If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
* This work was supported in part by a start-up grant from the Department of Biochemistry and Molecular Biology at Thomas Jefferson University. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The on-line version of this article (available at http://www.jbc.org) contains supplemental Fig. S1. 1 Supported by a fellowship from the Canadian Institutes of Health Research.
Poly(ADP-ribose) polymerase-1 (PARP-1) is a chromatin-associated enzyme with multiple cellular functions, including DNA repair, transcriptional regulation, and cell signaling. PARP-1 has a modular architecture with six independent domains comprising the 113-kDa polypeptide. Two zinc finger domains at the N terminus of PARP-1 bind to DNA and thereby activate the catalytic domain situated at the C terminus of the enzyme. The tight coupling of DNA binding and catalytic activities is critical to the cellular regulation of PARP-1 function; however, the mechanism for coordinating these activities remains an unsolved problem. Here, we demonstrate using spectroscopic and crystallographic analysis that human PARP-1 has a third zinc-binding domain. Biochemical mutagenesis and deletion analysis indicate that this region mediates interdomain contacts important for DNA-dependent enzyme activation. The crystal structure of the third zinc-binding domain reveals a zinc ribbon fold and suggests conserved residues that could form interdomain contacts. The new zinc-binding domain self-associates in the crystal lattice to form a homodimer with a head-totail arrangement. The structure of the homodimer provides a scaffold for assembling the activated state of PARP-1 and suggests a mechanism for coupling the DNA binding and catalytic functions of PARP-1.
). PAR synthesis is initiated on glutamate residues of target proteins, and subsequent polymerization of ADP-ribose units extends from the site of initiation. As a post-translational modification, PAR substantially changes the biophysical and electrostatic properties of a protein and can alter the DNA binding functions, the protein interaction properties, and the cellular location of target proteins (
). There are several factors that control PARP-1 activity, including self-association, interaction with histones and nucleosomes, NAD+ concentrations, structure-specific binding to DNA, and automodification (
). However, there are few molecular level insights into mechanisms that control PARP-1 activity.
PARP-1 is 113 kDa (in human) and has a modular architecture composed of multiple, independently folded domains (Fig. 1A). The PARP-1 polypeptide is generally described in three major segments that represent the biochemical activities and functional roles of the enzyme: the DNA-binding domain (DBD; residues 1–374), the automodification domain (residues 375–525), and the catalytic domain (residues 526–1014). The catalytic domain of PARP-1 is located at the C-terminal end of the protein. It is highly conserved in the PARP superfamily, particularly in a region called the PARP signature that is responsible for binding NAD+ (
NMR and x-ray models have structurally defined each of the independent domains of PARP-1 (Fig. 1A), except for the C-terminal region of the DBD, between the first two zinc fingers and the BRCT domain (Fig. 1B). We therefore undertook a structural analysis of this region to understand its role in PARP-1 function and to build upon our understanding of the multidomain structure of PARP-1 and how these domains assemble into an active DNA-dependent enzyme. The two N-terminal zinc fingers of PARP-1 bind to DNA structures to trigger activation of the C-terminal catalytic domain of PARP-1 (
). The molecular mechanism for coupling the DNA binding and catalytic functions of PARP-1 remains an open question. Based on our structural and biochemical analysis, we show that the DBD of human PARP-1 contains yet a third zinc-binding domain. One function of the new zinc-binding domain is to couple the DNA binding and catalytic activities of PARP-1.
Gene Cloning and Mutagenesis–The gene coding for full-length human PARP-1 (residues 1–1014) and for PARP-1 residues 1–234 were placed in the pET28 expression vector (Novagen) using restriction sites NdeI/XhoI such that full-length PARP-1 (PARP-1wt) and PARP-11–234 were each produced with an N-terminal hexahistidine tag and associated linker. The portions of the PARP-1 gene coding for residues 216–366, residues 1–366, and residues 379–1014 were placed in the pET24 expression vector (Novagen) using restriction sites NdeI/XhoI such that the protein fragments PARP-1216–366, PARP-11–366, and PARP-1379–1014 were each produced with a C-terminal hexahistine tag and associated linker. Mutagenesis was performed using the QuikChange Protocol (Stratagene).
Protein Expression and Purification–PARP-1wt, PARP-1216–366, and PARP-11–366 were expressed in Escherichia coli strain BL21(DE3)RIPL (Stratagene). PARP-1C298A, PARP-11–234, and PARP-1379–1014 were expressed in E. coli strain BL21(DE3)Rosetta2 (Novagen). The cells were grown in LB medium containing 100 μm ZnCl2 and induced with 200 μm isopropyl β-d-thiogalactopyranoside at 16 °C for 20 h. The cells were resuspended in 20 mm Hepes, pH 8.0, 500 mm NaCl, 0.5 mm Tris[2-carboxyethyl] phosphine (TCEP), 0.1% Nonidet P-40, 1 mm phenylmethylsulfonyl fluoride (and protease inhibitors) and lysed using a cell disrupter (Avestin). Cell debris were pelleted for 2 h at 40,000 × g, and the supernatant was then loaded onto a 5-ml HP chelating column (GE Healthcare) charged with Ni(II) and pre-equilibrated in lysis buffer without Nonidet P-40. The column was washed in lysis buffer containing 20 mm imidazole and eluted with 400 mm imidazole. The proteins were then loaded onto a 5-ml HP heparin column (GE Healthcare), except for PARP-1379–1014 (see below). For full-length PARP-1 constructs and the PARP-11–366 and PARP-11–234 fragments, the heparin column was equilibrated at 50 mm Tris-HCl, pH 7.0, 150 mm NaCl, 0.1 mm TCEP, and 1 mm EDTA and eluted with a gradient from 150 mm to 1 m NaCl. For PARP-1216–366, the heparin column was equilibrated with only 50 mm NaCl, and the elution gradient was 50 mm to 500 mm NaCl. PARP-1379–1014 was loaded onto a 5-ml HP Q column (GE Healthcare) equilibrated in 50 mm Tris-HCl, pH 7.0, 50 mm NaCl, 0.1 mm TCEP, and 1 mm EDTA and eluted with a gradient from 50 mm to 1 m NaCl. The proteins were next passed over a Sephacryl S200 gel filtration column (GE Healthcare) in 20 mm Hepes, pH 8.0, 150 mm NaCl, 0.1 mm TCEP, and 0.1 mm EDTA. Full-length PARP-1wt did not require the gel filtration purification step. Proteolytic degradation removed a small portion of the N terminus of PARP-11–366 and the C terminus of PARP-11–234 during protein purification. The degradation is noticed as a doublet of bands on SDS-PAGE. For Co(II)-substituted PARP-1216–366 (Co(II)-PARP-1216–366), E. coli were grown in minimal medium supplemented with 30 μm CoCl2. For selenomethionine (SeMet)-containing PARP-1216–366, BL21(DE3)RIPL E. coli were grown in defined medium (
). The purification protocol for Co(II)- and SeMet-PARP-1216–366 was as described for native PARP-1216–366.
Atomic Absorption and Amino Acid Composition Analysis–Atomic absorption analysis of zinc content was performed at Galbraith Laboratories, Inc. (Knoxville, TN). Amino acid composition analysis was performed at the Molecular Biology Core Facilities of The Dana Farber Cancer Institute (Boston, MA).
Absorbance Spectra of PARP-1216–366 and Co(II)-PARP-1216–366–Native PARP-1216–366 and Co(II)-PARP-1216–366 were diluted in 20 mm Tris, pH 8.0, and 100 mm NaCl to a final concentration of 35 μm. Absorbance spectra were recorded on a Shimadzu UV-2401PC spectrophotometer in the wavelength ranges of 800–500 and 500–250 nm. For metal competition experiments, 125 μm ZnCl2 or MgCl2 was added to 20 μm of Co(II)-PARP-1216–366. The mixture was incubated for 15 min and then rescanned from 500 to 250 nm.
CD Spectroscopy–A JASCO J-810 spectropolarimeter was used to record CD spectra. Wavelength scans (280–200 nm) were performed at 4 °C in a quartz cuvette of 1-mm path length. The protein was at 5 μm in a buffer composed of 5 mm Na,KPO4, pH 7.5, 50 mm Na2SO4, and 0.1 mm TCEP. The final spectra represent the averages of three scans. The data were analyzed using the JFIT program (Dr. B. Rupp).
Fluorescence Polarization DNA Binding Assay–A 42-base pair duplex DNA was assembled to contain the chicken cTnT promoter sequence from -102 to -59, including an MCAT-1 element (
). The MCAT DNA probe carried a 5′ FAM on the bottom strand. The binding reactions were performed in 12 mm Hepes, pH 8.0, 60 mm KCl, 0.12 mm EDTA, 5.5 μm β-mercaptoethanol, 8 mm MgCl2, 0.05 mg/ml bovine serum albumin, 4% glycerol. The reactions contained 5 nm DNA probe and various concentrations of protein. The reactions were incubated at room temperature (22 °C) for 30 min and then at 4 °C for 90 min. Fluorescence polarization data were collected on a Polarstar Optima microplate reader (BMG Labtech). The observed binding constant (KD) was obtained by fitting the data to a two-state binding model.
DNA-dependent Automodification Assay–The automodification reactions were performed at room temperature in 20 mm Tris-HCl, pH 7.5, 50 mm NaCl, 7.5 mm MgCl2, and 0.2 mm TCEP. PARP-1wt or PARP-1C298A (0.62 μm) was first preincubated with 1 μm nonfluorescent MCAT DNA (same sequence as referenced above (
)) for 20 min. 5 mm NAD+ was then added to the reaction, and the mixture was incubated for various times. Under these conditions, automodification results in a visible shift in the electrophoretic mobility of PARP-1 on SDS-PAGE (
). For the complementation assay, PARP-1379–1014 (0.93 μm) was incubated with a 1:1 molar ratio (0.93 μm) or 2:1 molar ratio (1.85 μm) of PARP-11–366, or PARP-11–234. The reactions were incubated for 10 min, after the addition of DNA and NAD+ as specified above. In each of the experiments described above, the reactions were stopped by the addition of SDS-loading buffer containing 0.1 m EDTA. The samples were resolved on a 7.5% SDS-PAGE, and the gel was treated with Imperial protein stain (Pierce).
Crystallization and Structure Determination of PARP-1216–366–Native and SeMet crystals of PARP-1216–366 were grown in sitting drops at 4 °C. The protein at 30–60 mg/ml was mixed with an equal volume of 20% ethanol and 100 mm Tris-HCl pH 8.5. Before flash-cooling in liquid nitrogen, the crystals were briefly (<1 min) transferred to 15% propanediol, 25% glycerol, 50 mm Tris-HCl, pH 8.5, 25 mm NaCl, and 0.5 mm TCEP. X-ray diffraction data were collected at Beamline X-12C at the National Synchrotron Light Source (Brookhaven National Laboratory, Upton, NY). X-ray data were collected at two wavelengths for both native and SeMet-PARP-1216–366 crystals, and the data were processed using HKL2000 (
) using the data collected at the zinc and at the selenium absorption edge for SeMet-PARP-1216–366 crystals. The model was primarily built into experimental electron density maps calculated with phases from the zinc SAD experiment (supplemental Fig. S1). The SeMet SAD data were mostly used to locate methionine residues to confirm the register of the amino acid sequence. Even so, electron density maps calculated with SeMet SAD phases were of good quality and helped confirm difficult areas of the map. The model was constructed using the molecular graphics program COOT (
) (Table 1). The final model includes residues 225–359, one zinc atom, 160 water molecules, one ethanol molecule, and three glycerol molecules. N-terminal residues 216–224 and C-terminal residues 360–366 were poorly represented in electron density maps and therefore were excluded from the model. The side chain atoms of residues 225–228, 233, and 337 were not clearly represented in electron density maps; therefore the side chains of these residues were truncated after the β-carbon. The illustrations were made using Pymol (DeLano Scientific), Photoshop (Adobe Systems), and Illustrator (Adobe Systems).
TABLE 1Crystallographic data collection and refinement statistics
PARP-1216–366 Is a Zinc-binding Domain–We were interested in determining the functional role of the C-terminal region of the PARP-1 DBD, that is, the portion of the PARP-1 polypeptide between the N-terminal zinc fingers and the central BRCT domain (residues 200–400) (Fig. 1). Several PARP-1 fragments were produced that sampled potential domain boundaries at the N-terminal and C-terminal ends of this region. Limited proteolysis of the PARP-1 fragments and mass spectrometry of the proteolytic products were used to determine the minimal boundaries of a domain that was resistant to further proteolysis (data not shown). A PARP-1 fragment resulting from this analysis included residues 216–366 of PARP-1 with a C-terminal hexahistidine affinity tag, PARP-1216–366. CD analysis of PARP-1216–366 showed that this fragment of PARP-1 had secondary structure elements (peaks at 220 and 210 nm) (Fig. 2A, Native curve) (
) that underwent similar cooperative transitions during thermal melts (data not shown), consistent with the denaturation of a folded protein domain. The CD spectrum was analyzed using JFIT, which predicted that the domain is composed of α-helices and coils (52.1 and 47.9%, respectively; R factor: 16.5%). The N terminus of PARP-1216–366 was sensitive to trypsin digestion because of a patch of Lys residues between residues 221 and 226; therefore the true boundary of this domain of PARP-1 is likely to be after this Lys-rich region. However, other fragments that sampled the N-terminal boundary gave the same CD profile. Further structural analysis focused on the PARP-1216–366 fragment of PARP-1.
There are no strong amino acid sequence homologs for PARP-1216–366 that can be identified through comparison with structure databases. Through visual inspection of the multiple sequence alignment of PARP-1 homologs (Fig. 1B), we noted four strictly conserved Cys residues (Cys295, Cys298, Cys311, and Cys321, in humans) within the PARP-1216–366 domain that had spacing reminiscent of the zinc finger protein fold (
). The spacing between the four Cys residues is strongly conserved among all organisms, following the pattern CX2CX11/12CX9C where C is cysteine and Xn is the number of amino acids between the Cys residues. We speculated that PARP-1216–366 might bind zinc as part of the structure of this domain. Atomic absorption analysis indicated that zinc co-purified with PARP-1216–366 through three chromatographic steps and extensive dialysis. Based on this analysis, there were 1.5 mol of zinc/mol of protein, whereas no zinc was detected in dialysis buffer. We therefore concluded that PARP-1216–366 binds at least one molecule of zinc.
Co(II)-PARP-1216–366 was produced to gain more insight into the metal binding properties of the new zinc-binding domain. Co(II) can frequently substitute for zinc-binding sites, and the spectroscopic properties of a protein-Co(II) complex can provide structural insights into a metal-binding protein (
). The absorbance spectra of native PARP-1216–366 and Co(II)-PARP-1216–366 (both at 35 μm) were analyzed from 800 to 500 nm (Fig. 2B) and from 450 to 280 nm (Fig. 2C). The absorbance peak at 340 nm for Co(II)-PARP-1216–366 is consistent with a charge transfer between a thiol ligand and Co(II) (Fig. 2B). This absorbance peak is not present in native PARP-1216–366. The absorbance measurement at 340 nm estimated that there are four to six Co(II)-thiol bonds/molecule of PARP-1216–366 (A340 = 0.175; ϵ = 1300–900 m-1 cm-1 (
), indicating that there are four ligands to Co(II). The absorbance at 685 nm estimated that there were 1.4 Co(II)/molecule of PARP-1216–366 (A685 = 0.038; ϵ = 780 m-1 cm-1). Collectively, the spectroscopic analysis of Co(II)-PARP-1216–366 indicated that there was one Co(II) coordinated by four Cys residues.
The absorbance peak at 340 nm disappears with the addition of a 6-fold excess of Zn(II) to Co(II)-PARP-1216–366 (Fig. 2D). In contrast, the same concentration of Mg(II) did not change the absorbance spectrum between 450 and 300 nm (Fig. 2D, inset). This result indicated that PARP-1216–366 preferentially binds zinc and that Co(II) occupies the same metal-binding site as zinc. Importantly, the secondary structure of Co(II)-PARP-1216–366, before and after treatment with excess zinc, was observed to be the same as native PARP-1216–366 by CD analysis (Fig. 2A), indicating that Co(II)- and Zn(II)-bound PARP-1216–366 have the same overall fold.
The Third Zinc-binding Domain Mediates Interdomain Contacts Important for PARP-1 Activation–Spectral analysis of the metal binding properties of PARP-1216–366 in solution strongly suggested that four Cys residues coordinated a zinc atom. The strictly conserved Cys residues (Cys295, Cys298, Cys311, and Cys321) were deemed the most likely zinc ligands. Each of these Cys residues was changed to Ala in PARP-1216–366 to test their potential roles as zinc-binding ligands and structural elements of the protein fold. However, each of these mutations made PARP-1216–366 difficult to produce in E. coli, and the extremely low yield of protein prevented further analysis of the mutants in this context. Wild-type PARP-1216–366 was abundantly produced in E. coli; therefore it is likely that the Cys to Ala mutations compromised the capacity of PARP-1216–366 to form a stable fold.
The Cys to Ala point mutants were also created in full-length human PARP-1 to test the functional role of the third zinc-binding domain. Each of these single amino acid substitutions led to a drastic decrease in protein production in E. coli as compared with full-length PARP-1 (PARP-1wt), further suggesting that the conserved Cys residues are important for efficient folding of the third zinc-binding region of PARP-1. For the mutant PARP-1C298A, it was possible to purify an amount of protein sufficient for biochemical analysis by using 10 times the number of cells used for PARP-1wt purification. PARP-1C298A had an elution profile identical to PARP-1wt over gel filtration, suggesting that the global structure of PARP-1C298A is similar to PARP-1wt and that the mutation has most likely introduced a localized change in PARP-1 structure.
The PARP-1C298A mutant was first tested in a DNA-dependent automodification assay. PARP-1wt and PARP-1C298A were each incubated with DNA, and NAD+ (5 mm) was added to start the automodification reaction. Time points of the reaction were analyzed by SDS-PAGE to resolve the modified and unmodified forms of the PARP-1. Under these conditions, poly(ADP-ribose)ylation and not mono(ADP-ribose)ylation of PARP-1 was observed (
), resulting in a large change in PARP-1 migration on SDS-PAGE. For PARP-1wt, the automodified form of the protein was apparent after only 10 s of incubation with NAD+, and the majority of the input protein was modified after 1 min (Fig. 3A, left panel). Further incubation of PARP-1wt led to a greater shift in electrophoretic mobility, the result of longer or more highly branched polymers of ADP-ribose. In contrast, the time course for automodification of the PARP-1C298A was much slower (Fig. 3a, right panels). The amount of automodified PARP-1C298A after 10 min of incubation was similar to that obtained for PARP-1wt after only 30 s of incubation. PARP-1C298A was able to produce long polymers of ADP-ribose but only after 60 min of incubation. Therefore, PARP-1C298A is an active enzyme but with slower kinetics for automodification compared with PARP-1wt.
The N-terminal zinc fingers of PARP-1 provide the major DNA binding activity of the enzyme, and the zinc fingers alone bind to DNA with similar affinity as full-length PARP-1 (data not shown). Therefore we expected that the C298A mutation would not affect DNA binding. We compared PARP-1wt and PARP-1C298A binding to DNA using a fluorescence polarization assay (Fig. 3B). PARP-1 bound DNA with an approximate KD of 17 nm (Fig. 3B), consistent with the results of others (
). PARP-1C298A bound to DNA with the same apparent KD of 17 nm (Fig. 3B). Thus, the third zinc-binding domain of PARP-1 does not contribute significantly to the DNA binding activity that triggers PARP-1 activity. This result indicates that the automodification defect of PARP-1C298A was not due to reduced DNA binding affinity. Furthermore, the ability of PARP-1C298A to bind DNA with the same affinity as PARP-1wt supports the conclusion that the C298A mutant has not affected the global structure of PARP-1.
We reasoned that the PARP-1C298A defect was due to a disruption of PARP-1 interdomain interactions that promote the active state of the enzyme. The importance of the new zinc-binding domain for interdomain communication was demonstrated using a variation of the automodification assay that depends on protein-protein interactions to activate PARP-1. The DNA-dependent automodification reaction can be reconstituted using two proteolytic fragments of PARP-1 (
). We performed this type of complementation experiment using cloned and purified constructs of PARP-1 that approximate the protease-generated fragments: PARP-11–366 and PARP-1379–1014. PARP-1379–1014 alone does not display DNA-dependent automodification activity, that is, there is no visible change in electrophoretic mobility of the protein fragment on SDS-PAGE (Fig. 4). However, in the presence of PARP-11–366, DNA-dependent automodification was observed, with the electrophoretic mobility of the PARP-1379–1014 fragment shifting because of the addition of ADP-ribose polymers (Fig. 4). Notably, both PARP-11–366 and PARP-1379–1014 are modified with ADP-ribose in this reaction, with the modification of PARP-11–366 resulting in a broad smear on SDS-PAGE. Importantly, PARP-1379–1014 does not bind significantly to the DNA used in this assay; therefore the complementation that restores automodification is not mediated by both protein fragments associating on DNA. Rather, the complementation of these two fragments of PARP-1 is mediated through protein-protein interactions. A fragment of PARP-1 that lacks the new zinc-binding domain, PARP-11–234, did not restore DNA-dependent automodification when added to the reaction with PARP-1379–1014 (Fig. 4). This result demonstrates that the third zinc-binding domain of PARP-1 mediates a protein-protein interaction with the C-terminal fragment of PARP-1, and this interdomain contact is necessary for DNA-dependent automodification of PARP-1.
Collectively, our mutagenesis and deletion analysis support a role for the new zinc-binding domain (PARP-1216–366) in forming interdomain contacts important for PARP-1 activation. The results suggest that the new zinc-binding domain of PARP-1 relays the DNA binding signal from the first two zinc fingers to the catalytic C terminus by helping to establish the active form of the enzyme. The C298A mutation disrupts the local structure of the third zinc-binding domain and thereby alters the efficiency of transmitting this signal.
Crystal Structure of PARP-1216–366–Subsequent to the structural and functional investigations described above, we crystallized and determined the structure of PARP-1216–366, the third zinc-binding domain of PARP-1. The crystal structure of PARP-1216–366 was determined using SAD experimental phases calculated with x-ray data collected at the zinc absorption edge from crystals grown with SeMet-PARP-1216–366. The atomic model was refined to 1.7 Å with a crystallographic R and Rfree of 18.1% and 23.0%, respectively (Table 1) (Fig. 5A). The crystal structure is consistent with the structural analysis of PARP-1216–366 in solution, namely, one zinc atom is coordinated by the four strictly conserved Cys residues (Fig. 5B). Likewise, the α-helical and coil content of the crystal structure are in very good agreement with the results of the CD spectra analysis of PARP-1216–366 in solution.
The fold of PARP-1216–366 consists of an N-terminal helical region (residues 225–289), a central zinc ribbon fold (residues 290–332), and a C-terminal tail (residues 333–359) (Fig. 5A). Three α-helices form a subdomain at the N terminus, with the first helix extending away from the subdomain. The zinc-binding region forms a separate subdomain, making primarily water-mediated contacts with the N-terminal helical subdomain. The zinc-binding subdomain resembles a zinc ribbon fold with four Cys ligands (
). Cys295 and Cys298 form a compact zinc knuckle, residues between Cys298 and Cys311 form a two-stranded β-sheet, and residues between Cys311 and Cys321 form a loop that extends away from the body of the zinc ribbon subdomain (extended loop; Fig. 5B). An α-helix on the C-terminal tail of PARP-1216–366 contributes to the fold of the N-terminal helical region, and the remainder of the C terminus extends away from the N-terminal subdomain (Fig. 5A).
PARP-1216–366 self-associates in the crystal lattice, forming an extensive homodimer interface that buries 2031 Å2 of accessible surface area on each monomer (Fig. 5C). The extended C-terminal tail lies across portions of the first and the third α-helix of the N-terminal subdomain and contacts the β-sheet of the zinc ribbon subdomain. Additionally, the α-helix of the C-terminal tail rests against the same α-helix of the second molecule of the homodimer. The residues that form specific contacts at the dimer interface are well conserved among PARP-1 homologs (Gln241, Asp285, Phe289, Phe304, Tyr309, Arg355, and Phe357) (Fig. 5C). Together with the expanse of the buried surface area, the self-association of PARP-1216–366 observed in our crystal structure is very likely to have functional relevance.
Identification of a Third Zinc-binding Domain–The discovery of a third zinc-binding domain of PARP-1 was surprising. Studies that identified PARP-1 as a zinc-binding protein (
) did not indicate three zinc-binding sites. In one study, full-length PARP-1 was estimated by atomic absorption analysis to bind one zinc atom. In a second study, full-length PARP-1 was found by energy dispersive x-ray fluorescence to bind two zinc atoms. These investigations used the same method of enzyme preparation (calf thymus PARP-1), and the conflicting results most likely reflect the difficulty in determining zinc content for a large protein. Following these studies, the sequence of human PARP-1 suggested two zinc fingers in the N terminus of the enzyme, and subsequent biochemical and structural analysis confirmed the presence of these two zinc fingers (
). Interestingly, amino acid sequence analysis programs that identify the CCHC motif of the N-terminal zinc fingers of PARP-1 do not indicate a zinc-binding motif in PARP-1216–366. This further indicates why the new zinc-binding site has been overlooked until now.
The initial indications that PARP-1216–366 was a zinc-binding domain (atomic absorption and cobalt studies) motivated us to target the proposed Cys residues for mutagenesis. These residues were reasoned to be important to the fold of PARP-1216–366; therefore mutating them would provide a strategy to probe the functional role of this domain. Indeed, the Cys to Ala substitutions altered the domain to the point that the mutants were mostly unsuitable for further analysis. Our crystal structure clearly demonstrates the importance of these residues for stabilizing the structure of the zinc ribbon subdomain (Fig. 5B). Despite the limitations of the mutational analysis, we were able to gain considerable insights into the function of the new zinc-binding domain through biochemical evaluation of the PARP-1C298A mutant (Fig. 3) and the complementation assay for DNA-dependent automodification (Fig. 4).
The Third Zinc-binding Domain Mediates Interdomain Communication–DNA-dependent automodification, using the fragments PARP-11–366 and PARP-1379–1014, demonstrates that protein-protein interactions between PARP-1 domains are involved in the mechanism of enzyme activation (Fig. 4). By deletion analysis, we showed that the third zinc-binding domain of PARP-1 is necessary for this interdomain communication. The delayed kinetics of PARP-1C298A automodification can be rationalized in this context. The C298A mutation compromises the structure for the third zinc-binding domain and thereby reduces the capacity to form interdomain contacts important for enzyme activation, resulting in a longer time course for automodification. Trucco et al. (
) showed that two other mutations in this region of PARP-1 (G313E and K249E), identified through random mutagenesis, effectively eliminated DNA-dependent enzymatic activity while not affecting DNA binding. Our crystal structure shows that each of these mutations introduces a negative charge that most likely disrupts the fold of the zinc ribbon subdomain (G313E) or the N-terminal helical region (K249E) (Fig. 5A). These mutations are consistent with our studies and support the role for the third zinc-binding domain in communicating between the DNA binding and catalytic portions of PARP-1.
A likely candidate for mediating these interactions is the extended loop of the zinc ribbon, which is surface-exposed and available for contacts and contains two absolutely conserved residues (Trp318 and Lys320). We propose that the interdomain interaction described above is formed between this extended loop of the zinc ribbon fold and the BRCT domain, because the BRCT domain mediates protein-protein interactions between PARP-1 domains (
). Structure-guided mutagenesis of the third zinc-binding domain will allow us to determine the precise region that mediates interdomain contacts. This will be a great advance in understanding how PARP-1 domains assemble and communicate with each other.
Although the two N-terminal zinc fingers of PARP-1 are the major contributors to the DNA binding activity of PARP-1, other portions of PARP-1 also bind to DNA. For example, a 36-kDa fragment of PARP-1 (residues 233–525) resulting from plasmin digestion interacts with DNA (
); however, unlike the N-terminal zinc fingers, this DNA binding activity is not structure-specific (e.g. strand breaks or hairpins) and requires long segments of DNA (∼222 base pairs). The major contributor to the DNA binding properties of the 36-kDa plasmin fragment was localized to its N terminus (
), which is part of the third zinc-binding domain. For this reason, the region containing the third zinc-binding domain is frequently described as part of the DBD. The calculated electrostatic surface potential of PARP-1216–366 does not indicate an obvious DNA-binding surface (not shown), but this does not exclude the possibility of DNA interactions with the third zinc-binding domain. The DNA binding properties of this domain may only be pertinent in the context of the functional interaction of PARP-1 with nucleosome structure (
Self-association of the Third Zinc-binding Domain Provides a Structural Scaffold for PARP-1 Assembly–The crystal structure of PARP-1216–366 completes a set of NMR and x-ray models that cover the entire PARP-1 polypeptide. The task now is to determine how these independently folded domains assemble in three dimensions. Identifying the structural details of key interdomain contacts will provide necessary constraints for constructing models that reflect the domain architecture of PARP-1.
An important aspect of PARP-1 activation is enzyme self-association. PARP-1 self-association is linked to maximal DNA-dependent enzyme activation (
). However, the molecular details of PARP-1 dimerization have not been established, so the mechanisms of PARP-1 self-assembly and activation are poorly understood.
The PARP-1216–366 homodimer formed in the crystal lattice of our x-ray structure provides a compelling model for the mechanism of PARP-1 self-association. The extent of buried surface at the dimer interface and the large number of conserved residues that mediate the interaction strongly argue that this is an important feature of the third zinc-binding domain and that it is functionally relevant. The 2-fold symmetry of the PARP-1216–366 homodimer juxtaposes the N terminus of one molecule and the C terminus of the second molecule. The precise anchoring of the two molecules could serve as a molecular cross-brace that organizes the three-dimensional arrangement of PARP-1 domains (Fig. 6). In this homodimer configuration, the BRCT domain is held near the zinc ribbon domain of a neighboring molecule, perhaps enforcing the proposed interaction between the third zinc-binding domain and the BRCT domain. Notably, this interaction is intermolecular, that is, the BRCT domain of one PARP-1 molecule contacts the zinc-binding domain of the second PARP-1 molecule.
Regulation of PARP-1 Activity through Protein-Protein Interactions–PARP-1 is inactivated by caspase-3 cleavage during certain apoptotic programs. Caspase-3 cuts the PARP-1 polypeptide within the DBD, separating the N-terminal zinc fingers from the new zinc-binding domain and reducing PARP-1 activity to a low basal level (
). Thus, caspase-3 uncouples the DNA binding and catalytic activity of PARP-1 by severing the physical link between the DNA-binding zinc fingers and the third zinc-binding domain that is responsible for communicating DNA binding to the catalytic domain. Given the importance of the third zinc-binding domain in activating PARP-1, it will be interesting to see whether modifications to the third zinc-binding region might provide an allosteric mechanism for controlling PARP-1 activity. Phosphorylation of residues adjacent to the zinc-binding domain (Thr372 and Ser374) modulates the activity of human PARP-1 in vivo (
). These residues are located on the linker between the third zinc-binding domain and the BRCT domain. This may well prove to be a cellular mechanism for controlling PARP-1 activity through regulation of interdomain contacts.
PARP-1 plays a pivotal role in the cellular response to stress, and regulation of PARP-1 activity is an important mechanism for coping with stress. Inhibition of PARP-1 has emerged recently as a successful strategy for moderating the cellular response to stress (e.g. stroke and inflammation), sensitizing cells to chemotherapy, and specifically killing cancer cells (e.g. BRCA-deficient cells) (
). Understanding the interdomain contacts that are unique to the activation of PARP-1 might provide novel strategies for specifically controlling PARP-1 activity for therapeutic benefit.
In summary, our structural and biochemical studies have revealed a third zinc-binding domain in human PARP-1 that is conserved at the sequence level and most likely at the structural level among all PARP-1 homologs. Unlike the other two zinc-binding domains of PARP-1, the new zinc-binding domain is not essential for the DNA binding activity that stimulates poly(ADP-ribose) synthesis. Rather, the third zinc-binding domain is involved in protein-protein interactions that orchestrate PARP-1 activation and are critical to the DNA-dependent stimulation of PARP-1. The crystal structure of the third zinc finger is the last piece of the PARP-1 puzzle. The goal now is to combine structural and biochemical methods to assemble these pieces into a working model for full-length PARP-1 that can define the molecular mechanism for enzyme activation.
We thank Dr. G. de Murcia for providing the human PARP-1 gene, Drs. Y.-M. Hou and C. P. Scott for critical reading of the manuscript, and the staff at Beamline X-12C for assistance with x-ray data collection.
The atomic coordinates and structure factors (code 2riq) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org/).