Polynucleotide Kinase and Aprataxin-like Forkhead-associatedProtein (PALF) Acts as Both a Single-stranded DNAEndonuclease and a Single-Stranded DNA 3� Exonuclease andCan Participate in DNA End Joining in a Biochemical System*□S

Received for publication, August 1, 2011, and in revised form, August 31, 2011 Published, JBC Papers in Press, September 1, 2011, DOI 10.1074/jbc.M111.287797

Sicong Li‡, Shin-ichiro Kanno§, Reiko Watanabe§, Hideaki Ogiwara¶, Takashi Kohno¶, Go Watanabe‡, Akira Yasui§,and Michael R. Lieber‡1

From the ‡Departments of Pathology, Biochemistry and Molecular Biology, Biological Sciences, and Molecular Microbiology andImmunology, Norris Comprehensive Cancer Center, Los Angeles, California 90089-9176, the §Division of Dynamic Proteome,Institute of Development, Aging, and Cancer, Tohoku University, Sendai 980-8575, Japan, and the ¶Division of Genome Biology,National Cancer Center Research Institute 1-1, Tokyo 104-0045, Japan

Background: PALF binds to the NHEJ protein XRCC4.Results: PALF is a single-stranded DNA endonuclease and 3� exonuclease. PALF can coordinate with known NHEJ proteins toachieve ligation.Conclusion: In vitro and in vivo, PALF is able to cooperate with NHEJ proteins to join double-stranded DNA breaks.Significance: A second nuclease, in addition to Artemis, can function with NHEJ proteins to achieve DNA end joining.

Polynucleotide kinase and aprataxin-like forkhead-associ-ated protein (PALF, also called aprataxin- and PNK-like factor(APLF)) has been shown to have nuclease activity and to use itsforkhead-associated domain to bind to x-ray repair comple-menting defective repair in Chinese hamster cells 4 (XRCC4).Because XRCC4 is a key component of the ligase IV complexthat is central to the nonhomologous DNA end joining (NHEJ)pathway, this raises the possibility that PALFmight play a role inNHEJ. For this reason, we further studied the nucleolytic prop-erties of PALF, and we searched for any modulation of PALF byNHEJ components. We verified that PALF has 3� exonucleaseactivity. However, PALF also possesses single-stranded DNAendonuclease activity. This single-stranded DNA endonucleaseactivity can act at all single-stranded sites except those withinfour nucleotides 3�of a double-strandedDNA junction, suggest-ing that PALF minimally requires approximately four nucleo-tides of single-strandedness. Ku, DNA-dependent proteinkinase catalytic subunit, and XRCC4-DNA ligase IV do notmodulate PALF nuclease activity on single-stranded DNA oroverhangs of duplex substrates. PALF does not open DNA hair-pins.However, in a reconstituted end joining assay that includesKu, XRCC4-DNA ligase IV, and PALF, PALF is able to resect 3�

overhanging nucleotides and permit XRCC4-DNA ligase IV tocomplete the joining process in a manner that is as efficient asArtemis. Reduction of PALF in vivo reduces the joining ofincompatible DNA ends. Hence, PALF can function in concertwith other NHEJ proteins.

PALF2 (also called APLF, C2orf13, and Xip1) is a recentlydescribed nuclease that was named for the fact that it sharesproperties with polynucleotide kinase (PNK) and aprataxin,hence the name PNK- and aprataxin-like factor (also APLF foraprataxin- and PNK-like factor) (1–4). The shared property isthe binding to XRCC4, a key component of the NHEJ ligasecomplex. PALF, PNK, and aprataxin also all have a forkhead-associated (FHA) domain. The FHA domain is the site of bind-ing to XRCC4 (5). The kinase CK2 phosphorylates XRCC4 (5),and this phosphorylated form of XRCC4 is necessary for theFHA domain of PALF, PNK, and aprataxin to bind XRCC4 (1,3). PALF has also been reported to bind to Ku (1, 3) as well asDNA-PKcs (1) and is important for cell survival after low-doseionizing radiation (4).XRCC4 is a key component of the XLF-XRCC4-DNA ligase

IV complex, the primary ligase for vertebrateNHEJ (6, 7). NHEJis the major pathway for repair of double-strand breaks andincludes the factors Ku, DNA-PKcs, Artemis, polymerase �,and polymerase � as well. The nuclease activities of NHEJ arethought to be largely provided by Artemis (8). Artemis hasendonucleolytic activity at all double- to single-strand transi-tions (9). For repair of double-strand breaks, 5� overhangs, 3�overhangs, and hairpin structures are key substrates ofArtemis.Artemis also appears to have 5� exonuclease activity, althoughmutagenesis has not yet definitively demonstrated this activityto be intrinsic toArtemis (10, 11). Artemis andDNA-PKcs forma tight complex, and Artemis endonucleolytic activity at dou-ble- to single-strand transitions requires activation of DNA-PKcs (8, 12). DNA-PKcs is a serine/threonine protein kinase

* This work was supported, in whole or in part, by a National Institutes of Healthgrant (to M. R. L.).

□S The on-line version of this article (available at http://www.jbc.org) containssupplemental Figs. 1– 4.

1 To whom correspondence should be addressed. Tel.: 323-865-0568; E-mail:lieber@usc.edu.

2 The abbreviations used are: PALF, polynucleotide kinase and aprataxin-likeforkhead-associated; PNK, polynucleotide kinase; NHEJ, nonhomologousDNA end joining; FHA, forkhead-associated; XRCC4, x-ray repair comple-menting defective repair in Chinese hamster cells 4; PKcs, DNA-dependentprotein kinase catalytic subunit; EGFP, enhanced green fluorescentprotein.

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 286, NO. 42, pp. 36368 –36377, October 21, 2011© 2011 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

36368 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 42 • OCTOBER 21, 2011

that is only activated by duplex DNA ends.WhenDNA-PKcs isactivated by duplex DNA ends, it uses ATP to autophosphoryl-ate itself andArtemis (8, 13). The autophosphorylation causes aconformational change inDNA-PKcs (14, 15), and this presum-ably causes a conformational change in Artemis to account forthe activation of the endonucleolytic properties ofArtemis (12).When the C-terminal portion of Artemis is removed, Artemisacquires endonucleolytic properties independent of DNA-PKcs (16, 17).The nucleolytic properties of PALF have been described as a

3� exonuclease and a double-stranded DNA endonuclease (1).PALF was also found to have nicking activity against variousDNA molecules containing abasic sites (1). Here we confirmthe 3� exonuclease activity of PALF (1). The endonucleolyticproperties of PALF appear to be more general than describedoriginally. PALF is able to endonucleolytically act on single-stranded DNA. The action described previously at double-stranded DNA ends may simply reflect breathing of those endsinto transient single-stranded tails. In contrast to Artemis,PALF is unaffected in its nucleolytic properties by DNA-PKcs.PALF is also unaffected by Ku or by XRCC4-DNA ligase IV.However, at incompatible ends with overhangs, we show thatpurified PALF is able to function in concert with purified Kuand XRCC4-DNA ligase IV in end joining reactions at an effi-ciency that is equivalent toArtemiswhenused in place of PALF.Finally, we find that in vivo siRNA knockdown of PALF resultsin a significant drop in the joining of incompatible DNA ends.Hence, PALF appears to be capable of participating in end join-ing with other NHEJ proteins.

EXPERIMENTAL PROCEDURES

Oligonucleotides and DNA Substrates—Oligonucleotidesused in this studywere synthesized byOperon Biotechnologies,Inc. (Huntsville, AL) and Integrated DNA Technologies, Inc.(SanDiego, CA).We purified the oligonucleotides using 12% or15% denaturing PAGE and determined the concentrationspectrophotometrically.DNA substrate 5� end labeling was done with [�-32P]ATP

(3000 Ci/mmol) (PerkinElmer Life Sciences, Boston, MA) andT4 polynucleotide kinase (New England Biolabs) according tothe manufacturer’s instructions. Substrates were incubatedwith [�-32P]ATP and T4 PNK for 30 min at 37 °C. T4 PNK wassubsequently inactivated by incubating samples at 72 °C for 20min. Unincorporated radioisotope was removed by using G-25Sephadex (Amersham Biosciences, Inc.) spin-column chroma-tography. For the hairpin substrate, YM164-labeled oligonu-cleotide was diluted in a buffer containing 10 mM Tris-hydro-chloride (pH 8.0), 1 mM EDTA (pH 8.0), and 100 mM NaCl,heated at 100 °C for 5min, allowed to cool to room temperaturefor 3 h, and then incubated at 4 °C overnight.The sequences of the oligonucleotides used in this study are

as follows: JG68, 5�-GAT CCT TCT GTA GGA CTC ATG-3�;JG169, 5�-TTT TTT TTT TTT TTT TTT TTT TTT TTTTTT-3�; YM164, 5�-TTT TTG ATT ACT ACG GTA GTAGCT ACG TAG CTA CTA CCG TAG TAA T-3�; YM130,5�-TTTTTTTTTTTTTTTACTGAGTCCTACAGAAGGATC-3�; YM149, 5�-ACT GAG TCC TAC AGA AGG ATCTTT TTT TTT TTT TTT-3�; YM8, 5�-AGG CTG TGT TAA

GTA TCT GCG CTC GCC CTC AGA GG-3�; YM9, 5�-CCTCTGAGGGCGAGCGCAGATACTTAACACAGCCT-3�;JG258, 5�-CGA GCC CGA TCC GCT TGA CCA GTA GTCTAG CAC GTG ACG ATT GCA TCC GTC AAG TAA GATGCA GAT ACT TAA CGG GG-3�; SL11, 5�-GTT AAG TATCTG CAT CTT ACT TGA CGG ATG CAA TCG TCA CGTGCT AGA CTA CTG GTC AAG CGG ATC GGG CTC GCCCCA AAA AA-3�; and SL15, 5�-ACT GAG TCC TAC AGAAGG ATC TTT TTT TTT TTS SSS. The sequence of siRNAfor PALF, 5�-CCA GAU GAC UCC CAC AAA UAG, was syn-thesized, annealedwith a complementary strand, and usedwitha final concentration of 20 �M. Antibody against PALF wasprepared as reported previously (1).Protein Expression and Purification—N-terminal His-tagged

PALF cloned into a pET-16b (Novagen) vector has beendescribed previously (1). Soluble His-PALF was expressed andpurified from pLysE BL21(DE3)-competent cells (Invitrogen).Cells were precultured in ampicillin untilA600 of 0.5. Cells werethen induced with isopropyl 1-thio-�-D-galactopyranoside (1mM) and cultured for an additional 3 h before harvesting. Har-vested BL21(DE3) cells expressingHis-PALFwere resuspendedin Ni-NTA binding buffer (50 mM NaH2PO4 (pH 7.8), 0.5 M

KCl, 2 mM �-mercaptoethanol, 10% glycerol, 0.1% TritonX-100, and 20 mM imidazole (pH 7.8)) supplemented with pro-tease inhibitors and lysed by sonication. The cell lysate wasapplied to a Ni-NTA-agarose resin (Qiagen). Resin was washedwith 35 mM imidazole. His-PALF was eluted off with bindingbuffer plus 500 mM imidazole. Eluted fractions were dialyzedagainst Hi-Trap heparin binding buffer (50 mM Tris-HCl (pH7.5), 10% glycerol, 2 mM EDTA, 1 mM DTT, 100 mM NaCl,0.02% Nonidet P-40), loaded onto a pre-equilibrated Hi-Trapheparin column, and eluted with a linear gradient to 1 M NaClover 20 ml. We were able to obtain a yield of 200 �g from a1-liter starting culture. His-PALF-containing fractions werereconcentrated onto Ni-NTA-agarose and eluted in bindingbuffer plus 500mM of imidazole to a final elution volume of 200�l. His-PALF was then applied to a Superose 12 gel filtrationcolumn (GE Healthcare) and eluted with 250 mM NaCl, 10%glycerol, 50 mM Tris-HCl (pH 7.5), and 1 mM DTT. PALF sam-ples were stored at 4 °C and used within 2 weeks because thenuclease activity of PALF is highly sensitive to freezing, aspointed out previously (1). The expression and purification ofDNA-PKcs from HeLa cells has been described previously (8).In Vitro Nuclease and Ligation Assays —We extended the

PALF nuclease optimizations performed previously (1) by fur-ther testing activity in a range of pH (6.5 to 8), divalent cationconcentration (0 to 20 mM MgCl2), and monovalent salt con-centration (1 mM to 100 mM KCl). On the basis of this, in vitroDNAnuclease assays were performed in a total volume of 10 �lwith a buffer composition of 25 mM Tris-HCl (pH 7.5), 10 mM

KCl, 10 mM MgCl2, 1 mM DTT and 50 ng/�l BSA. In the reac-tion, 50 nM single-stranded DNA substrate with an overhang(3� or 5�) or 20 nMhairpin substratewere incubatedwith 125 nMPALF, and, in specified cases, one or more of the following: 126nM DNA-PKcs, 100 nM Ku, 75 nM XRCC4/ligase IV, or a com-bination of the three proteins. In reactions containing XRCC4/ligase IV, the XRCC4 and ligase IV were prephosphorylated byCK2 according to the manufacturer’s instructions (Sigma-Al-

PALF Nuclease Activity and DNA End Joining

OCTOBER 21, 2011 • VOLUME 286 • NUMBER 42 JOURNAL OF BIOLOGICAL CHEMISTRY 36369

drich). Unless specified otherwise, when DNA-PKcs was pres-ent, 0.5 mM ATP and 0.5 �M 35-bp blunt-end DNA (YM 8/9)were also included in the specified reactions. Reactions werethen incubated at 37 °C for 1 h. After incubation, reactionswerestopped and analyzed on 12% denaturing PAGE gels.In vitro ligation assays were performed in a total volume of 10

�l with a buffer composition of 25mMTris (pH7.5), 2mMDTT,0.025% Triton X-100, 0.1 mM EDTA, 10% PEG, 50 ng/�l BSA,and 5% glycerol. In the in vitro ligation assay, a two-step reac-tion was performed. In the reaction, 50 nM double- strandedsubstrate (SL11/JG25)was incubatedwith 125nMPALF, and, inspecified cases, one or more of the following: 126 nM DNA-PKcs, 100 nM Ku, 75 nM XRCC4/ligase IV, or a combination ofthe three proteins. Unless specified otherwise, when DNA-PKcs was present, 0.5 mM ATP and 0.5 �M 35-bp blunt-endDNA (YM 8/9) were also included in the specified reactions. Atwo-step ligation reaction was carried out in the followingorder: PALF, substrate, and DNA-PKcs were first added to thereaction and incubated at 37 °C for 30 min. Ligase IV-XRCC4and Ku were then added, and the reaction was incubated foranother 30min at 37 °C. The reactions were stopped, and DNAwas phenol-extracted and analyzed on 8% denaturing PAGEgel. Gels were dried, exposed in a phosphorimager cassette, andscanned. Bands were cut out at the dimer position on ligationgels, and we used TOPO TA to clone this DNA into the Invit-rogen pCR2.1-TOPO vector. The insert was subsequentlysequenced using a Li-Cor 4200 sequencer, and the junctionswere analyzed.NHEJ Assay in Living Human Cells—As reported previously

(20), the I-SceI expression plasmid, pCMV-3xNLS-I-SceI, wasintroduced by transfection with Lipofectamine 2000 reagentinto 1.5 � 105 H1299dA3–1#1 cells harboring two I-SceI siteslocated 1.3 kb apart. The cells were pretransfected with siRNAfor 48 h using Lipofectamine RNAiMAX. For FACS analysis,cells were harvested by trypsinization, washed with PBS, andapplied to the FACS Calibur apparatus (BD Biosciences).EGFP-positive cells were counted using the Cellquest software.

RESULTS

PALF has 3� Single-stranded Exonuclease Activity—Previousdata has demonstrated that PALF has endonuclease activitywhich is dependent on the presence of abasic sites (1).We puri-fied PALF using Ni-NTA, HiTrap heparin, and Superose 12columns (Fig. 1A) and found it to have nuclease activity acrossan elution peak in proportion to the amount of protein visual-ized using Coomassie Blue staining (Fig. 1, A and B).Analysis of individual domains of PALF shows that theC-ter-

minal region, spanning amino acids 360–511, is necessary andsufficient to impart single-strand nuclease activity on the pro-tein (supplemental Figs. 1 and 2). The amino acid 360–511fragment alone contains all of the nucleolytic activity of thePALF protein (supplemental Fig. 1). A truncation mutant lack-ing the amino acid 378–511 domain has no detectable nucleaseactivity in our assay (supplemental Fig. 2). The amino acid 378–511 region contains two CYR (cysteine-tyrosine-arginine)motifs, also called PBZ motifs (poly [ADP-ribose] binding zincfinger).

We tested the activity of PALF on a single-stranded DNAsubstrate to determine whether PALF has nuclease capabilitiesbeyond those described originally. We designed a single-stranded poly dT 30mer labeled at the 5� end to investigate this(Fig. 2A, lane 1). On this substrate, PALF generates a series ofcleavage products (Fig. 2A, lane 2). The addition of DNA-PKcsand ATP does not increase or decrease the nuclease activity ofPALF on single-stranded DNA (Fig. 2A, lane 3). Stimulation ofDNA-PKcs autophosphorylation by the addition of unlabeleddouble-strandedDNA, YM8/YM9, also did not affect the activ-ity of PALF (Fig. 2A, lane 4).Artemis was used to generate a ladder starting at the 1 nucle-

otide position (Fig. 2A, lane 5). PALF does not have any 5�exonuclease activity. Unlike Artemis, no distinct product isformed at the 1 nucleotide position (Fig. 2A, lane 2 versus lane

FIGURE 1. Purification of PALF. A, PAGE gel on Superose 12 fractions. AfterNi-NTA and Hi-Trap heparin purification, Superose 12 fractions of PALF areshown on an 8% SDS-PAGE gel stained with Coomassie Blue on which PALFhas a gel mobility position at 81 kDa. Ladder designates the protein markerlane, and the fraction numbers are above each lane. B, nuclease activity ofPALF corresponding to Superose 12 fractions. Fractions across the Superose12 elution peak were assayed for nuclease activity using poly(dT) substrate(JG169). Each reaction consists of 50 nM single-stranded DNA substrate(JG169) and 50 nM PALF. Reactions were incubated for 2 h at 37 °C. Afterincubation, reactions were stopped and analyzed by 12% denaturing PAGE.

PALF Nuclease Activity and DNA End Joining

36370 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 42 • OCTOBER 21, 2011

5). The nucleolytic cleavage products that we do see can beexplained by a combination of 3� exonuclease activity and sin-gle-stranded endonuclease activity. In the presence of Mn2�,Artemis does not require DNA-PKcs to function endonucleo-lytically (10, 17). MnCl2 does not affect the activity of PALF(Fig. 2A, lane 6).To differentiate between these two activities, we labeled the

same substrate at the 3� end (Fig. 2B, lanes 1–3). Products gen-erated by PALF from this substrate include a prominent band atthe 1 nucleotide position (Fig. 2B, lanes 2 and 3), demonstratingdefinitive 3� exonuclease activity.However, a full range of largerproducts with a weaker profile is also present (Fig. 2B, lanes 2and 3). The amount of each species in this distribution of prod-ucts increases relatively equally as the amount of enzyme wasincreased (twice as much PALF in lane 2 than in lane 3). Giventhat the substrate was labeled at the 3� terminus, these weakerproducts can only be generated by the endonuclease activity ofPALF acting on the single-stranded substrate. Hence, we con-

clude that PALF has endonuclease activity on single-strandedDNA in addition to 3� exonuclease activity.PALF Has Single-stranded DNA Endonuclease Activity—If

PALF has both 3� exonuclease activity and single-strandedendonuclease activity, then the endonuclease activity couldpotentially be masked on a single-stranded DNA substrate.To test for endonuclease activity that does not require sub-strates with abasic sites, we designed substrates containingeither a 15-nucleotide 3� overhang or a 14-nucleotide 5�overhang, both labeled at the 5� end (Fig. 3A, lane 1 for the 5�overhang substrate, and B, lane 1 for the 3� overhang sub-strate). This endonuclease activity cleaves randomly withinthe 5� overhang. Unlike Artemis, the addition of DNA-PKcsdoes not further stimulate the endonuclease activity (Fig. 3a,lane 2 versus lane 3). Addition of unlabeled dsDNA to DNA-PKcs (to stimulate autophosphorylation) also has no effecton the activity of PALF at the 5� overhang (Fig. 3a, lane 2versus lane 3). Hence, we conclude that unlike Artemis,

FIGURE 2. 3� to 5� exonuclease activity of PALF on single-stranded DNA. A, exonuclease activity of PALF on single-stranded DNA. In the reaction, 50 nM

single-stranded DNA substrate (JG169) was incubated with the protein(s) indicated above the lane in a 10-�l reaction for 60 min at 37 °C. After incubation,reactions were stopped and analyzed by 12% denaturing PAGE. Protein concentrations are as follows: PALF, 125 nM and DNA-PKcs, 126 nM. As specified, 0.5 mM

ATP and 0.5 �M YM8/9 were also included in designated reactions. YM8/9 is a 35-bp blunt-ended double-stranded DNA that is used as DNA-PKcs cofactor.B, 3� exonuclease monitored with 3�-labeled substrate. In the reaction, 50 nM single-stranded DNA substrate (JG169) labeled at its 3� end was incubated withPALF in a 10-�l reaction for 60 min at 37 °C. Concentrations are as follows: 250 nM PALF (lane 2) and 125 nM PALF (lane 3). After incubation, reactions werestopped and analyzed by 12% denaturing PAGE. The asterisk represents the radiolabel in all figures. The bold arrow on the DNA substrate diagram beside thegel represents the site of DNA cleavage in all figures.

PALF Nuclease Activity and DNA End Joining

OCTOBER 21, 2011 • VOLUME 286 • NUMBER 42 JOURNAL OF BIOLOGICAL CHEMISTRY 36371

PALF does not require DNA-PKcs for endonuclease activityat a 5� overhang.Because of its ability to bind to Ku through an intermediate

domain between the FHA domain and the CYR motif (poly[ADP-ribose] binding zinc finger motif) of PALF (1, 3), wetested whether the addition of Ku has an effect on nucleaseactivity of PALF at a 5� overhang. The addition of Ku alongwithDNA-PKcs does not appear to modify the activity of PALFactivity (data not shown).Earlier, we saw that the single-stranded DNA endonuclease

activity of PALF is less efficient when 5� labeled single-strandedDNA substrates are cleaved to lengths smaller than 4 nt. Thesingle-stranded endonuclease activity also does not extend intothe double-stranded portion of the DNA. This is indicated bythe relative lack of bands above the 14th nucleotide, countingfrom the bottom of the gel (Fig. 3A, lane 2). (The very weakbands present between the 14th and 18th nucleotide may bedue to PALF action at sites of breathing of the single strand/double strand DNA junction.) Hence, PALF does not appear tohave significant activity on double-stranded DNA.On a 5�-labeled duplex DNA substrate with a 15-nt poly T

3� overhang, nine to ten distinct products are seen (Fig. 3B,

lane 2). PALF demonstrates endonuclease activity but doesnot cleave at positions within 4 or 5 nts from the double-stranded portion of the substrate (Fig. 3B, lane 2). The 3�overhang activity is not stimulated or blocked by the addi-tion of DNA-PKcs and ATP (Fig. 3B, lane 3). Further addi-tion of unlabeled double-stranded DNA, YM8/YM9, withDNA-PKcs also had no effect on the nuclease activity ofPALF (Fig. 3B, lane 3). The addition of Ku did not yield anychanges when compared with the basal level with PALFalone (data not shown). Like the 5� overhang substrate, noone cleavage product is dominant over any other. Hence, thenuclease activity on the single-stranded 3� overhang appearsto be random in its position. PALF also has 3� exonucleaseactivity. Therefore, the activity we see at the 3� overhang is acombination of both endonuclease as well as 3� exonucleaseactivity on single-stranded regions. When we position fourphosphothioester linkages at the 3� end of the substrate,much less product derives from the 3� overhang, indicatingdiminished 3� exonuclease activity (supplemental Fig. 3, lane2 versus lane 4). However, some product still arises, attrib-utable to the endonuclease activity, which can nick upstreamof the phosphothioester linkages.

FIGURE 3. Endonuclease activity of PALF on overhangs. A, in specified reactions, 50 nM 5�-labeled double-stranded DNA substrate, YM130/YM68 (5�overhang), was incubated with 125 nM PALF, 126 nM DNA-PKcs, 0.5 mM ATP, and 0.5 �M YM8/9 in a 10-�l reaction for 60 min at 37 °C. After incubation, reactionswere stopped and analyzed on 12% denaturing PAGE. B, in specified reactions, 50 nM 5�-labeled double-stranded DNA substrate, YM149/YM68 (3� overhang),was incubated with 125 nM PALF, 126 nM DNA-PKcs, 0.5 mM ATP, and 0.5 �M YM8/9 in a 10-�l reaction for 60 min at 37 °C. After incubation, reactions werestopped and analyzed by 12% denaturing PAGE.

PALF Nuclease Activity and DNA End Joining

36372 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 42 • OCTOBER 21, 2011

For 3� overhang substrates, PALF seems to functionallyoccupy approximately four or five nucleotides 3� to the duplexportion of the substrate, as indicated by termination of nickingat length 26 nt (rather than 21 nt) (Fig. 3B). The approximatelyfour or five nucleotides may reflect the minimum size requiredfor PALF to bind to single-stranded DNA and act upon sub-strates. This is the same behavior as is seen for Artemis endo-nuclease action on 3� overhangs (Fig. 4, lane 5, and Ref. 8). Weconclude that PALF has single-stranded endonuclease activityand that this is distinct from its 3� exonuclease activity.Because of the fact that PALF has endonuclease activity, we

further investigated the possibility that PALF could potentiallyopen a hairpin substrate like Artemis. A hairpin substrate witha 20-nucleotide duplex portion and a 6-nucleotide 5� overhangwas used for this experiment (Fig. 4, lane 1). When PALF aloneis incubatedwith the substrate, four endonuclease products canbe seen at the bottomof the gel (Fig. 4, lane 2). This is consistent

with the endonuclease activity we observed before for the 5�overhang substrate, YM130/YM68. Once again we used Arte-mis-DNA-PKcs to generate a ladder to show the 1 nucleotideposition and the hairpin opening product at the 28-nt position(Fig. 4, lane 5). This product is absent in the reaction wherePALF alone is added (Fig. 4, lane 2). The hairpin opening activ-ity of Artemis is dependent upon DNA-PKcs, and we exploredthis possibility with respect to PALF. DNA-PKcs andATPwereadded to the reaction in addition to PALF. This did not stimu-late PALF to cleave the hairpin, and no additional endonucleaseactivity is seen (Fig. 4, lane 3). Further stimulation of DNA-PKcs by unlabeled DNA YM8/9 also did not stimulate PALF tocleave a hairpin substrate (Fig. 4, lane 4). Therefore, unlikeArtemis, PALF does not show DNA-PKcs-dependent ability inopening hairpin substrates (Fig. 4, lane 4 versus lane 5). Theendonuclease activity further supports our previous findings ofPALF activity at a 5� overhang.The Nuclease Activity of PALF Is Not Affected by Ku or

XRCC4-DNA Ligase IV—Previous data have demonstratedbinding of PALF to XRCC4 via the FHA domain of PALF andbinding to Ku through a region in PALF between the FHAdomain and the two CYR (poly [ADP-ribose] binding zinc fin-ger) motifs. We wanted to investigate the possibility that theXRCC4-DNA ligase IV complex orKu can enhance or block theexonuclease or endonuclease activity of PALF. As above, a sin-gle-stranded poly T substrate labeled at the 5� end was used inincubationswith PALF for 1 hwith orwithout purifiedXRCC4-ligase IV or Ku. The basal activity of PALF generates a combi-nation of endonuclease products and 3� exonuclease products(Fig. 5, lane 2). XRCC4-DNA ligase IV, after being phosphoryl-ated by CK2, was run alone to rule out the possibility of con-tamination (Fig. 5, lane 3). XRCC4-DNA ligase IV withoutphosphorylationwas also used as a negative control (Fig. 5, lane4). The addition of XRCC4-ligase IV did not produce a signifi-cant change over basal PALF activity (Fig. 5, lane 2 versus lane5). A combination of Ku and XRCC4-ligase IV also did notchange the product profile (data now shown). Finally, a combi-nation of PALF, XRCC4-ligase IV, and DNA-PKcs failed tohave any effect relative to PALF alone (Fig. 5, lane 2 versus lane6). This is noteworthy, given that bothKu andXRCC4-ligase IVare known to bind to PALF and that DNA-PKcs is suggested tointeract with PALF according to proteome analysis (1). There-fore, the interactions do not affect the nuclease activity ofPALF.PALF Is Able to Cooperate with Other NHEJ Factors to Pro-

mote Ligation in Vitro and in Vivo— Even though the coreNHEJ factors did not stimulate the nuclease activity of PALF orpermit it to open hairpins, we were interested in whether PALFcould cooperate with other NHEJ factors to stimulate ligationin an in vitro assay. In particular, we wondered if the nucleaseactivities of PALF could substitute for those of Artemis inresecting overhangs.We tested this by using an oligonucleotide substrate (SL11/

JG258) with four nucleotides of microhomology that had sixadditional A nts attached to the 3� end that would requirenucleolytic resection before the 4 nts of microhomology couldbe utilized for ligation. Successful ligation would result in thesubstrate ligated to another identical molecule in a head-to-tail

FIGURE 4. Lack of endonuclease activity of PALF on hairpin substrates. Inspecified reactions, 20 nM hairpin DNA substrate, YM164, was incubated with125 nM PALF, 50 nM Artemis, 126 nM DNA-PKcs, 0.5 mM ATP, and 0.5 �M YM8/9in a 10-�l reaction for 60 min at 37 °C. After incubation, reactions werestopped and analyzed by 12% denaturing PAGE.

PALF Nuclease Activity and DNA End Joining

OCTOBER 21, 2011 • VOLUME 286 • NUMBER 42 JOURNAL OF BIOLOGICAL CHEMISTRY 36373

fashion.When the DNA substrate (Fig. 6A, lane 1) is incubatedwith XRCC4-ligase IV complex and Ku, it results in a smallbasal amount of ligation versus substrate alone (Fig. 6A, lane 2versus lane 1). In the absence of known NHEJ factors, whenPALF alone is added to the reaction (Fig. 6A, lane 3), no ligationproducts are formed. However when PALF, Ku, and XRCC4-ligase IV are all present, ligation products are formed (Fig. 6A,lane 4). PALF is able to significantly increase the amount ofproduct formed versus the basal level (Fig. 6A, lane 4 versus lane2). The addition of DNA-PKcs to the reaction hinders the liga-tion process (Fig. 6A, lane 5). It is thought that DNA-PKcsmustdissociate before ligation of DNA ends can occur (8, 14, 15).This process is thought to be facilitated by autophosphoryla-tion of DNA-PKcs. However, the precisemechanism is not wellunderstood. In vitro, this processmay be too slow in our ligationreactions and hence may inhibit the overall ligation efficiency.The ligation product was subsequently cut out from the gel,cloned into a TA cloning vector, and then sequenced (see“Experimental Procedures”).We found that all ten of the joinedproductmolecules had the six As resected, leaving only the fourCs at the 3� overhang (Fig. 6C). This allows for efficient joiningbecause of the microhomology between the 4 Cs on the topstrand at the 3� terminus of one duplex substrate and the 4Gs ofthe bottom strand of a second duplex substrate. This is also

consistent with the results above showing that the PALFnuclease activity terminates its cleavage 3 to 4 nucleotides awayfrom the junction between the overhang and the double-stranded portion of a substrate.We also compared the efficiency of ligation of PALF versus

that of Artemis in an in vitro assay. Using the same substrate(SL11/JG258), we added Ku and XRCC4-DNA ligase IV toArtemis or PALF. Artemis is able to resect the ends in a similarmanner as PALF, resulting in ligation using terminal micro-homology (Fig. 6B, lane 1). Under the same conditions, PALF isalso able to ligate the substrate with a similar efficiency as Arte-mis (Fig. 6B, lane 3, 12%, versus lane 1, 16%, and supplementalFig. 4, lane 4 versus lane 1). In some assays, we also addedNHEJfactors, XLF and polymerase �, and observed indistinguishableresults (supplemental Fig. 4).In vivo, we find that knockdown of PALF using siRNA

reduces rejoining of two incompatible I-SceI-generated DNAends by 50% (18) (Fig. 7). This is consistent with recent studiesof others that also suggest a role for PALF in NHEJ (19).

DISCUSSION

PALF Is Both a Single-stranded DNA 3� Exonuclease and aSingle-stranded DNA Endonuclease—The Yasui laboratory(1–3) was the first to demonstrate that PALF (also called APLF)has nuclease activity. Here we confirm that PALF has 3� exonu-clease activity on single-stranded DNA.We have also found that PALF has endonuclease activity on

single-stranded DNA. This activity is capable of cleaving at allsites of a single strand except for the last few nucleotides at the5� end. In addition, this activity is able to cleave at all positionsof 5� overhangs of duplex DNA except for the 5�-most fewnucleotides of the overhang. Finally, this activity cleaves at allpositions of 3� overhangs except for the approximately fournucleotides 3� to the boundary of the overhang with the duplexDNA (designated the “no cleavage zone” in Fig. 8).Unification of the PALF Nuclease Activities—We believe that

the double-strandedDNAendonuclease activity described pre-viously might be accounted for by the broader single-strandedDNA endonuclease activity described here. The transientbreathing of a double-stranded DNA end into single-strandedflaps, which is substantial (20), could serve as a single-strandedsubstrate for this single-strandedDNAendonuclease activity ofPALF. Given the action of PALF on all single-stranded DNA,the endonuclease of PALF has a broader range of substratesthan appreciated previously, thereby increasing its potentialimportance within the cell.In Vitro and in Vivo Function of PALF with Other NHEJ

Proteins—We did not see any stimulation or modification ofPALF 3� exonuclease and single-stranded DNA endonucleaseactivity byKu,DNA-PKcs, orXRCC4-DNA ligase IV.Althoughstudies indicate that Ku and XRCC4-DNA ligase IV each bindto PALF (1, 3), these interactions do not appear to affect PALFenzymatically.Importantly, we find that in an in vitro system consisting of

PALF, Ku, and XRCC4-DNA ligase IV is indeed able to joinincompatible DNA ends as efficiently as a system consisting ofArtemis-DNA-PKcs, Ku, and XRCC4-DNA ligase IV. Specifi-cally, PALF is able to resect the incompatible portion of an

FIGURE 5. Effect of Ku, XRCC4-DNA Ligase IV, and DNA-PKcs on PALFnuclease activity. In specified reactions, 50 nM 5�-labeled single-strandedDNA substrate (JG169) was incubated with 125 nM PALF, 126 nM DNA-PKcs,0.5 mM ATP, 0.5 �M YM8/9, 100 nM Ku, and 75 nM XRCC4-DNA ligase IV in a10-�l reaction for 60 min at 37 °C. After incubation, reactions were stoppedand analyzed by 12% denaturing PAGE.

PALF Nuclease Activity and DNA End Joining

36374 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 42 • OCTOBER 21, 2011

overhang to a point at which XRCC4-DNA ligase IV is able tosupport efficient ligation. Interestingly, at the DNA sequencelevel, we did not observe any ligation products other than thosewith the four junctional nucleotides that provide the maximal

terminal microhomology. (This does not preclude the possibil-ity that some junctions do contain only three nucleotides andmake up a small minority of the products formed.) Use of max-imal microhomology could occur by iterative removal of the As

FIGURE 6. PALF can cooperate with Ku and XRCC4-DNA ligase IV in double-strand DNA end ligation. A, in specified reactions, 50 nM 5�-labeled double-stranded DNA substrate (SL11/JG258) was incubated with 125 nM PALF, 126 nM DNA-PKcs, 0.5 mM ATP, and 0.5 �M YM8/9 for 30 min at 37 °C and followed bythe addition of 100 nM Ku and 75 nM XRCC4-DNA ligase IV for 30 min at 37 °C. After incubation, reactions were stopped and analyzed by 8% denaturing PAGE.Positions of the dimerized and trimerized DNA duplex products from the monomeric ligations were determined on the basis of duplex DNA markers not shownon the gel (also see sequencing results in C, which confirm the dimer junctions). B, in specified reactions, 50 nM 5�-labeled double-stranded DNA substrate(SL11/JG258), was incubated with 75 nM Artemis or 125 nM PALF, 126 nM DNA-PKcs, 0.5 mM ATP, and 0.5 �M YM8/9 for 30 min at 37 °C and followed by theaddition of 100 nM Ku and 75 nM XRCC4-DNA ligase IV for 30 min at 37 °C. After incubation, reactions were stopped and analyzed by 8% denaturing PAGE. C,dimer products from A, lane 4, were cut out of the gel, extracted, PCR-amplified, TA-cloned, and sequenced. The junctional sequences are shown. PALF removedthe AAAAAA (in italics) from each 3� overhang, thus allowing the 3�-CCCC overhang of one duplex substrate to anneal to the 3�-GGGG on another duplexsubstrate molecule. These proceeded to ligation by XRCC4-DNA ligase IV to yield the dimer product.

PALF Nuclease Activity and DNA End Joining

OCTOBER 21, 2011 • VOLUME 286 • NUMBER 42 JOURNAL OF BIOLOGICAL CHEMISTRY 36375

by the 3� exonuclease activity of PALF or by the endonucleasePALF activity.We favor the latter possibility becausewe did notsee a more diverse set of products.Could PALF and Artemis both function as nucleases in

NHEJ? This appears likely, on the basis of our in vivo siRNAknockdown of PALF (Fig. 7). Additionally, PALF suppressionresults in cells that are sensitive to methyl methane sulphonate(1) as well as sensitive to double-strand break agents.3 Artemis(in complex with DNA-PKcs) may bemore efficient in cleavinglong overhangs in an endonucleolytic manner (9), whereasPALF has an active 3� exonuclease activity. Artemis alone(without DNA-PKcs) has a weak single-stranded endonucleaseactivity (10), and PALF appears to be substantially stronger in

this activity. Therefore, in several ways, the nucleolytic activi-ties of Artemis and PALF are complementary to one another.

REFERENCES1. Kanno, S., Kuzuoka, H., Sasao, S., Hong, Z., Lan, L., Nakajima, S., and

Yasui, A. (2007) EMBO J. 26, 2094–21032. Iles, N., Rulten, S., El-Khamisy, S. F., andCaldecott, K.W. (2007)Mol. Cell.

Biol. 27, 3793–38033. Macrae, C. J., McCulloch, R. D., Ylanko, J., Durocher, D., and Koch, C. A.

(2008) DNA Repair 7, 292–3024. Bekker-Jensen, S., Fugger, K., Danielsen, J. R., Gromova, I., Sehested, M.,

Celis, J., Bartek, J., Lukas, J., and Mailand, N. (2007) J. Biol. Chem. 282,19638–19643

5. Koch, C. A., Agyei, R., Galicia, S.,Metalnikov, P., O’Donnell, P., Starostine,A., Weinfeld, M., and Durocher, D. (2004) EMBO J. 23, 3874–3885

6. Lieber, M. R. (2008) J. Biol. Chem. 283, 1–57. Budman, J., Kim, S. A., and Chu, G. (2007) J. Biol. Chem. 282,

11950–119593 S. Li, S. Kanno, R. Watanabe, H. Ogiwara, T. Kohno, G. Watanabe, A. Yasui, and

M. R. Lieber, unpublished data.

FIGURE 7. siRNA directed against PALF mRNA reduces NHEJ in vivo. A, assay for NHEJ of DSBs in vivo. Two I-SceI sites located in the reverse direction toproduce incompatible ends in the substrate DNA are shown as arrowheads. CMV, cytomegalovirus promoter/enhancer; HSV-TK, herpes simplex virus-thymi-dine kinase; pA, poly(A) signal. Ligation of two broken DNA ends generated by I-SceI digestion results in deletion of the HSV-TK open reading frame and leadsto production of a transcript that enables translation of EGFP instead of HSV-TK protein (for details, see Ref. 18). B, Western blot analysis of suppression for PALFexpression by siRNA in H1299dA3–1#1 cells. C, PALF is required for end joining of I-SceI-induced double strand breaks in H1299dA3–1#1 cells.

FIGURE 8. Summary of exo- and single-stranded endonuclease activities of PALF. PALF has known 3� exonuclease activity. Here we describe the single-strand DNA endonuclease activity of PALF, which can act at any position within a single-stranded region except within approximately 4 nts of the junction withdouble-stranded DNA (designated no cleavage zone). The diagonal slash marks on the DNA substrate diagram represent PALF cleavages.

PALF Nuclease Activity and DNA End Joining

36376 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 42 • OCTOBER 21, 2011

8. Ma, Y., Pannicke, U., Schwarz, K., and Lieber, M. R. (2002) Cell 108,781–794

9. Ma, Y., Schwarz, K., and Lieber, M. R. (2005) DNA Repair 4, 845–85110. Gu, J., Li, S., Zhang, X., Wang, L. C., Niewolik, D., Schwarz, K., Legerski,

R. J., Zandi, E., and Lieber, M. R. (2010) DNA Repair 9, 429–43711. Pawelczak, K. S., and Turchi, J. J. (2010) DNA Repair 9, 670–67712. Goodarzi, A. A., Yu, Y., Riballo, E., Douglas, P.,Walker, S. A., Ye, R., Härer,

C., Marchetti, C., Morrice, N., Jeggo, P. A., and Lees-Miller, S. P. (2006)EMBO J. 25, 3880–3889

13. Ma, Y., Pannicke, U., Lu, H., Niewolik, D., Schwarz, K., and Lieber, M. R.(2005) J. Biol. Chem. 280, 33839–33846

14. Meek, K., Dang, V., and Lees-Miller, S. P. (2008)Adv. Immunol. 99, 33–5815. Mahaney, B. L., Meek, K., and Lees-Miller, S. P. (2009) Biochem. J. 417,

639–65016. Niewolik, D., Pannicke, U., Lu, H., Ma, Y., Wang, L. C., Kulesza, P.,

Zandi, E., Lieber, M. R., and Schwarz, K. (2006) J. Biol. Chem. 281,33900–33909

17. Huang, Y., Giblin, W., Kubec, M., Westfield, G., St Charles, J., Chadde, L.,Kraftson, S., and Sekiguchi, J. (2009) J. Exp. Med. 206, 893–908

18. Ogiwara, H., Ui, A., Otsuka, A., Satoh, H., Yokomi, I., Nakajima, S., Yasui,A., Yokota, J., and Kohno, T. (2011) Oncogene 30, 2135–2146

19. Rulten, S. L., Fisher, A. E., Robert, I., Zuma, M. C., Rouleau, M., Ju, L.,Poirier, G., Reina-San-Martin, B., and Caldecott, K. W. (2011) Mol. Cell41, 33–45

20. Tsai, A. G., Engelhart, A. E., Hatmal, M. M., Houston, S. I., Hud, N. V.,Haworth, I. S., and Lieber, M. R. (2009) J. Biol. Chem. 284, 7157–7164

PALF Nuclease Activity and DNA End Joining

OCTOBER 21, 2011 • VOLUME 286 • NUMBER 42 JOURNAL OF BIOLOGICAL CHEMISTRY 36377