 5OH-5Me-HydNA oligonucleotide duplexes in HeLa cells extracts under the BER+EDTA condition ( Figure 7A, lanes 2 and 9). It should be noted that NEIL1catalyzed cleavage fragments migrate faster in denaturing PAGE as compared to fragments generated by NTH1 and APE1 because they contain 39-phosphate (39-P) residues ( Figure 7A, lanes 6 and 13 versus 5, 7, 12 and 14). This suggests that either 39-P residues are removed by a 39-phosphatase activity or that NTH1 could be a major hydantoin-DNA glycosylase activity in human cell extracts. To test this hypothesis we down-regulated NTH1 expression using specific siRNA duplex [42]. Two concentrations of the siRNA duplex were used to down-regulate the NTH1 protein level in HeLa cells ( Figure 7B). Depletion of NTH1 significantly reduces BER activity on both 5OH-HydNG and 5OH-5Me-HydNA substrates ( Figure 7A, lanes 3-4 and 10-11) as compared to the control non-silenced cell extracts (lanes 2 and 9) suggesting that indeed NTH1 is a major detectable DNA glycosylase activity towards pyrimidine-derived hydantoins in HeLa cells. Similar results were obtained when using the 59-[ 32 P]-labelled DHUNG oligonucleotide duplex as a substrate ( Figure 7C).
Fanconi anemia (FA) is a recessive cancer prone syndrome featuring bone marrow failure, hypersensitivity to interstrand DNA cross-links (ICLs) and also to ionizing radiation and oxidative stress [43,44]. Recently we demonstrated that the immortalized lymphoid cells of FA complementation Group A, C and D2 have decreased level of the NEIL1 protein, whereas, the cellular level of APE1 was similar to that in normal healthy cells [45]. In light of these observations we examined the incision activities of FA complementation Group C (FA-C) cell-free extracts towards 59-[ 32 P]-labelled 5OH-HydNG oligonucleotide duplex. As a control for normal non-FA cells we used AHH1 (WT) cells and FA-C cells complemented with a plasmid expressing FANCC protein (FA-C+FANCC). FA-C extracts exhibit somewhat decreased DNA glycosylase activity on 5OH-HydNG as compared to AHH1 and FA-C+FANCC extracts suggesting that FA-C cells are impaired in the repair of 5OH-Hyd residues (Supporting Information Figure S5). Since NTH1 is a major DNA glycosylase for 5OH-Hyd residues, we measured the NTH1 protein level in FA-C cells. Western blot analysis did not reveal any difference in NTH1 level in extracts from AHH1, FA-C and FA-C+FANCC cells (data not shown). At least three DNA glycosylases can initiate BER for 5OH-Hyd residues in human cells: NTH1, NEIL1 and NEIL2. The cleavage fragments generated by the extracts under BER condition contain 39-OH group ( Figure 7A) and do not migrate as b and b,d-elimination products generated by pure NTH1 and NEIL1 (lanes 5-6). We may propose that cleavage fragment with 39-OH group occur due to robust 39-diesterase and 39-phosphatase activities in human cellfree extracts. Taken together, these results may suggest deficiency in the repair of oxidative DNA damage in FA cells, however, the observed difference between FA and normal cells was very small (Supporting Information Figure S5).

Discussion
Oxidized bases are the major endogenous DNA lesions that can accumulate during aging. Purine and pyrimidine moieties of the respective nucleosides undergo oxidative degradation, resulting in a number of modified bases that can be highly mutagenic when present in DNA. 5OH-Hyd and 5OH-5Me-Hyd residues have been shown to be major oxidation decomposition products of cytosine and thymine, respectively [23,46,47]. Cells evolved several repair mechanisms to remove oxidized bases from the genome. In the present study, we investigated whether the AP endonucleases involved in the NIR pathway recognize the pyrimidine-derived hydantoins in duplex DNA. The results show for the first time that Nfo, Apn1 and APE1 can incise, in a DNA glycosylase-independent manner, duplex DNA containing both 5OH-Hyd and 5OH-5Me-Hyd residues.
Previous studies of the mechanism of the AP endonucleasecatalyzed nucleotide incision activities were mainly based on the analysis of migration pattern of cleavage fragments in denaturing PAGE [3,41]. Here, to provide insight into the mechanism of NIR activity we analyzed the AP endonuclease-generated cleavage fragments by MALDI-TOF mass spectrometry. The advantage of MALDI-TOF MS analysis is that it permits simultaneous measurements of every DNA products including non-labelled complementary strand, upstream and downstream cleavage fragments. As expected, the results obtained by MALDI-TOF MS perfectly confirmed those obtained by the denaturing PAGE separation technique: all AP endonucleases tested (i) incise the oligonucleotide duplexes 59 next to 5OH-Hyd and 5OH-5Me-Hyd residues generating 39 downstream cleavage fragments still containing 59-terminal damaged nucleotide; (ii) degrade 59 upstream cleavage fragments by their non-specific 39R59 exonuclease activity ( Figure 2). Importantly, the MALDI-TOF MS analysis shed new light on the mechanism of nucleotide incision activity on the hydantoins by demonstrating that 5OH-Hyd and 5OH-5Me-Hyd residues in DNA undergo degradation into ureido residues during and/or after incubation with the AP endonucleases. MS data revealed that cleavage of 5OH-HydNG and 5OH-5Me-HydNA oligonucleotide duplex by all AP endonucleases tested generate DNA fragments containing 59-terminal ureido residues. Indeed, the denaturing PAGE analysis demonstrated that 39 downstream cleavage fragments, derived from the treatment of 5OH-HydNG, migrate faster than 14-mer size marker fragment but still slower than 13-mer size marker and DNA glycosylase-generated fragments suggesting that 59-terminal hydantoin residue may undergo partial decomposition ( Figure 1B). Formation of ureido residues during AP endonuclease treatment does not depend on reaction condition and incubation time. Furthermore, the co-incubation of 5OH-5Me-HydNA duplex with Nfo and APE1 did not increase yield of cleavage fragments containing ureido residues (Supporting Information Figure S1). Interestingly, the oxidized pyrimidine bases can undergo ringchain tautomerism at C6-N1 or C5-N1 bond resulting in formation of acyclic linear structures which could be chemically less stable [48,49]. Loss of the base stacking stabilization after duplex incision next to damaged base might change the equilibrium of hydantoins ring-chain tautomerism into the less stable open form. However, MS analysis of P1 nuclease digestion of DNA containing both 5OH-Hyd and/or 5OH-5Me-Hyd residues did not reveal any modification of the hydantoin moiety in the nucleosides [30,36].
Previously, we proposed that the NIR activity requires a more tight binding of the AP endonucleases to DNA substrate containing an oxidatively damaged base, as a consequence APE1 has low turnover rate on adA-containing DNA substrate as compared to AP site DNA [13,50]. This tight mode of binding may enable recognition of oxidized bases in duplex DNA by the AP endonucleases by creating specific interactions of active site amino acid residues with a damaged base. Interestingly, when acting upon 5OH-5Me-HydNA, APE1 generates cleavage fragment containing only 59-ureido nucleotides whereas Nfo and Apn1 produce two fragments containing either 59-5OH-5Me-Hyd or 59ureido nucleotides (Figure 1). These results indicate that in contrast to Nfo and Apn1, APE1 cannot incise 5OH-5Me-HydNA duplex but rather ureido-containing oligonucleotide duplex. Since, NIR-deficient APE1 mutants cannot cleave the hydantoincontaining duplexes (Supporting Information Figure S1A) and that ureido residue is not present in the non-treated oligonucleotides it is tempting to speculate that under NIR condition APE1 may promote the conversion of 5OH-5Me-Hyd to ureido residue via interactions between its active site amino acids and the damaged pyrimidine. When APE1 binds to DNA it may convert part of 5OH-5Me-HydNA to UreidoNA duplex, this would enable APE1 to cut 59 next to ureido residue generating the observed cleavage fragment with 59-terminal ureido nucleotide. This is not possible under the BER+Mg 2+ condition (in the presence of 5 mM MgCl 2 ) since under this condition APE1 cannot bind to DNA substrate in the tight manner and catalyze the NIR activity (Supporting Information Figure S2A). Nfo and Apn1 could also promote the conversion of hydantoin to ureido residue by binding to 5OH-5Me-HydNA duplex since they also generate ureido residue after reaction. Interestingly, the co-incubation of 5OH-5Me-HydNA duplex with Nfo and APE1 did not increase yield of cleavage fragments containing ureido residues (Supporting Information Figure S1). Furthermore, ureido residues can be detected by MS after incubation of the 5OH-HydNG duplex with all AP endonucleases tested which may suggest conversion of 5OH-Hyd to ureido residue upon enzyme binding to DNA (Figure 2A,C). Hence, we may speculate that the formation of ureido residues in DNA might be a consequence of both chemical instability of the hydantoins and non-covalent interactions of a damaged base with active site amino acid residues upon AP endonuclease binding. Nevertheless, it should be noted that the degradation of the hydantoins to ureido during or after AP endonuclease-catalyzed cleavage of duplex DNA substrate does not affect removal of the dangling nucleotide residue during reconstitution of the NIR pathway in vitro, which leads to the restoration of a damage-free duplex oligonucleotide.
In previous studies, we characterized substrate specificities of the bacterial, yeast and human AP endonucleases towards damaged pyrimidines such as DHU, DHT and 5OHC and demonstrated that in vitro the AP endonucleases are more efficient than the DNA glycosylases/AP lyases [13,14,51]. In the present work, analysis of kinetic parameters showed that incision of 5OH-HydNG by Nfo, Apn1 and APE1 are highly efficient implying that the NIR pathway can efficiently compete with BER in the removal of 5OH-Hyd residues in DNA in vivo. In contrast, the kinetics parameters of the cleavage of 5OH-5Me-HydNA by eukaryotic AP endonucleases Apn1 and APE1 were inefficient as compared to Nfo and DNA glycosylases suggesting that in eukaryotes the majority of 5OH-5Me-Hyd residues would be removed rather in the BER pathway. Interestingly, among all human DNA repair enzymes tested human NTH1 DNA glycosylase has the highest k cat /K M value for incision of 5OH-5Me-HydNA substrate. Therefore, excision of 5OH-5Me-Hyd residues by NTH1 would rather initiate short-patch BER pathway similar to excision of 8oxoG residues by hOGG1 [52]. Here, based on a new substrate specificity of APE1 we performed a complete in vitro reconstitution of the human NIR pathway for 5OH-HydNG duplex oligonucleotides using purified proteins. Incubation of a 5OH-HydNG duplex in the presence of APE1, FEN1, POLb and LIG1 generated a free of 5OH-Hyd residues, full-length oligonucleotide (Figure 4). Interestingly, we did not observed futile repair of 5OH-HydNG duplex in the presence of DNA ligase activity suggesting that the repair of APE1-generated single-strand breaks should be accomplished through the removal of 59-dangling nucleotide in the longpatch NIR pathway. Overall, these results demonstrate that 5OH-Hyd residues can be processed in a DNA glycosylase-independent manner via the NIR pathway.
Data obtained with the purified proteins support the physiological relevance of the AP endonuclease-catalyzed nucleotide incision activity on DNA containing pyrimidine-derived hydantoins. To further investigate the role of various DNA repair pathways, we measured the AP endonuclease and DNA glycosylase activities in cell-free extracts from E. coli and human cells. In E. coli cell-free extracts we detected mainly DNA glycosylase activities with Nth and Nei being major DNA glycosylases responsible for incision of 5OH-HydNG and 5OH-5Me-HydNA duplexes and little NIR activity ( Figure 5). Although, the Nfo-catalyzed NIR activity towards 5OH-HydNG and 5OH-5Me-HydNA can be strongly induced by paraquat up to the level similar to those observed for DNA glycosylases (Supporting Information Figure S4). Interestingly, it was shown that E. coli nth nei mutants are hypersensitive to the lethal effects of ionizing radiation [53] and hydrogen peroxide [54], implying potential role of pyrimidine-derived hydantoins as lethal oxidative lesions in DNA. In the case of human cell-free extracts, depending on the reaction conditions either NIR+Zn 2+ and/or BER+EDTA activities were detected ( Figure 6A,C). Using small RNA silencing we demonstrated that the alternative DNA glycosylase-independent repair of 5OH-Hyd and 5OH-5Me-Hyd residues in duplex DNA depends upon APE1 thus substantiating the biological role of APE1-catalyzed NIR pathway in human cells ( Figure 6).
Recently, it has been demonstrated that Nei and NEIL1 mediated excision of 5OH-5Me-Hyd can result in an unproductive DNAprotein covalent (DPC) complex which hides the lesion from repair and represents more complex bulky lesion [55]. This observation further substantiates the biological role of NIR as an alternative pathway which avoids the generation of genotoxic intermediates during repair of the hydantoin DNA lesions.
Under BER (BER+EDTA and BER+Mg 2+ ) conditions, three human DNA glycosylases can excise 5OH-Hyd and 5OH-5Me-Hyd residues hence contributing to the redundancy in DNA repair pathways that may back-up each other and/or act preferably depending on chromatin context, DNA damage signalling pathway and various cellular regulation mechanisms. Study of the BER activities in HeLa cell extracts demonstrated that NTH1 is a major detectable DNA glycosylase activity towards 5OH-Hyd and 5OH-5Me-Hyd residues in DNA ( Figure 7). Surprisingly, we were not able to detect NEIL1 and NEIL2 activities using our hydantoin-DNA substrates possibly due to a strong 39-repair diesterase activity present in human cell-free extracts. Human FA cells appear to be a highly valuable model to study cellular response to endogenous oxidative DNA damage. Ambient oxygen induces chromosomal instability in FA cells suggesting impaired cellular defence against oxidative DNA damage, furthermore we have recently shown that FA cells have reduced amounts of NEIL1 [45]. Interestingly, here we demonstrated that FA cell-free extracts have slightly reduced BER incision activity towards 5OH-HydNG duplex oligonucleotide implying that NEIL1 may serve as a backup DNA glycosylase to repair pyrimidine-derived hydantoins (Supporting Information Figure S5). Human NTH1 protein has been shown to be able to initiate BER in nucleosome protected DNA [56], while NEIL1 and NEIL2 proteins excise oxidative base lesions in single-stranded and bubble DNA structures, suggesting their functions are coupled to DNA replication and/or transcription processes [57,58,59]. 5OH-Hyd and 5OH-5Me-Hyd residues are major oxidative pyrimidine lesions that accumulate in ancient DNA and may also accumulate during long chronic exposure to oxidizing agents [29]. Therefore, it is tempting to speculate that pyrimidine-derived hydantoins in non-transcribed heterochromatin DNA regions are main targets to NTH1 and APE1 but not to NEIL1, suggesting biological function of the NTH1-catalyzed BER and the APE1-NIR in the global genome repair pathway for pyrimidine-derived hydantoins elsewhere in genome.
The siRNAs sequences used to decrease APE1 and NTH1 in HeLa cells have been taken from previously described studies [42,61]. The siRNA specific to mouse major AP endonuclease, APEX, was used as negative control in both cases. Collection of the purified DNA glycosylases and AP endonucleases and human FEN1 was from the laboratory stock. The purified human POLb and LIG1 proteins were purchased from Trevigen (Gaithersburg, USA) and Enzymax (Lexington, USA), respectively. Polyclonal rabbit antibodies against human APE1 and NTH1 proteins were from Eurogentec and Alpha Diagnostic (Interchim, Montlucon, France), respectively.

DNA repair assays
The standard reaction mixture (20 mL) contained 5 nM of 32 Plabelled 5OH-HydNG and 5OH-5Me-HydNA oligonucleotide duplexes. Assays for the DNA glycosylases activities were performed in a ''BER+EDTA'' buffer containing 50 mM KCl, 20 mM HEPES-KOH (pH 7.6), 0.1 mg?mL 21 BSA, 1 mM DTT and 1 mM EDTA and 2.5 nM of a pure protein, unless otherwise stated. The release of 5OH-Hyd and 5OH-5Me-Hyd base adducts was measured by the cleavage of the oligonucleotide containing a single lesion at a defined position. For the monofunctional DNA glycosylases, the abasic sites left after damaged base excision action were cleaved by light piperidine treatment (10% (v/v) piperidine at 37uC for 10 min).
To measure kinetic parameters, range of duplex oligonucleotide substrate concentrations varied from 0.1 to 10 times the K M (generally from 2.5 nM to 250 nM) were incubated under respective standard reaction conditions with limiting amounts of enzymes (0.25 nM Nfo, 0.5 nM Apn1, 0.5 nM APE1, 2.5 nM NEIL1 and 20 nM NTH1) for 5-10 min at 37uC, unless otherwise stated. For K M and k cat determination, the linear velocities were plotted against substrate concentration and the hyperbolic curve obtained fit to a rectangular hyperbola by least-squares non-linear regression method. Apparent values were obtained for the Michaelis constant, K M , and the V max for cleavage; k cat was calculated by dividing the V max by the enzyme concentration. At least three independent experiments were performed for all analyses.
The standard reaction mixture (20 mL) for the NIR activity contained 50 mM KCl, 20 mM HEPES-KOH (pH 7.6), 0.1 mgNmL 21 BSA and 1 mM DTT for the Nfo protein or the same buffer supplemented with 5 mM MgCl 2 for the Apn1 protein. The activity of APE1 protein was tested either in the ''BER+Mg 2+ '' buffer containing 100 mM KCl, 20 mM HEPES-KOH (pH 7.6), 0.1 mgNmL 21 BSA, 1 mM DTT and 5 mM MgCl 2 which is optimal for its AP endonuclease activity or in the ''NIR buffer'' containing 50 mM KCl, 20 mM HEPES-KOH (pH 6.9), 0.1 mgNmL 21 BSA, 1 mM DTT and 0.1 mM MgCl 2 which is optimal for its nucleotide incision activity. When measuring APE1-NIR activity in the cell-free extracts MgCl 2 was replaced with ZnCl 2 [14].
The standard incision assay in cell-free extracts (20 ml) was performed with 5 nM 32 P-labelled DNA substrate either in the standard DNA glycosylase BER+EDTA reaction buffer: 50 mM KCl, 20 mM HEPES-KOH (pH 7.6), 0.1 mgNmL 21 BSA, 1 mM DTT and 1 mM EDTA or in the zinc-containing ''NIR+Zn 2+ buffer'' optimal for the APE1-catalyzed NIR activity in extracts: 50 mM KCl, 20 mM HEPES-KOH (pH 6.8), 0.1 mgNmL 21 BSA and 1 mM DTT and 0.1 mM ZnCl 2 [14]. Reaction mixtures were supplemented with either 0.5 mg of human cell-free extract or 3 mg of E. coli cell extracts and incubated for 60 min at 37uC, unless otherwise stated.

MALDI-TOF mass spectrometry analyses of the NIR pathway
Typically, 40 pmol of lesion containing oligonucleotide duplexes (in 20 mL) were incubated with the AP endonucleases (100 ng of APE1 and/or Nfo) in the appropriate ''NIR reaction buffer'' (see above) at 37uC for 1 h. The products were desalted on a MicroSpin G-25 column (GE Healthcare), prior subjection to the MALDI-TOF MS measurements. The latter MALDI mass spectra in the negative mode were obtained on a time-of-flight Biflex mass spectrometer (Bruker, Wissembourg, France) equipped with a 337 nm nitrogen laser and pulsed delay source extraction. The matrix was prepared by dissolving 3-hydroxypicolinic acid in 10 mM ammonium citrate buffer and a small amount of Dowex-50 W 5068-200 cation exchange resin (Sigma). Sample (1 mL) was added to matrix (1 mL) and the resulting solution was made homogeneous by stirring. The resulting sample was placed on the target plate and allowed to dry. Spectra were calibrated using reference oligonucleotides of known masses.
Cell-free extracts were prepared as previously described [3]. Briefly, E. coli cultures were grown to an OD 600 of 1.0 in the presence or absence of 0.25 mg/mL paraquat (+P) to induce Nfo [62]. Following cells were harvested by centrifugation, washed and lysed with 10 mg/mL lysozyme in a buffer containing 0.1 M HEPES-KOH (pH 7.5), 400 mM KCl, 0.1 mM EDTA. Cell lysates were clarified by centrifugation at 14000 rpm for 15 min at 4uC and then aliquoted and stored at -80uC until use.