The Enzymatic and DNA Binding Activity of PARP-1 Are Not Required for NF- (cid:1) B Coactivator Function*

Poly(ADP-ribose) polymerase 1 (PARP-1)-deficient mice are protected against septic shock, diabetes type I, stroke, and inflammation. We report that primary cells from PARP-1 (cid:2) / (cid:2) animals are impaired in (cid:1) B-dependent transcriptional activation induced by different stimuli involved in inflammatory and genotoxic stress signaling. PARP-1 was also required for p65-mediated transcriptional activation. PARP-1 enzymatic inhibitors did not inhibit the transcriptional activation of a (cid:1) B-de-pendent reporter gene in wild type cells. Remarkably, neither the enzymatic activity nor the DNA binding activity of PARP-1 was required for (cid:1) B-dependent transcriptional activation in PARP-1 (cid:2) / (cid:2) cells complemented with different PARP-1 mutants. However, PARP-1 interacted in vitro directly with both subunits of NF- (cid:1) B (p50 and p65), and mapping of the interaction domains revealed that both subunits bind to different PARP-1 domains. Furthermore, a PARP-1 mutant lacking the enzymatic and DNA binding activity interacted comparably to the wild type PARP-1 with p65 or p50. Finally, we showed that PARP-1 is activating the natural inducible nitric-oxide synthase and P-selectin promoter in a (cid:1) B-dependent manner upon stimulation of the cells with

(PC1) (3), is the best understood example of these (4). So far four additional PARPs were identified as follows: PARP-2, PARP-3, vault-PARP, and the tankyrase (5)(6)(7). Strikingly, the homology between these proteins is limited to the C-terminal half of PARP-1, whereas they all differ in their N-terminal portion. PARP-1 is a nuclear chromatin-associated protein of which one molecule is present per 1000 base pairs of DNA that detects specifically DNA strand breaks (4). In response to DNA strand damage caused by environmental genotoxic agents and endogenous cellular reactions, PARP-1 is activated by DNA strand breaks and initiates an energy-consuming cycle by transferring ADP-ribose units from NAD ϩ to nuclear proteins, including PARP-1 itself (4). The result of this process is a rapid depletion of the intracellular NAD ϩ and ATP, which leads to cellular dysfunction and death (8). In vitro studies have demonstrated that such a cellular suicide mechanism is responsible for cellular injury in response to oxygen-derived free radicals NO and peroxynitrite (9,10). The physiological function of PARP-1 is still under heavy debate. From studies using pharmacological inhibitors of PARP, poly(ADP-ribosyl)ation has been suggested to regulate gene expression and gene amplification, cellular differentiation, malignant transformation, cellular division, DNA replication, as well as apoptotic death (1,4,11). However, recent studies using cells from PARP-1 Ϫ/Ϫ mice have failed to demonstrate a role for PARP-1 in most of the suggested cellular processes, although it appears that PARP-1 has an important role in maintaining genomic stability (12,13). In addition, studies using these PARP-1 Ϫ/Ϫ mice showed that PARP-1 Ϫ/Ϫ mice were protected against myocardial infarct, stroke, streptozotocin-induced diabetes, neuronal damage induced by 1-methyl-4-phenyl-1,2,3,6 tetrahydropyridine, as well as LPS-induced septic shock, indicating that PARP-1 is involved in the regulation of the pathogenesis of these events (14 -16). However, the clarification of the physiological role of PARP-1 requires further study.
NF-B encompasses a family of inducible transcription factors including p50 (NF-B1), p52 (NF-B2), p65 (RelA), c-Rel, and RelB (reviewed in Ref. 17). These proteins share a conserved 300 amino acid region within their amino termini, termed Rel homology domain (RHD), which is responsible for DNA binding, dimerization, nuclear translocation, and interaction with heterologous transcription factors. They exist as homo-or heterodimers with a wide range of DNA binding and activation potentials. Although all Rel family members bind to DNA, only p65, c-Rel, and RelB contain a transactivation domain. The prototypical form, NF-B, is a heterodimer consisting of the two subunits p50 and p65 (18). In most unstimulated cells, NF-B is sequestered in the cytoplasm as an inactive transcription factor in a complex with one of the several inhibitors of NF-B (IB␣, -␤, or -⑀) (19,20). Treatment of cells with extracellular stimuli leads to the rapid phosphorylation of IB that results in ubiquitination of IB and subsequent degrada-tion by the 26 S proteasome pathway. Dissociation of NF-B unmasks nuclear localization sequences of the p65 and p50, which leads to nuclear translocation and binding of NF-B to specific B consensus sequences in the chromatin and activation of specific subsets of genes (20). The stimuli include cytokines such as interleukin-1 or TNF-␣, bacterial lipopolysaccharides (LPS), phorbol esters, UV light, ␥-irradiation, or potent oxidants (21). The NF-B response occurs in virtually all cell types, in combination with a variety of coactivators. Therefore, it is no surprise that the genes activated by NF-B will also vary depending on which other activators are present in the cell (22). NF-B plays a crucial role in the regulation of genes involved in immune and inflammatory responses (23). NF-B has additionally been implicated as an important regulator of cellular events such as cell proliferation and apoptosis and to be associated in neurodegenerative processes (24,25). Interestingly, in p50 Ϫ/Ϫ mice ischemic damage is drastically reduced (26), which leads to the idea of a cell death-promoting role of NF-B in ischemia and injury. On the other hand, there is emerging data that NF-B may counteract cell death in global ischemia (24,27).
Our group and others (28,29) have provided earlier evidence that PARP-1 might be required in vivo for specific B-dependent gene expression. We showed that B-dependent transcriptional activation was severely affected in immortalized PARP-1 Ϫ/Ϫ cells after treatment of the cells with TNF-␣, etoposide, and UV-C. We have found that the nuclear translocation was not affected in PARP-1 Ϫ/Ϫ cells and that PARP-1 was forming an immunoprecipitable nuclear complex with the large subunit p65 of NF-B. The molecular mechanism of PARP-1 contribution for B-dependent gene expression was not clear.
In the present study, we extended the list of substances unable to induce NF-B in primary PARP-1 Ϫ/Ϫ fibroblasts. Expression of p65 in primary PARP-1 Ϫ/Ϫ fibroblasts resulted in severely reduced activation of a B-dependent promoter. Furthermore, we investigated to what extent the enzymatic and the DNA binding activity of PARP-1 are required for B-dependent transcriptional activation. Experiments with enzymatic inhibitors and complementation experiments of primary PARP-1 Ϫ/Ϫ fibroblasts with expression plasmids expressing different forms of PARP-1 surprisingly revealed that the transcriptional coactivation of PARP-1 was taking place independently of its enzymatic activity or its ability to bind DNA. However, biochemical experiments confirmed that PARP-1 interacts directly with both subunits of NF-B, and mapping of the interaction domains revealed that both subunits are interacting with different domain of PARP-1. Finally, we provide evidence that PARP-1 regulates together with NF-B the gene expression of the natural iNOS and P-selectin promoter. Together our results provide strong evidence that PARP-1 is truly functioning as a specific transcriptional coactivator and thus provide an explanation of why PARP-1 Ϫ/Ϫ animals are protected against different NF-B-dependent diseases such as septic shock and epithelial dysfunction in diabetes type I.
Cell Culture, Transient Transfection, and Nuclear Extracts-HeLa cells, Hodgkins-L445, Jurkat T cells, primary wt (PARP-1 ϩ/ϩ ), and PARP-1 Ϫ/Ϫ fibroblasts were grown as described (33). Nuclear extracts were prepared as described previously (32). Cells were transfected as previously described (31,33). The amount of DNA indicated in the figure legends was calculated for 10 ml of medium. Total amounts of DNA and equal molar ratios of promoters were kept constant in all set ups by using empty vectors. Only cell passages 2-4 were used for transfection experiments. Due to differences in transfection efficiencies, an expression plasmid of ␤-galactosidase (pph-NGVL-RSV-nt-␤-Gal) was cotransfected as a transfection efficiency control, and luciferase activities were normalized based on ␤-galactosidase activity. Recombinant TNF-␣ was obtained from R & D Systems; LPS (0128 B12), MNNG, and CPT were from Sigma; 3-AB was from ICN Biomedicals Inc.; and the potent PARP-1 inhibitors 2,3-Bis(2-pyridyl)quinoxaline and 2-(methoxyphenyl)-benzimidazole-4 carboxamide were obtained from Novartis, Basel, Switzerland. Luciferase activity was measured as described previously (33).
In Vitro Transcription/Translation and GST Pull-down Assays-Proteins fused to GST expressed in Escherichia coli were purified using glutathione-Sepharose beads (Amersham Pharmacia Biotech) according to the manufacturer's protocols. His-Myc-tagged p50, p65, and Histagged PARP-1 expressed in E. coli were purified according to protocols from Novagen. All purified proteins were confirmed by Western blot analysis using the corresponding antibodies. Coupled in vitro transcription-translation reactions were carried out using the TNT T7-Quick System (Promega) according to the manufacturer's protocol. GST pulldown assays were performed in the presence of 100 -150 mM NaCl as described previously (33).

Inability of Proinflammatory or Genotoxic Substances to Activate B-dependent Gene Expression in Primary PARP-1 Ϫ/Ϫ
Fibroblasts-We reported recently (28) that TNF-␣, etoposide, and UV-C treatments of immortalized PARP-1 Ϫ/Ϫ cells were not able to activate B-dependent gene expression. To investigate whether the described effect was due to the immortalized status of the cells, the same experiments were repeated with primary PARP-1 Ϫ/Ϫ fibroblasts using different NF-B inducers. Primary wild type (PARP-1 ϩ/ϩ ) or PARP-1 Ϫ/Ϫ fibroblasts were transfected with a B-dependent reporter plasmid containing the luciferase reporter gene under the control of two B-binding sites and subsequently treated with different proinflammatory or genotoxic substances such as human TNF-␣, lipopolysaccharides (LPS), H 2 O 2 , N-methyl-N-nitro-N-nitrosoguanidine (MNNG), and camptothecin (CPT). All tested substances were able to induce B-dependent gene expression in the wild type fibroblasts, which resulted in an increase of luciferase activity (Fig. 1A, left panel). The same transfection experiments with a reporter gene under the control of mutated B-binding sites revealed that the observed induction was NF-B-specific. However, none of the tested inducers could activate the reporter gene in primary PARP-1 Ϫ/Ϫ fibroblasts (Fig. 1A, right panel). Moreover, basal transcriptional activation (unin-duced NF-B) was severely reduced in PARP-1 Ϫ/Ϫ cells (data not shown). These experiments confirmed the inability of different stimuli to induce NF-B in PARP-1 Ϫ/Ϫ fibroblasts. Since the tested substances represent different classes of NF-B inducers (21), PARP-1 seemed to be an important downstream cofactor used by all these tested signaling pathways.
The functional inability of induced NF-B to activate gene expression in primary PARP-1 Ϫ/Ϫ fibroblasts was further investigated by studying p65-mediated transcription. Wild type (PARP-1 ϩ/ϩ ) or PARP-1 Ϫ/Ϫ fibroblasts were cotransfected with expression vectors for p65 together with the reporter plasmid containing the luciferase reporter gene under the control of two wild type or mutated B-binding sites. Whereas expression of p65 alone in wild type cells led to a significant activation of reporter gene expression in a B-dependent manner (24.5-fold), only a slight activation (3.5-fold) of the reporter gene could be observed in PARP-1 Ϫ/Ϫ cells (Fig. 1B). The expression of p65 in PARP-1 Ϫ/Ϫ cells was confirmed by Western blot analysis (data not shown). These results suggested that PARP-1 would act as a coactivator for B-dependent gene expression.

PARP-1 Inhibitors Do Not Inhibit NF-B-mediated Gene
Expression-To investigate whether the enzymatic activity of PARP-1 is required for the activation of B-dependent genes in vivo, different PARP-1 enzymatic inhibitors were tested in transient transfection assays using a reporter plasmid. Wild type (PARP-1 ϩ/ϩ ) fibroblasts were transfected with a B-dependent reporter plasmid and subsequently treated with 3-aminobenzamide (3AB) or two specific enzymatic PARP-1 inhibitors alone or together with TNF-␣ or LPS as indicated in the figure legend ( Fig. 2A). None of the tested inhibitors were able to inhibit induction of the luciferase activity after treatment of the cells with TNF-␣ or LPS. Repeated experiments with a mutated reporter plasmid showed that the TNF-␣ or LPS-induced reporter gene activity was B-specific. To exclude the possibility that the lack of inhibition was due to incomplete chromatin formation of the transiently transfected reporter plasmid, the same inhibitors were tested in fibroblasts containing a stable integrated B-dependent reporter gene. As with the transient transfection experiments, no inhibition was observed with the tested inhibitors (data not shown), suggesting that the PARP-1 enzymatic activity was not necessary for Bdependent gene expression after stimulation of the cells with TNF-␣ and LPS.
The Enzymatic and DNA Binding Activity of PARP-1 Are Not Required for Basal B-dependent Gene Expression in Vivo-Next we investigated which activity of PARP-1 would be necessary for the transcriptional activation. An enzymatic inactive mutant of PARP-1 (E988K (34)), a DNA-binding mutant (C21G/C125G (35)), and a double mutant of PARP-1 lacking both the DNA binding and the enzymatic activity were generated and cloned into expression vectors. All forms of PARP-1 could be detected in PARP-1 Ϫ/Ϫ cells by a Western blot analysis using a PARP-1-specific antibody (Fig. 2B). Transient cotransfections of these different mutants with a B-dependent reporter plasmid revealed that the wild type as well as all generated PARP-1 mutants are able to restore basal B-dependent gene expression in primary PARP-1 Ϫ/Ϫ fibroblasts (Fig. 2C). Together, these results indicated that the enzymatic and DNA binding activity of PARP-1 are not required for the reconstitution of the basal B-dependent gene expression in PARP-1 Ϫ/Ϫ cells.
Synergistic Activation of the Reporter Gene by p65 with Wild Type or Double Mutated PARP-1-Next we examined whether wild type and the double mutated PARP-1 lacking the enzymatic and DNA binding activity (E988K/C21G/C125G) would also be able to activate p65-mediated gene expression. Both forms of PARP-1 and a p65 expression plasmid were cotransfected with a B-dependent reporter plasmid into primary PARP-1 Ϫ/Ϫ fibroblasts. p65 alone activated the reporter gene only slightly (compare Fig. 1A), whereas wild type or the double mutated PARP-1 alone were able to increase the basal level of B-dependent gene expression only to a certain extent (compare Figs. 2C and 3A). However, when the PARP-1 forms were cotransfected with p65, they were able to activate the reporter gene in a synergistic manner (Fig. 3A). Experiments repeated with a reporter plasmid containing mutated B-binding sites confirmed that the observed effect was B-specific. Similar results were obtained in wild type (PARP-1 ϩ/ϩ ) cells, although the synergistic activation was not to the same extent due to high activation of the reporter plasmids by p65 alone (data not shown). Together these results indicated that wild type or the double mutated PARP-1 are both sufficient for p65-mediated gene expression and that the enzymatic and DNA binding activities of PARP-1 are not required for this activation.
The Enzymatic and the DNA Binding Activity of PARP-1 Are Not Required for NF-B Induced by Proinflammatory or Genotoxic Substances-Next we tested whether wild type or double mutated PARP-1 was able to restore physiologically induced NF-B transcriptional activation in primary PARP-1 Ϫ/Ϫ fibroblasts. PARP-1 Ϫ/Ϫ cells were cotransfected with one of the two PARP-1 forms and a B-dependent reporter plasmid and subsequently stimulated either with TNF-␣, LPS, etoposide, UV-C, or camptothecin. Stimulation of PARP-1 Ϫ/Ϫ cells resulted in a synergistic activation of the reporter gene in the presence of PARP-1 independently whether the wild type or the double mutated PARP-1 was tested (Fig. 3, B-F). These results indicated that neither the enzymatic nor the DNA binding activity of PARP-1 seem to play an essential role for the transcriptional activation through NF-B in response to proinflammatory or genotoxic substances.

PARP-1 and NF-B Form a Complex in Vivo-
The observation that the enzymatic and DNA binding activity of PARP-1 is not required for transcriptional activation implicated that PARP-1 might influence B-dependent gene expression through direct protein-protein contact. To investigate whether NF-B and PARP-1 would form such a complex, coimmunoprecipitation experiments were performed. Nuclear extract of TNF-␣ stimulated HeLa cells were immunoprecipitated with either an anti-p65, an anti-p50, or a control antibody (Fig. 4A,  lanes 1-4). Bound protein were subsequently tested by Western blot analysis using a specific antibody against PARP-1 and both subunits of NF-B p65 and p50, respectively (Fig. 4A). PARP-1, p65, and p50 were all detected when immunoprecipitated with an antibody against p65 and p50 but not with the control antibodies, indicating that endogenous PARP-1 and NF-B formed a tight complex. These experiments confirmed the already obtained interaction between p65 and PARP-1 and confirmed that also p50 is complexing after TNF-␣ stimulation with PARP-1 in vivo.
p50 and p65 Bind Directly to PARP-1-To investigate whether PARP-1 would directly interact with both subunits of NF-B p50 or p65, PARP-1 was expressed as GST fusion protein (Fig. 4B, left panel) and bound to glutathione beads followed by incubation with in vitro translated p50 or p65. After extensive washes of the beads, bound proteins were analyzed by SDS-PAGE and subsequent autoradiography. Both subunits were able to bind independently of each other to GST-PARP-1 (Fig. 4B, right panel), indicating that the observed binding of the heterodimer NF-B to PARP-1 might be mediated by both subunits separately. The same experiments with bacterially expressed and purified subunits of NF-B revealed the same result (see Fig. 5C).
Identification of the PARP-1 Interaction Domain in p65 and p50 -To map the interaction domain within p65, different GST fusion proteins were generated expressing either the complete Rel homology domain (RHD) (amino acids 1-300), a truncated RHD expressing amino acids 1-275, or the transactivation domain of p65 expressing amino acids 300 -551. GST pull-down experiments with TNF-␣ stimulated HeLa nuclear extracts, and subsequent Western blot analysis for bound PARP-1 revealed that the last 25 amino acids of the p65 RHD were responsible for the binding to endogenous PARP-1 (Fig. 4C,  compare lanes 4 and 5). The transactivation domain of p65 was not able to interact with PARP-1 under the tested conditions (Fig. 4C, lane 3). To map the PARP-1 interaction domain within p50, wild type p50 and a C-terminal deletion mutant expressing only amino acids 1-341 (p50⌬) were bacterially expressed as His-Myc-tagged proteins and included in GST pull-down experiments with full-length PARP-1. Although full-length p50 was able to bind to PARP-1, the deletion mutant did not bind longer under the tested conditions (Fig. 4D), indicating that the interaction domain in p50 was lying between amino acid residues 341 and 433. Together, these results indicated that p65 is interacting with PARP-1 through the amino acids 275-300, whereas p50 interacts with PARP-1 through the amino acids 341-433 containing the "glycine-rich hinge" region.
Identification of the p50/p65 Interaction Domains in PARP-1-In order to map the p50 or p65 interaction domains in PARP-1, PARP-1 deletion mutants were cloned and expressed as GST fusion proteins in bacteria. GST pull-down experiments with these GST-PARP-1 fusion proteins together with bacterially expressed and purified His-Myc-tagged p50 and His-Myc-tagged RHD of p65 revealed that full-length PARP-1 was able to interact with full-length p50 and the RHD of p65 (Fig. 5A,  lane 7), confirming that the interactions observed above were very likely direct and not mediated by other proteins. Experiments with different mutants expressing amino acids 1-683, 341-531, and 341-1014 revealed that His-Myc-tagged p65 would interact with those fusion proteins to the same extent as with the wild type protein (Fig. 5A, lanes 4 -6). The fusion protein expressing the first 140 amino acids encompassing the zinc finger I and a part of the zinc finger II in the DNA binding domain of PARP-1, however, did not bind to p65 (Fig. 5A, lane  3). Remarkably, except for the negative control, GST, all tested PARP-1 deletion mutants interacted with p50, although to a different extent (Fig. 5A, lanes 3-6). The mutant expressing the amino acids 1-683 bound p50 comparable to wild type fusion protein, whereas all other mutants (1-140, 341-531, and 341-1014) interacted to a weaker extent (Fig. 5A). These results indicated that p50 and p65 would preferentially interact with a region between amino acids 341 and 531 of PARP-1. p50 would, in addition, interact with a region between amino acids 1 and 140.
To strengthen the observation that p50 and p65 would bind to different domains within PARP-1, we investigated whether supplemented DNA would affect these interactions. GST pulldown experiments using bacterially expressed and purified p50 and p65 (amino acids 1-300) fusion proteins with HeLa nuclear FIG. 3. The enzymatic activity and the DNA binding activity of PARP-1 are not required for B-dependent activation upon stimulation. A, primary PARP-1 Ϫ/Ϫ fibroblasts were cotransfected with pBII-Luc or pf-Luc (2 g, see Fig. 1A) and RSV-nt-␤-galactosidase (200 ng) together with RSV-p65 (0.5 g, see Fig. 1B, left panel), CMV-PARP-1 (5 g) (wt or double mutant C21G/C125G/ E988K (mut)), RSV-empty, or CMVempty as indicated (see "Materials and Methods"). Cells were harvested 36 h after transfection, and B-dependent gene expression was determined as described in Fig. 1A. Error bars indicate S.E. of three independent experiments. B-F, induction of NF-B in primary PARP-1 Ϫ/Ϫ fibroblasts after transient complementation with wild type or double mutated PARP-1. PARP-1 Ϫ/Ϫ fibroblasts were cotransfected with pBII-luc or pf-luc (2 g, see Fig. 1A), CMV-PARP-1 (5 g, wt or mut C21G/C125G/E988K(mut)), or CMVempty vector together with RSV-nt-␤-galactosidase (200 ng). Cells 4 -30 h after transfection were treated for 4 h either with human TNF-␣ (10 ng/ml, B) or LPS (10 g/ml, C), etoposide (10 M, D), UV-C (25 mJ/m 2 , E), and CPT (15 M, F), respectively. The indicated activation was determined by the ratio of the relative luciferase activity measured for the pBIIluc and for pf-luc reporter plasmid after the indicated stimulation. The obtained ratio for untreated cells was arbitrarily set as 1. Error bars indicate S.E. of three independent experiments. extracts were repeated in the presence of supercoiled plasmid DNA or double-stranded (ds) oligonucleotides. Interestingly, the interaction of PARP-1 with p50 and p65 was independent of supercoiled plasmid DNA (Fig. 5B, lanes 3 and 5 in upper and  lower panel). Surprisingly, ds oligonucleotides inhibited the interaction of PARP-1 with p50 but not with p65 (Fig. 5B, lane  7). The addition of ethidium bromide restored as expected the inhibitory effect of ds oligonucleotide on the interaction of PARP-1 with p50 (Fig. 5B, lane 8). These results indicated that the ds oligonucleotides were competing with p50 but not with p65 for PARP-1 binding. Whether ds oligonucleotides could induce conformational changes in PARP-1 and thus influence p50 binding is currently under investigation.
The Double Mutant PARP-1 Lacking Enzymatic and DNA Binding Activity Still Interacts with p65 and p50 -To understand further the regulation of NF-B transcriptional activity by PARP-1, wild type PARP-1 and the double mutant C21G/ C125G/E988K were expressed as His-tagged proteins and purified and incubated in a GST pull-down experiments with GST-p50 and GST-p65 bound to glutathione beads. Bound proteins were analyzed by SDS-PAGE and subsequent Western blot analysis using a PARP-1-specific antibody (Fig. 5C). Both subunits were able to bind independently of each other to PARP-1 wild type or the double mutant of PARP-1 (Fig. 5C, lanes [5][6][7][8], indicating that the mutations on PARP-1 are not interfering with its binding to p50 and p65 and that PARP-1 is functioning as a bridging factor between NF-B and other factors of the transcription machinery.

PARP-1 Transcriptionally Coactivates the Natural iNOS and P-selectin Promoters in B and T Cells in a B-dependent
Manner-Finally we tested whether PARP-1 would activate the natural promoter of the iNOS and P-selectin gene. The iNOS gene was reported to play a critical role in the pathophysiology of the previously described diseases (9,10). A reporter plasmid expressing the luciferase gene under the regulatory control of the mouse iNOS promoter was transfected into Hodgkins-L445 B cells. Stimulation of B cells with a TNF-␣/LPS combination resulted in a 7-fold induction of the reporter plasmid. Cotransfection of wild type PARP-1 with the reporter resulted in a 3-fold induction, probably due to the high endogenous level of PARP-1 in these cells (Fig. 5A). However, subsequent activation of these cells expressing PARP-1 with the TNF-␣/LPS combination resulted in a synergistic activation of the reporter gene (Fig. 6A). The same experiments with the reporter plasmid containing mutated B sites in the iNOS promoter confirmed that the observed effect was NF-B-specific and not mediated by other transcription factors possibly induced by TNF-␣ or LPS. Experiments were repeated in Jurkat T cells with the same reporter plasmids but cotransfected with p65. As seen with the Hodgkins B cells, PARP-1 and p65 activated the iNOS controlled reporter gene synergistically in a B-dependent manner (Fig. 6B). Replacing the iNOS promoter with the P-selectin promoter confirmed that PARP-1 would also synergistically activate this promoter in a p65-mediated manner (Fig. 6C). DISCUSSION Growing experimental evidence suggests that PARP-1 can function as a repressor or coactivator of transcription factors. Our initial observations revealed that immortalized PARP-1 Ϫ/Ϫ cells show a defect in B-specific transcriptional activation in vivo and that PARP-1 is forming a complex with NF-B (28). Here we have corroborated the molecular details of this initial observation. We showed that different proinflammatory or genotoxic substances lacked the ability to induce NF-B in primary PARP-1 Ϫ/Ϫ cells. Simultaneous expression of PARP-1 and p65 in primary PARP-1 Ϫ/Ϫ fibroblasts results in synergis-

FIG. 4. Both subunits of NF-B interact directly with PARP-1 in vitro and mapping of the interaction domains of p65 and p50.
A, both subunits of NF-B, p50 and p65, interact with PARP-1 in vivo. Endogenous nuclear p65, p50, and PARP-1 of hTNF-␣-stimulated HeLa cells were immunoprecipitated (IP) in the presence of 100 mM NaCl by using an anti-p65 IgG, an anti-p50 IgG, or control IgGs. Bound proteins were resolved by SDS-PAGE and subsequently detected by Western blot (WB) analysis for p50, p65, and PARP-1. Lane 1 represents 10 g (5%) of the input protein. B, PARP-1 binds directly to p50 and p65 in vitro. GST (0.8 g) or GST-PARP-1 (0.8 g) were incubated with in vitro translated p50 and p65 in the presence of 120 mM NaCl as described under "Materials and Methods." Bound complexes were resolved by SDS-PAGE and visualized by autoradiography. C, PARP-1 binds to the C terminus of the RHD domain of p65. Pull-downs with the indicated proteins fused to GST (1.5 g) and nuclear extract from HeLa cells (200 g) were treated with hTNF-␣. GST was used as a control. Bound proteins were resolved by SDS-PAGE followed by subsequent Western blot (WB) analysis for PARP-1. Input lane represent 10% of the input (20 g). D, PARP-1 binds in vitro directly to the C-terminal domain of p50. 0.8 g of GST or GST-PARP-1 was incubated with 0.5 g of His-Myc-tagged p50 full-length or His-Myc-tagged ⌬p50 (amino acids 1-341) in the presence of 150 mM NaCl. Bound complexes were resolved by SDS-PAGE followed by Western blot analysis for Myc-tagged p50 (upper panel) and Myc-tagged p50⌬ (lower panel). Input lane represents 20% of the input. tic activation of a B-dependent reporter gene. Transfection experiments with inhibitors of the enzymatic activity of PARP-1 confirmed that the enzymatic activity of PARP-1 is not required in vivo for B-dependent gene expression from a transiently transfected or integrated reporter gene. Complementation experiments of PARP-1 Ϫ/Ϫ cells with a PARP-1 expression plasmid lacking both the DNA binding and enzymatic activity confirmed that PARP-1 does not have to bind to DNA for the activation of basal B-dependent gene expression. Neither were both PARP-1 activities required when B-dependent gene expression was activated by p65 overexpression or induced by proinflammatory (such as TNF-␣ and LPS) or genotoxic substances (etoposide, UV-C, or CPT). We further showed that PARP-1 interacts in vitro directly with both subunits of NF-B through different domains. Finally, we showed that PARP-1 is necessary and sufficient for B-dependent transcriptional activation of the iNOS and P-selectin promoter induced either by p65 overexpression or in response to TNF-␣ and LPS.
The observations that complementation of PARP-1 Ϫ/Ϫ cells with wild type PARP-1 restored the basal as well NF-B-dependent gene expression induced by different proinflammatory or genotoxic substances implicate that PARP-1 is playing a critical and central role downstream of two completely different signaling pathways. The hypothesis is further strengthened by the fact that p65-mediated gene expression is also dependent on PARP-1 (Fig. 1B) and exclude that PARP-1 is involved in the activation of the NF-B signaling cascade as recently proposed (36). Furthermore, these experiments provide evidence that PARP-1 cannot be replaced by the other isoforms of PARP-1 such as PARP-2 or PARP-3.
PARP-1 was recently identified by Meisterernst and co-workers (3) as an active component of upstream factor stimulatory activity-derived cofactor PC1 that could increase the specificity of the initiation of RNA polymerase II transcription (37). Earlier studies (38,39) have suggested that full activation of transcription by NF-B in cell-free systems required a crude upstream factor stimulatory activity coactivator fraction in addition to general initiation factors. PARP-1/PC1 was thought to provide an architectural function together with the other upstream factor stimulatory activity-derived positive cofactors PC3/Dr2/topoisomerase I and PC4/single-stranded DNA-binding protein in stabilizing the preinitiation complex (40,41). p65 has also been reported previously to interact directly through its transactivation domain with other coactivators (42), the TATA box-binding protein, and several TATA box-binding protein-associated factors (39,43). These interactions of p65 might also explain why overexpression of p65 in PARP-1Ϫ/Ϫ cells resulted in reduced but still detectable gene expression (see Fig. 1B). Furthermore, PARP-1 is found to be associated with regions actively transcribed by RNA polymerase II in undamaged cells (44). PARP-1 increases the transcriptional activity of several transcription factors (39,43) including TAX and AP-2 FIG. 5. Mapping of the interaction domains on PARP-1 and the double mutated PARP-1 are still able to interact with p65 and p50. A, mapping of the p65 and p50 interaction domain on PARP-1. Bacterially expressed and purified His-Myc-tagged p50 and p65 RHD (0.5 g) were pulled down using the indicated PARP-1 fragments (0.8 g) fused to GST in the presence of 100 (p65) or 150 mM (p50) NaCl. Bound complexes were resolved by SDS-PAGE followed by Western blot (WB) analysis for Myc-tagged p65 (upper panel) and Myc-tagged p50 (lower panel). Input lane represents 20% of the input. B, p50 and p65 bind to different domains of PARP-1. 0.8 g of GST, GST-p50, or GST-p65 were incubated with HeLa nuclear extracts in the presence of supercoiled (sc) plasmid (500 ng), double strand (ds) oligonucleotides (100 ng), or ethidium bromide (100 g/ml) as indicated. Bound proteins washed 3 times in the presence of 100 mM NaCl were resolved by SDS-PAGE followed by Western blot analysis for PARP-1, GST-p50, and GST-p65-RDH. Lane 1 represents 5% (10 g) of the input. C, p65 and p50 are both able to bind to wild type and the double mutant of PARP-1. GST (1 g), GST-p50, or GST-p65 (1 g) were incubated with purified His-tagged wild type and double mutant of PARP-1 (mut ϭ C21G/C125G/ E988K) (0.8 g) in the presence of 120 mM NaCl as described under "Materials and Methods." Bound complexes were resolved by SDS-PAGE and visualized by Western blot (WB) analysis. (45,46). Transcription factors such as retinoid X receptor and AP-2 were reported to tightly associate with PARP-1 in vivo (46,47). Other reports (3,4,48) showed that the presence of both single strand breaks in DNA and physiological concentrations of NAD ϩ in cell-free transcription systems resulted in auto-modification of PARP-1 which abrogated its stimulatory activity on gene expressions. Poly(ADP-ribosyl)ation of transcription factors could prevent both their binding to DNA and the formation of active transcription complexes (4). NF-B was recently reported to be modified by PARP-1, and this modification was negatively affecting its DNA binding activity (49). We could, however, neither confirm ribosylation of NF-B in vitro or in vivo using different preparations of PARP-1 or an antibody against ADP-ribose polymers (data not shown) nor could we observe a decreased NF-B DNA binding activity after treatment of primary fibroblasts with PARP-1 inhibitors such as 3AB and 2,3-Bis(2-pyridyl)quinoxaline (data not shown). Our observation is furthermore strengthened by a recent report (50) showing that a specific PARP-1 inhibitor PJ34 did not affect DNA binding of NF-B.
Reports (36) have also described an inhibitory effect of benzamide and nicotinamide on the expression of other B-dependent genes in PARP-1 Ϫ/Ϫ mice, implying that PARP-1 enzymatic activity is playing a role in B-dependent gene expression. Since our results do not confirm that the enzymatic activity of PARP-1 is required for B-dependent transcriptional activation, we hypothesize that the inhibitory effect observed with benzamide and nicotinamide on B-dependent gene expression may be mediated by other proteins, since these substances exert also other effects (51).
The observation that PARP-1 DNA binding activity is not required for B-dependent gene expression provides additional evidence that transcriptional coactivation by PARP-1 is not caused by DNA nicks in the transfected plasmids or near the integrated reporter gene, which would tether PARP-1 randomly to the enhancer elements, but that PARP-1 is fulfilling its coactivator function through specific protein-protein contact to both subunits of NF-B. Together, PARP-1 might therefore not only act as a bridging factor but may also function synergistically with other cofactors in stabilizing the interactions between NF-B and the basal transcription machinery, thereby facilitating the formation and subsequent activation of the preinitiation complex in vivo. Whether PARP-1 is interacting directly with components of the basal transcription machinery or other coactivators of NF-B remains to be determined.
Our results implicated that NF-B together with PARP-1 is responsible in vivo for the activation of the iNOS and P-selectin which has been implicated in several diseases (9,10). This observation is supported by the observation of different groups (16,29,52) that showed that gene expression of different Bdependent genes such as iNOS, P-selectin, the intercellular adhesion molecule-1, and TNF-␣ is suppressed in PARP-1 Ϫ/Ϫ mice. The fact that PARP-1 Ϫ/Ϫ mice do not show the same phenotype as p65 Ϫ/Ϫ animals indicate that only a subset of B-dependent genes are PARP-1-dependent and that the requirement of PARP-1 for B-dependent gene expression may be dependent on the tissue and development stage-specific expression of PARP-1.
Bacterial membrane component LPS, when injected into mice, causes a shock-like state leading to death. The mechanism by which LPS induces endotoxic shock is related to its ability to activate the NF-B/Rel family of transcription factors, enabling the expression of several critical genes involved in the pathogenesis of septic shock (23). These results explain the extreme resistance of PARP-1 Ϫ/Ϫ mice to lethality induced by LPS (14,29). Brain injury induces a cascade of signaling events that stimulate NF-B activation in injured neurons and in injury-responsive glial cells. Data from studies using mice lacking the p50 subunit of NF-B suggest that, overall, NF-B is playing a dual role in ischemic neuronal death. Whereas activation of NF-B in neurons increases their survival after a stroke, activation of NF-B in microglia promotes ischemic neuronal degeneration (53).
Diabetic patients frequently suffer from endothelial dysfunction and vascular alterations. Endothelial cells incubated in FIG. 6. PARP-1 coactivates the iNOS and P-selectin promoter in a B-dependent manner. A, Hodgkins-L445 B cells were transfected with pGL2-miNOS (Ϫ1485/ϩ31) with either wild type or mutant B sites (2 g) and RSV-nt-␤-galactosidase (200 ng) together with CMV-PARP-1 (5 g) (wt) or CMV-empty as indicated (see "Materials and Methods"). Cells treated with TNF-␣ (10 ng/ml) and LPS (10 g/ml) were harvested 36 h after transfection, and B-dependent gene expression was determined as described in Fig. 1A. Error bars indicate S.E. of three independent experiments. B and C, Jurkat T cells were transfected with pGL2-miNOS (Ϫ1485/ϩ31) or pGL2-mP-selectin with either wild type or mutant B sites (2 g) and RSV-nt-␤-galactosidase (200 ng) with CMV-PARP-1 (5 g) (wt) or CMV-empty together with RSV-p65 (0.5 g, left panel) or RSV-empty as indicated (see "Materials and Methods"). Transcriptional activation was determined as described in Fig. 1A. Error bars indicate S.E. of three independent experiments. high glucose exhibited production of reactive nitrogen and oxygen species by the nitric oxidase that gene expression is regulated by NF-B. The high level of oxygen species consequently resulted in single strand DNA breaks, PARP-1 activation, and associated metabolic impairment (50).
Our results provide evidence that PARP-1 is a novel and essential transcriptional coactivator of NF-B in vivo in response to different stimuli and implicate the following model (Fig. 7). Upon activation of cells after septic shock, stroke, or streptozotocin-induced diabetes type I, NF-B is induced and activates in concert with PARP-1 certain genes (such as iNOS and P-selectin). The enzymatic and DNA binding activities of PARP-1 are not required for this process. The ability of several expressed gene products to activate NF-B results in a persistent activation of NF-B (Fig. 7A). The cellular iNOS induction results on the other hand in an increase of intracellular NO up to a micromolar concentration (54). In comparison, the two other isoforms, neuronal nitric-oxide synthase and endothelial nitric-oxide synthase which are constitutively activated in neuronal and endothelial cells, synthesize NO only in the nanomolar concentration (55). NO is converted into a cytotoxic derivative, peroxynitrite. The high concentration of peroxynitrite induces a high number of DNA damage that leads to an excessive activation of PARP-1 and a depletion of cellular energy resulting in mitochondrial free radical generation and cell necrosis that finally leads to enhanced systemic inflammation, endothelial dysfunction, or neurodegeneration (Fig. 7B). Thus, PARP-1 may regulate cell necrosis at two levels, first through its coactivator function for NF-B and second by depleting the intracellular NAD ϩ and ATP. Pharmacological inhibition of PARP-1 improves the adverse clinical effects in different pathologies associated with inflammation after cell death (56). Since the enzymatic and DNA binding activities are not required for B-dependent transcriptional activation after treatment of cells with proinflammatory or genotoxic substances, we propose that the observed anti-inflammatory effects of the PARP-1 inhibitors do not influence PARP-1 coactivator function but inhibit the NAD ϩ and ATP depletion and subsequently also cell necrosis and tissue damage.