Catalytic Mechanism of the Bacterial Non-Heme Fe(II) and 2-Oxoglutarate Dependent Enzyme AlkB with Single-Stranded DNA Containing Complex Guanine Adducts

Abstract

The bacterial non-heme Fe(II)/2-Oxoglutarate (2OG)-dependent enzyme AlkB repairs alkylation damages in single-stranded DNA (ss-DNA) nucleotide bases. This study examines for the first time the reaction mechanism of the AlkB-catalyzed repair of alkylated and exocyclic guanine adducts (GAs) in single-stranded DNA induced by everyday chemical exposures associated with cancers and other genetic disorders. The studied substrates include N2-furfurylguanine (FF-dG), N2-tetrahydrofuran-2-yl-methylguanine (HF-dG), 3-(2’-deoxy-β-D-erythro-pentofuranosyl)-5,6,7,8-tetrahydro-6-hydroxypyrimido[1,2-α]purin-10(3H)-one (α-OH-PdG), 3-(2’-deoxy-β-D-erythro-pentofuranosyl)-5,6,7,8-tetrahydro-8-hydroxypyrimido[1,2-α]purin-10(3H)-one (γ-OH-PdG), and 3-(2’-deoxy-β-D-erythro-pentofuranosyl) pyrimido[1,2-α]purin-10(3H)-one (M₁dG). Using molecular dynamics based combined quantum mechanics/molecular mechanics (QM/MM) and QM calculations, we provide unique mechanistic insights into AlkB’s catalytic reaction pathways with ss-DNA containing complex alkylated/exocyclic GAs in strong correlation to experimental studies. While HF-dG, FF-dG, α-OH-PdG, and γ-OH-PdG are repaired through C-H hydroxylation, M₁dG follows epoxidation. The study elucidated that the repair mechanism favors the open tautomer of γ-OH-PdG and the closed tautomer of α-OH-PdG, respectively, in agreement with experimental studies, due to the preferable SCS interactions and the catalytic domain’s loop L1 and L4 dynamics. Our study also elucidated that the post-hydroxylation/post-epoxidation steps proceed in water rather than the enzyme. The results reveal the unique catalytic mechanism of AlkB with ss-DNA containing complex GAs, which can be used in drug design and metalloenzyme redesign.

Graphical Abstract

1. Introduction

The non-heme Fe(II)/2-oxoglutarate (2OG) dependent oxygenases AlkB play a pivotal role in directly repairing damaged DNA in Escherichia coli (E. coli)^1–8 and are interlinked with the adaptive responses^9–11 to cytotoxic and mutagenic DNA/RNA lesions.^4,12–14 AlkB oxidizes aberrant groups in damaged DNA, thus facilitating their restoration to an undamaged state. In comparison to the other DNA damage repair enzymes, AlkB possesses a very broad substrate scope.^4,12–14 Nine human AlkB homologs, ALKBH1-ALKBH9 (with ALKBH9, also known as the fat mass and obesity-associated protein FTO), have been identified; however, they usually have a narrower substrate scope than AlkB.^4,15–17

AlkB contains a double-stranded β-helix (DSBH) motif - a structural characteristic for the Fe(II)/2OG enzymes, where eight β-strands form a jellyroll fold,^18,19 containing the active site Fe(II) center (Figure 1).^6,7 AlkB has an N-terminal region with 90 residues, unique to the AlkB subfamily, known as the “nucleotide recognition lid” (NRL) subdomain, which is essential for substrate recognition and positioning the damaged nucleotide in the active site.^20,21 Experimental studies show that AlkB pulls the damaged base into its active site through the base-flipping process, which is driven by the movement of adjacent base pairs to maintain the continuous stacking of the bases.^7,22 In general, AlkB prefers single-stranded DNA (ss-DNA) comparison to double-stranded DNA (dsDNA), where the base flipping would require a higher energy as the complementary strand can act as a non-competitive inhibitor.⁷

Figure 1.

The crystal structure of the AlkB complexed with m1G substrate (PDB ID: 3KHC) features the C-terminal extension (CTE) in sky blue, the N-terminal extension (NTE) in cornflower blue, the ss-DNA in corol, active site residues in yellow, and the nucleotide recognition lid in light pink.

The AlkB catalytic strategy follows the typical mechanistic pathways of non-heme Fe(II)/2OG enzymes (Figure 2).^23–25 The catalytic strategies of Fe(II)/2OG enzymes have been intensively studied by experimental (crystallographic, spectroscopic, biochemical) methods^23,26–29 and computational (quantum mechanical (QM), quantum mechanics/molecular mechanics (QM/MM), and molecular dynamics (MD)) approaches.^30–33 In the resting state, Fe(II) adopts a six-coordinate (6C) geometry with two histidines (H131, H187), one aspartate (D133), and three water molecules forming its first coordination sphere (FCS). The subsequent bidentate coordination of 2OG replaces two water molecules, but the Fe(II) center retains its 6C state. The following binding of the main substrate in the close vicinity of the Fe(II) center, however, leads to a 5C state by displacing a water molecule from Fe(II), thus providing an open coordination site for dioxygen binding to the Fe(II)leading to the formation a ferric-superoxo (Fe(III)-OO^.−) complex (Figure 2). The next steps of the catalytic mechanism of AlkB enzymes have been comprehensively studied computationally using QM, MD, and QM/MM methods.^34–40 The studies predict that the Fe(III)-OO^{. −} intermediate undergoes decarboxylation of 2OG to form succinate and a high-spin (HS) Fe(IV)=O (ferryl) intermediate. The ferryl complex is the active oxidant that performs hydrogen atom abstraction (HAT) from the aberrant alkyl group in DNA to form a Fe(III)-hydroxo complex, followed by a rebound hydroxylation that leads to dealkylation of the substrate and regeneration of Fe(II) resting state.

Figure 2.

The general catalytic mechanism of non-heme Fe(II)/2OG enzymes is shown, with the studied reactions highlighted in the dotted box.

AlkB is primarily known as a demethylase acting on a broad variety of methylated DNA bases such as 1-methyladenine (m1A),⁴¹ 1-methylguanine (m1G),⁴¹ N,²-methylguanine (m2G),¹³ N,²-ethylguanine (e2G),¹³ 3-methylthymine (m3T),⁴¹ 3-methylcytosine (m3C),⁴¹ N,⁶-methyladenine (m6A)⁴² and preferably in ss-DNA.⁷ However, AlkB is also capable of catalyzing much more complex dealkylation reactions on substrates, including i) complex alkylated guanines such as N²-furfurylguanine (FF-dG),¹³ N²-tetrahydrofuran-2-yl-methylguanine (HF-dG) (Figure 3),¹³ and ii) exocyclic bridged adducts such as hydroxy-3,N⁴-propanocytosine,^43,44 3,N⁴-ethenocytosine (εC),^43,45 1,N⁶-ethenoadenine (εA),^42,46 hydroxy-3,N⁴-ethanocytosine,^43,45 1,N⁶-ethanoadenine (EA),^42,46 3-(2’-deoxy-β-D-erythro-pentofuranosyl)-5,6,7,8-tetrahydro-6-hydroxypyrimido[1,2-α]purin-10(3H)-one (α-OH-PdG), 3-(2’-deoxy-β-D-erythro-pentofuranosyl)-5,6,7,8-tetrahydro-8-hydroxypyrimido[1,2-α]purin-10(3H)-one (γ-OH-PdG), and 3-(2’-deoxy-β-D-erythro-pentofuranosyl) pyrimido[1,2-α]purin-10(3H)-one (M₁dG) (Figure 3).¹⁴ Such complex alkylated and exocyclic adducts of DNA bases are results from various environmental and endogenous agents, and some of these damages are linked to cancer^47–51 and other diseases such as hyperglycemia⁵², obesity⁵³, and neurological disorders like Parkinson’s disease,^54,55 Huntington’s disease,⁵⁶ Rett syndrome,⁵⁷ autism spectrum disorders,⁵⁸ and many more.^59–62 For example, acrolein, a reactive α,β-unsaturated aldehyde, is found in environments polluted with tobacco smoke and automobile emissions and can also be produced endogenously through lipid peroxidation. It can damage DNA base pairs,^63–66 disrupt mitochondrial function, affect cellular respiration,^67,68 and can be related to Alzheimer’s disease in humans.^68–71 Often, the interaction of acrolein with nucleotides builds cytotoxic, mutagenic exocyclic adducts like α-OH-PdG and γ-OH-PdG lesions.^14,64,72 Analogous to acrolein, malondialdehyde (MDA) is primarily formed endogenously through cellular metabolism and dietary intake, contributing to DNA damage and disease pathology.⁷³ MDA interacts with the DNA, which leads to the formation of M₁dG lesions.^14,73,74 Similarly, the formation of the N²-methyl furan adducts (FF-dG) by nitrofurazone (NFZ),¹³ a commonly used antimicrobial drug for animals, has been reported to cause tumors in animals and is a potential human carcinogen.^75,76

Figure 3.

AlkB substrates studied in this work. a) α-OH-PdG, b) γ-OH-PdG, c) FF-dG, d) HF-dG, e) M₁dG.

Experimental studies using Q-TOF mass spectrometry revealed that AlkB efficiently repairs bulky lesions like γ-OH-PdG, M₁dG, and α-OH-PdG through complex and overlapping pathways, targeting both the open-ring and closed-ring forms of these three carbon-bridge exocyclic adducts (Figures 3 and 4).¹⁴ The AlkB-catalyzed repair of most of the adducts follows the standard hydroxylation mechanism; however, the M₁dG adduct deviates from this pathway by forming an epoxide intermediate,¹⁴ similar to the repair of etheno lesions by AlkB.^{16,34,38,43,45}

Figure 4.

The proposed mechanism for the repair of the complex alkylated – a) HF-dG b) FF-dG and exocyclic adducts – c) α-OH-PdG, d) γ-OH-PdG e) M₁dG by AlkB.

Experimental studies also showed that AlkB repairs FF-dG and HF-dG by the hydroxylation of the methylene group bonded to the N2 position of guanine (Figures 4a and 4b).¹³ Subsequently, it is proposed that proton transfer from the hydroxyl group on the methylene carbon to the N2 guanine restores the regular DNA and releases tetrahydrofuran (THF) aldehyde for HF-dG and furfuryl aldehyde for FF-dG. However, the hydroxylated intermediate of the substrate can undergo non-enzymatic dehydration to form HF-2H and FF-2H, which are Schiff’s base dehydration products observed experimentally.¹³

Based on experimental observations, two tautomeric forms of α-OH-PdG (open α-OH-PdG or OαG and closed α-OH-PdG or CαG forms) are proposed (Figures 3a and 4c). It is proposed that in both forms, the reaction starts with hydroxylation of the γ-carbon position of guanine (Figure 4c). In the CαG, the hydroxylated intermediate is proposed to undergo ring-opening facilitated by a proton transfer from the γ-hydroxyl group to the N1 of guanine, resulting in an aldehydic intermediate, followed by a second proton transfer from the α-hydroxyl group to the N2 nitrogen to yield the undamaged guanine. In the OαG, the hemiaminal intermediate undergoes proton transfer from the γ-hydroxyl group to the N1 position of guanine, ultimately forming undamaged guanine and MDA.¹⁴

For the open-ring (OγG) and the closed forms (CγG) of the γ-OH-PdG substrate, the α-carbon position of guanine is proposed to undergo hydroxylation.¹⁴ In the CγG, the hydroxylated substrate then undergoes ring opening at the N1 position, resulting in the same intermediate as the hydroxylated open ring aldehydic form (Figure 4d). Further, the proton transfer to the N2 guanine from the hydroxyl group on the α-carbon coordinated to N2 will result in the generation of undamaged guanine and MDA.¹⁴

Unlike the other substrates studied, the AlkB catalysis with the M₁dG substrate is proposed to form an epoxide intermediate at the β,γ-carbon unsaturated bond. Further, the two hydration steps lead to the hydroxylation of α, β, γ positions; then, two consecutive proton rearrangements result in the conversion to guanine and 2-OH MDA (Figure 4e).¹⁴

Although the biological effects of the alkylating and exocyclic agents and their chemical transformations have been studied experimentally, the catalytic reaction mechanism of AlkB with these chemically complex adducts is completely unexplored. There is currently no information on the energies and geometries of key transition states (TSs) and intermediates (IMs), nor on how the accommodation of complex substrates in the active site influences protein dynamics. Furthermore, the role of the second-coordination sphere (SCS), long-range (LR) interactions, and correlated motions in the catalytic process remains undefined. It is unclear whether the same SCS/LR residues involved in repairing methylated bases also contribute to the catalysis of complex adducts or if there are substrate-specific variations in their roles. The mechanism of the post-hydroxylation reaction is also unexplored. Although studies on the FTO⁷⁷ and TET2⁷⁸ suggest that the post-hydroxylation reaction can occur outside the enzyme in an aqueous environment, the AlkB process with alkylated/exocyclic adducts remains unclear. Altogether, the complex alkylated/exocyclic adducts introduce multiple levels of complexity over simple methylation, such as size, increased steric on the active site, ring-chain tautomeric forms variability, and alternative reaction pathways. These effects are entirely unexplored, although they are vitally important for understanding complex AlkB-catalyzed dealkylations.

To complement the missing knowledge, we performed Molecular Dynamics (MD) - based hybrid quantum mechanics/ molecular mechanics (QM/MM) and cluster quantum mechanical (QM) calculations of the catalytic mechanism of AlkB with ss-DNA containing alkylated and exocyclic GA, including the α-OH-PdG, γ-OH-PdG, M₁dG, FF-dG, and HF-dG. The studies reveal the mechanistic catalytic strategies of AlkB with ss-DNA containing complex alkylated/exocyclic GAs in correlation to the experimental data and identified the key SCS/LR residues that contribute specifically to the reactions with each of the adducts.

2. Methodology

2.1. System preparation

The X-ray crystal structure of E. coli AlkB complexed with ss-DNA containing a m1G lesion (PDB ID: 3KHC)⁷⁹ was used as the starting structure. The active site Co(II) used for crystallization was replaced with Fe(IV) using GaussView 6.0, and 2OG was modified to succinate. The metal-coordinated water molecule was also substituted with an oxygen atom to model the ferryl complex. Hydrogen atoms were added using the Leap module in Amber, and standard protonation states were assigned at pH 7.0. The Amber force field FF14SB⁸⁰ was used to generate the parameters for standard amino acid residues. The parameters for the HS ferryl complex, where Fe is coordinated by two histidine residues, one aspartic acid residue, succinate, and an oxygen atom, were prepared using the Metal Center Parameter Builder (MCPB.py)⁸¹ in the Amber22 suite.⁸² The Seminario method⁸³ was applied to calculate bond and angle force constants, with the point charges for the electrostatic potential determined via RESP charge fitting (ChgModB method).⁸⁴ The necessary substrates were derived from m1G using GaussView, and the parameters were generated using the Antechamber⁸⁵ module in Amber22. Similarly antechamber is adopted for generating the parameters for Fe(IV) coordinated succinate and oxygen. Transferable Intermolecular Potential 3-Point (TIP3P)⁸⁶ water molecules were used to solvate the system, ensuring the water box had a minimum distance of 10 Å between the protein surface and its boundary. Na⁺ counter ions were used to neutralize the charge of the protein system.

2.2. Molecular dynamics simulations

After solvating the system, a two-step minimization was performed. The initial step optimizes the water molecules and Na+ ions while restraining the solute (protein) with a 500 kcal mol⁻¹ Å⁻² harmonic potential. The second step involves optimizing all atoms without restraints, effectively removing clashes and unwanted contacts. The 5000 steps of steepest descent and conjugate gradient energy minimization are adopted in each of the two minimization steps. Post-minimization, the system was heated from 0 to 300 K across 50 ps in a canonical (NVT) ensemble using the Langevin thermostat⁸⁷ with the solute molecules restrained by a 50 kcal mol⁻¹ Å⁻² harmonic potential. Following heating, the system was kept at a constant 300K in an NPT ensemble for 1 ns, applying a 5 kcal mol⁻¹ Å⁻² restraint on the solute molecules while maintaining a pressure of 1 bar. Further, the system is equilibrated at a fixed temperature of 300K and 1 bar pressure (NPT ensemble) without restraining the solute molecule for 3ns. Productive MD calculations were performed using the GPU version of Amber22 for 1 μs at 1 bar, with a pressure coupling constant of 2 ps. The simulations were conducted using the FF14SB force field⁸⁰ under periodic boundary conditions. The pressure was maintained with a Berendsen barostat,⁸⁸, and hydrogen bonds were constrained using the SHAKE algorithm.⁸⁹ The LR electrostatic interactions were calculated using the Particle Mesh Ewald (PME).⁹⁰ The CPPTRAJ module⁹¹ from AmberTools was used to perform Root Mean Square Deviation (RMSD), Root Mean Square Fluctuations (RMSF), electrostatic interactions, and hydrogen bond analysis. VMD⁹² and Chimera⁹³ were used for MD trajectory analysis. Dynamic cross-correlation analysis (DCCA) and principal component analysis (PCA) were performed using the Bio3D package in R.⁹⁴

2.3. QM/MM calculations

ChemShell⁹⁵ was used to perform QM/MM calculations, integrating Turbomole⁹⁶ for the QM region and DL_POLY⁹⁷ for the MM region. The water molecules and ions beyond 12 Å of the protein were stripped. The QM region consisted of Fe(IV) metal, imidazole groups of histidine H131 and H187, succinate, the side chain carboxylate of D133, the oxo group, and the guanine base of the substrates. The atoms within the 8 Å of the active site are considered the flexible MM region, and the positions of molecules beyond that are fixed to their positions. The Amber FF14SB force field was used to describe the MM region, and the QM part was accounted for using density functional theory (DFT) with the unrestricted B3LYP (UB3LYP) functional. UB3LYP functional has been used in numerous computational studies of metalloenzymes and has been shown to be reliable in predicting the electronic structure and reaction path calculations involving Fe(IV)-oxo systems.^{25,78,98–106} Experimental and computational studies have demonstrated that dioxygen activation and substrate oxidation in non-heme Fe(II)/2OG enzymes occur via the HS quintet state (S = 2, M = 5).^37,107–112 So, all mechanistic studies involving the active ferryl intermediate were carried out on the quintet state. Geometric optimizations were conducted using DFT with the UB3LYP functional and def2-SVP basis set (B1) for all atoms. The DL_find optimizer¹¹³ performed potential energy scans along the reaction coordinate with 0.1 Å increments from the reactant complex (RC) to obtain TS, IM, and products. The TS complexes were optimized without constraints using the dimer method¹¹⁴ in ChemShell. Single-point (SP) calculations on the geometries were performed using the UB3LYP functional and def2-TZVP basis set (labeled B2). Further, frequency calculations were carried out on all optimized geometries to confirm the TS and minima. Zero-point energy (ZPE) corrections from the frequency calculations were added to the SP energy to obtain the final reported B3 energies (B2+ZPE). Spin Natural Orbital (SNO) analysis¹¹⁵ was used to examine the molecular orbitals involved in the reaction pathway. Energy decomposition analysis (EDA), developed by Cisneros et al.^116–118 was conducted on the optimized RCs and TSs to determine the total energy contributions/interactions (ΔE), including van der Waals and electrostatic interactions of individual residues. The differences in interaction energies between the TS and RC structures (ΔΔE) were calculated to assess the contributions of individual residues to TS stabilization. Non-covalent interaction (NCI) analysis^119–121 was performed to visualize the non-covalent interactions.

2.4. QM calculations

The reaction paths in the solution were calculated using hybrid cluster-continuum (HCC) calculations with Gaussian16,¹²² including the substrate part corresponding to the QM region in the QM/MM calculations. The solvation effects were incorporated using the Solvation Model Based on Density (SMD).¹²³ Additionally, explicit water molecules were included around the substrate to assist proton transfer.

3. Results and Discussion

3.1. How do the ss-DNA substrates containing different GAs influence the interactions and dynamics in the enzyme-substrate complexes?

3.1.1. Do the active site interactions change upon the binding of different GAs?

At present, there are no crystallographic or NMR structures of the AlkB-DNA complexes with the complex alkylated and exocyclic GAs, which are subjects of the present study. The chemical nature of the alkylations/exocycles is more complicated than that of a single methyl group; therefore, it is expected that the complex alkylations/exocycles will influence the geometries, local SCS interactions, and LR correlated dynamics of the enzyme-substrate (ES) complexes. To provide these insights, we performed MD simulations of the AlkB-ss-DNA ferryl complexes for all the proposed substrates (the alkylated FF-dG, HF-dG, and exocyclic α-OH-PdG, γ-OH-PdG, and M₁dG). All complexes were well equilibrated and showed stable RMSDs (Figures S1–S7). The average distance between the ferryl oxygen O_p and the substrate carbon (or hydrogen) undergoing the initial HAT step varied between 3.48 Å and 4.65 Å, and the average angle between the Fe, O_p, and C varied between 118.5 and 137.2 degrees across the ferryl complexes (Figures S1–S7).

The MD simulations show that in all the AlkB-substrate complexes, the substrate is enclosed within a hydrophobic pocket in the active site, involving residues M49, T51, M57, V59, L118, and L128 (Figure 5). The interactions with the hydrophobic pocket vary between the different substrates (Table S1). Interactions with M57 and L128 are observed in all ferryl complexes, while L118 forms part of the hydrophobic pocket in all complexes but FF-dG. V59 is present in HF-dG, while M49 and T51 are in FF-dG (Figures 6a, 6b, and S8, Table S1). There are also subtle differences in the interactions between CγG and OγG forms. In the CγG, interactions with V59 and L128 are present; however, the OγG interacts with T51 and L128 instead (Figures 6c, 6d, and S9). In the case of OαG and CαG, interaction with M49 is absent in comparison to other substrates (Figures 7a, 7b, and S10, Table S1), indicating that there are fine and delicate differences in the SCS interactions for the different structural modifications of the substrate, which could be important for the substrate specificity (Table S1).

Figure 5.

Residues forming the hydrophobic pocket (shown in green) surrounding the overlaid substrates in AlkB-Fe(IV)=O-GA complex. Substrates are colored as follows:

Figure 6.

The SCS interactions in a) HF-dG b) FF-dG c) CγG d) OγG substrates from 1μs MD simulation of the AlkB- ferryl complex. The frames are selected based on the average structure from the equilibrated portion of the MD simulations and capture the representative interactions observed throughout the simulation.

Figure 7.

The SCS interactions in a) CαG, b) OαG, c) M₁dG substrates from 1μs MD simulation of the AlkB-ferryl complex.The frames are selected based on the average structure from the equilibrated portion of the MD simulations and capture the representative interactions observed throughout the simulation.

Apart from the hydrophobic interactions, the guanine ring of the substrates forms π-stacking interactions with H131 and W69, stabilizing the substrate orientation within the active site (Figures 6 and 7). However, in the HF-dG and FF-dG complexes, the W69 interaction with the guanine is interfered with by the HF and FF rings, respectively, but the stacking interaction with the H131 is maintained in both (Figures 6a and 6b). In the FF-dG complex, the aromatic FF ring adduct forms a T-shaped π-stacking interaction with the indole side chain of W69. Additionally, a hydrogen bond between the side chain of the S129 and the DNA ribose’s O4’ oxygen is observed in FF-dG, OγG, OαG, and AlkB-M₁dG complexes (Figures 6 and 7). While in the CγG complex, the side chain of S58 forms a hydrogen bond with the ribose’s hydroxyl group. Moreover, the side chain’s hydroxyl group of Y76 forms hydrogen bonds with the phosphate oxygen in OγG, CαG, and OαG substrate. Y76 plays a crucial role in stabilizing all the substrates in the AlkB ferryl complexes through interactions with the sugar-phosphate backbone, except for HF-dG and FF-dG. In the HF-dG, a unique hydrogen bond between the guanine oxygen of the substrate and the backbone of A135 is observed (Figures 6 and 7). The NCI plot (Figure S11) illustrates key non-covalent interactions stabilizing the complex.

The tautomeric OγG and CγG forms of γ-OH-PdG are characterized by different structural features. For example, the OγG has a larger bond angle between the Fe, oxygen, and the targeted hydrogen atom for abstraction, averaging 137.2° compared to 119.2° in the CγG (Figures S3 and S4). The CγG exhibits an increased distance of the targeted carbon from the ferryl oxygen, with average distances of 4.7 Å and 3.6 Å for the CγG and OγG in the MD trajectory, respectively.

The structural features along the MDs are consistent with the crystallographic structure of the AlkB-m1G complex (PDB ID: 3KHC),⁷⁹ highlighting the role of SCS residues in substrate binding. For example, the role of Y76 in AlkB catalysis shown in both the crystallographic studies and the present MD simulations is aligned with biochemical studies, which showed that the Y76A mutant showed only 20% of the wild-type enzyme’s activity.⁷⁹ Similarly, the stacking interaction of W69 was observed in the crystal structure and is crucial for proper substrate positioning in the active site, which is also reproduced in the present MD. The hydrogen bonds between the hydroxyl group of T51 and the backbone amide group of G53 with 5’ phosphate of nucleotide adjacent to the aberrant base observed in the crystal structure disappear in the MDs of all AlkB-substrate complexes. However, T51 interacts with the substrate’s sugar-phosphate backbone and, along with Y76 and S129, plays a key role in substrate stabilization in the active site, as observed in the crystal structure. Similarly, in most of the AlkB-substrate complexes, the C4 carboxylate of the succinate is stabilized by a salt bridge with the sidechain of R204, and the C1 carboxylate is stabilized by a hydrogen bonding interaction with R210; also, the metal-coordinated aspartate forms a hydrogen bonding interaction with R210 (Figure S12).

3.1.2. Do the correlated motions change upon binding of different GAs?

In all the AlkB-Fe(IV)=O-GA complexes in this work, we observed consistent correlated motions between the substrates and the SCS residues surrounding them (L118, L128, W69, M49, M57, V59, and P52) (Figure S13). In AlkB-FF-dG, HF-dG, M₁dG, CγG, OγG, CαG and OαG complexes, the Fe center shows correlated motions with the β-sheets- β-C2 to β-C7 (Figure S14). While the OαG complex exhibits a correlation only with the β-C7 sheet. Generally, all the substrates show correlated motions with the loop connecting the β1 and β2 sheets of the DSBH. These correlated motions might play a role in stabilizing and orienting the substrate towards the Fe(IV)-oxo moiety for the HAT reaction.

The overlaid average MD structures of AlkB-Fe(IV)=O-GA complexes reveal variations in the conformation of flexible regions in AlkB, suggesting differential interactions with the substrates (Figure 8). These flexible regions include the loop between the N-terminal and β1 sheet of the NRL (M49-S58, L1), the loop connecting the β1 and β2 sheets of the NRL (T71-Y76, L2), the loop linking α-C1 and the β1 sheet (the secondary structures are labeled in Figure 1 for the reference) of the catalytic domain (Y109-P114, L3), and the loop between the β2 and β3 sheets of the catalytic domain (Q132-P142, L4) (Figure 8). The PCA further highlights the significance of these flexible loops in the dynamics of the ferryl complex (Figure S15). The L4 loop (between the β2 and β3 sheets of the catalytic domain) consistently emerges as a flexible region in PCA analysis for most substrates (Figure S15).

Figure 8.

Overlaid average MD structure of AlkB-Fe(IV)=O-GA complexes for each substrate with the flexible loop labeled. The AlkB ferryl complex are colored as follows: .

Correlated motions with the participation of the loop regions show variations between the different substrates, which suggest the importance of the loops in the modulation of adduct-selective catalysis by AlkB, and their potential implementation in AlkB redesign. More details are provided in the SI (Figures S13–S15 and pages S18–S19)

3.2. Catalytic mechanism of AlkB with ss-DNA containing alkylated/exocyclic GAs

To reveal the reactivity of AlkB with the alkylated/exocyclic adducts, we first explored the HAT reaction in each AlkB-Fe(IV)=O-GA complex, followed by the reaction of rebound hydroxylation, and finally, we analyzed the proton transfer reactions leading to the final products.

3.2.1. How does HAT proceed in the different AlkB-alkylated/exocyclic complexes?

3.2.1.1. HAT in the alkylated substrates FF-dG and HF-dG.

In the alkylated adducts FF-dG and HF-dG, the HAT takes place from the methylene group bound to the N2 atom of the alkylated guanine (Figures 4a and 4b). The lowest HAT barriers are 10.7 kcal/mol and 14.9 kcal/mol (Figure 9), respectively, with Boltzmann average values of 11.6 kcal/mol (FF-dG) and 15.6 kcal/mol (HF-dG) across five different starting structures for each complex (RCs) (Tables S2–S4). The RC and TS geometries look similar in both the FF-dG (TSFF1) and HF-dG (TSHF1) complexes (Figure 9b and 9d). The average O_p-H distances and Fe-O_p-H angles are 1.31 Å and 133.4° (FF-dG) and 1.30 Å and 143.0° (HF-dG). For FF-dG, the radical delocalization in the aromatic furan ring provides greater stability to the intermediate IMFF1 compared to IMHF1 in HF-dG, which is evident from the reaction energy of −11.9 kcal/mol compared to the 0.4 kcal/mol for HF-dG (Figures 9a and 9c).

Figure 9.

Analysis of FF-dG and HF-dG substrates. (a,c) QM/MM energy profile at the QM(B3)/MM B3LYP level of theory (in kcal/mol) for the HAT and rebound hydroxylation steps of FF-dG and HF-dG, respectively. (b,d) Geometries of the stationary points in the HAT and rebound steps of FF-dG and HF-dG, respectively, with structural parameters shown in black (distance in Å, angle in degree) and spin population shown in blue.

The SCS residues that stabilize the TS in both FF-dG (TSFF1) and HF-dG (TSHF1) include W69 (via a stacking interaction) and L128 ( via hydrophobic interaction); in TSFF1 -; M49, T51 (via hydrophobic interactions) and S129 (via a hydrogen bond); and in TSHF1 - M57, V59, L118 (via hydrophobic interactions), and A135 (via a hydrogen bond) (Figures 6a, 6b and S8). In TSFF1, the aromatic side chain of the W69 is oriented perpendicular to the furan ring of FF-dG, forming a T-shaped π-π stacking interaction, while in TSHF1, the W69 forms an edge-to-edge π-interaction with the guanine ring, with smaller π-electron cloud overlap. Therefore, the low HAT barrier for the FF-dG compared to HF-dG could be correlated to the stronger TS stabilizing π-π interaction between the W69 and the furan ring of FF-dG. In addition, the furan ring of the FF-dG interacts with the side chain of the M49, which can also contribute to the proper orientation of the substrate in the active site, while this interaction is absent for HF-dG in both RC and TS complexes for HAT. Together, these stabilizing interactions could drive AlkB’s higher efficiency in repairing FF-dG compared to HF-dG (Figures 6a, 6b and S8).¹³

3.2.1.2. HAT in exocyclic substrates γ-OH-PdG and α-OH-PdG

For the γ-OH-PdG and α-OH-PdG substrates, we first examined the transition from closed to open conformation in the enzyme and in water. In the enzyme environment, the activation barriers of 65.4 kcal/mol for γ-OH-PdG and 66.5 kcal/mol for α-OH-PdG, leading to endothermic open conformations (9.8 and 20.4 kcal/mol, respectively), suggest that tautomerization is unlikely to proceed inside the enzyme. Therefore, we explored the possibility of these ring-opening transformations to be performed in solution. Our QM calculations (QM(B3)/B3LYP) showed barriers of 20.7 kcal/mol (γ-OH-PdG) and 27.6 kcal/mol (α-OH-PdG) and led to the corresponding open forms with endothermicity of 3.2 kcal/mol (γ-OH-PdG) and 8.0 kcal/mol(α-OH-PdG) (Figure S16). The calculations further suggest that the ring opening in the γ-OH-PdG could be more favorable than that in the α-OH-PdG. Since both the open and closed forms are substrates for AlkB, we propose that the tautomerization happens before binding to the AlkB protein.

Then, we explored the HAT reaction from the RC complexes in both the open and the closed forms. The lowest activation barrier for HAT from the N2 guanine-bonded exocyclic carbon in the OγG is 10.2 kcal/mol (TSγO1), which is lower than that of the closed configuration (28.4 kcal/mol, TSγC1). The Boltzmann average barriers for the five RCs for OγG and CγG are 11.1 and 29.5 kcal/mol, respectively (Table S2). This substantial barrier difference indicates AlkB repairs the OγG more efficiently than the CγG, in agreement with experimental studies (Figures 10a and S17a, S18).¹⁴

Figure 10.

Analysis of OγG and CαG substrates. (a,c) QM/MM energy profile at the QM(B3)/MM B3LYP level of theory (in kcal/mol) of the HAT and rebound hydroxylation steps of OγG and CαG, respectively. (b,d) Geometries of the stationary points in the HAT and rebound steps of OγG and CαG, respectively, with structural parameters shown in black(distance in Å, angle in degree) and spin population shown in blue.

Comparing the geometric parameters in RCs shows a substantial difference in the distance between the ferryl oxygen and the hydrogen atom on the Cα for the HAT, with values of 4.94 Å and 2.33 Å for the CγG and OγG, respectively (Figures 10b and S17b). The SCS residues could also influence the differential catalytic efficiency of AlkB towards the γ-OH-PdG tautomers. For example, to properly orient for HAT, rotation of the exocyclic adduct region towards ferryl oxygen in CγG is essential. However, this movement is hindered by the hydrophobic residues surrounding this region, leading to a higher HAT barrier. The strong hydrophobic interactions with the residues in the region between β1 and β2, particularly T51, M57, V59, and S58 (steric interaction), result in restricted motion of the CγG, thereby affecting substrate orientation. The residues stabilizing the γ-OH-PdG substrate are M57, S58, V59, W69, Y76, L118, and L128 for the CγG and T51, M57, W69, Y76, L118, and L128 for the OγG form (Figure 6c and 6d). Substrate orientation and the SCS stabilization favor the OγG for the efficient HAT.

For the α-OH-PdG substrate, the HAT barriers for the CαG and OαG forms are 18.0 kcal/mol and 23.6 kcal/mol, respectively (Figures 10c and S17c), which agrees with the experimentally observed higher efficiency of AlkB in repairing the CαG. The Boltzmann average barriers for the five RCs for CαG and OαG are 18.9 and 23.6 kcal/mol, respectively (Table S2). The geometries in both RCs are comparable (Figures 10d and S17d), which indicates that the SCS residues are crucial in controlling the reactivity with α-OH-PdG, similar to γ-OH-PdG. The bulkiness of the adducts at the guanine N1 position in α-OH-PdG could enhance steric clashes in the active site, and the disruption of the guanine ring stacking interaction with W69 could lower the repair efficiency in the OαG. Furthermore, the following SCS interactions orient the substrate effectively for the repair in the CαG. i) the π-π interaction between W69 and the guanine ring; ii) the hydrogen bond between the hydroxyl group on the exocyclic adduct and N120; and iii) the hydrophobic interaction with V59 (Figures 7a and 7b). Similar to γ-OH-PdG substrates, the same residues show strong hydrophobic interaction in the closed form; however, the reactive groups are oriented more favorably to the ferryl oxygen, leading to an effective HAT in the α-OH-PdG. Dehydrated closed form of α-OH-PdG could serve as an alternative substrate for the AlkB enzyme, and the additional details on the study are provided in SI, Figure S19 asnd page S21.

Our calculations confirm that AlkB repairs the OγG and CαG forms more efficiently than their respective tautomers, in agreement with experimental studies (Figures 11 and S18).¹⁴

Figure 11.

The relative reaction profiles at the QM(B3)/MM B3LYP level of theory of the substrates repaired by AlkB enzyme. Each of the reaction paths starts on its on RC1 and the energies (in kcal/mol) of all RC1s are set to zero.

3.2.2. Molecular orbital mechanism of the HAT reactions

In all the AlkB-substrate complexes studied in this work, HAT follows a σ channel mechanism, in which an α electron (ms = +1/2) from the substrate’s σ_C-H orbital transfers to the Fe(IV)-O_p σ*_dz2 antibonding orbital(Figure 12).^{111,124–126} In a few snapshots, the π channel mechanism is observed; however, it has a higher barrier compared to the σ channel (e.g., HF-dG substrate’s snapshot 3 – HAT barrier is 27.5 kcal/mol and shows a π channel mechanism, Table S3), which indicates that the σ channel could be the predominant pathway. The Fe-O_p-H angles in the TSs vary from 116.9° to 175.7° (Table S3). The MO pathway is further confirmed by Mulliken spin population and SNO analysis (Figures S20–S25). The spin population on Fe varies from 3.93 to 4.08 in the HAT TSs (σ channel), compared to 3.11 to 3.21 in the RCs, indicating an increase in the number of unpaired electrons in Fe(IV)=O during the TSs. Also, in the HAT TSs, the spin population on the target carbon atom in the substrate varies from −0.21 to −0.27, indicating that an α electron is transferred to Fe-oxo while a β electron is retained on the carbon atom. A representative SNO analysis (Figure 12b) and the visualization of the spin density (Figure S26) for the OγG illustrate the electron transfer between the Fe-oxo σ*z² antibonding orbital and the substrate’s σ_C-H orbital.

Figure 12.

a) The electron transfer pathway involved in the HAT by ferryl complex. b) Spin natural orbitals (SNOs) and their occupancies (in bracket) in the HAT TSγO1 for AlkB-OγG complex.

3.2.3. How do SCS and LR interactions stabilize the HAT TS?

We further explored the SCS and LR residues that energetically contribute to stabilizing the HAT TSs using EDA (Figures S27–S33).^116–118 The analysis reveals that SCS residues are invariably involved in the TS stabilization for all substrates, but it also indicates residues that participate in TS stabilization for specific substrates. W69, N120, R210, and E136 were found invariably for all TSs. The residues contributing to the stabilization of the TS of FF-dG are M49, F185, and R210, while for the TS stabilization of HF-dG- are involved R204, N120, R210, F154, and W69 (Figure S34). For γ-OH-PdG, W69 and R210 are found to stabilize the HAT TSs in both the CγG and OγG forms. Y76 specifically stabilizes the TS in the OγG, while S145, R204, L118, L128, and Y122 stabilize the TS in the CγG. For α-OH-PdG - N120 stabilizes the TSs for both the OαG and CαG forms, F185 is specific to the OαG, while L118, R210, S129, and W69 are specific to the CαG. The study revealed the combination of invariant and substrate-specific SCS residues that are involved in the TS stabilization that, in the long term, can be used for substrate-specific AlkB redesign.

3.2.4. How does the flexibility of loop regions assure the proper RC orientation for the HAT?

MD studies showed that loops L1, L2, L3, and L4 differentially modulate each substrate, thus assuring effective binding (Figures 8 and S15). The substrate-specific differential positioning of these four loops is also maintained in both the optimized RC complexes and TS structures, thus confirming their differential conformational adjustment for each substrate is conserved along the HAT reaction path, further affirming the role of the loops in catalysis for each substrate. An interesting observation from PCA is the reduced flexibility of loops L1 to L4 in the CγG and OαG substrates, which are less efficiently repaired compared to their open and closed counterparts. This reduced flexibility might reflect the enzyme’s inability to optimally accommodate these forms, supporting the preference for the OγG and the CαG by AlkB. In the OαG substrate, the rigidity of the L4 loop arises from a hydrogen bond between Q132 and K134 (Figure S35). In the CαG, K134 instead stabilizes the backbone phosphate of the adjacent adenosine nucleotide in the 5’ end. A similar trend is observed in γ-OH-PdG, where the CγG shows stronger interactions between Q132 and K134 compared to the OγG. In the OγG, K134 also interacts with the adenosine nucleotide for up to 250 ns in the MD simulation, but this interaction is absent in the CγG.

The elucidation of the invariant and substrate-specific SCS interactions and substrate-specific loop interactions provides a deeper understanding of the origins of the catalytic process in AlkB with diverse alkylated/exocyclic substrates and provides in the long-term background for modulation of catalysis of AlkB with a broad number of substrates, but also for selective modulation with individual substrates.

3.3. Mechanism of the rebound hydroxylation

The HAT results in the formation of Fe(III)-OH intermediate and a C-centered substrate radical. In the next step, the OH group from Fe(III)-OH transfers to the C radical to form a C-OH bond (rebound hydroxylation), while the electron from the C-radical transfers to Fe(III) and reduces to Fe(II).^111,125,127 The hydroxylated substrate then undergoes a series of rearrangements to yield undamaged DNA.^13,14 The DFT study on the alkane hydroxylation process in non-heme iron(IV) complexes shows a concerted barrierless transition in the quintet state.¹²⁸ In our study, the formation of hydroxylated substrates proceeds either without an energy barrier or with barriers ranging from very low up to 23.0 kcal/mol. Specifically for the OγG, substrates require no rebound barrier to become IM2 in the B3 energy, consistent with similar computational studies.^37,129–131 However, in the B2 energy profile (without ZPE correction), small rebound barriers of 1.5 are observed for OγG(Table S4). An interesting observation is the presence of hydrogen bonding between the hydrogen of ferric hydroxo and the succinate oxygen in the three efficiently repaired substrates in the study, namely CαG, FF-dG, and HF-dG, and they also have low rebound barriers. For example, the barrier is 0.4 kcal/mol for FF-dG, 0.1 kcal/mol for HF-dG, and 0.7 kcal/mol for CαG, which also agrees with similar non-heme Fe(II)/2OG studies.^128,132,133 Compared to their tautomers, OαG and CγG exhibit a higher rebound barrier, corresponding with AlkB’s lower repair efficiency for these tautomers.¹⁴ The higher barrier likely results from the steric interactions, which hinder proper substrate orientation towards the Fe center. The same residues that stabilize the HAT TSs were also found to stabilize the rebound TSs.

3.4. A differential repair of M₁dG via epoxidation

In the AlkB-M₁dG ferryl complex, the Fe(IV)-oxo moiety aligns towards the Cγ=Cβ bond of the substrate. Therefore, epoxidation via oxygen atom transfer (OAT) is observed rather than the typical HAT.^34,38,134 The OAT leads to the formation of a Fe(III)-alkoxide intermediate (IMM₁1, Figure 13) with a radical character on the Cγ atom. The reaction proceeds via TSM₁1, which requires a barrier of 20.9 kcal/mol. In TSM₁1 geometry, the Op-Cβ distance gets shorter to 1.90 Å from 3.16 Å. The spin population on the Cγ carbon at TS and IM states is −0.37 (TSM₁1) and −0.75 (IMM₁1), respectively, and the charges (close to zero) confirm the radical character (Figure 13c). In the next step, the electron from the Cγ radical transfers to Fe(III) to form the Fe(II) state and the epoxide product, which occurs rapidly with a barrier of 4.1 kcal/mol. The results are consistent with the prior experimental and computational studies on the epoxidation mechanism by the non-heme Fe(II)/2OG-dependent enzymes.¹³⁵ The stacking interaction of the substrate’s guanine ring with W69 and H131 plays a major role in orienting the substrate’s Cγ=Cβ bond in the active site. Further, the SCS residues M57, W69, Y76, L128, and S129 are involved in stabilizing the epoxidation TSs as similarly observed in the HAT studies.

Figure 13.

Analysis of M₁dG substrate. a) QM/MM energy profile at the QM(B3)/MM B3LYP level of theory(in kcal/mol) of the epoxidation step. b) Geometries of the stationary points in the epoxidation step with structural parameters shown in black (distance in Å, angle in degree) and spin population shown in blue. c) The spin and the NBO values of the TSM₁1 and IMM₁1 structures of the M₁dG substrate prior to the epoxidation.

Experimental studies propose that the epoxide formation is followed by two hydration and two proton transfers, leading to the formation of undamaged guanine and aldehyde products (Figure 4e).¹⁴ Similar to the post-hydroxylation, post-epoxidation steps are also likely to proceed in water rather than in the enzyme due to the absence of a residue to assist the hydration step and the lack of sufficient water molecules in the active site to assist the proton transfer reaction.

3.5. Post-hydroxylation reactions – do they proceed in the enzyme or in water?

Post-hydroxylation steps can proceed either in the enzyme environment or in the solvent, as in FTO⁷⁷ and TET2⁷⁸ enzymes. Post-hydroxylation steps in AlkB-Fe(IV)=O-GA complex repair were investigated using QM/MM and QM HCC calculations to determine whether these reactions proceed in the enzyme environment or solution. As a representative substrate, we studied the hydroxylated OγG, which undergoes proton transfer from the hydroxyl group to guanine’s N2 to become the undamaged state. Detailed analysis of the calculations is provided in the SI (Table S5, pages S35–S36). The results indicate that the post-hydroxylation steps to restore the undamaged substrate take place in a water environment and not in the enzyme and proceed via water chain-mediated proton transfers.

We further explored the post-hydroxylation steps in water for the other AlkB-product complexes-HF-dG, FF-dG, OγG, and OαG substrates. The OγG, CγG, and CαG substrates share similar intermediates after hydroxylation, so we studied the undamaged guanine formation in OγG and obtained the activation barrier of 16.5 kcal/mol (QM(B3) B3LYP level of theory). However, the OαG substrate has a different hydroxylated intermediate structure, so we simulated the post-hydroxylation step separately. For the OαG, HF-dG, and FF-dG, the barriers are 20.4, 16.8, and 18.3 kcal/mol, respectively (Figures S36–S40). The hydroxylated FF-dG and HF-dG products can either directly convert into undamaged DNA or form an imine intermediate by losing water - a process that requires 20.3 kcal/mol for FF-dG and 20.7 kcal/mol (QM(B3) B3LYP level of theory) for HF-dG (Figure S41 and page S40 for more details).

3.6. Post-epoxidation reactions for M₁dG

Following the epoxidation of M₁dG, the formed epoxide undergoes a water-mediated ring opening (IMM₁-HYD1) followed by hydration with barriers of 24.2 kcal/mol and 14.3 kcal/mol, respectively, forming a triol intermediate IMM₁-HYD2 (Figure S42). In the next step, a proton transfer from water to N1 of guanine with a similar barrier of 24.9 kcal/mol leads to IMM₁-PT1. Subsequent proton transfer to N2 guanine with a barrier of 14.8 kcal/mol forms the undamaged guanine and MDA aldehyde, with a reaction energy of –25.8 kcal/mol relative to the epoxide intermediate (PDM₁).

4. Conclusions

The molecular mechanism of AlkB-mediated repair of damaged alkylated and exocyclic ss-DNA guanine lesions was investigated using MD, QM/MM, and cluster QM methods in strong correlation to earlier experimental studies.^13,14 The reaction follows the HAT and rebound hydroxylation pathway in most of the substrate systems, however, an epoxidation is favored when the substrate contains an unsaturated Cγ=Cβ group oriented towards the Fe(IV)-oxo moiety. The epoxidation follows a step-wise mechanism involving the formation of Fe(III)-alkoxide radical species. The removal of GA from HF-dG, FF-dG, γ-OH-PdG, and α-OH-PdG is initiated via C–H activation, consistent with experimental observations, and highlights the critical role of second coordination sphere (SCS) residues such as W69, M57, L118, Y76 and L128 in stabilizing the hydrogen atom transfer transition states; while residues including M57, W69, Y76, L128, and S129 play central role in stabilizing the epoxidation mechanism of M₁dG. The stacking interactions of W69 and H131 with the substrate are critical for substrate positioning and catalysis. Both the HAT and epoxidation follow a σ channel pathway.

The study distinguishes the preference for ring-chain tautomers, showing that AlkB exhibits differential repair efficiency toward open and closed ring tautomers. The AlkB enzyme shows strong hydrophobic interactions with closed-ring tautomers, which play a key role in modulating HAT reactivity. In CγG, these interactions hinder HAT by misorienting the reactive site away from the ferryl species. In contrast, in α-OH-PdG, the reactive region is favorably aligned toward the ferryl oxygen, promoting HAT. These differences arise mainly from hydrophobic residues in the L1 region, namely T51, M57, V59, and S58 (steric interaction), and the better conformational adaptability of the L4 loop to accommodate the closed-ring form. Our calculations also reveal that AlkB repairs FF-dG more efficiently than HF-dG, which agrees with experimental results due to more preferable SCS interactions (such as stacking interaction with W69). PCA analysis and structural overlays of the average MD structures of the ferryl complexes further support these substrate-specific preferences. These reveal that the flexible loop regions, particularly L1, L2, L3, and L4, adapt to accommodate and stabilize each substrate within the enzyme’s active site. The post-hydroxylation and post-epoxidation processes proceed through water-assisted proton transfer reactions in the water solvent rather than the enzyme. A key correlated motion involving the loop between the β1 and β2 regions of the catalytic domain with each substrate is observed, which might be instrumental for the flexible tuning of the GAs orientation in the enzyme active site. Our computational approach extends these experimental findings by exploring the detailed reaction mechanisms and elucidating the geometries and relative energies of all key intermediates and transition states along the catalytic pathways. Notably, the predicted structures of the reaction intermediates are in excellent agreement with the experimental data. Furthermore, our simulations accurately captured the enzyme’s differential preference for the ring chain tautomers, thereby clarifying the conformational and atomistic determinants of this selectivity. In addition, in agreement with experimental observations, our study reveals the atomistic basis for why the post-hydroxylation reactions proceed more effectively in aqueous solution than within the enzyme active site. The results reveal the unique catalytic reaction mechanism of the Fe(II)/2OG enzyme AlkB with ss-DNA containing complex GAs, which, in the long term, can be used in drug design and metalloenzyme redesign.

Supplementary Material

6. Associated Content

Supporting Information

The QM geometries of QM/MM optimized structures, spin densities, and supporting data on QM/MM and MD results are included in the Supporting Information. (PDF).

Synopsis.

This study unravels the catalytic mechanism of the Fe(II)/2OG-dependent enzyme AlkB of the repair of complex alkylated and exocyclic guanine lesions in single-stranded DNA, using MD, QM/MM, and QM methods.

Data Availability Statement

The data underlying this study are available in the published article and its Supporting Information.