INTRODUCTION

Poly(ADP-ribose) polymerase 1 enzyme (PARP1) plays an important role in regulating gene expression and maintaining genome integrity. PARP1 is activated under conditions of stress and synthesizes a negatively charged polymer, poly(ADP-ribose) (PAR), that performs signaling functions [1-6]. PARP1 is the most abundant of the PARP family members (PARPs 1-4, tankyrases 1 and 2, PARPs 6-16) and accounts for approximately 90% of PARP catalytic activity [7-9]. Pathogenesis of the diseases of cardiovascular, nervous, immune, respiratory, and other body systems is often associated with PARP1 activation and PAR synthesis [10-13]. Furthermore, preclinical data on the PARP1’s involvement in DNA repair led to the development of a series of novel anticancer inhibitors (including olaparib, rucaparib, niraparib, etc.) [14-17].

PARP1 utilizes NAD+ as a donor of ADP-ribose units, releasing nicotinamide as a by-product. PAR synthesis involves three ADP-ribosylation reactions: (i) initiation (attachment of the first ADP-ribose unit to an acceptor protein), (ii) elongation, and (iii) branching (Fig. 1) [18-21]. In the elongation reaction, the O-glycosidic bond is formed between the adenine ribose of the terminal PAR residue and nicotinamide ribose of the metabolized NAD+, with the 2′A-OH group acting as an attacking nucleophile. In the branching reaction, the bond is formed between two nicotinamide ribose rings, with 2′N-OH acting as a nucleophile (subscripts A and N denote adenine and nicotinamide ribose, respectively). Formation of the negatively charged PAR polymers (up to 200 units in size) results in modulation of chromatin structure and recruitment of a number of cellular proteins [22]. In particular, activity of PARP1 at the sites of DNA damage recruits base excision repair proteins XRCC1, DNA polymerase β, and DNA ligase III [23, 24].

Fig. 1.
figure 1

Chemical structure of PAR composed of ADP-ribose units. Branches are formed every 40-50 units.

Full size image

Despite its biological significance, catalytic mechanism of PAR synthesis remains underexplored because of the complex substrate/product structure. Some assumptions on the PAR and NAD+ binding can be made based on the crystal structure of the complexes of PARP1 with inactive substrate mimics [18, 25, 26]. The Glu988 residue in the active site is critical for polymer growth and likely provides the required orientation of PAR and NAD+ by forming hydrogen bonds with their 3′A-OH and 2′N-OH groups [18, 25, 27, 28]. Some authors consider that Glu988 also forms a hydrogen bond with the 2′A-OH group of PAR and acts as a proton acceptor upon nucleophilic attack by the SN2 mechanism [18, 27]. However, in our molecular dynamics (MD) simulation of the enzyme–substrate complex with terminal fragment of PAR and NAD+ we failed to observe direct interaction between the Glu988 and 2′A-OH as well as reactive in-line configuration of the O2′A, C1′N, and N1N atoms, typical for SN2 [29]. The present paper is a further attempt to provide details concerning PARP1 catalysis through molecular modeling and metadynamics simulations. We describe the possible SN1-like mechanism for the elongation reaction involving formation of an intermediate oxocarbenium ion.

MATERIALS AND METHODS

Starting model of the PARP1 enzyme-substrate complex was constructed based on the catalytic domain coordinates extracted from the 4dqy crystal structure (residues 662-1011) [30]. N- and C-terminal ends of the protein were capped with ACE (acetyl) and NME (N-methylamide) groups, respectively. Coordinates of NAD+ were transferred from the 6bhv structure [26]. Coordinates of an ADP moiety as a terminal PAR fragment were transferred from the 1a26 structure [18]; its diphosphate chain was capped with a methyl group (Fig. S1 in the Online Resource 1).

Next, the model was optimized using AmberTools20 and Amber20 [31-33] installed on the Lomonosov-2 supercomputer [34]. The protein molecule was described with the ff14SB force field [35]. The NAD+ molecule was described with parameters from the Amber Parameter Database (http://amber.manchester.ac.uk) [36, 37]. Parameters for the methyl-ADP molecule were derived as follows. Force constants, equilibrium bond lengths/angle values, and van der Waals parameters were taken from the NAD+ parameter set. Partial atomic charges were taken from the NAD+ set except for the methoxy group. Charges for this group were determined using the R.E.D.-III.5 and RESP programs [38, 39], as shown in Fig. S2 in the Online Resource 1. Hydrogen atoms were added to the protein structure considering ionization properties of amino acid residues, and then it was surrounded by a layer of TIP3P water (12 Å).

Energy minimization included two stages: the first one with positional restraints on heavy atoms of the protein and substrates (2500 steepest descent steps + 2500 conjugate gradient steps), and the second one without restraints (5000 steepest descent steps + 5000 conjugate gradient steps). The obtained system was heated up from 0 to 300 K with positional restraints on the protein and substrate atoms (1000 ps, constant volume) and equilibrated at 300 K (1000 ps, constant pressure); time step was 0.002 ps. Finally, a 100 ns trajectory of the equilibrium MD simulation was calculated. All protocols are described in detail in our previous work [40]. The simulation trajectories were analyzed using cpptraj 5.1.0 [41] and VMD 1.9 [42].

To perform metadynamics calculations, 4 frames were selected from the equilibrium MD trajectory in which positions of the substrates were close to the reactive configuration. Each selected frame was then simulated using a hybrid quantum mechanics/molecular mechanics (QM/MM) functionality in Amber20 [43]; detailed protocols are given in Table S1 in the Online Resource 1. The QM region included essential parts of the substrates and of Lys903 and Glu988 residues, which were treated using the PM6-D semi-empirical Hamiltonian [44, 45], and the MM region included the remaining atoms of the system (Fig. S3 in the Online Resource 1). Prior to metadynamics, 25 ps simulations with 0.001-ps time steps were carried out to adjust structures of the active site after switching from the MM to QM/MM approximation. Next, well-tempered metadynamics implemented in Plumed 2.6.2 [46-50] was used to explore free energy surface of the PARP1-catalyzed elongation reaction. Free energy landscape was reconstructed by 8 simulations (2 for each starting structure) run in parallel with shared metadynamics potential (so-called “walkers”) [51]. Duration of each simulation was 300 ps, time step was 0.0005 ps.

The first collective variable (CV1) was defined as d1-d2, where d1 – distance between the C1′N and N1N atoms, and d2 – distance between the O2′A and C1′N atoms (Fig. 2a). The second collective variable (CV2) was defined as d3-d4, where d3 – distance between the O2′A and O2′A:H atoms, and d4 – distance between the O2′A:H atom and either the O3′A or O2′N atom (Fig. 2b). CV1 describes nucleophilic substitution at the C1′N atom, and CV2 – deprotonation of the nucleophilic 2′A-OH group. Gaussian potentials of an initial height of 6 kJ/mol and width of 0.1 Å (CV1) and 0.075 Å (CV2) were added every 100 simulation steps, bias factor was set to 16. Upper and lower walls were applied to d1, d2, d3, and d4 (to limit exploration in the CV1/CV2 space), as well as to some other distances to prevent unwanted proton transfer scenarios (e.g., transfer between O2′A and O2′N in system A, or between O2′A and O3′A in system B). Gaussian potentials were summed with metadynminer 0.1.7 (“fes2” function) [52], minimum energy paths were calculated using the nudged elastic band method [53].

Fig. 2.
figure 2

Collective variables used for metadynamics simulations. a) CV1 describes nucleophilic substitution at the C1′N atom. b) CV2 describes deprotonation of the nucleophilic 2′A-OH group. This group is activated by Glu988, and the corresponding proton relay system could include either the 3′A-OH (system A) or 2′N-OH (system B) group.

Full size image

RESULTS

Elongation of PAR polymers is the most common reaction catalyzed by PARP1. Two substrates are required for elongation activity: NAD+ (ADP-ribose donor) and terminal PAR fragment (ADP-ribose acceptor). We have decided to use methyl-ADP molecule as a structural analogue of the acceptor substrate terminus in molecular modeling experiments. A 100-ns MD simulation of the PARP1 enzyme-substrate complex demonstrated that NAD+ forms two hydrogen bonds with Gly863 and π-stacking with Tyr907, which confirms the results of previous studies [25, 29]. The catalytic Glu988 residue forms hydrogen bonds with the 2′N-OH group of NAD+ and 3′A-OH group of methyl-ADP, providing reactive orientation of the substrates. The neighboring Lys903 residue forms hydrogen bonds with both Glu988 side chain (stabilizing its conformation) and diphosphate moiety of methyl-ADP (Fig. 3).

Fig. 3.
figure 3

Important molecular interactions observed in the modeled enzyme–substrate complex of PARP1. Hydrogen bonds of Gly863, Lys903, and Glu988 are shown as dotted lines. Phenyl group of Tyr907 forming π-stacking is shown in yellow. For clarity, non-polar hydrogen atoms are omitted.

Full size image

Nucleophilic 2′A-OH group of the acceptor substrate is presumably activated by the Glu988 carboxyl group [18, 27]. However, according to the MD modeling data, these groups do not interact with each other directly (Fig. 3). We therefore considered two possible proton relay pathways for further metadynamics simulations: in system A proton transfer is mediated by the 3′A-OH of the acceptor substrate, and in system B – by the 2′N-OH of the donor substrate (Fig. 2b).

QM/MM metadynamics allowed us to reconstruct 2D free-energy landscapes of the PARP1-catalyzed elongation reaction, in which CV1 describes nucleophilic substitution at the C1′N atom, and CV2 – deprotonation of the nucleophilic 2′A-OH group (Fig. 4). The energy surfaces demonstrate that in the case of both proton relay systems, A and B, the reaction proceeds via the SN1-like mechanism: an intermediate (oxocarbenium ion) is formed and next it is attacked by the 2′A-OH group of the acceptor substrate, with concomitant proton transfer to Glu988. It is worth mentioning, however, that the proton transfer is conducted by the 3′A-OH (System A) rather than by 2′N-OH (System B), as indicated by the corresponding free-energy barriers (Fig. 5). Discrepancy in the P (reaction products) energies is likely due to the proton transfer to different oxygen atoms of the Glu988 carboxyl group in systems A and B (see Fig. 2).

Fig. 4.
figure 4

2D free-energy landscapes of the PARP1-catalyzed elongation reaction. a) Proton relay system A (includes 3′A-OH). b) Proton relay system B (includes 2′N-OH). Collective variable CV1 describes nucleophilic substitution at the C1′N atom, and CV2 – deprotonation of the nucleophilic 2′A-OH group. An oxocarbenium ion intermediate (I) is formed along the minimum energy pathway between reactants (R) and products (P).

Full size image
Fig. 5.
figure 5

Comparison of the free energy profiles of elongation reaction reconstructed using proton relay systems A and B (R, reactants; I, intermediate; P, products). The reaction coordinate represents progress along the minimum energy pathways shown in Fig. 4. Calculated errors for R, I, and P states are given in Table S2 in the Online Resource 1.

Full size image

Figure 6 shows the obtained structures of the reactants, oxocarbenium ion intermediate, and products in system A. The reactants are properly oriented with respect to each other by the Glu988 residue to maintain the reactive conformation (Fig. 6a). The oxocarbenium ion intermediate is formed with the release of nicotinamide and is stabilized due to the negative charge of Glu988. Planar configuration of the reactive center (C1′N atom) facilitates subsequent nucleophilic attack by the 2′A-OH group (Fig. 6b). A new O-glycosidic bond is formed between the adenine ribose and nicotinamide ribose to produce the elongation reaction product; the Glu988 side chain becomes protonated due to the proton transfer from 2′A-OH (Fig. 6c).

Fig. 6.
figure 6

Structures of reactants (a), oxocarbenium ion intermediate (b), and products (c) obtained from modeling of the PARP1-catalyzed elongation reaction.

Full size image

DISCUSSION

Metadynamics is a powerful method to investigate free energy landscapes of enzymatic reactions [54-58]. It biases the system evolution by a potential constructed as the sum of Gaussian functions deposited along the trajectory in the collective variable space [46]. The presented metadynamics study describes in detail the PARP1 elongation reaction mechanism, summarized in Fig. 7. It shows attachment of the ADP-ribose unit to the growing PAR chain accompanied with proton transfer to Glu988 via the 3′A-OH group. Formation of the intermediate furanosyl oxocarbenium ion confirms our hypothesis about the SN1-like mechanism [29]. The main limitation of our methodology is that the acceptor substrate is modeled using the methyl-ADP molecule. The negatively charged PAR polymer may electrostatically contribute to formation of the oxocarbenium ion, whereas methyl-ADP mimics only the terminal PAR unit.

Fig. 7.
figure 7

Proposed SN1 mechanism of the PARP1 elongation reaction. a) Substrate binding. b) Oxocarbenium ion intermediate. c) Product formation.

Full size image

Participation of 3′A-OH in catalysis is consistent with the results of an experimental study of NAD+ analogues lacking hydroxyl groups at positions 3′A and/or 2′A [59]. Qualitative analysis (SDS-PAGE/Western blotting with anti-PAR antibodies) showed that absence of the 3′A-OH group leads to the less efficient PAR formation, but not completely prevents the reaction (which points to the existence of an alternative 2′N-OH proton relay system). The preferable proton transfer via 3′A-OH could presumably be explained by a shorter distance between the 2′A-OH and 3′A-OH groups in the enzyme-substrate complex: the distance O2′A ∙∙∙ O3′A calculated from the 100 ns MD trajectory was 2.7 ± 0.1 Å, whereas the distance O2′A ∙∙∙ O2′N was 3.2 ± 0.5 Å. It might be interesting to design NAD+ analogues lacking hydroxyl group at the position 2′N or at both positions 3′A and 2′N to confirm our hypothesis on the proton relay.

Using molecular modeling and metadynamics simulations, we have demonstrated for the first time that the PARP1-catalyzed elongation reaction proceeds via the SN1-like mechanism involving formation of an intermediate oxocarbenium ion. Intriguingly, our results show that the nucleophilic 2′A-OH group of the acceptor substrate (terminal PAR unit) is activated by the general base Glu988 not directly but through the proton relay system including the adjacent 3′A-OH group. The findings of this study shed light on the PAR synthesis machinery involving complex polymer substrates and PARP1, a major sensor of DNA damage, and may be used in further design of new PARP1 competitive inhibitors mimicking the substrate and/or oxocarbenium intermediate structure.