PARP-2, A Novel Mammalian DNA Damage-dependent Poly(ADP-ribose) Polymerase*

Poly(ADP-ribosylation) is a post-translational modification of nuclear proteins in response to DNA damage that activates the base excision repair machinery. Poly(ADP-ribose) polymerase which we will now call PARP-1, has been the only known enzyme of this type for over 30 years. Here, we describe a cDNA encoding a 62-kDa protein that shares considerable homology with the catalytic domain of PARP-1 and also contains a basic DNA-binding domain. We propose to call this enzyme poly(ADP-ribose) polymerase 2 (PARP-2). The PARP-2 gene maps to chromosome 14C1 and 14q11.2 in mouse and human, respectively. Purified recombinant mouse PARP-2 is a damaged DNA-binding protein in vitro and catalyzes the formation of poly(ADP-ribose) polymers in a DNA-dependent manner. PARP-2 displays automodification properties similar to PARP-1. The protein is localized in the nucleusin vivo and may account for the residual poly(ADP-ribose) synthesis observed in PARP-1-deficient cells, treated with alkylating agents or hydrogen peroxide.

In response to DNA-strand breaks introduced either directly by ionizing radiation or indirectly following enzymatic incision at a DNA lesion, the immediate poly(ADP-ribosylation) of nuclear proteins converts the DNA ends into intracellular signals that modulate DNA repair and cell survival programs. At the sites of DNA breakage, poly(ADP-ribose) polymerase (PARP) 1 (EC 2.4.2.30) catalyzes the transfer of the ADP-ribose moiety from its substrate NAD ϩ , to a limited number of proteins involved in chromatin architecture, DNA repair, or in DNA metabolism including PARP itself (1)(2)(3)(4). Recently, the generation of PARP-deficient mice by homologous recombination (5,6) has clearly demonstrated the involvement of PARP in the maintenance of the genomic integrity due to its role during base excision repair (7)(8)(9)). An substantial delay in DNA strandbreak repair was observed following treatment of PARP-deficient cells with monofunctional alkylating agents (10). This severe DNA repair defect appears to be the primary cause for the observed cytotoxicity of N-methyl-N-nitrosourea, methylmethanesulfonate (MMS), or ␥-rays leading to cell death occurring after a G 2 /M block (10).
It was assumed for many years that PARP activity was associated with a single protein displaying unique DNA damage detection and signaling properties. This assumption was challenged by the recent discovery in Arabidopsis thaliana of a gene coding for a PARP-related polypeptide of a calculated molecular mass of 72 kDa (11). It then became evident that two structurally different PARP proteins, both possessing DNA-dependent poly(ADP-ribose) activities, were present in both A. thaliana as well as in maize (12)(13)(14). 2 Furthermore, it has been reported recently that mouse embryonic fibroblasts derived from PARP knockout are capable of synthesizing ADP-ribose polymers in response to DNA damage (15), suggesting that in mammals, like in plants, at least one additional member of the PARP family may exist in addition to the classical zinc finger containing PARP.
This work describes the cloning of a human and a murine cDNA coding for a new member of the PARP family, based on its similarity with the A. thaliana 72-kDa poly(ADP-ribose) polymerase (12), which those authors named APP. We denote this new protein PARP-2, to differentiate from the classical PARP protein renamed PARP-1. We demonstrate that PARP-2 is a nuclear DNA-dependent poly(ADP-ribose) polymerase catalyzing the formation of ADP-ribose polymers in the presence of damaged DNA, thus suggesting a biological role in the cellular response to DNA damage.

Determination of Poly(ADP-ribose) Synthesis in Cell Extracts-To
determine poly(ADP-ribose) synthesizing activity in cells extracts, equal amounts of protein from various cell extracts were incubated in standard conditions with 200 M [␣-32 P]NAD ϩ (30.0 nCi/nmol), 2 g/ml histones, and 10 g/ml DNA activated by DNase I at 25°C for 10 min. Activity is expressed as the ratio between the radioactivity of the acid insoluble material produced by PARP-1 ϩ/ϩ and PARP-1 Ϫ/Ϫ cell extracts.
EST Searches and Cloning of cDNA Encoding Murine PARP-2-Sequence analyses were performed using the GCG sequence analysis package (Wisconsin package version 8.1, Genetic Computer Group, Madison, WI). The primary sequence of the A. thaliana PARP homologue (APP, GenBank 228 accession number Z48243) was used to search the complete EST data base (release 107/55 of GenBank 228 data base) using TBLASTN of the BLAST program (16). Two murine and one human EST (GenBank 228 accession numbers AA608364, AA529426, and AA595596, respectively) encoding APP homologues but distinct from the classical PARP-1 of 113 kDa were chosen and purchased from the IMAGE consortium (Research Genetics, Huntsville, AL). EST clone number AA529426 contained the cDNA (1.2 kilobase) encoding the catalytic domain of mouse PARP-2 (aa 199 -559) in the pCMV SPORT vector (Research Genetics, Huntsville, AL). In order to express the catalytic domain of PARP-2, a SmaI site and an ATG surrounded by the translation initiation sequence (17) of PARP-1 (18) were created by PCR, using the sense primer AGGCCCCGGGGGGAG-GATGGCGACTTTGAAGCCTGAGTCTCAG and the reverse primer CCAGCAAGGTCATAGTATAGTCC. The PCR fragment (550 bp) was restricted with SmaI-AccI, and used to replace the wild type sequence in clone AA529426. The entire catalytic fragment (homologous to domain F in Fig. 3) was thereafter excised with SmaI-NotI and cloned in the baculovirus recombination vector pVL 1393, giving pVL-mPARP-2-F.
In order to generate a full-length mPARP-2 cDNA, a SalI-NotI fragment of AA529426 was used to screen a random-primed ZAP II cDNA library (kindly provided by J. M. Garnier, IGBMC, Illkirch, France) established from 10-day-old mouse embryos. One of the positive clones selected contained part of the murine PARP-2 cDNA, starting 45 bp upstream of the ATG and encoded aa 1 to 527 of the protein. In order to obtain good expression in the baculovirus expression system, the region around the ATG was replaced. A SmaI site and an ATG surrounded by the translation initiation sequence of PARP-1 were introduced by PCR. The primers used were, sense primer: AGGCCCCGGGGGGAGGATG-GCGCCGCGGCGGCAGAGATCAGGCTCTGG, reverse primer: AT-CATCATTTCTTCCATGG. After purification, the PCR product (712 bp) was restricted by SmaI-NcoI and cloned in pVL-mPARP-2-F, to generate the full-length mPARP-2 baculovirus recombination vector pVL-mPARP-2 vector. The amplified fragments were sequenced to ensure that no mutation has been introduced by the PCR.
A human PARP-2 EST clone was identified in the LifeSeq TM data base of Incyte Pharmaceuticals using the human PARP-1 sequence for a data base search. The clone (2286233) was ordered and analyzed. It started 608 bp downstream of the start codon and ended 130 bp downstream of the stop codon. A digoxigenin-labeled probe (digoxigenin labeling kit, Roche Molecular Biochemicals, Mannheim, Germany) derived from this clone was used to screen a human total brain cDNA library (CLONTECH, Palo Alto, CA). Several overlapping positive clones were isolated and sequenced. One of them had an insert length of 857 bp starting 2 nucleotides upstream of the start codon and covering 855 bp of the human PARP-2 open reading frame. These clones were used to construct a full-length human PARP-2 cDNA.
Fluorescence in Situ Hybridization (FISH) Analysis-Human chromosomes were prepared from human peripheral blood lymphocytes immediately after incorporation of bromodeoxyuridine. Mouse chromosomes were prepared from normal mouse fibroblast cultures. The mouse PARP-1 gene was identified with a 5-kilobase clone (of the mouse PARP-1 gene), which was labeled by nick-translation with biotin-11-dUTP (Sigma, France). The PARP-2 gene was identified in the mouse by a probe biotinylated by PCR consisting of the full-length gene; and in the human by a 1.3-kilobase clone of human PARP-2 cDNA, namely, EST AA595596, which was labeled by nick-translation with biotin-11-dUTP (Sigma, France). A standard hybridization procedure was used as described previously (19). The mouse PARP-1 probe was used at a concentration of 15 ng/ml in the presence of a 100-fold excess of mouse Cot-1 DNA; the human PARP-2 probe was used at a concentration of 20 ng/ml and the PCR probe of mouse PARP-2 at a concentration of 1 ng/ml in presence of 600-fold excess of mouse Cot-1 DNA, in 15 ml of hybridization buffer for each slide. Direct banding of bromodeoxyuridinesubstituted human chromosomes was obtained by incubation in an alkaline solution of p-phenylenediamine (PPD11) (20) and staining with propidium iodide. Mouse chromosomes were stained with DAPI and identified by computer-generated reverse DAPI banding. Immunochemical detection of hybridization was performed using goat antibiotin antibodies (Vector laboratories, Burlingame, CA) and rabbit FITC-conjugated anti-goat antibodies (Biosys, Compiègne, France). Metaphases were observed under a fluorescent microscope (DMRB, Leica, Germany). Images were captured using a cooled photometrics CCD camera and Quips-smart capture software (Vysis).
RNA Preparation and Northern Blot Analysis-Poly(A) ϩ mRNA was purified from PARP-1 ϩ/ϩ and PARP-1 Ϫ/Ϫ 3T3 cells using the messenger RNA isolation kit from Stratagene (La Jolla, CA). Total mouse tissues RNA was purchased from Ambion (Austin, TX). Two micrograms of poly(A) ϩ mRNA and 10 g of total RNA were fractionated on 1% agarose, 2.2 M formaldehyde gels and transferred to Hybond N nylon membrane (Amersham). The murine PARP-2 probe corresponded to the 800-bp internal EcoRI fragment. The blot was hybridized for 16 h with probe labeled by the random priming method (1 ϫ 10 6 cpm/ml) (21) in ExpressHyb mixture (CLONTECH, Palo Alto, CA), washed, and autoradiographed at Ϫ80°C for 48 h.
Generation of Antibodies against PARP-2-Two rabbits were immunized by intramuscular injections of 200 g of purified mPARP-2 catalytic domain (aa 204 -559) in the presence of complete Freund's adjuvant for the first inoculation (day 0) and incomplete Freund's adjuvant for subsequent inoculations on days 15, 30, 45, and 60. The rabbits were bled every fortnight until week 14, beginning a week after the second injection. This antibody named YUC, raised against the PARP-2 catalytic domain, recognizes both human and murine PARP-2, but not PARP-1.
Cloning and Expression of Nter-PARP-2 Fused to GST or GFP-A 180-bp EcoRI-EcoRI fragment corresponding to aa 1-69 of mouse PARP-2, and named mouse Nter-PARP-2, was generated by PCR from pVL-mPARP-2-F with the 5Ј-oligonucleotide GGATCCCGGGAATTCG-GATGGCGCCGCGGCGGC and the 3Ј-oligonucleotide GGCTTTGC-CCGAATTCTTTAACAGCAAGGTCT and cloned either into pGEX-2T expression vector or into pEGFP-C3 in-frame with the GST or GFP reading frame, respectively. The GST fusion protein was overexpressed and purified using glutathione-Sepharose 4B (Pharmacia Biotech Inc.) according to the specifications of the manufacturer. The pEGFP-Nter-PARP-2 construction was used to transfect HeLa cells, and transient expression of the GFP-Nter-PARP-2 fusion was monitored by fluorescence microscopy, 24 -48 h after transfection.
Overproduction and Purification of the Murine Recombinant PARP-2-pVL PARP-2 and the Baculogold 228 linearized baculovirus DNA (Pharmingen) were co-transfected into Sf9 cells according to the manufacturer's instructions. Cell propagation and protein production was performed as described previously (22). The identity of the purified protein was confirmed by Western blot with the polyclonal antibody against PARP-2 (1:2000 dilution). Purification of PARP-2 was performed as described previously (23) for the purification of chicken PARP-1 catalytic domain with some modifications. Briefly, the cell pellet (1.5 ϫ 10 9 cells) was homogenized in 75 ml of 100 mM Tris-HCl, pH 7.5, 0.2% Tween 20, 0.2% Nonidet P-40, 14 mM ␤-mercaptoethanol, 10 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 M NaCl and an anti-protease mixture (complete 228, Mini, Roche Molecular Biochemicals). After sonication, (1 min at 30% of the maximum output using a Branson SONIFIER large probe), the cell lysate was cleared by centrifugation at 50,000 ϫ g for 90 min. After precipitation with protamine sulfate (1 mg/ml), and clarification by centrifugation at 50,000 ϫ g for 25 min, the proteins precipitating at 35% ammonium sulfate saturation were eliminated by centrifugation (20,000 ϫ g for 20 min). Further ammonium sulfate was added to the supernatant to 70% saturation and the pellet was collected after a 20-min centrifugation at 20,000 ϫ g. The proteins were resuspended in 150 ml of 100 mM Tris-HCl, pH 7.5, 14 mM ␤-mercaptoethanol, 10 mM EDTA, 1 mM phenylmethylsulfonyl fluoride containing the antiprotease mixture. This sample was subjected to a 3-aminobenzamide Affi-Gel 10 column chromatography. The elution was performed with 3-methoxybenzamide and fractions containing the polypeptide were concentrated on an ultrafiltration membrane Diaflo YM30 (Amicon, Inc.). A sample of each step of the purification was stored for further analysis on SDS-PAGE and quantification of proteins by the method of Bradford (24).
Western Blot, Southwestern Blot, and Estimation of Amount of hPARP-2-Immunoblots and Southwestern blot were carried out as described previously (25). The number of PARP-2 molecules per cell was estimated in a Western blot experiment by comparing the intensity of the immunoreaction of anti-PARP-2 antibody on PARP-2 (YUC) from 400,000 HeLa cells with increasing amounts of purified PARP-2.
Stimulation by DNA Strand Breaks-Samples were incubated for 10 min at 25°C in assay buffer (100 l) consisting of 50 mM Tris-HCl, pH 8.0, 4 mM MgCl 2 , 0.2 mM dithiothreitol, with or without 200 ng of calf thymus DNA previously treated with DNase I and 400 M [␣-32 P]NAD ϩ (100 nCi/nmol). The reaction was stopped by the addition of 5% (w/v) trichloroacetic acid containing 1% (w/v) inorganic phosphate, and the acid-insoluble radioactivity was washed 3 times in the same solution and once in 95% EtOH and the radioactivity was measured. The stimulation of PARP activity by DNA treated with DNase I was expressed as the ratio of the activity of PARP with activated DNA to the activity of PARP without DNA.
Auto Poly(ADP-ribosylation) Reaction-800 ng of purified PARP-2 was incubated with 200 ng of calf thymus DNA previously treated with DNase I or without DNA, in 100 l containing 100 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 10 mM dithiothreitol, and 800 nM [␣-32 P]NAD ϩ (100 nCi/nmol) (26). After the indicated time of incubation at 25°C, the reaction was stopped with ice-cold acetone (80% v/v) and incubated for 30 min at Ϫ20°C. Insoluble material was pelleted by centrifugation for 20 min at 4°C, washed once with 100% acetone, once with watersaturated ether, and dried. The pellet was resolubilized in 50 l of 1 ϫ Laemmli buffer and analyzed on 8% SDS-PAGE. Gels were stained with Coomassie Blue, destained, dried, and autoradiographed on Kodak Bio-Max MS film.
Polymer Analysis-To analyze the polymer synthesized by either PARP-1 or PARP-2, samples were incubated in a standard enzymatic assay with 400 M [␣-32 P]NAD ϩ . Radioactive proteins were treated with 100 mM NaOH, 20 mM EDTA for 1 h at 60°C (27). The solution was then neutralized with 100 mM HCl. One volume of phenol/chloroform (1:1) was added and remaining traces of phenol/chloroform were extracted from the aqueous layer three times with diethyl ether. The polymer was then precipitated twice with ethanol and the pellets were dissolved in water. The polymer was either analyzed on a sequencing gel or treated with snake venom phosphodiesterase and analyzed by two-dimensional thin-layer chromatography according to Keith et al. (28). The radioactive spots on the TLC were scraped from the thin layer and 32 P label was determined by scintillation counting. The average polymer size and the branching frequency were calculated according to Miwa and Sugimura (29).
Determination of Kinetic Parameters-Samples were incubated for 10 min at 25°C in assay buffer (100 l) consisting of 50 mM Tris-HCl, pH 8.0, 4 mM MgCl 2 , 200 M dithiothreitol, 200 ng of calf thymus DNA previously treated with DNase I and various concentrations of [␣-32 P]NAD ϩ (100 nCi/nmol). The reaction was stopped by the addition of 5% (w/v) trichloroacetic acid, containing 1% inorganic phosphate, and the acid-insoluble radioactivity was washed 3 times in the same solution and once in 95% EtOH and the radioactivity measured.

RESULTS
Poly(ADP-ribose) Synthesis in PARP Ϫ/Ϫ Mouse Embryonic Fibroblasts-In a previous paper (5) we described the inactiva-tion of the PARP-1 gene, in mouse by homologous recombination. Gene disruption was assessed by Southern blotting and by Western blotting using a panel of specific monoclonal and polyclonal antibodies that failed to detect the full-length PARP-1 or any of its functional NH 2 -terminal or COOH-terminal domains. Using low doses of damaging agents (H 2 O 2 or MMS) that are known to trigger PARP-1 activity in wild-type (wt) cells, no ADP-ribose polymers were detected in PARP-1 Ϫ/Ϫ cells (data not shown). However, high doses of damaging agents were able to trigger poly(ADP-ribose) synthesis in PARP-1 ϩ/ϩ cells (Fig.  1, G) as well as in PARP-1-deficient cells (Fig. 1, C and E), as shown by immunofluorescence. This result suggests that there is a poly(ADP-ribose) polymerase activity distinct from PARP-1, that is activated by DNA damage.
To confirm that poly(ADP-ribose) polymerase activity was present in PARP-1 Ϫ/Ϫ cells and evaluate its contribution, a quantitative and qualitative analysis of the reaction products was performed. Whole cell extracts were prepared from spleen, testis, primary or 3T3 embryonic fibroblasts of PARP-1 ϩ/ϩ and PARP-1 Ϫ/Ϫ mice, and tested for PARP activity. The results displayed in Fig. 2A show that all PARP-1 Ϫ/Ϫ cells tested display 5 to 10% of total PARP activity stimulated by DNA strand breaks, compared with wt cells. This residual activity is inhibited by 2 mM 3-aminobenzamide supporting the idea of a new PARP enzyme activity.
The reaction products synthesized in the presence of [␣-32 P]NAD ϩ were characterized by removing the radiolabeled material from the acceptor proteins and fractionating by electrophoresis on long 20% denaturating polyacrylamide gels. The distribution of ADP-ribose polymers synthesized in wt and PARP-1 Ϫ/Ϫ cells are similar (Fig. 2B). The reaction products were further characterized by two-dimensional TLC after polymer hydrolysis by snake venom phosphodiesterase (28). No significant differences were observed in the products, PRAMP, (PR) 2 AMP, and AMP synthesized by each cell genotype (data not shown). These results demonstrate that PARP-1 Ϫ/Ϫ cells exhibit a bona fide poly(ADP-ribose) polymerase activity presumably associated with novel PARP protein(s), confirming a previous report by Shieh et al. (15).
Cloning of the Full-length Murine PARP-2 (mPARP-2)-In plant cell nuclei, PARP activity has been reported to be associated with a protein of approximately 113 kDa (30). However, in A. thaliana, the cloning of a PARP homologue (11) revealed a protein, called APP, with a theoretical M r of 72,000. This protein showed a high similarity (60%) to the catalytic domain  3T3 cells (A-F). Cells were mock exposed (A and B) or exposed to 1 mM H 2 0 2 for 10 min (C and D), or to 1 mM MMS for 30 min (E-H). Poly(ADP-ribose) was detected with the 10H monoclonal antibody (43) followed by the fluorescein isothiocyanate-conjugated anti-mouse antibody (A, C, E, and G). Nuclei were stained with DAPI (B, D, F, and H).
of vertebrate PARP, but was completely divergent in its NH 2terminal extremity where it harbored a helix-loop-helix domain. The existence of two distinct PARP genes in plants was definitely confirmed by the cloning of two different maize genes coding for a classical PARP with zinc fingers of 110 kDa (ZAP) and a structurally non-classical PARP (NAP, homologous to APP) of 72 kDa, respectively (12,14). Therefore, we undertook a search for the mammalian equivalent of APP and NAP. Expressed sequence tags (EST) of the GenBank/EMBL data bases (dbEST) were screened with the APP primary sequence. Several EST from mouse and human were identified which shared homology with the catalytic domain of APP, while distinct from PARP-1. These sequences were candidates for a new mammalian PARP homologue, possibly responsible for the poly(ADP-ribosylation) activity detected in PARP-1 Ϫ/Ϫ cells. The murine EST were used to screen a mouse ES cell cDNA library. We thus were able to construct the complete cDNA (GenBank™ accession number AJ007780). Simultaneously, the cloning of the homologous human cDNA (GenBank™ accession number AJ236912) was undertaken following a similar strategy. We propose to name this new gene poly(ADP-ribose) polymerase-2 (PARP-2).
Chromosomal Localization-The chromosomal localization of the human PARP-2 gene was identified by FISH on human chromosomes using a human PARP-2 probe. Consistent signals on chromosome 14, band 14q11.2 were identified (Fig. 3A); 100% of 25 metaphases showed at least one signal on chromosome 14 at 14q11.2; 24% of the metaphases showed signals on both chromosomes 14 on the same position. Similarly, the murine PARP-2 gene was mapped using a murine PARP-2 probe. FISH on mouse chromosomes exhibited consistent signals on chromosome 14, band 14C1 (Fig. 3B); 70% of 25 metaphases showed signals on 14C1; 10% had double signals on chromosome 14 band 14C1, and 20% showed signals on the two homologues at the same position. As a control, chromosomal localization of the murine PARP-1 gene was performed. FISH on mouse chromosomes exhibited consistent signals on chromosomes 1, band 1 H5 (Fig. 3C); 30 metaphases were observed, of which 74% showed signals in this position; 41% showed one signal on both chromosomes 1, 26% showed double signals on one of the two homologues. Altogether, these results confirm the existence of two distinct and unique genes coding for two different PARP molecules. A strict synteny was observed between man and mouse in the two chromosome regions contain-ing the PARP-1 gene as well as those containing the PARP-2 gene.
Expression of mPARP-2 mRNA-Northern blot analysis performed on poly(A) ϩ mRNAs of 3T3 cells derived from PARP-1 ϩ/ϩ and PARP-1 Ϫ/Ϫ mice and on total RNAs from mouse tissues (Fig. 4) revealed a PARP-2 transcript at an apparent molecular size of 2.0 kilobase. The same amount of transcript was detectable in both PARP-1 ϩ/ϩ and PARP-1 Ϫ/Ϫ cell lines, indicating that there is no compensation for the PARP-1 deficiency in PARP-1 Ϫ/Ϫ cells by up-regulation of PARP-2 gene expression. The tissue distribution of PARP-2 showed at least a basal expression in all tissues and higher expression in germline. Moreover, Northern blot analysis on total mRNA from HeLa cells treated or untreated with genotoxic agents such as UV-B (500 J/m2), UV-C (20 J/m 2 ), N-methyl-NЈnitro-N-nitrosoguanidine (50 M), and H 2 O 2 (0.5 mM) revealed that the level of PARP-2 mRNA is not increased following genotoxic stress (data not shown), like for PARP-1.
Structural Organization of mPARP-2-The complete nucleotide sequence of murine PARP-2 predicts a full-length cDNA clone of 1,707 bp containing 16 bp of 5Ј-noncoding sequence which has been confirmed by sequencing the 5Ј region of the genomic DNA, 3 a single open reading frame of 1,680 bp, and a 3Ј-noncoding region of 125 bp. The deduced amino acid sequence predicts a protein of 559 amino acids and a molecular mass of 62 kDa. The amino acid sequences of human and murine PARP-2s (hPARP-2 and mPARP-2, respectively) can be aligned with human PARP-1 and the two plant PARP enzyme sequences APP and NAP (12,14) (Fig. 5A). The alignment shows that the carboxyl-terminal regions of PARP-2 are highly conserved in comparison with the sequence of PARP-1 and the plant enzymes at their catalytic domain (43% identity between hPARP-1 and hPARP-2 catalytic domains). Moreover, it contains the PARP signature corresponding to the ADP-ribose donor site (aa 401-450) in mPARP-2 as well as the crucial residues forming the acceptor site (E534, K445, Y532, N448, H370, L530, M432), including the catalytic glutamate at position 534 (31). Interestingly, the blocks of sequence conservation (Fig. 5A) correspond strictly to the secondary structure of the PARP-1 catalytic domain, whereas, as frequently observed, sequence variability occurs mainly in loops (data not shown). A short variable region (aa 195-207) contains glutamate residues that may be potential polymer acceptors sites. This new PARP is then expected to have similar catalytic properties to PARP-1. It is noted that, the NH 2 -terminal part of PARP-2 (aa 1-64) has no significant homology with any other previous PARP. Interestingly, the NH 2 terminus region of human and mouse PARP-2 shows a higher sequence variability compared with the highly conserved COOH terminus catalytic region (62% identity between the NH 2 terminus of mPARP-2 and hPARP-2) and no homology with APP. The catalytic domains of mPARP-2 and hPARP-2 or APP share 87 or 47% identity, respectively. However, this region contains basic residues that could bear potential DNA-binding properties. On the other hand, these basic residues could be involved in the nuclear and/or nucleolar targeting of the protein (32). Therefore the structural organization of PARP-2 is reminiscent of PARP-1, in the sense that it is modular, made up of two distinct domains, one responsible for the catalytic function, the other involved in putative DNA binding function (Fig. 5B).
polypeptides of 62 and 40 kDa, respectively. These apparent molecular masses corresponded with the predicted values deduced from the respective cloned mPARP-2 cDNA. These proteins were purified following a procedure similar to that used for chicken PARP-1 overexpressed in baculovirus (23) (Fig. 6, A  and B). Typically, 1 liter of cell culture yielded 10 mg of highly purified 40-kDa catalytic domain or 62-kDa full-length mPARP-2. The purified catalytic domain was used to raise a polyclonal antibody (YUC) that did not cross-react with PARP-1 despite the high homology between their catalytic domains (data not shown). A Western blot analysis using this antibody showed that the apparent molecular weight of the purified full-length mPARP-2 was identical to that of the endogenous protein from 3T3 cell extract (Fig. 6C). This result demonstrated that the size of the endogeneous protein of 3T3 cells is identical to the protein encoded by the open reading frame from mPARP-2 cDNA. In addition, the abundance of PARP-2 in 3T3 and HeLa cells was determined (see "Experimental Procedures") and indicated about 200,000 molecules per cell (data not shown) which is comparable to that of PARP-1 (33,34).
Nuclear Location of mPARP-2-In PARP-1-deficient 3T3 cells, the poly(ADP-ribose) synthesized after DNA damage is found in the nucleus (Fig. 1) favoring a nuclear location for the PARP(s) involved in that process. The cellular distribution of mPARP-2 was addressed by immunofluorescence using the polyclonal anti-mPARP-2 antibody (YUC) in PARP-1 Ϫ/Ϫ 3T3 cells. As expected PARP-2 accumulated in the cell nucleus with an apparent peripheral location (Fig. 7A). In the NH 2 -terminal part of PARP-2, the presence of basic motifs could be responsible for nuclear targeting of the protein. To address this possibility, aa 1-69 were fused to GFP and the fusion protein was expressed in HeLa cells. While GFP is cytoplasmic (Fig. 7C) the addition of aa 1-69 of mPARP-2 is sufficient to target the fusion protein to the nucleus (Fig. 7E). This confirmed that the nuclear location module of mPARP-2 is in a domain distinct from the catalytic domain and located in the NH 2 -terminal part of the protein, similar to the structure previously described in PARP-1 (35).
mPARP-2 Binds To and Is Activated by DNase I-damaged DNA-The activity of PARP-1 is dependent on the presence of DNA. To test the hypothesis that PARP-2 shares a similar property, purified protein was incubated with [␣-32 P]NAD ϩ in the absence or presence of calf thymus DNA treated with DNase I. Fig. 8A shows that the activity of PARP-2 is stimulated by a factor of 15, while PARP-1 incubated in the same conditions, is stimulated by a factor of 50. Even though the activation of PARP-2 is lower than for PARP-1, this result shows clearly that PARP-2 is also activated by DNA which has been treated by DNase I.
Since, in all the previously known PARP-1 proteins the DNAbinding function is harbored in a domain distinct from the catalytic domain, we hypothesized that the NH 2 -terminal part of PARP-2 might contain a DNA-binding property. Thus, the NH 2 -terminal domain of mPARP-2 (aa 1-69) was expressed as a fusion product with GST (GST-NterPARP-2) in Escherichia coli, purified, and tested for its DNA binding capacity by Southwestern analysis (Fig. 8B). Both the purified full-length mPARP-2 and the GST-NterPARP-2 proteins bound 32 P-labeled DNA activated by DNase I confirming that PARP-2 is a DNA-binding protein and that aa 1-69 are sufficient for this function, in addition to the function of a nuclear location signal. Thus, this small domain carries two features important for the biological role of PARP-2: targeting to the nucleus and binding to DNA.
Autopoly(ADP-ribosylation) of mPARP-2-To demonstrate the ability of mPARP-2 to synthesize ADP-ribose polymers, the autopoly(ADP-ribosylation) of mPARP-2 was examined by following the electrophoretic mobility of the enzyme incubated with [␣-32 P]NAD ϩ and DNA activated by DNase I. Fig. 9A shows the autoradiographic profile of the PARP-(ADP-ribose) conjugates generated at the indicated times. PARP-2 is effi- ciently automodified in the presence of DNA strand breaks and the electrophoretic mobility of the conjugates decreases with the time of incubation. This presumably reflects the increase of the polymer size and complexity as the incubation progresses.
When the automodification reaction is performed using higher NAD ϩ concentrations (close to the K m value or above,) most of the resulting reaction products remain concentrated in the wells of the gel (data not shown). To characterize the product of reaction, the radioactivity associated with the protein was resolved on a 20% denaturating polyacrylamide gel. The "ladder" characteristic of ADPribose polymers was observed for mPARP-2 and the polymer lengths were comparable to that of hPARP-1. The structure of the product was further characterized by two-dimensional TLC following polymer digestion by snake venom (28). The three products expected, AMP, PRAMP, (PR) 2 AMP, were obtained indicating the similarity between the polymer synthesized by mPARP-2 and hPARP-1 in terms of size, chain length, and branching frequency (29).
Kinetic Parameters of mPARP-2-To compare the catalytic properties of mPARP-2 with that of hPARP-1, their enzymatic activities were assayed under the standard conditions. Analysis by the Lineweaver-Burk plot estimated a K m of 130 M for mPARP-2, which represents an affinity for NAD ϩ 2.6-fold lower than hPARP-1 (50 M). A k cat of 42 ϫ 10 Ϫ3 s Ϫ1 was calculated assuming first-order kinetics, where the only variable was the substrate NAD ϩ , because PARP-2 is both enzyme and second substrate. PARP-2 is then a much less efficient enzyme in terms of catalysis than hPARP-1; the k cat /K m ratio for mPARP-2 is 323 s Ϫ1 M Ϫ1 which is 18 times lower than that of hPARP-1 (6,000 s Ϫ1 M Ϫ1 ). Thus, this activity measured in vitro compares well with the 5-10% residual activity found in PARP-1 Ϫ/Ϫ cell extracts stimulated by DNA strand breaks ( Fig. 2A). DISCUSSION In this work, which was initiated to resolve the intriguing persistence of poly(ADP-ribose) formation in PARP-1 Ϫ/Ϫ cells, we have identified a new member of the PARP family, PARP-2. Human and mouse PARP-2 genes were mapped to chromosome 14q11.2 and 14C1, respectively, which are distinct from PARP-1 loci, supporting the conclusion that PARP-2 is coded by a different gene.
PARP-2, together with the plant enzymes APP and NAP, belongs to a class of PARP molecules characterized by (i) a short, variable and basic NH 2 -terminal domain, ranging from 64 in PARP-2 to 140 residues in length in APP and NAP, bearing both the DNA-binding element and a nuclear location signal and (ii) a highly conserved COOH terminus domain highly homologous to the PARP-1 catalytic domain.
The alignment of the sequences shows that in the catalytic domain all the residues crucial for the activity (initiation, elongation, and branching) are conserved suggesting that the catalytic function should also be conserved in PARP-2. Indeed, the analysis of the enzyme products demonstrated this functional conservation between both enzymes (PARP-1 and PARP-2). Looking for protein substrates of PARP-2, we noticed that purified histones were not polymer acceptors in heteromodification reactions (data not shown). However, we cannot rule out the existence of physiological acceptors for PARP-2. PARP-2 automodifies itself and possesses glutamate residues that could play the role of acceptor sites. In PARP-1 the ADP-ribose acceptor sites are mainly located in the automodification domain D. Although this domain is absent in PARP-2, automodification takes place efficiently, indicating that this mode of regulation has also been conserved. In PARP-1, domain D contains a BRCT motif (36 -38) that is, together with domain A, the major protein-protein contact interface for interaction with partners (8,39). The absence of these motifs in PARP-2 suggests that if it interacts with protein partner(s), this interaction should be driven by some other mechanisms involving a binding module that still has to be discovered.
Perhaps the most intriguing feature of PARP-2 is its binding to DNA and its activation by DNA that has been treated with DNase I, despite the absence of the characteristic zinc finger module which acts as a nick sensor in PARP-1 (5,40). The DNA-binding domain in mPARP-2 encompasses aa 1-64 but does not present any obvious DNA binding motif. However, it is rich in basic amino acids (27% Lys or Arg), which are likely to be involved in this function. Given that DNA treated with DNase I contains a complex mixture of DNA ends, footprinting experiments with defined DNA probes will be necessary to further understand the DNA-dependent activation of PARP-2. Even though, if PARP-1 and PARP-2 respective DNA-binding domains interact in a different manner with different DNA structures, we may speculate on a possible conservation of the activation mechanism based presumably on a conformational change of the active site loop induced by DNA binding (31).
Despite major differences between PARP-1 and PARP-2 including the smaller size, the absence of zinc fingers and BRCT domains, PARP-2 like PARP-1 is targeted to the nucleus, binds to and is activated by DNA which has been treated by DNase I which in turn stimulates poly(ADP-ribose) synthesis. While the catalytic function is structurally and mechanistically conserved between the two enzymes, their physiological role most probably differs. Their functional specificity is presumably determined by their variable NH 2 -terminal modules.
While this work was in progress, a third member of the PARP family, tankyrase, was identified and localized to human telomeres (41). Tankyrase is a 142-kDa protein having similarity to ankyrins and to PARP-1 catalytic fragment. Tankyrase is able to synthesize poly(ADP-ribose) but apparently independently of the presence of DNA. TRF1 which is a negative regulator of telomere maintenance can be modified in vitro by tankyrase. Again, one can imagine that the function of tankyrase on telomeres is regulated by modules distinct from the catalytic region.
The existence of a family of PARP proteins raises a number of important questions with regards to their specific functions. We cannot exclude the possibility that they could function as backups in the same cell survival pathway. One has also to realize that the global PARP activity measured in a damaged cell may in fact represent the addition of at least two distinct enzymatic activities (PARP-1, PARP-2, and perhaps more). The disruption of the PARP-2 gene will be necessary to elucidate both its physiological role during DNA damage and repair, and a possible functional redundancy.
All these new PARP proteins share a similar catalytic site and were shown to be inhibited by 3-aminobenzamide (41,42) (this work, Fig. 2A). This common inhibition may explain some "side effects" while attempting to inhibit PARP-1 with inhibitors that were thought to be specific for this enzyme, possibly by interfering with different biological functions related to the other PARP. Under these conditions, the pharmacological inhibition of PARP activity under pathological conditions will certainly require detailed understanding of the specific role of the different PARP family members in vivo as well as detailed crystal structures of their catalytic sites.