N-Glycosyl Bond Formation Catalyzed by Human Alkyladenine DNA Glycosylase

The equilibria for hydrolysis of the N-glycosyl bonds of nucleotides and nucleosides favor their nucleobase and ribose products (1, 2). Indeed, de novo and salvage pathways for nucleotide biosynthesis typically use 5-phosphoribosyl-α-pyrophosphate (PRPP) as a substrate for N-glycosyl bond formation, as its utilization is driven by the subsequent hydrolysis of released PPi. Base excision repair DNA glycosylases hydrolyze the analogous N-glycosyl bond between a damaged base and C1 of a deoxyribosyl moiety, giving rise to a free base and an abasic site within DNA (3). Although the overall equilibrium for this reaction favors hydrolysis, human alkyladenine DNA glycosylase (AAG) unexpectedly synthesizes 1, N⁶-ethenodeoxyadenosine (εdA) within DNA when supplied with 1, N⁶-ethenoadenine (εA) and abasic DNA. The internal equilibrium constant (K_int) for this enzymatic transformation is near unity.

Mutagenic εdA lesions form when adenine bases in DNA react with certain lipid-peroxidation products or pollutants (4, 5). AAG is the only human glycosylase known to excise εA from εdA-containing DNA, and its mechanism has been proposed to include fast two-step DNA binding, rate-limiting N-glycosyl bond hydrolysis, and rapid dissociation of εA (6). We explored possible inhibition of the AAG reaction by its εA product by adding the free base to AAG•εdA-DNA, where εdA-DNA is a 25 bp oligonucleotide that contains a centrally located εdA (Scheme 1). The single turnover rate constant of 0.04 min⁻¹ obtained for the AAG-catalyzed hydrolysis of εdA-DNA to ab-DNA in the absence of added εA matched the value of k_hyd measured previously (Figure 1A, filled squares) (6). Addition of 0.6 mM εA to an otherwise identical reaction resulted in faster, but only partial (~70%), conversion of εdA-DNA to ab-DNA (Figure 1A, open squares), whereas a slower reaction that proceeded to completion was expected for simple product inhibition of AAG by εA.

We tested the possibility that εA reacts reversibly with ab-DNA to regenerate εdA-DNA by adding free εA to AAG•ab-DNA. Reactions containing 0.6 mM εA and initiated with either AAG•ab-DNA or AAG•εdA-DNA proceeded to the same endpoint of ~0.3 (Figure 1A, open symbols), which shows the same equilibrium between AAG-bound species being approached from opposite directions. To verify that the plateau reached in the AAG•ab-DNA reaction represented the equilibrium between AAG•ab-DNA and newly synthesized AAG•εdA-DNA, we tried to perturb it predictably. We first increased [AAG•ab-DNA] while maintaining [εA] at 0.6 mM and followed the conversion of ab-DNA to εdA-DNA for 100 min, until the expected endpoint of ~0.3 had been attained. The reaction was then diluted 25-fold, resulting in a mixture containing only 24 µM εA, and the decay of the putative εdA-DNA was followed to its new equilibrium position (Figure 1B). The rates of decay of the putative εdA-DNA and authentic εdA-DNA were the same within error (Figure 1B, inset), consistent with the model that AAG converts ab-DNA and εA into εdA-DNA via direct reversal of the glycosylase hydrolysis step (Figure 2, k_syn).

We next investigated the dependence of the synthesis reaction on the concentration of εA. As shown in Figure 3A, reactions of AAG•ab-DNA with εA proceeded with faster initial rates to higher endpoints as εA concentration increased. The reaction endpoints are plotted as a function of εA concentration in Figure 3B, and saturation with εA is observed with a K_1/2 of 0.9 mM. At saturating concentrations of εA the fractional concentration of AAG•εdA-DNA is 0.7 (the maximum endpoint), and the fractional concentration of AAG•ab-DNA•εA is correspondingly 0.3, indicating that εdA-DNA is more stable than its hydrolysis products when bound to AAG. The internal equilibrium constant for these AAG-bound species is the ratio of their fractional concentrations, K_int = 0.3/0.7 = 0.4. K_int is also equivalent to k_hyd/k_syn, the ratio of the respective rate constants for formation of AAG•ab-DNA•εA and AAG•εdA-DNA from one another, so a value of 0.1 min⁻¹ for k_syn can be calculated from K_int and the reported value of 0.04 min⁻¹ for k_hyd (Figure 1A; 6). The best global fit of the data in Figure 3A to the model in Figure 2 gives slightly different values of K_int = 0.2 and k_syn = 0.2 min⁻¹ (see Supporting Information), so we report an average value of 0.3 ± 0.1 for K_int.

To our knowledge the conversion of ab-DNA and εA into εdA-DNA by AAG is the first example of N-glycosyl bond synthesis by a DNA glycosylase. We assessed the generality of this reverse reaction by investigating several related reactions. AAG did not insert either adenine or hypoxanthine into ab-DNA, despite extended incubation times with millimolar concentrations of each free base, nor was any reaction observed when a different DNA glycosylase, 3-methyladenine DNA glycosylase II (AlkA), was incubated with ab-DNA and εA (data not shown). Thus, the ability of AAG to stably synthesize εdA-DNA appears to be unusual. A related enzyme activity that inserts purines into depurinated DNA has been reported previously (7, 8), and kinetic isotope effects for the DNA glycosylase MutY are consistent with a dissociative mechanism in which reversible C-N bond cleavage precedes irreversible nucleophilic attack by water (9).

Interestingly, the structure of AAG crystallized in the presence of an εdA-containing DNA substrate revealed an intact N-glycosyl bond (10). Although the authors proposed that the high concentration of MgCl₂ present during crystallization may have inhibited the glycosylase activity of AAG, our results raise the possibility that the substrate may have instead persisted due to the internal equilibrium that favors synthesis of εdA-containing DNA over its hydrolysis by AAG. The crystal structure shows εdA flipped into the AAG active site, securely held by aromatic stacking interactions and a hydrogen bond between the backbone amide of His-136 and N⁶ of the nucleobase. This snug fit helps to account for the observed tight binding of εdA-containing DNA by AAG (6, 11, 12) and the highly favorable equilibrium for flipping εdA into the active site (K_flip ~ 1300; 6). The unusual stability of εdA-DNA when bound to AAG, which enabled us to detect its synthesis and to measure the internal equilibrium constant for its hydrolysis, presumably arises from its tight binding to the active site coupled with slow subsequent hydrolysis of its N-glycosyl bond. In contrast, human uracil DNA glycosylase (UDG) has been crystallized bound to its uracil and abasic DNA products (13, 14), consistent with an internal equilibrium for UDG that favors hydrolysis of uracil-containing DNA (15).

As described above, results from reactions at saturating εA concentrations allowed determination of a value of 0.3 for K_int, the equilibrium constant for hydrolysis of the N-glycosyl bond of AAG-bound εdA-DNA. K_int values near unity are considered to be optimal for enzymes that operate reversibly (16). While it stands to reason that certain metabolic enzymes that act at the N-glycosyl bond may benefit from reversibility (e.g., nucleosidases, nucleoside phosphorylases), synthesis of εdA-containing DNA by AAG is expected to be negligible under typical conditions, due to low cellular concentrations of abasic DNA and εA. In principle, however, ready reversibility provides a potential mechanism for direct repair by DNA glycosylases, in which excision and dissociation of an aberrant base from DNA are followed by binding and stable incorporation of the appropriate base. Such a repair mechanism, if present in nature, would be reminiscent of the overall base exchange processes catalyzed by tRNA guanine transglycosylases, nucleotide deoxyribosyl-transferases, and nucleoside ribosyltransferases (17, 18).