Gp

Ja

b

a

ARRAA

KADOABD

1

DdtpDfoaltbic

mt1sMd

1d

DNA Repair 9 (2010) 690–699

Contents lists available at ScienceDirect

DNA Repair

journa l homepage: www.e lsev ier .com/ locate /dnarepai r

enetic interactions between HNT3/Aprataxin and RAD27/FEN1 suggestarallel pathways for 5′ end processing during base excision repair

ames M. Daleya, Thomas E. Wilsonb, Dindial Ramotara,∗

Centre de Recherche, Hôpital Maisonneuve-Rosemont, Université de Montréal, Montréal, QC H1T 2M4, CanadaDepartment of Pathology, University of Michigan Medical School, Ann Arbor, MI 48109, USA

r t i c l e i n f o

rticle history:eceived 5 January 2010eceived in revised form 26 February 2010ccepted 22 March 2010vailable online 15 April 2010

eywords:

a b s t r a c t

Mutations in Aprataxin cause the neurodegenerative syndrome ataxia oculomotor apraxia type 1.Aprataxin catalyzes removal of adenosine monophosphate (AMP) from the 5′ end of a DNA strand, whichresults from an aborted attempt to ligate a strand break containing a damaged end. To gain insightinto which DNA lesions are substrates for Aprataxin action in vivo, we deleted the Saccharomyces cere-visiae HNT3 gene, which encodes the Aprataxin homolog, in combination with known DNA repair genes.While hnt3� single mutants were not sensitive to DNA damaging agents, loss of HNT3 caused syner-

prataxinNA repairxidative DNA damagetaxia oculomotor apraxia type 1ase excision repairNA ligase

gistic sensitivity to H2O2 in backgrounds that accumulate strand breaks with blocked termini, includingapn1� apn2� tpp1� and ntg1� ntg2� ogg1�. Loss of HNT3 in rad27� cells, which are deficient inlong-patch base excision repair (LP-BER), resulted in synergistic sensitivity to H2O2 and MMS, indicatingthat Hnt3 and LP-BER provide parallel pathways for processing 5′ AMPs. Loss of HNT3 also increased thesister chromatid exchange frequency. Surprisingly, HNT3 deletion partially rescued H2O2 sensitivity inrecombination-deficient rad51� and rad52� cells, suggesting that Hnt3 promotes formation of a repair

ved b

intermediate that is resol

. Introduction

In all living cells, genome integrity is constantly challenged byNA damaging agents including reactive oxygen species generateduring metabolism, ultraviolet light from the sun, ionizing radia-ion, and various environmental chemicals. A variety of DNA repairathways exist to repair the diverse lesions caused by these agents.NA ligases are a common feature of all DNA repair pathways, per-

orming the final nick-sealing step in repair to restore the integrityf the chromosome. In eukaryotes, DNA ligation is initiated whendenosine monophosphate (AMP) from ATP is covalently linked to aysine residue in the active site of the ligase. AMP is then transferred

o the 5′ phosphate of the DNA strand break, and nucleophilic attacky the adjacent 3′ hydroxyl breaks the 5′ pyrophosphate bond, seal-

ng the break and releasing AMP [1]. DNA damaging agents oftenreate single- and double-strand breaks with blocked ends that

Abbreviations: AOA1, ataxia oculomotor apraxia type 1; AMP, adenosineonophosphate; BER, base excision repair; AT, ataxia telangiectasia; ATLD, ataxia

elangiectasia-like disorder; SCAN1, spinocerebellar ataxia with axonal neuropathy; HIT, histidine triad; NHEJ, nonhomologous end joining; MMS, methylmethaneulfonate; 4-NQO, 4-nitroquinoline-1-oxide; SCE, sister chromatid exchange; MRX,re11-Rad50-Xrs2; dRP, deoxyribosephosphate; SSB, single-strand break; DSB,

ouble-strand break.∗ Corresponding author. Tel.: +1 514 252 3400x4684; fax: +1 514 252 3430.

E-mail address: dramotar.hmr@ssss.gouv.qc.ca (D. Ramotar).

568-7864/$ – see front matter © 2010 Elsevier B.V. All rights reserved.oi:10.1016/j.dnarep.2010.03.006

y recombination.© 2010 Elsevier B.V. All rights reserved.

must be processed prior to ligation [2–6]. An end with a damaged3′ terminus adjacent to a 5′ phosphate may be a substrate for pre-mature attempted ligation, in which the absence of a 3′ hydroxylcauses the reaction to stall prior to the final step, leaving an adeny-lated 5′ phosphate.

Aprataxin reverses abortive ligation by removing the 5′ AMP,creating a 5′ phosphate and effectively resetting the break fora future ligation attempt [7]. Aprataxin was identified as thegene mutated in ataxia oculomotor apraxia type 1 (AOA1), aspinocerebellar ataxia that shares characteristics with knownDNA repair disorders such as ataxia telangiectasia (AT), ataxiatelangiectasia-like disorder (ATLD), and spinocerebellar ataxia withaxonal neuropathy 1 (SCAN1), caused by mutations in ATM, MRE11,and TDP1, respectively [8,9]. Aprataxin is a member of the histidinetriad (HIT) superfamily of nucleotide hydrolases and transferases,named for their characteristic H�H�H�� catalytic motif, where� is a hydrophobic amino acid [9]. It also contains an N-terminalforkhead associated (FHA) domain, which is thought to mediateinteractions with the ligase cofactors XRCC1 and XRCC4, and a C-terminal zinc finger domain implicated in DNA binding [10–12]. Inaddition to removing 5′ AMPs, Aprataxin has also been reported to

possess 3′ processing activity [13]. This activity is only observedat higher enzyme concentrations, however, suggesting that 5′ AMPremoval is the primary function of Aprataxin.

While the biochemical activity of Aprataxin has been well-characterized in vitro, how the protein functions in the context of

Repa

DatahaticCsAc

yfAfi[oiBcbipabatrrTtrt

2

2

ohaY(pfiJccbcFi

2

wpv

J.M. Daley et al. / DNA

NA repair in vivo is less clear. Whether cells lacking Aprataxinre sensitive to DNA damaging agents is controversial, indicatinghat redundant mechanisms may cooperate to repair abortive lig-tion intermediates [14–17]. Recent reports in which Aprataxinas been inactivated in addition to other repair deficiencies havettempted to address this issue. Caldecott and co-workers showedhat treatment of Aprataxin null cells with aphidicolin, whichnhibits DNA synthesis in long-patch base excision repair (LP-BER),auses a delay in repair of H2O2- and MMS-induced damage [16].ells mutant for both Aprataxin and Tdp1, a 3′ processing enzyme,howed a similar phenotype [18]. These data support the idea thatprataxin cooperates with other DNA repair mechanisms in pro-essing 5′ AMPs.

We sought to investigate the role of Aprataxin in vivo using theeast Saccharomyces cerevisiae as a model system, which allowedor rapid construction of mutants in multiple DNA repair pathways.prataxin-like proteins exist in lower eukaryotes, and of the HIT

amily members present in yeast, only Hnt3 can remove 5′ AMPsn vitro, suggesting that it is the functional homolog of Aprataxin7]. Here, we tested multiple DNA damaging agents and did notbserve sensitivity in hnt3� single mutants, but deletion of HNT3n strains defective for 3′ end processing (apn1� apn2� tpp1�), LP-ER (rad27�), and glycosylase/lyase activity (ntg1� ntg2� ogg1�)aused synergistic sensitivity to agents that cause single-strandreaks. The pattern of sensitivity in these strains allowed us to

dentify potential repair intermediates where abortive ligation isredicted to occur. Mutation of conserved residues in the Hnt3 HITnd Zn-finger domains enhanced H2O2 sensitivity in the rad27�ackground as much as HNT3 deletion, and human Aprataxin wasble to substitute for Hnt3 in yeast. We also report increased sis-er chromatid exchange in hnt3� cells. HNT3 deletion unexpectedlyescued the H2O2 sensitivity of recombination mutants rad51� andad52�, but enhanced H2O2 sensitivity in the rad50� background.o explain this, we propose that Hnt3 might promote recombina-ional repair of strand breaks with 5′ AMP ends. Loss of Hnt3 couldescue rad51� and rad52� H2O2 sensitivity by preventing forma-ion of a toxic intermediate that must be resolved by recombination.

. Materials and methods

.1. Yeast strains

S. cerevisiae strains were isogenic derivatives of the previ-usly described wild-type strains YW388 (MATa-inc ade2�0is3�200 leu2 met15�0 trp1�63 ura3�0) and YW465 (MAT˛de2�0 his3�200 leu2 met15�0 trp1�63 ura3�0) [19].W778 (apn1�::HIS3 apn2�::kanXM4 tpp1�::MET15), YW1715rad27�::KanMX4 carrying pTW572) and YW1807 (YW389 carryingTW572) were previously described [20,21]. All strains createdor this study are listed in Supplemental Table 1. Deletions werentroduced via the one-step gene replacement technique [22].DY102 (hnt3-H130A) and JDY103 (hnt3-CCYC(188:191)AAYA) wereonstructed by replacing URA3 in JDY76 using a tailed PCR productontaining the mutant allele, and the mutations were confirmedy sequencing. YW1621 was created by amplifying the HIS3 sisterhromatid exchange allele at the can1 locus (obtained from Mikeasullo) with flanking primers and transforming the PCR productnto YW465 [23].

.2. Survival curves and spotting assays

For survival curves, cells growing exponentially in YPAD mediaere washed twice with water, resuspended in 20 mM potassiumhosphate buffer (pH 7.5) at an OD600 of 0.500, and exposed toarying concentrations of H2O2, MMS or bleomycin at 30 ◦C with

ir 9 (2010) 690–699 691

shaking for 1 h. Survival was determined by plating serial dilu-tions to synthetic complete plates. In spotting assays, overnightYPAD cultures were diluted to an OD600 of 0.500 and 10-foldserial dilutions were spotted to YPD plates containing the indi-cated concentration of the agent. Briefly, cultures were diluted to10,000 cells/ml in YPAD containing the indicated drug concentra-tion. After 28 h of growth, survival was determined by plating tononselective media. At this time, untreated wild-type cultures werenearing the end of the exponential growth phase.

2.3. Sister chromatid exchange frequency

A previously described HIS3-based assay was used to measurethe sister chromatid exchange frequency [23]. Cells were eithermock-treated or treated with H2O2 as described above and platedto synthetic complete media or media lacking histidine. The SCE fre-quency was calculated as the fraction of cells that grow on plateslacking histidine.

2.4. Construction of Aprataxin and Hnt3 expression plasmids

HNT3 or human APTX coding sequences were amplified fromwild-type yeast genomic DNA or a lambda library of human cDNA[24], respectively, using primers with tails homologous to pTW438.The PCR products were then gap repaired into the 2 � plas-mid pTW438 digested with HindIII (NEB) in the rad27� hnt3�yeast strain JDY107. This produced plasmids expressing C-terminalNLS-Myc tagged Hnt3 or Aprataxin from the ADH1 promoter.The plasmids were sequenced to verify the lack of mutations.Sequencing was done by the McGill University and Genome QuebecInnovation Centre. Primer sequences are available upon request.

2.5. NHEJ oligonucleotide-modified plasmid assay

OMPs were created as previously described [25]. Briefly, twopairs of annealed oligonucleotides designed to restore the ADE2coding sequence were ligated onto the ends of pTW423 digestedwith BglII and XhoI. Plasmids were purified by agarose gelelectrophoresis and the concentration was quantified by UV spec-trometry. For DSBs with 3′ hydroxyls, 5′ phosphates were added toOMPs with T4 polynucleotide kinase (NEB) followed by a secondround of purification. For construction of DSBs containing 3′ phos-phates, oligonucleotides were synthesized with 5′ phosphates toprevent removal of 3′ phosphates by the 3′ phosphatase activity ofT4 polynucleotide kinase. Oligonucleotide ligation was monitoredby primer extension, and 70–90% of the plasmid was typically lig-ated (data not shown). All strains used in the NHEJ assays carriedpTW572, a HIS3-marked RNH35 overexpression plasmid that com-plements the Okazaki fragment processing defects and reduces thetoxicity of the transformation procedure in rad27� cells [21,26].For joints with compatible overhangs (Fig. 6A), 100 ng OMP wasco-transformed with 100 ng of the ClaI-digested (Fig. 6) or uncut(Supplemental Fig. 1) control plasmid pTW571 [21]. For gappedjoints (Fig. 6B), which are joined less efficiently, 600 ng OMP wasused.

3. Results

3.1. Deletion of HNT3 increases sensitivity to H2O2 and MMSwhen 3′ processing is impaired

Like Aprataxin, Hnt3 can process stalled ligation intermediatesin vitro [7]. As a starting point for our analysis, we hypothesized thatabortive ligation might occur at BER intermediates that contain 5′

phosphates, but cannot be ligated due to a damaged 3′ terminusor base. Models for three such intermediates are shown in Fig. 1.

692 J.M. Daley et al. / DNA Repair 9 (2010) 690–699

F H2O2-b

Tticpp

ig. 1. Potential roles for Hnt3 in processing stalled ligation intermediates after (A)y NIR.

o determine whether Hnt3 is indeed important for processing of

hese substrates in vivo, we constructed an hnt3� strain and exam-ned sensitivity to the DNA damaging agents H2O2 and MMS. H2O2reates strand breaks with blocked 3′ termini, such as 3′ phos-hates, which might undergo abortive ligation if a 5′ phosphate isresent (Fig. 1A). MMS produces alkylated bases, which are rapidly

induced SSBs, (B) AP sites induced by MMS, and (C) oxidative base damage repaired

processed by glycosylases to yield AP sites [27]. AP sites are cleaved

by the AP endonucleases Apn1 and Apn2, forming strand breakswith 3′ hydroxyls and 5′ deoxyribosephosphates (dRPs) [28,29].Despite the presence of a 3′ hydroxyl and a 5′ phosphate, thisintermediate has been shown to undergo abortive ligation in vitrobecause it lacks an intact base [12] (Fig. 1B). As shown in Figs. 2A,

J.M. Daley et al. / DNA Repair 9 (2010) 690–699 693

Fig. 2. HNT3 deletion causes synthetic sensitivity to various DNA damaging agents in the apn1� apn2� tpp1� and ntg1 ntg2 ogg1 backgrounds. (A), (B), (D) and (E) Cellsw etermt (C) 10c

3ai

awpeTs(Heeatawarsasa

ere treated with H2O2 for 1 h at the indicated concentrations, and survival was dhe mean ± standard deviation of three independent experiments for each strain.oncentration of bleomycin.

B and E, hnt3� cells were not sensitive to either H2O2 or MMS, ingreement with recent reports showing wild-type levels of repairn Aprataxin-deficient human cells exposed to these agents [15,16].

The lack of sensitivity in hnt3� cells might be explained by rapidction of DNA repair enzymes on blocked 3′ ends. To test this idea,e deleted HNT3 in an apn1� apn2� tpp1� strain, in which 3′ endrocessing is strongly compromised due to the lack of 3′ phosphodi-sterase activities of Apn1 and Apn2 and 3′ phosphatase activity ofpp1 [20]. This strain would be expected to accumulate persistenttrand breaks with blocked 3′ termini upon treatment with H2O2Fig. 1A). The apn1� apn2� tpp1� strain was highly sensitive to2O2, in agreement with previous results, and this sensitivity wasxacerbated upon additional deletion of HNT3 (Fig. 2A) [20]. Theffect was relatively modest, however. We reasoned that loss of thebility to remove 5′ AMPs might not have a strong effect becausehe cells in which abortive ligation occurs are already likely to dies a result of their 3′ processing defect. We therefore investigatedhether HNT3 deletion would instead have a greater effect in the

pn1� apn2� double mutant, which possesses functional Tpp1 and

etains the ability to process 3′ phosphates. Indeed, loss of HNT3ensitized these cells to H2O2 (Fig. 2B). This was especially evidentt the highest dose, where the triple mutant was 18-fold more sen-itive than the double (Fig. 2B). In agreement with previous data,pn1� apn2� tpp1� cells were sensitive to bleomycin, another

ined relative to the untreated control as described in Section 2. Curves represent-fold serial dilutions of cells were spotted to YPD plates containing the indicated

agent that causes strand breaks with blocked 3′ termini, but lossof HNT3 had no additional effect (Fig. 2C) [20]. This implies that aspecific type of damage induced by H2O2 but not bleomycin is asubstrate for abortive ligation.

Aprataxin has been shown to possess 3′ processing activity inaddition to its 5′ AMP removal activity, but whether yeast Hnt3 alsohas this activity is unknown [13]. To further investigate whetherloss of this activity might explain the enhanced H2O2 sensitivityobserved in 3′ processing-deficient strains, we introduced an Hnt3overexpression plasmid into apn1� apn2� tpp1� cells. If Hnt3is indeed a 3′ processing enzyme, overproducing it might restoreresistance to H2O2 in this strain. This was not the case, however,as the Hnt3 plasmid had no effect compared to the vector alone(Fig. 2D). This result is consistent with the 5′ AMP processing activ-ity of Hnt3 primarily accounting for the phenotypes we observed.

3.2. HNT3 deletion sensitizes ntg1� ntg2� ogg1� cells to H2O2,but not MMS

In the nucleotide incision repair (NIR) pathway, the endonucle-ase activity of Apn1 cleaves upstream of an oxidized base, leavinga strand break with a 3′ hydroxyl adjacent to a flap containing a 5′

phosphate and the damaged base, a potential substrate for abortiveligation (Fig. 1C) [30]. To test this possibility, we deleted HNT3 in an

694 J.M. Daley et al. / DNA Repair 9 (2010) 690–699

Fig. 3. rad27� hnt3� strains show synergistic sensitivity to H2O2, t-BH, and MMS, but not other DNA damaging agents. Survival after (A) H2O2, (B) MMS, (C) t-BH and (D)bleomycin A5 treatment was determined as in Fig. 2A. (E) Cells were spotted to YPD plates containing the indicated concentration of the agent.

Repair 9 (2010) 690–699 695

nttwsa

cchoaraAstcF

3a

ahdgouo(vwpaaiHa

lDtasttnse

3c

mcraiitrtd

Fig. 4. Hnt3 point mutations and complementation with human Aprataxin. (A) Pointmutations in the Hnt3 HIT and Zn-finger domains cause a null phenotype. Survival

J.M. Daley et al. / DNA

tg1� ntg2� ogg1� background. These cells lack the glycosylaseshat cleave the N-glycosidic bond of oxidized bases, forcing the cello utilize the NIR pathway. The ntg1� ntg2� ogg1� hnt3� strainas indeed more sensitive to H2O2 than the ntg1� ntg2� ogg1�

train (Fig. 2E), supporting the idea that abortive ligation can occurt NIR intermediates.

Because the glycosylases also act as AP lyases, their deletionould drive repair of MMS-induced damage down the AP endonu-lease pathway, leading to accumulation of 5′ dRPs adjacent to 3′

ydroxyls and increased dependence on Hnt3. The ntg1� ntg2�gg1� strain was no more sensitive to MMS than the parent,nd loss of HNT3 had no further effect (data not shown). Thisesult is consistent with the idea that the AP lyases process onlysmall fraction of AP sites in vivo, and the 5′ dRP ends created byP endonucleases can be efficiently processed by LP-BER. In thiscenario, the 3′ hydroxyl terminus serves as a primer for DNA syn-hesis, and the resulting 5′ flap containing the AMP could then beleaved by the flap endonuclease Rad27, the homolog of humanEN1 [31] (Fig. 1).

.3. Hnt3 cooperates with long-patch BER in repair of oxidationnd alkylation damage

To directly test the possibility that Hnt3 and LP-BER act in par-llel to process 5′ AMP-containing SSBs, we constructed rad27�nt3� strains and treated these with H2O2 and MMS. Indeed, theouble mutant was more sensitive to both agents than either sin-le mutant (Fig. 3A, B and E), implying that abortive ligation canccur during processing of both H2O2- and MMS-induced lesionsnder conditions where 3′ processing is not impaired. While lossf HNT3 caused robust sensitivity to H2O2 in a rad27� backgroundFig. 3A), the effect was less pronounced with MMS when testedia both transient and chronic exposure (Fig. 3B and E). To testhether loss of HNT3 affects activation of the DNA damage check-oint, we measured Rad53 phosphorylation by Western blottingnd cell cycle progression by FACS in wild-type, hnt3�, rad27�nd rad27� hnt3� cells after H2O2 treatment. rad27� cells showedncreased Rad53 phosphorylation and a prolonged S-phase, butNT3 deletion had no additional effects, indicating that checkpointctivation occurs normally in the absence of Hnt3 (data not shown).

To gain further insight into the types of lesions at which abortiveigation occurs, we treated rad27� hnt3� cells with a variety ofNA damaging agents. The double mutant was synthetic sensitive

o tert-butyl hydroperoxide, an oxidizing agent thought to cre-te damage similar to H2O2 (Fig. 3C) [32,33]. We did not observeynthetic sensitivity to bleomycin, 4-NQO, a UV-mimetic agenthat produces base adducts, or the topoisomerase 1 poison camp-othecin (Fig. 3D and E). These data suggest that these agents doot produce DNA structures that undergo failed ligation at a sub-tantial level, or that any such intermediates are rapidly processedven in the absence of HNT3 and LP-BER.

.4. Point mutations in the HNT3 catalytic and Zn-finger domainsonfer a null phenotype

Human Aprataxin contains an N-terminal FHA domain thatediates interactions with other DNA repair proteins, a central

atalytic HIT domain, and a C-terminal zinc finger domain that isequired for DNA binding [8,12]. The HIT and Zn-finger domainsre conserved in S. cerevisiae Hnt3, but not the FHA domain. Wentroduced HNT3 point mutations at the chromosomal locus to

nvestigate the roles of these domains in DNA repair in vivo. Inhe HIT domain, histidine 130 was mutated to alanine. The cor-esponding mutation in Aprataxin eliminates catalytic activity ofhe enzyme in vitro [7]. Alanine substitutions were also intro-uced for conserved cysteines 188, 189 and 190, which presumably

after H2O2 treatment was determined as in Fig. 2A. (B) Expression of Myc-NLS-tagged human Aprataxin from a plasmid complements HNT3 deletion. Survival afterH2O2 treatment was determined as in Fig. 2A, except the strains were grown insynthetic complete media lacking uracil to maintain selection for the plasmid.

coordinate the Zn2+ ion in the Zn-finger. Analogous Aprataxin Zn-finger mutations eliminate DNA binding [12]. In combination withrad27�, both hnt3-H130A and hnt3-CCYC(188:191)AAYA were assensitive to H2O2 as hnt3� (Fig. 4A), indicating that the HIT andZn-finger domains are required for Hnt3 function in vivo.

3.5. Human Aprataxin can complement hnt3� in yeast

To determine whether human Aprataxin can substitute for Hnt3in yeast, we introduced a plasmid into the rad27� hnt3� strain inwhich C-terminal Myc-NLS-tagged Aprataxin is expressed from theADH1 promoter. An HNT3 expression plasmid with the same tagwas used as a control. Expression of Aprataxin restored the H2O2resistance to a similar extent as HNT3, indicating that Aprataxin cansubstitute for Hnt3 in vivo (Fig. 4B). These data further establish thatHnt3 and Aprataxin are indeed homologs.

3.6. HNT3 deletion does not affect the mutation rate, butincreases spontaneous and induced sister chromatid exchange

Loss of DNA repair genes often causes a mutator phenotype dueto increased use of error-prone lesion bypass and backup repairmechanisms. We therefore used the canavanine forward mutationassay to measure the spontaneous mutation rate in hnt3� andrad27� hnt3� strains. As previously reported, the mutation rate

696 J.M. Daley et al. / DNA Repair 9 (2010) 690–699

Fig. 5. hnt3� cells show increased sister chromatid exchange, but this effect is lost in rad27� and apn1� apn2� tpp1� backgrounds. SCE frequencies were calculated asdescribed in Section 2 after mock-treatment (A) or treatment with the indicated concentrations of H2O2 (B). (C)–(E) HNT3 deletion sensitizes rad50� cells to H2O2, but rescuesr ig. 2Ab on of

wemTpR

wbTwIet(

ad51� and rad52� strains. Survival after H2O2 treatment was determined as in Fackgrounds. Cells were spotted to YPD plates containing the indicated concentrati

as increased in rad27� yeast, but HNT3 deletion had no furtherffect (Supplemental Table 2) [34]. In addition, the H2O2-inducedutation frequency was not affected by loss of HNT3 (Supplemental

able 2). These data indicate that the mechanisms utilized to repairersistent adenylated strand breaks in the absence of HNT3 andAD27 are not mutagenic.

SSBs with 5′ AMPs could be converted to DSBs upon collisionith replication forks. Such DSBs would be expected to be repaired

y recombination with the newly replicated intact sister chromatid.o test this hypothesis, we used a previously described system in

hich two partial fragments of HIS3 are placed on chromosome

V, and full-length HIS3 can only be restored by a sister chromatidxchange (SCE) event [23]. In this assay, the hnt3� strain spon-aneously gave rise to 4.8-fold more His+ colonies than wild-typeFig. 5A). Upon treatment with H2O2, the SCE frequency was higher

. (F) HNT3 deletion does not affect MMS sensitivity in rad50�, rad51�, or rad52�MMS.

in both wild-type and hnt3�, and as above, deletion of HNT3 led toa 3.6-fold increase in His+ colonies (Fig. 5B). Therefore, loss of HNT3leads to higher levels of recombination, likely caused by persistentSSBs.

We next measured SCE events in rad27� and apn1� apn2�tpp1� strains. While the spontaneous and H2O2-induced SCE fre-quencies were elevated in both compared to wild-type, additionaldeletion of hnt3� had no further effect (Fig. 5A and B). Thus, loss ofLP-BER or 3′ processing masks the hnt3� effect that is observed ina wild-type background (see Section 4). To gain further insight into

the connection between Hnt3 and recombination, we attempted toquantify Rad52-GFP foci in wild-type, hnt3�, rad27� and rad27�hnt3� cells, but found that H2O2 did not induce Rad52 foci evenat lethal doses and after prolonged incubation (data not shown).Spontaneous Rad52 foci were visible in ∼30% of rad27� cells as

Repair 9 (2010) 690–699 697

ce

3as

towwpaft[rtbSitDfactitoiaA

3e

kitpaoiloAmwBsLuhccaaRHitgg

Fig. 6. NHEJ of DSBs with ligation-blocking 3′ phosphates or gaps does not requireHnt3. (A) and (B) OMPs were used to create DSBs with the indicated structures(see Methods). Repair is expressed as the ratio of Ade+ colonies, which arise fromaccurate repair of the OMP, to Leu+ colonies, which result from co-transformationof a control plasmid digested with ClaI. A reduction in this ratio indicates a defect

J.M. Daley et al. / DNA

ompared to ∼1% of wild-type cells, but loss of HNT3 had no furtherffect in either background (data not shown).

.7. Loss of HNT3 partially rescues H2O2 sensitivity in rad51�nd rad52� mutants, but increases the sensitivity of a rad50�train

The observation that SCE is increased in hnt3� yeast suggestshat recombination acts as an alternative mechanism for repairf 5′ adenylated strand breaks. To further test this hypothesis,e deleted HNT3 in rad50�, rad51�, and rad52� mutants, inhich different steps of recombination are impaired. Rad50 isart of the Mre11/Rad50/Xrs2 (MRX) complex, which has nucle-se activity implicated in 5′ resection of DSBs and is also importantor checkpoint activation [35–38]. After resection, Rad51 coatshe newly generated 3′ ssDNA tails, promoting strand exchange39]. Rad52 promotes pairing of complementary strands and isequired for all forms of recombination [40]. We predicted thathe double mutants would show increased sensitivity to H2O2,ut this was only observed for the rad50� hnt3� strain (Fig. 5C).urprisingly, HNT3 deletion partially rescued sensitivity to H2O2n the rad51� and rad52� backgrounds (Fig. 5D and E, respec-ively). These data imply that Hnt3 activity at replication-associatedSBs is toxic in the absence of Rad51 or Rad52. Parallel roles

or Hnt3 and MRX-catalyzed resection in 5′ AMP removal couldccount for the synthetic sensitivity observed in rad50� hnt3�ells. To explain the rescue of rad51� and rad52�, we speculatehat Hnt3 may act to promote recombination upstream of strandnvasion and exchange catalyzed by Rad51 and Rad52 (see Sec-ion 4). Loss of HNT3 had no effect on MMS sensitivity in rad50�r rad52� backgrounds (Fig. 5F), indicating that this phenomenons specific to H2O2. Thus, Hnt3 may affect recombination medi-ted by H2O2-induced single-strand breaks but not MMS-inducedP sites.

.8. Lack of evidence for Hnt3 involvement in nonhomologousnd joining

The nonhomologous end joining (NHEJ) pathway has long beennown to rejoin DSBs with damaged or mismatched termini, mak-ng it a likely scenario for abortive ligation [21,41]. The observationhat Aprataxin interacts with XRCC4, which forms a stable com-lex with DNA ligase IV and is required for NHEJ, also suggestsrole for Aprataxin in NHEJ [11]. To test this directly, we used

ligonucleotide-modified plasmids (OMPs) to ask whether Hnt3s required for NHEJ of DSBs that are likely substrates for failedigation [25]. In this system, annealed oligonucleotides are ligatednto the ends of a linearized plasmid in an essential region of theDE2 gene, allowing selection for a specific joint. A control plas-id digested with ClaI within the LEU2 gene is co-transformedith the OMP and is repaired by “simple religation” NHEJ [21].ecause this DSB does not require end processing and is not aubstrate for abortive ligation, a reduction in the ratio of Ade+ toeu+ colonies indicates a specific defect in processed NHEJ. Wesed this system to create a panel of DSBs with 4-base 3′ over-angs in which the annealed bases were kept constant. DSBs wereonstructed to contain 5′ phosphates adjacent to each possibleombination of 3′ phosphates and 1-nt gaps (Fig. 6). Premature 5′

denylation would be expected to occur only if the presence ofgap or 3′ phosphate causes an abortive ligation attempt. Since

ad27 has been implicated in NHEJ and functions in parallel with

nt3 in BER, we tested hnt3�, rad27� and hnt3� rad27� strains

n addition to wild-type and pol4� controls [21,42,43]. None ofhe strains we tested showed a significant defect at DSBs with noaps, regardless of the presence of a 3′ phosphate (Fig. 6A). Atapped DSBs, the rad27� strain showed a partial defect when a

in processed NHEJ. Each bar represents the mean ± standard deviation from threeindependent transformations.

3′ phosphate was present, but loss of HNT3 had no additional effect(Fig. 6B). Additive effects at processed joints were not observedwhen the experiment was repeated using a supercoiled plasmidcontrol, although these data were complicated by the low andhighly variable rate at which rad27� strains are transformed withcircular plasmids (Supplemental Fig. 1). These data indicate thatDSBs with blocked termini can be efficiently rejoined in the absenceof both Rad27 and Hnt3. Thus, while this does not rule out a role forHnt3 in NHEJ, it shows that redundant mechanisms exist for 5′ endprocessing.

4. Discussion

The ability of Aprataxin to remove 5′ AMPs has been well-characterized in vitro, but how this activity functions in the contextof DNA repair in vivo remains largely unknown [7,12,44]. To addressthis issue, we used yeast genetics to characterize the role of Hnt3,

the homolog of Aprataxin, in DNA repair. We find that HNT3 dele-tion increases sensitivity to DNA damaging agents in mutants thataccumulate potential abortive ligation substrates. Hnt3 cooperateswith Rad27-dependent LP-BER in repair of oxidative and alkylationdamage, an idea supported by recent mammalian data [16]. Finally,

6 Repa

wr

4p

dbrtTcbn(cflwpeaBfcs

4l

aoshatpa(ac

Sse(anU[aca

tch[Ab5o3oa

98 J.M. Daley et al. / DNA

e present genetic evidence that Hnt3 also affects recombinationalepair of H2O2-induced damage.

.1. Hnt3 and Rad27-dependent LP-BER provide parallelathways for processing adenylated strand breaks

It was recently shown that repair of H2O2- and MMS-inducedamage is delayed when long-patch DNA synthesis is inhibitedy aphidicolin in cells lacking Aprataxin [16]. Our finding thatad27� hnt3� double mutant yeast show synergistic sensitivityo both of these agents (Fig. 3) extends and confirms this result.ogether, these data support a model in which adenylated 5′ endsan be processed either via direct reversal by Aprataxin/Hnt3 ory flap cleavage in LP-BER (Fig. 1A). Overexpression of Hnt3 didot increase the H2O2 resistance above the level of a rad27� strainFig. 4), suggesting that Rad27 performs functions that cannot beomplemented by Hnt3, such as cleavage of non-adenylated 5′

aps. Our finding that hnt3� cells are not hypersensitive to H2O2as unexpected given that extracts from hnt3� yeast cannot repairlasmids damaged by direct treatment with H2O2 [7]. A likelyxplanation for this discrepancy is that LP-BER is active in vivo, butplasmid with numerous SSBs is probably a poor substrate for LP-ER. A shattered plasmid may lack stably paired 3′ hydroxyl termini

rom which to initiate repair synthesis, and yeast extracts may notontain active DNA replication machinery required for long-patchynthesis.

.2. Identification of DNA lesions that are substrates for abortiveigation in vivo

If the primary role of Aprataxin/Hnt3 is indeed processing ofdenylated 5′ ends, synthetic sensitivity upon loss of Hnt3 shouldnly occur when cells are treated with agents that induce DNAtructures that undergo abortive ligation. Any strand break thatas a 5′ phosphate but cannot be ligated due to the absence of3′ hydroxyl or an intact base could potentially be a failed liga-

ion substrate. These could include 3′ phosphates, aldehydes andhosphoglycolates, as well as 5′ dRPs. Our genetic data implicatebortive ligation at single-strand breaks with damaged 3′ terminiFig. 1A), strand breaks with 3′ hydroxyls and 5′ dRPs that occurs intermediates in AP site processing (Fig. 1B), and oxidized basesleaved by Apn1 during NIR (Fig. 1C).

The Schizosaccharomyces pombe homolog of Aprataxin,PCC18.09c, was recently identified in a screen for 4-NQOensitivity, but we did not observe sensitivity to this agent inither the hnt3� single mutant or in combination with rad27�Fig. 3B) [45]. Therefore, processing of 4-NQO damage may producen intermediate that can undergo abortive ligation in S. pombe butot S. cerevisiae. The presence of the UV damage endonucleaseve1 in S. pombe but not S. cerevisiae might explain this difference

46,47]. Uve1 incises 5′ of UV-induced photoproducts and bulkydducts created by 4-NQO, yielding a 3′ hydroxyl adjacent to a flapontaining the lesion and a 5′ phosphate, a potential substrate forbortive ligation.

In addition to removing 5′ AMPs, Aprataxin has been reportedo process 3′ phosphate and 3′ phosphoglycolate ends [13]. Pro-essing of adenylated 5′ ends by Aprataxin is much more efficient,owever, likely indicating that this is its physiological function7]. Whether Hnt3 has 3′ processing activity is currently unknown.lthough many of the phenotypes we observed could be explainedy 3′ processing defects, we favor the interpretation that impaired

′ AMP processing is principally responsible for these data, sinceverexpression of Hnt3 had no effect on the H2O2 resistance of′ processing-deficient apn1� apn2� tpp1� cells (Fig. 2D). Thebservations that apn1� tpp1� rad1� yeast are slow growing andpn1� apn2� rad1� cells are inviable due to accumulation of 3′

ir 9 (2010) 690–699

blocking lesions also argue against Hnt3 providing additional 3′

processing activity in yeast [48].

4.3. HNT3 deletion increases the SCE frequency in wild-type butnot repair-deficient backgrounds

hnt3� Yeast showed increased spontaneous and H2O2-inducedSCEs, but this effect was masked in rad27� and apn1� apn2�tpp1� backgrounds (Fig. 5A and B). The SCE frequency in bothof these backgrounds is increased relative to wild-type (Fig. 5Aand B), likely the result of persistent SSBs stalled at 5′ or 3′ pro-cessing steps, respectively. We reasoned that HNT3 deletion maynot further increase the SCE frequency in these strains if Hnt3acts at a subset of strand breaks that can be repaired by BER ina wild-type background, but are unrepairable except by recom-bination in the absence of 5′ flap endonuclease or 3′ processingactivities. Such breaks are therefore committed to recombinationalrepair regardless of whether abortive ligation occurs, explainingwhy HNT3 deletion does not further increase the SCE frequency inthese backgrounds. The observations that RAD52 deletion is syn-thetic lethal with both rad27� and apn1� apn2� tpp1� supportthis idea [20,34].

4.4. Genetic interactions between HNT3 and recombination genes

The inviability of rad52� rad27� and rad52� apn1� apn2�tpp1� yeast implies that recombination serves as a backup path-way for repair of oxidative DNA damage, likely after SSBs areconverted to DSBs by replication fork collision [20,34]. We there-fore predicted that HNT3 deletion would sensitize recombinationmutants to oxidizing agents such as H2O2. Our observation that lossof HNT3 partially rescued rad51� and rad52� yeast while sensitiz-ing rad50� cells to H2O2 was therefore surprising (Fig. 5C–E). Thisdifference may be explained by the step in recombination affectedby each mutation. Rad50 is part of the Mre11-Rad50-Xrs2 (MRX)complex, which is involved in 5′ resection of DSBs, the first step inrecombination and a likely mechanism by which the 5′ AMP couldbe removed [35–37]. Thus, synthetic sensitivity in hnt3� rad50�yeast may be explained by the loss of two parallel pathways for 5′

AMP removal. After resection generates 3′ ssDNA tails, Rad51 andRad52 catalyze strand invasion and exchange with the templateduplex [49]. Removal of the 5′ AMP by Hnt3 might commit the DSBto repair by recombination, perhaps by generating a DNA structurethat requires the activity of Rad51 and Rad52 for resolution. In thisscenario, HNT3 deletion would rescue rad51� and rad52� cells bypreventing formation of this intermediate. Alternative mechanismsof 5′ AMP removal might allow recombination-independent path-ways such as post-replication repair or NHEJ to resolve the damage.While we were unable to identify a requirement for Hnt3 in NHEJ(Fig. 6), it remains possible that Hnt3 participates in NHEJ in parallelwith unidentified nucleases, or at specific DSB structures that werenot included in our analysis. Future work will focus on identifyingand characterizing novel biochemical activities of Hnt3 that mightprovide a more detailed mechanistic explanation for these data.

In summary, the data presented here establish a role forHnt3 in processing abortive ligation intermediates in parallel withRad27-dependent LP-BER. We identify multiple BER intermediatesincluding SSBs with blocked 3′ ends and 5′ dRPs formed after APsite cleavage as potential abortive ligation substrates in vivo, andwe provide genetic evidence that Hnt3 affects downstream DSBrepair by affecting recombination. This work establishes roles for

Hnt3 in multiple DNA repair pathways in vivo.

Conflict of interest

There is no conflict of interest.

Repa

A

osT

A

t

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

J.M. Daley et al. / DNA

cknowledgments

We wish to thank Dr. M. Saparbaev, Dr. E.B. Affar and membersf the Ramotar laboratory for their helpful comments. J.M.D. wasupported by a fellowship from Maisonneuve-Rosemont Hospital.his work was supported by an NSERC grant (202432-01) to D.R.

ppendix A. Supplementary data

Supplementary data associated with this article can be found, inhe online version, at doi:10.1016/j.dnarep.2010.03.006.

eferences

[1] A.E. Tomkinson, et al., DNA ligases: structure, reaction mechanism, and func-tion, Chem. Rev. 106 (2) (2006) 687–699.

[2] T. Coquerelle, et al., Strand breaks and K′ end-groups in DNA of irradiatedthymocytes, Int. J. Radiat. Biol. Relat. Stud. Phys. Chem. Med. 24 (4) (1973)397–404.

[3] W.D. Henner, S.M. Grunberg, W.A. Haseltine, Sites and structure of gammaradiation-induced DNA strand breaks, J. Biol. Chem. 257 (19) (1982)11750–11754.

[4] E. Pastwa, et al., Repair of radiation-induced DNA double-strand breaks isdependent upon radiation quality and the structural complexity of double-strand breaks, Radiat. Res. 159 (2) (2003) 251–261.

[5] E.C. Friedberg, G.C. Walker, W. Siede, R.D. Wood, R.A. Schultz, T. Ellenberger,DNA Repair and Mutagenesis., ASM Press, Washington, DC, 2005.

[6] D. Ramotar, et al., Cellular role of yeast Apn1 apurinic endonuclease/3′-diesterase: repair of oxidative and alkylation DNA damage and control ofspontaneous mutation, Mol. Cell. Biol. 11 (9) (1991) 4537–4544.

[7] I. Ahel, et al., The neurodegenerative disease protein aprataxin resolves abortiveDNA ligation intermediates, Nature 443 (7112) (2006) 713–716.

[8] M.C. Moreira, et al., The gene mutated in ataxia-ocular apraxia 1 encodesthe new HIT/Zn-finger protein aprataxin, Nat. Genet. 29 (2) (2001)189–193.

[9] H. Date, et al., Early-onset ataxia with ocular motor apraxia and hypoalbumine-mia is caused by mutations in a new HIT superfamily gene, Nat. Genet. 29 (2)(2001) 184–188.

10] H. Date, et al., The FHA domain of aprataxin interacts with the C-terminal region of XRCC1, Biochem. Biophys. Res. Commun. 325 (4) (2004)1279–1285.

11] P.M. Clements, et al., The ataxia-oculomotor apraxia 1 gene product has a roledistinct from ATM and interacts with the DNA strand break repair proteinsXRCC1 and XRCC4, DNA Repair (Amst) 3 (11) (2004) 1493–1502.

12] U. Rass, I. Ahel, S.C. West, Actions of aprataxin in multiple DNA repair pathways,J. Biol. Chem. 282 (13) (2007) 9469–9474.

13] T. Takahashi, et al., Aprataxin, causative gene product for EAOH/AOA1,repairs DNA single-strand breaks with damaged 3′-phosphateand 3′-phosphoglycolate ends, Nucleic Acids Res. 35 (11) (2007)3797–3809.

14] P. Mosesso, et al., The novel human gene aprataxin is directly involved in DNAsingle-strand-break repair, Cell. Mol. Life Sci. 62 (4) (2005) 485–491.

15] N. Gueven, et al., Aprataxin, a novel protein that protects against genotoxicstress, Hum. Mol. Genet. 13 (10) (2004) 1081–1093.

16] J.J. Reynolds, et al., Defective DNA ligation during short-patch single-strandbreak repair in ataxia oculomotor apraxia 1, Mol. Cell. Biol. 29 (5) (2009)1354–1362.

17] N. Gueven, et al., A subgroup of spinocerebellar ataxias defective in DNA dam-age responses, Neuroscience 145 (4) (2007) 1418–1425.

18] S.F. El-Khamisy, et al., Synergistic decrease of DNA single-strand break repairrates in mouse neural cells lacking both Tdp1 and aprataxin, DNA Repair (Amst)8 (6) (2009) 770–776.

19] J.R. Vance, T.E. Wilson, Uncoupling of 3′-phosphatase and 5′-kinase func-tions in budding yeast. Characterization of Saccharomyces cerevisiae DNA3′-phosphatase (TPP1), J. Biol. Chem. 276 (18) (2001) 15073–15081.

20] J.R. Vance, T.E. Wilson, Repair of DNA strand breaks by the overlapping functionsof lesion-specific and non-lesion-specific DNA 3′ phosphatases, Mol. Cell. Biol.21 (21) (2001) 7191–7198.

21] J.M. Daley, T.E. Wilson, Evidence that base stacking potential in annealed 3′

overhangs determines polymerase utilization in yeast nonhomologous endjoining, DNA Repair (Amst) 7 (1) (2008) 67–76.

[

[

ir 9 (2010) 690–699 699

22] C.B. Brachmann, et al., Designer deletion strains derived from Saccharomycescerevisiae S288C: a useful set of strains and plasmids for PCR-mediated genedisruption and other applications, Yeast 14 (2) (1998) 115–132.

23] M. Fasullo, T. Bennett, P. Dave, Expression of Saccharomyces cerevisiae MATaand MAT alpha enhances the HO endonuclease-stimulation of chromosomalrearrangements directed by his3 recombinational substrates, Mutat. Res. 433(1) (1999) 33–44.

24] S.J. Elledge, et al., Lambda YES: a multifunctional cDNA expression vector for theisolation of genes by complementation of yeast and Escherichia coli mutations,Proc. Natl. Acad. Sci. U.S.A. 88 (5) (1991) 1731–1735.

25] J.M. Daley, T.E. Wilson, Rejoining of DNA double-strand breaks as a function ofoverhang length, Mol. Cell. Biol. 25 (3) (2005) 896–906.

26] J. Qiu, et al., Saccharomyces cerevisiae RNase H(35) functions in RNA primerremoval during lagging-strand DNA synthesis, most efficiently in cooperationwith Rad27 nuclease, Mol. Cell. Biol. 19 (12) (1999) 8361–8371.

27] S. Boiteux, M. Guillet, Abasic sites in DNA: repair and biological consequencesin Saccharomyces cerevisiae, DNA Repair (Amst) 3 (1) (2004) 1–12.

28] R.A. Bennett, The Saccharomyces cerevisiae ETH1 gene, an inducible homologof exonuclease III that provides resistance to DNA-damaging agents and limitsspontaneous mutagenesis, Mol. Cell. Biol. 19 (3) (1999) 1800–1809.

29] R.E. Johnson, et al., Identification of APN2, the Saccharomyces cerevisiae homologof the major human AP endonuclease HAP1, and its role in the repair of abasicsites, Genes Dev. 12 (19) (1998) 3137–3143.

30] A.A. Ischenko, M.K. Saparbaev, Alternative nucleotide incision repair pathwayfor oxidative DNA damage, Nature 415 (6868) (2002) 183–187.

31] K. Kim, S. Biade, Y. Matsumoto, Involvement of flap endonuclease 1 in baseexcision DNA repair, J. Biol. Chem. 273 (15) (1998) 8842–8848.

32] A. Jilani, et al., Molecular cloning of the human gene, PNKP, encoding a polynu-cleotide kinase 3′-phosphatase and evidence for its role in repair of DNAstrand breaks caused by oxidative damage, J. Biol. Chem. 274 (34) (1999)24176–24186.

33] S.A. Altman, et al., tert-Butyl hydroperoxide-mediated DNA base damage incultured mammalian cells, Mutat. Res. 306 (1) (1994) 35–44.

34] D.X. Tishkoff, et al., A novel mutation avoidance mechanism dependent onS. cerevisiae RAD27 is distinct from DNA mismatch repair, Cell 88 (2) (1997)253–263.

35] Z. Zhu, et al., Sgs1 helicase and two nucleases Dna2 and Exo1 resect DNA double-strand break ends, Cell 134 (6) (2008) 981–994.

36] A. Jazayeri, et al., Mre11-Rad50-Nbs1-dependent processing of DNA breaksgenerates oligonucleotides that stimulate ATM activity, EMBO J. 27 (14) (2008)1953–1962.

37] E.P. Mimitou, L.S. Symington, Sae2, Exo1 and Sgs1 collaborate in DNA double-strand break processing, Nature 455 (7214) (2008) 770–774.

38] J.H. Lee, T.T. Paull, ATM activation by DNA double-strand breaks through theMre11-Rad50-Nbs1 complex, Science 308 (5721) (2005) 551–554.

39] T. Ogawa, et al., Similarity of the yeast RAD51 filament to the bacterial RecAfilament, Science 259 (5103) (1993) 1896–1899.

40] U.H. Mortensen, et al., DNA strand annealing is promoted by the yeast Rad52protein, Proc. Natl. Acad. Sci. U.S.A. 93 (20) (1996) 10729–10734.

41] T.E. Wilson, M.R. Lieber, Efficient processing of DNA ends during yeast nonho-mologous end joining. Evidence for a DNA polymerase beta (Pol4)-dependentpathway, J. Biol. Chem. 274 (33) (1999) 23599–23609.

42] X. Wu, T.E. Wilson, M.R. Lieber, A role for FEN-1 in nonhomologous DNA endjoining: the order of strand annealing and nucleolytic processing events, Proc.Natl. Acad. Sci. U.S.A. 96 (4) (1999) 1303–1308.

43] H.M. Tseng, A.E. Tomkinson, Processing and joining of DNA ends coordinatedby interactions among Dnl4/Lif1, Pol4, and FEN-1, J. Biol. Chem. 279 (46) (2004)47580–47588.

44] U. Rass, I. Ahel, S.C. West, Molecular mechanism of DNA deadenylation bythe neurological disease protein aprataxin, J. Biol. Chem. 283 (49) (2008)33994–34001.

45] G.P. Deshpande, et al., Screening a genome-wide S. pombe deletion library iden-tifies novel genes and pathways involved in genome stability maintenance,DNA Repair (Amst) 8 (5) (2009) 672–679.

46] K.K. Bowman, et al., A new ATP-independent DNA endonuclease from Schizosac-charomyces pombe that recognizes cyclobutane pyrimidine dimers and 6-4photoproducts, Nucleic Acids Res. 22 (15) (1994) 3026–3032.

47] M. Takao, et al., Characterization of a UV endonuclease gene from the fissionyeast Schizosaccharomyces pombe and its bacterial homolog, Nucleic Acids Res.

24 (7) (1996) 1267–1271.

48] A.S. Karumbati, et al., The role of yeast DNA 3′-phosphatase Tpp1 andrad1/Rad10 endonuclease in processing spontaneous and induced base lesions,J. Biol. Chem. 278 (33) (2003) 31434–31443.

49] S.C. West, Molecular views of recombination proteins and their control, Nat.Rev. Mol. Cell Biol. 4 (6) (2003) 435–445.