p21CDKN1A Regulates the Binding of Poly(ADP-Ribose) Polymerase-1 to DNA Repair Intermediates

The cell cycle inhibitor p21CDKN1A was previously found to interact directly with DNA nick-sensor poly(ADP-ribose) polymerase-1 (PARP-1) and to promote base excision repair (BER). However, the molecular mechanism responsible for this BER-related association of p21 with PARP-1 remains to be clarified. In this study we investigate the capability of p21 to influence PARP-1 binding to DNA repair intermediates in a reconstituted BER system in vitro. Using model photoreactive BER substrates containing single-strand breaks, we found that full-length recombinant GST-tagged p21 but not a C-terminal domain truncated form of p21 was able to stimulate the PARP-1 binding to BER intermediates with no significant influence on the catalytic activity of PARP-1. In addition, we investigate whether the activation of PARP-1 through poly(ADP-ribose) (PAR) synthesis, is required for its interaction with p21. We have found that in human fibroblasts and in HeLa cells treated with the DNA alkylating agent N-methyl-N'-nitro-N-nitrosoguanidine (MNNG), the interaction of p21 with PARP-1 was greatly dependent on PAR synthesis. In fact, an anti-PAR antibody was able to co-immunoprecipitate p21 and PARP-1 from extracts of MNNG-treated cells, while blocking PAR synthesis with the PARP-1 inhibitor Olaparib, drastically reduced the amount of p21 co-immunoprecipitated by a PARP-1 antibody. Our results provide the first evidence that p21 can stimulate the binding of PARP-1 to DNA repair intermediates, and that this cooperation requires PAR synthesis.

Transfection of HeLa cells with selected plasmids was performed with Effectene reagent (Qiagen). GST-tagged constructs coding for full-length (FL) PARP-1, the N-terminal region encompassing the DNA binding and auto-modification domain (A-D) and the C-terminal region containing the catalytic domain (E-F) were obtained from Josiane and Gilbert de Murcia (CNRS-ESBS, Illkirch, France), as previously described [40]. The vector used for the expression of p21 fused to GFP (p21-GFP) has also been previously described [46]. Recombinant GSTtagged human p21 full length (p21FL), or a truncated form containing the C-terminal domain (p21Cter), were expressed in bacteria and purified, as previously reported [47].
The plasmids bearing cDNAs of human PARP-1 were kindly provided by Dr. M. Satoh (Laval University, Canada), the expression vectors for rat pol ß and human APE1 were a generous gift of Dr. S. H. Wilson (NIEHS, NIH, USA). The recombinant proteins PARP-1, APE1 and pol ß were overexpressed in E. coli and purified, as previously described [48].

DNA substrates and 5'-end labeling
Oligodeoxyribonucleotides were 5'-32 P-phosphorylated with T4 polynucleotide kinase and purified by 20% polyacrylamide/7.0 M urea gel electrophoresis as described [42], followed by electro-elution and precipitation with 2% solution of LiClO 4 in acetone. The precipitated oligodeoxyribonucleotides were dissolved in 10 mM Tris-HCl (pH 8.0) and 1 mM EDTA (TE buffer). Complementary oligonucleotides were annealed by heating (90°C, 5 min) a solution of equimolar amounts, followed by slow cooling-down to room temperature, to form the doublestranded DNA substrates containing nick (DNApF and DNAp), uracil (DNA-U) or 3-hydroxy-2-hydroxymethyltetrahydrofuran (DNA-F). The structures of DNA duplexes used are summarized in Table 1.

Preparation of photoreactive nicked DNA substrates
To generate photoactive DNA Ã pF or DNA Ã p (Table 1), the standard reaction mixtures (300 μl) contained 50 mM Tris-HCl (pH 8.0), 50 mM NaCl, 10 mM MgCl 2 , 0.1 μM pol ß and 70 μM FABGdCTP, 0.8 μM 5'-[ 32 P]-labeled DNA containing one nucleotide gap with 5'-phosphate or 5'-furanophosphate. The one nucleotide gapped DNA substrates were generated by treating DNA-F with 20 nM APE1, DNA-U with 20 nM UDG and 20 nM APE1 before the reaction. The reaction mixtures were incubated for 15 min at 37°C to allow pol ß-dependent elongation of the primers with FABGdCMP and the reaction was then stopped by adding EDTA to a 20 mM final concentration, followed by precipitation by adding 1/10 volume of 3 M Na Acetate (pH 5.0), and 2.5 volumes of 96% ethanol. The photoreactive DNAs (DNA Ã pF or DNA Ã p) were dissolved in TE buffer at a final concentration of 2 μM.

Photocrosslinking assay
Crosslinking of PARP-1 was carried out in reaction mixtures (20 μl) that contained 40 mM Tris-HCl (pH 7.5), 40 mM NaCl, 10 mM MgCl 2 , 0.1 μM (DNA Ã pF or DNA Ã p) and 0.1 μM human recombinant PARP-1 or BTNE (1.25 mg/ml). To these mixtures, 0.5 mM NAD + , 1.2 μM p21 or p21Cter were added, as indicated in the figure legends. The reaction mixtures were incubated at 37°C for 3 min then irradiated by UV light (λ = 312 nm, 1.5 J/cm 2 ) on ice. A Bio-Link BLX-312 cross-linker (Vilber-Lourmat) was used as light source in all experiments.  Reactions were stopped by adding SDS-sample buffer and heating for 5 min at 96°C. Photocrosslinking products were separated in a 10% SDS-PAGE. The gels were dried and subjected to phosphor imaging for quantification using Molecular Imager/Quantity One software (Bio-Rad, USA) or "Typhoon" (Amersham Pharmacia Biotech, USA).

Chromatin fractionation
To detect chromatin-bound PARP-1 in the presence of Olaparib, human p21 +/+ and p21 −/− fibroblasts, cells were first exposed to MNNG as indicated previously and then collected at the end of treatment (t 0) or re-incubated in the presence or in the absence of Olaparib for further 30 min. Finally, cells were separated in soluble and chromatin-bound fractions and proteins in the latter fraction were released by DNase I digestion [49].

p21 stimulates PARP-1 binding to DNA BER intermediates in vitro
Previous studies have shown that p21 and PARP-1 can cooperate in the regulation of BER [40], and that PARP-1 can be selectively photocrosslinked by BER intermediates containing a nick in mouse embryonic fibroblast (MEF), and in bovine testis nuclear extracts (BTNE) [12,42,51]. In this study, we have applied the photocrosslinking assay to investigate the influence of p21 on the interaction of PARP-1 with DNA repair intermediates in cell extracts. The photoreactive substrates were pre-synthesized by pol ß via incorporation of the FABGdCMP moiety onto the 3'-end of upstream primer and contained a nick with a 3-hydroxy-2-hydroxymethyltetrahydrofuran with 5-phosphate (5'-pF) (DNA Ã pF) or a 5-phosphate group (DNA Ã p) ( Table 1). Given that the 5'-pF group cannot be removed by the lyase activity of pol ß [51], the DNA Ã pF can be considered as the central intermediate of BER at the stage just prior to 5 0 -deoxyribose phosphate lyase activity of pol ß in the short patch BER, or as an early intermediate in the long patch BER. The DNA Ã p may be considered as the penultimate product in the BER pathways [42].
First, we investigated whether p21 could influence PARP-1 crosslinking to nicked DNA duplexes in BTNE. To this purpose, the photoreactive DNA Ã pF or DNA Ã p were incubated with BTNE proteins in the presence of p21 full-length (p21), or a truncated form lacking the first N-terminal 74 amino acids (p21Cter) (Fig 1A), and then exposed to UV light to induce crosslinking (Fig 1B). Addition of p21 to BTNE resulted in a noticeable increase in the level of endogenous PARP-1 crosslinking to both DNA Ã pF and DNA Ã p (Fig 1B, lanes 2 and 5). In contrast to p21, addition of p21Cter did not influence the labelling of PARP-1 with the DNA substrates (Fig 1B, lanes 3 and 6). An even more pronounced effect of p21 on labelling of PARP-1 was observed for purified PARP-1 protein (Fig 1C and 1D). P21 significantly increased the p21 CDKN1A Stimulates DNA Binding of PARP-1 crosslinking of purified PARP-1 to nicked DNA Ã pF, as well as to DNA Ã p, by more than 50% (P<0.05) (Fig 1D). These results indicate that p21 can influence the PARP-1 binding to nicked BER intermediates in vitro. In addition, this effect was more pronounced on the DNA substrate containing a nick with the 5'-pF group. On the whole, the photocrosslinking assay provides evidence that p21 stimulates the binding of PARP-1 to DNA nicks.  8). Then the mixtures were UV irradiated (lanes 2-4 and 6-8). Free DNA probe is also shown. The products were separated on 10% SDS-PAGE and analyzed by PhosphorImaging. The reaction conditions and analysis of the reaction products are described in "Materials and Methods". The nucleotide sequence and structure of (DNA*pF and DNA*p) is reported (Table 1). (D) Diagram showing the quantitative analysis results shown in panel C. The crosslinking of PARP-1 was defined as the percentage (%) of photoreactive DNA, which formed covalent adducts with PARP-1. Error bars represent relative mean ± SD, n = 4. Asterisks denote statistically significant difference (*, P<0.05, ns, not significant; t-test). Since the above experiments were performed in the absence of the PARP-1 substrate NAD + (i.e. without PARP-1 catalytic activation), we also tested the effect of p21 on PARP-1 activity on nicked photoreactive DNA in the presence of NAD + . As expected, when the incubation mixture was supplemented with NAD + before UV-light irradiation, PARP-1 labelling was reduced with a concomitant appearance of crosslinking products with slower electrophoretic mobility, and corresponding to labelling of PARylated PARP-1 (Fig 2A, lanes 2 and 6). When p21 was added to the reaction, the labelling of PARylated PARP-1 increased (Fig 2A, lanes 3  and 7), while the effect was not seen with p21Cter (Fig 2A, lanes 4 and 8). By monitoring the influence of p21 on the kinetics of PAR synthesis, in the presence of nicked DNA (DNApF or DNAp) and [ 32 P]-labelled NAD + , we observed that p21 had only a modest or no effect on PARP-1 catalytic activity (Fig 2B and 2C, respectively). Cumulatively, these results indicate that in vitro p21 can modulate PARP-1 binding to DNA BER intermediates, without a net influence on the following PAR synthesis.

Interaction of p21 with PARP-1 is mediated by PAR
Next, we wanted to determine whether in vivo PAR chains formed during BER-dependent SSB production mediated the interaction of p21 with PARP-1. To this purpose, we performed immunoprecipitation with an antibody recognizing PAR polymer in cell extracts obtained from quiescent human fibroblasts, untreated or treated for 30 min with the DNA alkylating agent MNNG. The results showed that in the sample exposed to MNNG, the antibody reacted with PAR polymers corresponding mainly to auto-PARylated PARP-1, detectable as a higher molecular weight band (about 250 kDa) compared to the canonical unmodified form (about 113 kDa) (Fig 3A). Interestingly, in the sample immunoprecipitated from MNNG-treated cells, a band reacting with a p21-specific antibody was also detected. This result prompted us to investigate whether p21 itself may be a PARP-1 substrate, rather than only binding to PAR, as has been suggested in the past [52]. Incubation of recombinant GST-tagged p21 with purified PARP-1 showed that PARylated p21 was barely detectable (Fig 3B, lane 2).
To better characterize the interaction between p21 and PARP-1 mediated by DNA damage-induced PAR synthesis, human fibroblasts were pre-incubated with the PARP-1 inhibitor Olaparib, before treatment with MNNG. Under these conditions, Olaparib blocked MNNG-induced PAR synthesis (Fig 4A, lane 4), as also observed by immunofluorescence staining of PAR polymers with anti-PAR antibody (S1 Fig). Cells were collected and lysed for immunoprecipitation with anti-PARP-1 antibody, followed by Western blot analysis of PARP-1, XRCC1, pol ß, p21 proteins and PAR levels in each immunoprecipitate. Olaparibmediated inhibition of PARP-1 not only prevented its association with p21, but also significantly reduced PARP-1 interaction with other important BER factors, such as XRCC1 and pol ß (Fig 4A, lane 2). To confirm these results, HeLa cells were co-transfected with vectors driving the expression of GST-tagged full-length PARP-1 (FL), the auto-modification and DNA-binding domain (AD), or the catalytic domain (EF), together with p21-GFP [40]. After affinity pull-down with GSH-agarose beads, the interaction between tagged PARP-1 proteins and p21-GFP was analysed by Western blot. The results, confirming the association of p21-GFP with the GST-PARP-1 FL or the AD domain [40], showed that their interaction was greatly abolished by the Olaparib-mediated inhibition of PARP-1 activity (Fig 4B). In agreement with the results obtained on fibroblasts, the interaction of PARP-1 with pol ß, and with XRCC1, was significantly reduced in the presence of Olaparib. To further confirm these results, the co-localization of p21 with chromatin-bound PARP-1 was investigated by immunofluorescence staining of exogenous PARP-1 FL, or the AD domain, with an antibody against the GST tag and by detecting p21-GFP autofluorescence (S2 Fig). In cells exposed to Olaparib before and during treatment with MNNG, the extent of co-localization was significantly reduced by about 50%, as compared with that observed in cells treated only with MNNG ( Fig 4C). Absence of p21 affects PARP-1 recruitment to DNA damage In order to obtain further insight into the requirement of p21 for PARP-1 binding to DNA after damage, both normal p21 +/+ and p21 −/− human fibroblasts were exposed for 30 min to MNNG and then immediately collected or incubated in drug-free medium for further 30 min, both in the presence or absence of Olaparib. The chromatin-bound fraction of each sample was then isolated and samples were analysed by Western blot for PARP-1 binding to DNA lesion sites. In p21 +/+ cells, PARP-1 was recruited to chromatin after DNA damage by MNNG in a time-dependent manner, and bands with slower mobility, relative to PARylated PARP-1, were detected only after long exposure of the blot (Fig 5A, long exp). In the presence of Olaparib, the high MW bands were less detectable (Fig 5A). In contrast, elevated levels of 113 kDa band representing PARP-1 were already observed in the untreated control p21 −/− cells, and further recruitment to chromatin after DNA damage was considerably limited, as compared with p21 +/+ cells. High MW bands could also be detected in p21 −/− (Fig 5A, long exp), but no clear difference could be observed when compared to p21 +/+ cells. Densitometric analysis of Western blot bands revealed that the amount of chromatin-bound PARP-1 (113 kDa band) increased by about 3 times immediately after MNNG treatment. During the post-treatment period, PARP-1 levels further increased in the presence of Olaparib, when compared to the levels observed in its absence in p21 +/+ cells but not in p21 −/− cells (Fig 5B).

Discussion
In the search for PARP-1 partners, we have previously demonstrated the physical association between p21 and the DNA binding domain of PARP-1 [40]. In order to investigate the possible modulation of the basic reactions of PARP-1 by p21 during DNA damage, in the present study we have carried out in vitro experiments using model DNA substrates mimicking BER intermediates which have been previously demonstrated to interact with PARP-1 in MEF extracts [42,53]. We have found that p21 increases the binding of PARP-1 to these BER intermediates, suggesting that p21 could exert a positive modulation of DNA binding by PARP-1 (Fig 1). This is an original observation, supported also by in vivo experiments, which confirmed that the interaction was not restricted to the purified enzyme, but could occur with the endogenous protein as well (Fig 1B).
We then addressed the possible modulation of PARP-1 catalytic activity and observed that p21 does not significantly influence the synthesis of PAR. This is in apparent contrast with the data we reported using a p21 C-terminal peptide, which was able to inhibit PARP-1 auto-modification on activated DNA [38]. However, the present evidence indicating that the full-length p21 protein does not have such an effect could be ascribed to the type of DNA substrate used (Fig 2). In fact, PARP-1 activity on DNA duplexes containing an abasic site, or a nick introduced by APE1 endonuclease, was considerably more efficient than on activated DNA, suggesting that PARP-1 prefers to bind to BER intermediates rather than SSBs.
To further investigate the impact of PAR synthesis on the p21-PARP-1 axis, we performed immunoprecipitation experiments with cell extracts from samples treated with the PARP inhibitor Olaparib. Given that under these conditions the inhibition of PARP-1 activity by Olaparib significantly reduced the association and the co-localization between the two proteins, we can assume that they may interact through PAR. This observation is in agreement with findings indicating that p21 recognizes PAR [52], although a direct physical interaction in the absence of PARP-1 activity may also occur [38,40]. The possibility that p21 itself might be PARylated was investigated, but the extent of this modification appeared to be irrelevant (Fig 4B).
In our previous report, we suggested that p21 could regulate the release of PARP-1 from DNA damage sites, because chromatin-bound PARP-1 persisted longer in MNNG-treated p21 −/− , than in p21 +/+ human fibroblasts [40]. The present results clearly show that p21 regulates chromatin binding of the main PARP-1 form (113 kDa), while no useful information could be obtained from the recruitment of high MW bands, due to their limited amounts in the nuclear extracts. The unusual high basal levels of PARP-1 in the absence of DNA damage could explain why PAR levels were elevated in p21 −/− cells [40,54]. Thus, it is conceivable that persistently high levels of chromatin-bound PARP-1 and PAR, as observed in p21 −/− cells, could be detrimental to further DNA repair events [40].
Finally, it has to be taken into account that, after initial binding of PARP-1 to DNA damaged sites, a second wave of PARP-1 association to DNA occurs, which is dependent on PARP-1 activity [55]. According to this mechanism, p21 could provide help in the feedback-regulated PARylation of PARP-1, through the interaction with auto-modified PARP-1. This hypothesis is in agreement with our present evidence that p21 interaction with PARP-1 is favoured by PAR. After this second wave of PARP-1 recruitment and auto-modification, the decreased affinity for DNA may induce the subsequent PARP-1 release, necessary to avoid the inhibition of pol ß [48].
In conclusion, the above results indicate that not only initial PARP-1 binding to damaged DNA is favoured by p21, but also PARP-1 persistent binding to DNA damaged sites is modulated by p21 thanks to PAR binding. This may help in the feedback-regulated PARylation of PARP-1 [55], which ultimately results in its dissociation from DNA. HeLa cells grown on coverslips were transfected with vectors for the expression of PARP-1-FL, or the AD domain, tagged with GST, and of p21-GFP. Forty-eight h later, cells were incubated with 100 μM MNNG for 30 min, after pre-incubation with 10 μM Olaparib for 4 h. Cells were then processed for hypotonic lysis in situ before fixation, as described [40]. Cells were stained with anti-GST antibody (1:100), and then labelled with a secondary antibody conjugated with Alexa 594 (red fluorescence); p21-GFP was detected by the green fluorescence. DNA was counterstained with Hoechst 33258 (Scale bar, 10 μm). (PDF)

Supporting Information
for FABGdCTP and S.H. Wilson for pol ß and human APE1 vectors. Thanks are also due to Dr. M. Parks for revision of English style.