DNA-catalyzed sequence-specific hydrolysis of DNA.

Deoxyribozymes (DNA catalysts) have been reported for cleavage of RNA phosphodiester linkages, but cleaving peptide or DNA phosphodiester linkages is much more challenging. Using in vitro selection, here we identified deoxyribozymes that sequence-specifically hydrolyze DNA with multiple turnover and with a rate enhancement of 108 (possibly as high as 1014). The new DNA catalysts require both Mn2+ and Zn2+, which is noteworthy because many natural DNA nucleases are bimetallic protein enzymes.

The first deoxyribozyme (DNA enzyme) was identified by in vitro selection in 1994 and cleaves an RNA phosphodiester bond1. Since that time, DNA has been shown to catalyze many reactions, including several preparatively useful transformations2. RNA cleavage is the most studied DNA-catalyzed reaction3, and deoxyribozymes with practical cleavage abilities for nearly all RNA dinucleotide target sequences have been developed4,5. However, DNA-catalyzed hydrolysis of other biologically relevant bonds, such as amide (peptide) or DNA phosphodiester linkages, has not been reported. Non-site-selective Cu2+-dependent oxidative cleavage of DNA has been described6. These reactions appear to proceed through a diffusible intermediate, resulting in nonhydrolytic and nonspecific DNA cleavage. Self-hydrolysis of G-quadruplex DNA has been reported, but the sequence scope appears to be limited7. Amide and especially DNA phosphodiester linkages are intrinsically more difficult to hydrolyze than RNA, by about a factor of 20 for amides8 and at least 400 and possibly much more for DNA. Estimates of the uncatalyzed half-life of DNA at near-physiological conditions range from ~4000 years9 to 30 million years10 to over 10 billion years11, which makes catalysis of DNA hydrolysis an especially challenging goal12–14.

In this study, we initially sought to use in vitro selection to identify deoxyribozymes that cleave amide bonds15 by presenting a tripeptide sequence to a random N40 DNA region in the structural arrangement of Fig. 1a. The desired reaction site was held across from the initially random DNA enzyme region by Watson-Crick interactions between fixed DNA sequences of the substrate and complementary binding arm sequences of the deoxyribozyme strand. This arrangement has been used to identify deoxyribozymes with a variety of catalytic activities including RNA cleavage2, and here it allows us to focus on the fundamental ability of DNA to catalyze cleavage of our substrate.

Figure 1

In vitro selection of deoxyribozymes that cleave a substrate with both amide and DNA phosphodiester linkages. (a) Key step of the selection strategy. The loop on the right enables PAGE separation during selection of the DNA sequences that cleave the substrate, but this loop is dispensable for catalysis. (b) Chemical composition of the substrate between the two unpaired T DNA nucleotides at the ends of the binding arms. (c) Observed cleavage sites within the substrate for nine unique new deoxyribozymes. All four sites correspond to hydrolysis at DNA phosphodiester linkages. 10MD30, 36, and 41 cleave at the same site as 10MD9; 10MD4 and 13 cleave at the same site as 10MD1. Sequences of all nine deoxyribozymes are provided in Supplementary Table 1.

The selection strategy was derived from two precedents: (1) many successful selections for deoxyribozymes that cleave RNA linkages3; and (2) our own extensive efforts to identify deoxyribo-zymes that ligate two RNA substrates.2,16 During the key bond-cleavage step of each selection round, the incubation conditions included 40 mM MgCl2, 20 mM MnCl2, and 1 mM ZnCl2 in 70 mM Tris, pH 7.5, and 150 mM NaCl at 37 °C. Three divalent metal ions were included to enable emergence of a potentially wide range of catalytic mechanisms. The substrate (Fig. 1b) incorporated an Ala-Ser-Ala tripeptide cleavage target between two DNA sequences designed to interact with fixed DNA elements (“binding arms”) that flank the N40 region. The C-terminal Ala of the tripeptide was connected by an amide bond to 5′-NH2-T of DNA, and the N-terminal Ala was joined to DNA via a short γ-hydroxybutyric acid (GHB) linker. This substrate offers for cleavage four amide bonds as well as numerous DNA phosphodiester bonds. At the outset, our expectation was that peptide bond cleavage should be favored over DNA hydrolysis, due to the intrinsic reactivity difference noted above.

After nine selection rounds with the substrate of Fig. 1b, the DNA pool achieved 35% substrate cleavage during the 14 h incubation period (Supplementary Methods). A tenth round with 2 h incubation led to 16% substrate cleavage, which did not increase after additional rounds. Individual round 10 deoxyribozymes were screened for activity, cloned, and sequenced essentially as described previously16,17. Of 44 clones, a total of four different cleavage sites in the substrate were used by nine unique deoxyribozymes (Fig. 1c). Surprisingly, all four cleavage sites corresponded to hydrolysis at specific DNA phosphodiester linkages, whereas no substrate amide bond was hydrolyzed by any deoxyribozyme. The precise location of each DNA cleavage site and the integrity of the hydrolyzed DNA termini were assigned and verified by MALDI mass spectrometry of PAGE-purified reaction products (Supplementary Fig. 2 and Supplementary Table 2), comparison with PAGE standards (Supplementary Fig. 3), and biochemical ligation assays using the reaction products and T4 DNA ligase (Supplementary Fig. 4).

Four new deoxyribozymes, named 10MD9, 10MD1, 10MD14, and 10MD5 (one for each observed cleavage site), were characterized further. All four deoxyribozymes require both Mn2+ and Zn2+ (Supplementary Fig. 5). Mg2+ is either dispensable (10MD9 and 10MD5), supportive of two-fold higher yield than with Mn2+ and Zn2+ alone (10MD1), or several-fold inhibitory (10MD14). While 10MD9, 10MD1, and 10MD14 function better in Tris than in HEPES buffer, 10MD5 has higher activity in HEPES and is inhibited by high concentrations of Tris (Supplementary Fig. 6). Standard incubation conditions of 70 mM Tris or HEPES (as appropriate), pH 7.5, 20 mM MnCl2, 1 mM ZnCl2, and 150 mM NaCl at 37 °C were used in further assays.

10MD9 and 10MD1 strictly require the presence of the tripeptide region in the substrate (Fig. 2a). 10MD14 retains some activity with an all-DNA substrate but is ~40-fold faster when the tripeptide region is included (Fig. 2a). The role of the tripeptide region in catalysis has not yet been examined. All three deoxyribozymes cleave a DNA phosphodiester linkage very close to the tripeptide region (additionally, 10MD9 appears to catalyze a secondary reaction of its product resulting in loss of the GHB moiety; Supplementary Fig. 7). In contrast, 10MD5 hydrolyzes a DNA linkage four nucleotides removed from the tripeptide region (Fig. 1c). 10MD5 cleaves an all-DNA substrate as efficiently as the original tripeptide-containing substrate (kobs 2.7 ± 0.3 h−1 and yield 66 ± 4% [n = 11]; Fig. 2a), with formation of 5′-phosphate and 3′-hydroxyl DNA termini as established above.

Figure 2

Hydrolysis of DNA phosphodiester bonds by the new deoxyribozymes. (a) Single-turnover data under standard incubation conditions (including Mn2+ and Zn2+) for each of 10MD9 (■), 10MD1 (▲), 10MD14 (▼;), and 10MD5 (●), with either the original tripeptide-containing substrate (open symbols) or the all-DNA analogue in which the tripeptide portion is replaced by deoxyadenosine (filled symbols). kobs values are in Supplementary Table 3. The PAGE image shows cleavage of the all-DNA substrate by 10MD5. (b) Summary of DNA substrate sequence requirements for 10MD5, which efficiently cleaves an all-DNA substrate with a modest recognition site.

We evaluated the 10MD5 deoxyribozyme for its tolerance of DNA substrate sequences (Fig. 2b; data in Supplementary Fig. 8). Outside of the ATG trinucleotide segment immediately 5′ of the hydrolysis site as well as the T immediately to the 3′-side, 10MD5 tolerates all substrate nucleotides, as long as Watson-Crick base-pairing is maintained. The detailed sequence requirements within this ATG^T “recognition site” are currently under investigation.

As a divalent metal ion cofactor, none of Cu2+, Fe2+, Co2+, or Ni2+ affects 10MD5 cleavage activity when included at either 0.1 μM or 1 μM along with Mn2+ and Zn2+. None of Cu2+, Fe2+, Co2+, or Ni2+ at 0.1 μM through 10 mM can substitute for either Mn2+ or Zn2+, whereas 1 mM Cd2+ can substitute for Mn2+ with ~105 lower DNA cleavage rate. In addition, 40 mM Ca2+ can substitute for Mn2+ with ~40-fold lower rate (Supplementary Fig. 9). When 10MD5 was tested with Mn2+ and Zn2+, the Kd,app was 5 ± 2 mM for Mn2+ (at 1 mM Zn2+); the optimal Zn2+ concentration was ~1 mM (at 20 mM Mn2+; Supplementary Fig. 10), with little activity below 0.5 mM Zn2+. The importance of monovalent ions was also examined (Supplementary Fig. 11). When Na+ was decreased from 150 mM to either 10 mM or 0 mM, activity was slightly suppressed, whereas increasing Na+ to 450 mM or replacing 150 mM Na+ with 150 mM K+ had little effect. Finally, the pH dependence of 10MD5 was examined (Supplementary Fig. 12). The deoxyribozyme has a rather sharp pH optimum near 7.5 (as measured for the 1 M buffer stock solution), with substantially reduced activity when the pH was increased or decreased by merely 0.1–0.2 pH units. Changing the sulfonic acid buffer from HEPES to MOPS led to comparable activity. Both the rather sharp Zn2+ optimum and the strong pH dependence of 10MD5 are similar to our observations for a different Zn2+-dependent deoxyribozyme that catalyzes an unrelated RNA ligation reaction18. Several other deoxyribozymes and ribozymes are known that either require or can Zn2+ as a cofactor19.

To validate that 10MD5 catalyzes hydrolytic rather than oxidative DNA cleavage, we performed additional experiments. When the 10MD5-catalyzed cleavage reaction was performed in 18O-water (18OH2), we observed by mass spectrometry that 18O was incorporated into the 5′-phosphate group (Supplementary Fig. 13). This observation is consistent with a hydrolysis reaction in which water is the source of the new oxygen atom but inconsistent with oxidative cleavage, in which O2 from the air would be the source of the oxygen atom. Of course, much work remains to investigate the mechanism(s) of DNA-catalyzed DNA hydrolysis, including dissection of structural and catalytic roles of Mn2+ and Zn2+. With unrelated inorganic catalysts, “formal hydrolysis” of DNA can result from an oxidative pathway.20 We also evaluated the effect of including Mn3+, H2O2, or (NH4)2S2O8 (ammonium persulfate) as potential oxidants, either in addition to Mn2+ and Zn2+ or in place of one of the two metal ions. In all cases, no effect of the added oxidant was observed (Supplementary Figure 14). Finally, we included either or both of catalase or superoxide dismutase; suppression of DNA cleavage was not observed (Supplementary Figure 14). Taken together, these experiments validate the conclusion that 10MD5 catalyzes DNA phosphodiester cleavage via a hydrolytic rather than an oxidative mechanism. In sharp contrast, the Cu2+-dependent deoxyribozymes reported by Breaker and coworkers involve diffusible intermediates, resulting in relatively nonselective DNA cleavage by “oxidative destruction” of a nucleotide6.

The new DNA-hydrolyzing deoxyribozymes are capable of multiple turnover. When the “binding arms” of 10MD5 were shortened, which is anticipated to promote product release4, at least 40 turnovers were observed (Supplementary Fig. 15). Similar results were found for the other three deoxyribozymes. When the 10MD5 deoxyribozyme was isolated after a multiple-turnover reaction and treated with piperidine to induce strand cleavage at any damage sites, strand-cleavage events were not observed (Supplementary Fig. 16), further validating nonparticipation of oxidizing species in 10MD5-catalyzed DNA hydrolysis.

In summary, we have identified numerous deoxyribozymes that require two different divalent metal ions, Mn2+ and Zn2+, to catalyze sequence-specific cleavage of DNA phosphodiester linkages. A requirement by a deoxyribozyme or ribozyme for two different metal ions simultaneously is rare but precedented19. The DNA-catalyzed heterobimetallic DNA hydrolysis activity was identified here despite the presence in the selection substrate of four nearby amide bonds, which have a much higher uncatalyzed hydrolysis rate. The reason that the peptide bonds in the selection substrate were not cleaved is currently unknown. One focus of our ongoing efforts is to identify deoxyribozymes that function with peptide substrates21. These efforts include revised selection strategies specifically seeking DNA-catalyzed amide bond cleavage.

The rate enhancement of the 10MD5 deoxyribozyme, which sequence-specifically cleaves an all-DNA substrate with multiple turnover, is at least 108 and possibly as high as 1014 over the uncatalyzed DNA hydrolysis reaction, as calculated from the corresponding half-lives. Mechanistic analyses of the new deoxyribozymes are warranted, especially regarding the roles of Mn2+ and Zn2+ because many natural enzymes are DNA or RNA nucleases that require multiple metal ion cofactors22,23. Considerable attention has been devoted to small model systems for DNA nuclease activity13,14. The identification of heterobimetallic DNA catalysts that sequence-specifically hydrolyze DNA phosphodiester linkages suggests that deoxyribozymes have significant potential as practical sequence-specific DNA cleavage reagents24,25.