The DNA-Binding Domain of Human PARP-1 Interacts with DNA Single-Strand Breaks as a Monomer through Its Second Zinc Finger

Poly(ADP-ribose)polymerase-1 (PARP-1) is a highly abundant chromatin-associated enzyme present in all higher eukaryotic cell nuclei, where it plays key roles in the maintenance of genomic integrity, chromatin remodeling and transcriptional control. It binds to DNA single- and double-strand breaks through an N-terminal region containing two zinc fingers, F1 and F2, following which its C-terminal catalytic domain becomes activated via an unknown mechanism, causing formation and addition of polyadenosine-ribose (PAR) to acceptor proteins including PARP-1 itself. Here, we report a biophysical and structural characterization of the F1 and F2 fingers of human PARP-1, both as independent fragments and in the context of the 24-kDa DNA-binding domain (F1 + F2). We show that the fingers are structurally independent in the absence of DNA and share a highly similar structural fold and dynamics. The F1 + F2 fragment recognizes DNA single-strand breaks as a monomer and in a single orientation. Using a combination of NMR spectroscopy and other biophysical techniques, we show that recognition is primarily achieved by F2, which binds the DNA in an essentially identical manner whether present in isolation or in the two-finger fragment. F2 interacts much more strongly with nicked or gapped DNA ligands than does F1, and we present a mutational study that suggests origins of this difference. Our data suggest that different DNA lesions are recognized by the DNA-binding domain of PARP-1 in a highly similar conformation, helping to rationalize how the full-length protein participates in multiple steps of DNA single-strand breakage and base excision repair.

0mM NaCl or at 200mM NaCl. At low ionic strength (0mM NaCl), where binding is expected to be strongest, a 1:1 mixture of nicked dumbbell DNA and F2 shows a well-resolved HSQC spectrum with properties consistent with a monodisperse solution of the expected 26kDa complex; this behavior was very similar to that seen for F1+F2 binding to the gapped DNA ligand. In contrast, the HSQC spectrum of a 1:1 mixture of gapped dumbbell DNA with F1 under the same conditions shows many much broader lines, consistent with an exchange process taking place on an intermediate rate on the chemical shift timescale. This exchange could either be between the free and bound states of the protein, if the concentration of free protein under these conditions is sufficiently high to cause appreciable broadening (i.e. > approx. 10%), or it could arise from conformational exchange within a dynamic protein-DNA interface, which would also be consistent with weaker binding (or from a combination of both processes).
At higher ionic strength (200mM NaCl), where binding is substantially weaker and off-rates faster, the appearance of the spectra appears at first sight to be reversed between the F1 and F2 cases. However, this is deceptive; in reality, the slow exchange behavior seen at low ionic strength for F2 is shifted to the intermediate regime at higher ionic strength as a result of weakened binding, while the intermediate regime behavior seen for F1 at low ionic strength is now shifted to the fast exchange regime at high ionic strength, again as a result of weakened binding. Consistent with this, the chemical shift perturbations (relative to the free protein) seen in the case of F1 at high ionic strength are very small relative to those seen for F2 at low ionic strength (these being the only two sets of conditions for which the comparison can be made, due to the low quality of spectra in which intermediate rate exchange is evident). IIIα. In each case, two orientations (related by a 180° rotation about the x axis) are shown, and corresponding cartoon views in chainbow coloring are shown above and below to show the orientation of the structure. Potentials were calculated using the program APBS 3 using a threshold value of ±10eV for the coloring, and visualized using the program pymol. 4 Disordered tails were removed from the display (this is why the surfaces have small missing regions).
The top views ( (a-f) each show the surface of the triple-stranded β-sheet and the long loop L1, which are the principle sites of DNA interaction as mapped by the NMR chemical shift perturbation experiments (see main text). Both PARP-1 F2 and the DL3 finger have a pronounced basic patch in this part of the surface, although its location differs somewhat between the two. In contrast, the electrostatic surface of PARP-1 F1 is less basic in this region, consistent with its markedly lower affinity for ligands mimicking damaged DNA. The lower views (g-m) show the opposite face of each structure. There are no pronounced basic patches, and in the case of the DL3 finger this part of the surface is predominantly acidic.   The uncertainty associated with the magnitude of the of the axial and rhombic component of the alignment tensor were estimated using a Monte-Carlo based error analysis implemented in program Module, setting the measurement error for experimental RDCs as 2Hz.

Supplementary Materials and Methods
All spectra were processed using the program TOPSPIN (Bruker GmbH, Karlsruhe) and analysed using either the program SPARKY 10

Structure Calculations:
Initial structures of PARP-1 fragments F1 and F2 were calculated using the program ATNOSCANDID, 13 for which the input comprises the respective protein sequences (F1 residues 1-108 and F2 residues 103-214), the full resonance assignment and the following 3D NOESY datasets: 15 N NOESY-HSQC (τ m = 150 ms), 13 C aliphatic region NOESY-HSQC (τ m = 150 ms) and 13 C aromatic region NOESY-HSQC (τ m = 150 ms). Dihedral restraints for the backbone were obtained from chemical shifts using the program TALOS. 14 Within ATNOS-CANDID, the internal generation of backbone dihedral restraints was suppressed and the TALOS restraints were specified as external input. During the ATNOSCANDID calculations no metal was represented explicitly, but the effect of metal binding was approximated by including inter-ligand distance constraints as follows: S γ to S γ , 3.7-4.0Å; S γ to histidyl-N, 3.4-3.8Å. At this stage all histidyl N atoms were assigned ambiguously as either N δ or N ε atoms.
In order to be able to employ explicit zinc bonding and geometry terms in the force-field for the calculations (including bond-angle and, for the histidines, in-plane constraints), we next calculated PARP-1 F1 and F2 structures using XPLOR-NIH. 15 As input, these calculations used the set of NOE restraints generated by the final (seventh) cycle of ATNOSCANDID, re-formatted for use in XPLOR-NIH. Since the XPLOR-NIH calculations employed r -6 summation for all groups of equivalent protons and non-stereospecifically assigned prochiral groups, and since no stereoassignments were made (and the assignment-swapping protocol within XPLOR-NIH for deriving stereoassignments indirectly during the structure calculation itself was not applied), the constraints for all such groups were converted to group constraints (i.e. such groups were specified using wildcards such as HB*). All lower bounds were set to zero. 16 Structures were calculated from polypeptide chains with randomized phi and psi torsion angles using a two-stage simulated annealing protocol within the program XPLOR-NIH, essentially as described in reference 17.
The structures calculated in XPLOR-NIH were finally subjected to a further stage of refinement using a full force field and an implicit water-solvent model as implemented in the program AMBER 9 (reference). Calculations comprised an initial minimization (200 steps steepest descent then 1800 steps conjugate gradient), then 20ps of molecular dynamics was repeated twice using a simulated annealing protocol (5000x1fs-steps heating from 0K to 500K; 13000x1fs-steps cooling to 100K; 2000x1fs-steps of final cooling to 0.0K) and a final minimization (200 steps steepest descent then 1800 steps conjugate gradient). The experimental distance and TALOS-derived torsional restraints were applied throughout, and force constants for the restraints were increased linearly during the simulated annealing stages to final values of 20 kcal mol -1 Å -2 for distance restraints and 100 kcal mol -1 rad -2 for torsion restraints. Implicit solvent representation using the generalized Born method was employed throughout (igb=1), and Langvin temperature control was used (ntt=3; gamma_ln=5).

Fluorescence Anisotropy:
After each addition of protein titrant, the solution was stirred for 30 s, and after 60 s the fluorescence and the fluorescence anisotropy was measured. Data were treated and analyzed essentially as described in Supplementary Reference 18.
In the experiments performed, changes in the intensity of fluoresceine fluorescence upon binding of the protein were less than 10% once corrected for dilution. As a result, the fractional fluorescence intensities were assumed to remain constant Supplementary References: