Structural Insight into the Discrimination between 8-Oxoguanine

Jan 10, 2018 - hOgg1 and FPG are the primary DNA repair enzymes responsible for removing the major guanine (G) oxidative product, namely, ...
7 downloads 0 Views 5MB Size
Download PDF
Article Cite This: Biochemistry 2018, 57, 1144−1154

pubs.acs.org/biochemistry

Structural Insight into the Discrimination between 8‑Oxoguanine Glycosidic Conformers by DNA Repair Enzymes: A Molecular Dynamics Study of Human Oxoguanine Glycosylase 1 and Formamidopyrimidine-DNA Glycosylase Shahin Sowlati-Hashjin and Stacey D. Wetmore* Department of Chemistry and Biochemistry, University of Lethbridge, 4401 University Drive West, Lethbridge, Alberta T1K 3M4, Canada S Supporting Information *

ABSTRACT: hOgg1 and FPG are the primary DNA repair enzymes responsible for removing the major guanine (G) oxidative product, namely, 7,8-dihydro-8-oxoguanine (OG), in humans and bacteria, respectively. While natural G adopts the anti conformation and forms a Watson−Crick pair with cytosine (C), OG can also adopt the syn conformation and form a Hoogsteen pair with adenine (A). hOgg1 removes OG paired with C but is inactive toward the OG:A pair. In contrast, FPG removes OG from OG:C pairs and also exhibits appreciable (although diminished) activity toward OG:A pairs. As a first step toward understanding this difference in activity, we have employed molecular dynamics simulations to examine how the anti and syn conformers of OG are accommodated in the hOgg1 and FPG active sites. When anti-OG is bound, hOgg1 active site residues are properly aligned to initiate catalytic base departure, while geometrical parameters required for the catalytic reaction are not conserved for syn-OG. On the other hand, the FPG catalytic residues are suitably aligned for both OG conformers, with anti-OG being more favorably bound. Thus, our data suggests that the differential ability of hOgg1 and FPG to accommodate the anti- and syn-OG glycosidic conformations is an important factor that contributes to the relative experimental excision rates. Nevertheless, the positions of the nucleophiles with respect to the lesion in the active sites suggest that the reactant complex is poised to initiate catalysis through a similar mechanism for both repair enzymes and supports a recently proposed mechanism in which sugar-ring opening precedes nucleoside deglycosylation. In the first BER step, a DNA glycosylase detects the damaged nucleobase, flips the damaged nucleotide into the active site, and cleaves the glycosidic bond between the nucleobase and 2′deoxyribose. A monofunctional glycosylase only catalyzes glycosidic bond cleavage, while the deglycosylation step facilitated by a bifunctional glycosylase occurs before or after cleavage of the C1′−O4′ bond in the 2′-deoxyribose ring. Moreover, all bifunctional glycosylases process the DNA backbone on the 3′-side of the damaged site (β-lyase activity), while some bifunctional glycosylases are also capable of nicking the DNA backbone on the 5′-side of the lesion (δ-lyase activity). After the glycosylase processes the substrate, an endonuclease nicks the 3′- and 5′-sides of the lesion in the case of a monofunctional glycosylase or the 5′-side in the case of a bifunctional glycosylase that only exhibits β-lyase activity. The BER process is complete when a polymerase inserts the complementary canonical nucleotide opposite the remaining base, and a ligase seals the DNA backbone.

T

he oxidation of guanine (G) at the C8 position by a hydroxyl radical (•OH) produces a radical intermediate, which can undergo further oxidation to form 7,8-dihydro-8oxoguanine (OG) or reduction to yield a ring-opened product (2,6-diamino-4-hydroxy-5-formamidopyrimidine or FapyG, Scheme 1).1,2 The formation of OG is the overall most common DNA damaging event in humans, occurring approximately 103 times per cell per day.3−5 OG may retain the preferred anti conformation of G (χ = ∠(O4′−C1′−N9− C4) = 180° ± 90°, Scheme 1) and Watson−Crick (WC) hydrogen bond with an opposing cytosine (C) in the DNA double helix (Scheme 2a). However, repulsive interactions between O8 and O4′, as well as the 5′-phosphate group, in antiOG can promote rotation about the glycosidic bond, resulting in the syn-OG conformation (χ = 0 ± 90°).6 In the syn orientation, OG preferentially pairs with adenine (A) through the Hoogsteen face of the lesion (Scheme 2b),7,8 which in turn leads to a G → T transversion mutation after two rounds of DNA replication.9−11 This alteration is one of the most common mutations found in cells associated with lung, breast, ovarian, colorectal, and renal cancers.12 Nonbulky nucleobase damage, including OG, is typically repaired through the base excision repair (BER) pathway.13−15 © 2018 American Chemical Society

Received: December 22, 2017 Revised: January 10, 2018 Published: January 10, 2018 1144

DOI: 10.1021/acs.biochem.7b01292 Biochemistry 2018, 57, 1144−1154

Article

Biochemistry Scheme 1. Proposed Mechanisms for the Formation of OG and FapyG from Natural Ga

a

The dihedral angle χ (∠(O4′−C1′−N9−C4)) is shown in bold blue for anti-G (defined as χ = 180° ± 90°).

Scheme 2. (a) Watson−Crick Base Pair between anti-OG and C and (b) Hoogsteen Base Pair between syn-OG and A

OG is removed by specific glycosylases in humans and bacteria, including human oxoguanine glycosylase 1 (hOgg1) and formamidopyrimidine-DNA glycosylase (FPG or MutM), respectively. An abundance of experimental work has been done to understand the activities of hOgg1 and FPG. While hOgg1 uses a Lys residue as the nucleophile and a catalytic Asp residue, which has several proposed roles (vide inf ra), FPG uses a terminal proline (Pro2) as the nucleophile and a catalytic Glu (Figure 1). For OG repair, both FPG and hOgg1 have been shown to exhibit an activity preference based on the nucleobase opposite OG. Specifically, hOgg1 efficiently removes OG paired with C, but the excision rate for OG paired with A is negligible, being 1000-16 to 3000-fold17 slower than OG excision opposite C. Similarly, FPG more efficiently processes OG opposite C than A, but only at an ∼18-17 to 35-fold18 greater rate, leading to appreciable activity regardless of the opposing (C or A) nucleobase. Although factors early in the recognition step have been proposed to play a role in dictating the observed hOgg1/FPG activity dependence on the opposing base (i.e., differences in base pair strengths,18,19 direct contacts between the enzyme and the opposing base,20−22 or disruption of the DNA−enzyme interface destabilizing the precatalytic complex23), differences in the glycosidic orientation of OG paired opposite C and A may influence how the substrate is bound in the active site and therefore the repair efficiency. Crystal structures of hOgg1 bound to OG-containing DNA reveal that OG adopts the anti conformation in the active site when paired opposite C in the DNA duplex.20,24,25 Furthermore, crystal structures show that OG from an OG:C pair in DNA binds in the anti conformation in the active site of Ogg from Methanocaldococcus jannaschii26

Figure 1. X-ray crystal structure of (a) 2′-fluoro-dOG (FdOG) bound in the hOgg1 active site (PDB ID 3KTU) and (b) OG bound in the active site of the Glu3Gln FPG mutant (PDB ID 1R2Y). Crystallographic water is shown as red spheres and Ca2+ ion as pink.

and Clostridium actetobutylicum.21 In contrast, crystal structures of FPG suggest that OG paired opposite C binds in the syn conformer in the active site.27 Nevertheless, the structurally similar FapyG lesion (Scheme 1) paired opposite C binds in the anti conformation in the FPG active site.28 These differences in the OG binding conformation could reflect the ability of OG to rotate about the glycosidic bond during the base flipping step or inherent differences in the capability of these enzymes to bind different OG glycosidic conformations. Computational studies have provided a wealth of information about hOgg129−40 and FPG.23,32,41−49 Previous molecular dynamics (MD) simulations on hOgg1 have compared the eversion of anti-OG and G from the DNA duplex into the active site to understand discrimination against G,40 examined 1145

DOI: 10.1021/acs.biochem.7b01292 Biochemistry 2018, 57, 1144−1154

Article

Biochemistry the dynamics of wild-type (WT)29 and mutant33 hOgg1 bound to DNA containing anti-OG to gain insight into the role of active site residues and considered the potential recognition and catalysis mechanisms.33 Nevertheless, to the best of our knowledge, the ability of the hOgg1 active site to accommodate different OG glycosidic conformations has not been addressed. Despite MD simulations being used to consider how anti- and syn-OG are bound in the FPG active site41−44 and identify potential catalytically important active site residues,41−44 the preferred OG binding conformation in FPG is a matter of increasing debate. Specifically, Amara et al.42 and Song et al.44 concluded that the syn conformer is better accommodated in the active site than anti-OG, while Zaika et al.41 suggest that anti-OG binding is preferred and Perlow-Poehnelt et al.43 revealed that syn-OG can rotate into an anti conformation in the FPG active site. To complement these works, a recent study23 reveals that initiating MD simulations from a high anti OG conformation (i.e., χ = −64°) in the Lactococcus lactis FPG active site results in spontaneous drift toward the anti-OG orientation. Unfortunately, there is currently no direct comparison between the abilities of hOgg1 and FPG to accommodate multiple OG binding orientations in the literature. In addition to the ambiguity surrounding the substrate binding orientation, the chemical reactions catalyzed by hOgg1 and FPG have not been completely explained despite computational studies investigating the enzymatic mechanisms using truncated models30−32,35,46 or complete DNA−enzyme complexes.34,37−39,45,48,49 Two SN2 mechanisms have been proposed for the first chemical step catalyzed by these enzymes. In the first mechanism,30−32,35,46,50 nucleophilic attack (Lys249 in hOgg1 and Pro2 in FPG, Figure 1) at C1′ of 2′-deoxyribose leads to base departure (deglycosylation, Scheme 3, left). In the second mechanism,38,48,49,51 an acid (Asp268 in hOgg1 and Glu3 in FPG, Figure 1) facilitates cleavage of the C1′−O4′ bond in the sugar ring upon nucleophilic attack (ring opening,

Scheme 3, right). Although quantum mechanical calculations have predicted that the second mechanism leads to a lower barrier for OG excision by both enzymes,38,48 more evidence is required to unequivocally identify the favored pathway. Furthermore, due to discrepancies in the proposed mechanisms, the exact roles of the catalytic Asp268 in hOgg1 and Glu3 in FPG remain elusive. For example, Asp268 has been proposed to initiate nucleoside hydrolysis52 or stabilize the Schiff base intermediate.53 Both Asp26820,53 and Glu351,54 have been suggested to deprotonate the nucleophile and transfer a proton to the 2′-deoxyribose of the damaged nucleotide to facilitate the ring-opening step. Proton transfer from the nucleophile to Glu3 was suggested to take place through a water chain in FPG54 but could possibly occur directly from Lys249 to Asp268 in hOgg1.53 In the absence of proton transfer, Asp268 and Glu3 must be in contact with bulk solvent in order to promote the ring-opening step. To shed light on how OG is bound and excised by hOgg1 and FPG, the present study uses molecular dynamics simulations to investigate changes in the active site configuration and dynamics with the OG glycosidic conformation. By performing a detailed analysis of the interactions between the substrate, active site residues, and solvent, the present work provides insight into how these enzymes accommodate different OG conformations and prevent mutations. Importantly, our data suggests that the hOgg1 active site can more accurately discriminate against syn-OG than FPG and thereby identifies the OG glycosidic bond orientation as one factor that contributes to the relative experimental excision rates. Furthermore, our detailed analysis of active site interactions affords proposals regarding the roles of various hOgg1 and FPG residues in binding or discriminating against OG conformers. Finally, our new structural data sheds light on the relative importance of the previously proposed SN2 catalytic mechanisms50,51 for the first chemical step catalyzed by both repair enzymes. Together, our simulations rationalize the relative activities of these critical repair enzymes and highlight similarities as well as key differences in how human and bacterial repair enzymes combat the major DNA oxidation product.

Scheme 3. Previously Proposed Mechanisms for the Formation of the DNA−Enzyme Cross-link Intermediate in the First BER Chemical Step Facilitated by a Bifunctional Glycosylase,50,51 which Involves Deglycosylation prior to (left) or after (right) Ring Opening

■

COMPUTATIONAL DETAILS Starting structures for MD simulations were built using high resolution X-ray crystal structures. In the case of hOgg1, a crystal structure of the enzyme bound to DNA containing OG fluorinated at C2′ (2′-fluoro-dOG, FdOG) was used (PDB ID 3KTU).20 Missing residues (Gln80, Asp81, and Lys82) were added using PyMol55 and GaussView,56 and the C2′-flourine atom of the OG nucleotide was replaced by a hydrogen atom. In the case of FPG, a crystal structure of the Glu3Gln mutant of Bacillus stearothermophilus FPG bound to OG-containing DNA was used (PDB ID 1R2Y),27 with Gln3 substituted by the native Glu. For both crystal structures, missing heavy atoms and all hydrogen atoms were added using the LEaP module of Amber 12.57 Protonation states of ionizable residues were assigned using PROPKA 3.0,58 with the exception of key active site residues. Specifically, for hOgg1, His270 was protonated since MD simulations suggest that this protonation state better maintains the active site orientation of the crystal structure.37 Additionally, Lys249 and Cys253 were modeled as neutral since quantum mechanical calculations show that Lys249+1Cys254−1 is energetically unfavorable compared to Lys249Cys254 regardless of the protonation state of Asp268.37,38 Moreover, 1146

DOI: 10.1021/acs.biochem.7b01292 Biochemistry 2018, 57, 1144−1154

Article

Biochemistry

no large deviations in the active site were observed. At this point, two 20 ns MD production simulations were completed using different initial velocities. For all systems, the two replicas led to highly similar structures (rmsd with respect to the crystal structure of ∼0.5−0.8 Å, Figures S1 and S2, Supporting Information). Subsequently, one trajectory was chosen for a further 80 ns of production simulation under similar conditions. For all systems, the representative structures from the 20 and 100 ns simulations show significant resemblance (Figures S3 and S4). In the main text and the remainder of the SI, the 100 ns simulation results are presented. Trajectory analysis was completed using the cpptraj module of Amber 12.57 The root-mean-square deviation (rmsd) of the protein and DNA backbone over the production phase was analyzed for each simulation to ensure the system was stable (Table S1). Each trajectory was saved every 2 fs over the course of the production simulation. Clustering was completed for the trajectories using the hierarchical agglomerative algorithm (with ε = 4) based on the configurations of OG and key active site residues, namely, Gly42, Lys249, Asp268, Gln315, and Phe319 for hOgg1 and Pro2 and Glu3 for FPG. Although a single representative structure is shown for the cluster with the highest occupancy (see Table S1 for the corresponding rmsd and occupancy), the geometrical and energetic analyses were completed over all structures in the highest occupied cluster and the corresponding dynamical information is provided in the Supporting Information (SI). Throughout the manuscript, key average geometrical parameters are reported (standard deviations are provided in the SI), including the nucleophilic distance (i.e., the distance between Lys249/Pro2 and C1′ of 2′-deoxyribose or d(Nζ/N··· C1′)), the distance between Asp268 (Oδ) or Glu3 (Oε) and O4′ of 2′-deoxyribose (i.e., d(Oδ/Oε···O4′)), and the relative orientation of the nucleophile (Lys249 or Pro2) and the OG nucleotide (i.e., the ∠(Nζ/N−C1′−O4′) and ∠(Nζ/N−C1′− N9) nucleophilic attack angles). A 120° angle cutoff and a 3.4 Å distance cutoff were imposed between (donor and acceptor) heavy atoms for all hydrogen-bonding active site interactions. The linear interaction energies (LIE) are reported for key active site stacking and hydrogen-bonding interactions. The distribution of water in the active site was examined using a threedimensional 20 Å × 20 Å × 20 Å grid centered on the OG nucleotide, with 0.5 Å spacing between grid points. In order to assess the possibility of proton transfer from bulk solvent to (anionic) Asp268 or Glu3, the average number of water molecules close to these residues was determined using a spherical solvation shell with a radius of 3.4 or 6.0 Å centered on one of three atoms (i.e., O4′ of OG and Oδ1 or Oδ2 of Asp268 for hOgg1 or Oε1 or Oε2 of Glu3 for FPG).

previous MD simulations revealed that the salt bridge increases the solvent density in the active site, which may prohibit catalysis. 37 Finally, although the dipole moment of Lys249+1Cys254−1 has been proposed to play an important role in the recognition step,24 reversing the dipole direction with a Lys249Cys/Cys253Lys (KCCK) double mutant does not affect recognition, which may indicate that these residues are neutral.52 Depending on the order of the deglycosylation and ring-opening steps (Scheme 3), Asp268 may be anionic or neutral upon substrate binding. Specifically, if deglycosylation occurs first, Asp268 is likely anionic in order to stabilize the positive charge developing on 2′-deoxyribose.52,53 Alternatively, catalysis of an initial ring-opening step requires proton transfer to O4′ of 2′-deoxyribose, which could be facilitated by Asp268 that has been neutralized by Lys249 or solvent. Therefore, models were considered with anionic and neutral Asp268, and the solvent distribution in the active site was carefully monitored. Similarly, Glu3 in FPG was considered to be either anionic or neutral, while the Pro2 nucleophile was neutral. Hydrogen atoms were placed on the Oδ or Oε atoms of Asp268 or Glu3 automatically using the LEaP module of Amber 12 and were visually inspected to ensure that the initial conformation adopted the most favorable proton orientation for an isolated amino acid. The OG substrate was bound in the active site in both the anti and syn conformations for each active site protonation state. Although the base opposite the lesion affects enzymatic activity17,18 and may play a direct role in enzyme function,23 crystal structures of FPG reveal OG bound in the syn orientation despite the opposing C in the DNA duplex.27 Therefore, to uncouple other contributing factors and thereby compare the effects of the glycosidic conformation on OG binding for hOgg1 and FPG, both OG conformers were considered and the opposing base maintained as C for both enzymes. All systems were assigned AMBER parm99SB parameters,59 with nonstandard OG parameters taken from a previous work37 assigned parm99SB and GAFF parameters,60,61 using the ANTECHAMBER module of Amber 12.57 In the case of FPG, parameters for the zinc ion were taken from ZAFF.62 Each system was neutralized with Na+ ions and solvated in an explicit TIP3P water box that was at least 8.0 Å from the edge of the DNA−enzyme complex. Constraints were imposed on covalent bonds involving hydrogen atoms using the SHAKE algorithm, while the particle mesh Ewald algorithm was used for long-range electrostatic interactions. The solvent molecules and ions were relaxed using 500 steps of steepest descent, and 500 steps of conjugate gradient minimization, while the protein and DNA were constrained using a 500.0 kcal mol−1 Å−2 force constant. The entire system was then minimized using 1000 steps of unrestrained steepest descent, followed by 1500 steps of unrestrained conjugate gradient minimization. Subsequently, the system was heated from 0 to 300 K over 20 ps with restraints on the solute (10 kcal mol−1 Å−2). Prior to the production phase, each system was equilibrated for 20 ps. The periodic boundary condition was employed for all MD simulations. Once the preliminary models were established, each system was used for 20 ns of unconstrained MD preproduction simulations under NPT conditions (1 atm and 300 K). From each resulting trajectory, representative structures were chosen for subsequent 20 ns preproduction simulations, which employed different initial velocities to enhance the sampling over the phase space. This iterative process was repeated until

■

RESULTS hOgg1. anti-OG. The representative MD structure of antiOG bound in the hOgg1 active site with anionic or neutral Asp268 (Figures 2 and S5) is similar to the starting crystal structure in terms of the relative orientation of key active site residues and OG (average active site backbone rmsd