[Univ. of Illinois]  [School of Chemical Sciences]  [Dept. of Chemistry]  [Organic Chemistry]  [Chemical Biology]

[Journals List]  [Group Meetings]  [Group Pictures]  [CV & Contact Info]

[SKS Lab Home]  [SKS Lab People]  [SKS Lab Publications]

DNA as a catalyst for bioorganic and biochemical reactions

This webpage provides a brief introduction to deoxyribozymes and describes our research efforts in the area. Many of our published papers (see Full publications listing) tell the full story of our ongoing work with deoxyribozymes. Here we provide an overview of the research along with key examples that highlight the chemical questions, challenges, and opportunities that arise when using DNA as a catalyst.

We have published several review articles in the research area of deoxyribozymes:
• S. K. Silverman, "Deoxyribozymes: DNA Catalysts for Bioorganic Chemistry", Org. Biomol. Chem. 2004, 2, 2701-2706 [#30]
• S. K. Silverman, "In Vitro Selection, Characterization, and Application of Deoxyribozymes that Cleave RNA", Nucleic Acids Res. 2005, 33, 6151-6153 [#43]
• S. K. Silverman, "In Vitro Selection and Application of Nucleic Acid Enzymes (Ribozymes and Deoxyribozymes)", Wiley Encyclopedia of Chemical Biology 2007, in press [#52]
• S. K. Silverman, "Artificial Functional Nucleic Acids: Aptamers, Ribozymes, and Deoxyribozymes Identified by In Vitro Selection", in Functional Nucleic Acids for Sensing and Other Analytical Applications, edited by Y. Lu and Y. Li; Springer (New York, NY), 2007 [#55]
• C. Höbartner and S. K. Silverman, "Recent Advances in DNA Catalysis", Biopolymers 2007, 87, 279-292 [#61]
• D. A. Baum and S. K. Silverman, "Deoxyribozymes: Useful DNA Catalysts In Vitro and In Vivo", Cell. Mol. Life Sci. 2008, 65, 2156-2174 [#65]
• S. K. Silverman, "Catalytic DNA for Synthetic Applications—Current Abilities and Future Prospects", Chem. Commun. 2008, 3467-3485 [#67]

Map of this webpage (click on any link to go to that section of the webpage):
• Introduction to deoxyribozymes
• In vitro selection of deoxyribozymes
• Overview of DNA-catalyzed RNA ligation reactions
• Deoxyribozyme-catalyzed RNA ligation by nucleophilic attack at a 2',3'-cyclic phosphate: Questions of regioselectivity and site-selectivity
• Deoxyribozyme-catalyzed RNA ligation by nucleophilic attack at a 5'-triphosphate: Formation of branched and linear RNA
• Branched RNA synthesized by deoxyribozymes: Implications for natural RNA splicing
• Lariat RNA synthesis by deoxyribozymes: Exquisite site-selectivities for reactions that are impossible to achieve by traditional organic synthesis approaches
• Using the RNA ligation products of deoxyribozymes to test specific biochemical hypotheses
• DECAL: Deoxyribozyme-catalyzed labeling of RNA
• Branched RNA-DNA conjugates applied to studies of DNA as a constraint
• Engineering a three-helix-junction deoxyribozyme to function with a small-molecule substrate
• Beyond RNA ligation: Deoxyribozymes that catalyze a variety of reactions
• Elucidating deoxyribozyme structures and reaction mechanisms
• Miscellaneous issues related to deoxyribozymes

Introduction to deoxyribozymes

Catalytic RNA molecules play important roles in fundamental biochemistry, including protein synthesis by the ribosome and RNA splicing by the spliceosome. As a means to explore the catalytic power of nucleic acids and as a practical experimental approach to catalyze desirable bioorganic reactions, we investigate deoxyribozymes, which are also called DNA enzymes, DNAzymes, or catalytic DNA. Deoxyribozymes are homogeneous DNA molecules—with particular, well-defined sequences and molecular weights—that can catalyze specific chemical reactions. In this regard they are directly analogous to protein enzymes and natural RNA enzymes (ribozymes). No naturally occurring DNA enzymes have been discovered, and it is an interesting but unanswered question if nature has ever used DNA for catalysis. Regardless of the existence or nonexistence of natural deoxyribozymes, since 1994 several research groups have reported artificial deoxyribozymes with various catalytic activities. The field of deoxyribozymes is an exciting new scientific area for at least three reasons: (1) Structural and mechanistic studies of deoxyribozymes help us to understand how natural ribozymes function. (2) DNA has been shown to be a useful practical catalyst for challenging bioorganic transformations that cannot be performed using more traditional synthetic approaches. (3) Deoxyribozymes represent a new form of catalysis, and chemists rarely encounter such an intellectual opportunity.

Nobody yet knows how to design rationally a deoxyribozyme; i.e., how to predict a DNA sequence that is able to catalyze a particular desired reaction. The same is largely true for ribozymes and protein enzymes, although the first studies along these lines are just appearing. In the absence of natural examples or artificial design principles, the technique of “in vitro selection” is a powerful approach to identify new catalysts with desired activities from random-sequence pools of candidate molecules. In our laboratory, we use in vitro selection to identify DNA enzymes with desired catalytic activities. We study the resulting deoxyribozymes to understand how they perform catalysis, and we apply the deoxyribozymes as practical reagents for bioorganic chemistry and biochemistry.

Click here to return to the top of this webpage

In vitro selection of deoxyribozymes

Largely for practical reasons, most of the deoxyribozymes that we have identified ligate (join) two RNA substrates. The use of nucleic acid substrates allows simple separation of the binding and catalytic functions of the DNA. The binding energy is provided by Watson-Crick "binding arms", which flank the "enzyme region" that performs the catalysis. Ideally, the sequences of the RNA substrates are completely variable, such that the deoxyribozyme retains catalytic activity as long as the DNA binding arm sequences are changed to maintain Watson-Crick complementarity with the RNA. In practice, such complete generality for the RNA substrates is often not achieved by any one deoxyribozyme. Fortunately, in many cases the requirements imposed on the RNA substrate sequences are not very extensive, such that the deoxyribozymes have a significant degree of general applicability to prepare useful and interesting RNA products. Furthermore, it is often reasonable to perform several in vitro selection experiments in parallel, each with a different set of RNA substrate sequences. The resulting deoxyribozymes can collectively be used with nearly any particular combination of RNA substrates.

Shown below is the selection scheme that we use to identify deoxyribozymes that ligate RNA. In step A of this three-step procedure, the right-hand RNA substrate (blue) is attached to the deoxyribozyme strand (brown), which incorporates an N₄₀ random region (green; 40 random DNA nucleotides in a row). In step B, the left-hand RNA substrate (L) is incubated with the pool of deoxyribozymes along with any chosen metal ions and at any desired pH and temperature (one specific set of conditions is shown in the scheme below). Only those DNA sequences that are competent for RNA ligation under these conditions during the allotted reaction time are able to join L and R. Because R is already covalently attached to the DNA, any catalytically active DNA sequence grows larger by the size of L, and this size difference allows straightforward separation of the "winning" DNA sequences by polyacrylamide gel electrophoresis (PAGE). Finally, in step C, the active DNA sequences are amplified by the polymerase chain reaction (PCR), and the desired single-stranded DNA product (separated by PAGE) is carried forward to another round of selection. After a suitable number of selection rounds (typically 5-12 rounds), catalytically active DNA sequences have come to dominate the pool, and individual DNA sequences are cloned, sequenced, and prepared independently by solid-phase DNA synthesis. Their ligation activities are then verified using the two RNA substrates, and their structures and mechanisms are studied in greater detail. It is important to note that during practical use of these deoxyribozymes to ligate RNA, the DNA strand is entirely separate from the two RNA substrates. The resulting trimolecular system formally satisfies a reasonable definition of "catalysis" and also allows the deoxyribozymes to be applied as useful laboratory reagents for RNA ligation.

An iterated selection procedure (i.e., with multiple selection rounds) is always required to identify highly active DNA sequences. This is because most of the astronomically large number of possible 40-mer DNA sequences are not catalytically active. (There are 4⁴⁰ = 10²⁴ possible DNA sequences of length 40, and about 10¹⁴ of these can be used in practice to initiate a selection experiment.) From any given sample of the initially random DNA pool, some DNA sequences may accidentally survive a specific round of selection not because they are truly catalytically active but because a small number of such molecules can cross the ligation energy barrier merely due to statistical fluctuations, and PCR is a highly sensitive procedure that amplifies these accidental cases. Only by requiring that certain DNA sequences reproducibly achieve catalysis in multiple selection rounds can truly active deoxyribozymes be identified.

The length of the random pool in most of our experiments is 40; i.e., we use an N₄₀ random region. The value of 40 is chosen as a compromise between the need for a larger random region, which allows more complex catalytic activities, and the need for a smaller random region, which allows exploration of a sufficient fraction of DNA "sequence space" in one experiment. With an N₄₀ region we examine only about 10¹⁴/10²⁴= 10^–10 [one ten-billionth of one percent!] of the possible DNA sequences of length 40 in any particular selection experiment. Nevertheless, our results demonstrate that interesting and useful deoxyribozymes can readily be identified from these N₄₀ pools. Other investigators working with catalytic nucleic acids have used random regions over 200 nucleotides in length, but in those cases the fraction of sequence space being examined is truly miniscule. For example, with a 200-nucleotide random region, there are 4²⁰⁰ = 10¹²⁰ possible sequences, for which only 10¹⁴/10¹²⁰ = 10^–106 (!!) of sequence space is explored. In some of our more recent experiments that seek catalytic activities other than RNA ligation, we are examining N₇₀ random regions, in which we examine only about 10¹⁴/10⁴² = 10^–28 of sequence space.

For finding DNA sequences that catalyze chemical reactions other than RNA ligation, we adapt the selection scheme described above with appropriate technical modifications. For example, we have identified deoxyribozymes that catalyze the Diels-Alder reaction between two small-molecule substrates using a related selection scheme (see Chandra & Silverman, J. Am. Chem. Soc. 2008, 130, 2936-2937 [#64]).

Click here to return to the top of this webpage

Overview of DNA-catalyzed RNA ligation reactions

As we began our studies of deoxyribozymes several years ago, we had in mind two parallel goals: (1) Learn the fundamental rules for catalysis by nucleic acids; and (2) obtain useful DNA reagents for synthetic reactions. To satisfy both of these goals, we chose functional group combinations for the two RNA substrates that reflect naturally occuring RNA functional groups that are also readily obtained by most biochemists, including those who may not wish to perform multistep organic synthesis. Because of this choice, we hope that our deoxyribozymes will be useful in practice to a broad range of researchers. An RNA ligation reaction necessarily requires a nucleophile and an electrophile (assuming that free radical reactions will not be used). In any particular RNA ligation reaction, the nucleophile will almost certainly be a ribose hydroxyl group at either a 5'-, 3'-, or 2'-position; the experiments described below are consistent with this expectation. Therefore, the key choice in designing a deoxyribozyme selection is to decide what RNA electrophile will be used. There are two reasonable choices that are consistent with both of our overall goals: a 2',3'-cyclic phosphate and a 5'-triphosphate. As described here, we have performed many in vitro selection experiments using RNA substrates with each of these electrophiles.

Click here to return to the top of this webpage

Deoxyribozyme-catalyzed RNA ligation by nucleophilic attack at a 2',3'-cyclic phosphate: Questions of regioselectivity and site-selectivity

A 2',3'-cyclic phosphate group is susceptible to attack by a nucleophile with opening of the cyclic phosphate ring. Our initial expectation was that the 5'-hydroxyl group of the right-hand RNA substrate would act as the nucleophile during ligation. If the 2'-oxygen of the cyclic phosphate serves as the leaving group with a 5'-hydroxyl nucleophile, then the product is a 3'–5' phosphodiester linkage, which is the "native" RNA linkage found in nature. On the other hand, if the 3'-oxygen serves as the leaving group, then the product is the "non-native" 2'–5' linkage, which is also found in nature (but only rarely and in very specific contexts). The regioselectivity in this ring-opening reaction was one of the major questions that would be answered experimentally by identifying and studying new deoxyribozymes.

In several different selection efforts, all of the resulting deoxyribozymes were found to create non-native 2'–5' linkages (see Flynn-Charlebois et al., J. Am. Chem. Soc. 2003, 125, 2444-2454 [#19], Flynn-Charlebois et al., J. Am. Chem. Soc. 2003, 125, 5346-5350 [#21], and Semlow & Silverman, J. Mol. Evol. 2005, 61, 207-215 [#33]). Currently we do not understand in a mechanistic sense why the DNA-catalyzed opening of a 2',3'-cyclic phosphate favors formation of 2'–5' linkages over 3'–5' linkages. For practical purposes (e.g., to prepare RNAs for structure-function studies), synthesis of non-native 2'–5' linkages is undesired, because their introduction may have unintended side effects. In contrast, the native 3'–5' linkage is sought because its formation leaves no trace of evidence that a ligation event ever occurred; it is a "traceless" ligation. Therefore, we are particularly interested to investigate why 2'–5' linkages are formed in these ligation reactions and to find ways to create 3'–5' linkages instead.

Our initial selection efforts as described above all used Mg²⁺ as the divalent metal ion cofactor, in part because natural ribozymes typically require Mg²⁺. However, with in vitro selection we are not restricted to using Mg²⁺, and we have investigated several other metal ions as well. Transition metal ions like Zn²⁺ have different ligand preferences, pK_a values for bound water molecules, and other chemical properties when compared with Mg²⁺. Therefore, we anticipated that Zn²⁺ could mediate different chemical reactions when used as the metal ion cofactor with DNA enzymes. When similar selections were performed using Zn²⁺ instead of Mg²⁺, indeed many RNA ligation products were observed other than the 2'–5' linkages formed with Mg²⁺ (see Hoadley et al., Biochemistry 2005, 44, 9217-9231 [#36]). The 2',3'-cyclic phosphate of the left-hand RNA substrate still served as the electrophilic reaction partner, but both types of ring-opening reaction were observed (i.e., either the 2'-oxygen or the 3'-oxygen acted as the leaving group). As the nucleophilic reaction partner, several different nucleophiles of the right-hand RNA substrate were used: the 5'-hydroxyl; the 2'-hydroxyl of the 5'-terminal nucleotide; and the 2'-hydroxyl of the second nucleotide counting from the 5'-terminus. In all cases, each deoxyribozyme created just one specific ligation product. The wide variety of RNA linkages formed by Zn²⁺-dependent deoxyribozymes highlights the versatility of transition metals like Zn²⁺ to mediate nucleic acid catalysis. This variety of products also demonstrates conclusively that deoxyribozymes can change reaction pathways, rather than merely increasing the rates of reactions that would have occurred in the absence of deoxyribozymes.

One particularly interesting observation from these Zn²⁺ selections was that certain Zn²⁺-dependent deoxyribozymes are able to create native 3'–5' RNA linkages using a 2',3'-cyclic phosphate RNA substrate, even though none of our Mg²⁺-dependent deoxyribozymes could accomplish this feat. In ongoing work, we are continuing to explore the possibility of using Zn²⁺-dependent deoxyribozymes as practical catalysts for 3'–5' RNA ligation using cyclic phosphate substrates. Note that each of these Zn²⁺-dependent deoxyribozymes creates just one type of RNA linkage, although the collection of deoxyribozymes can synthesize many linkages. These findings underscore that "DNA catalysis" is quite distinct from "DNA templating" as practiced by others, which would not be able to enforce such strict site-selectivity and regioselectivity in the reactions of the RNA substrates. See the section at the bottom of this page for more comments on the relationship between DNA catalysis and DNA templating.

Click here to return to the top of this webpage

Deoxyribozyme-catalyzed RNA ligation by nucleophilic attack at a 5'-triphosphate: Formation of branched and linear RNA

When the electrophilic reaction partner is a 5'-triphosphate on the right-hand RNA substrate rather than a 2',3'-cyclic phosphate on the left-hand RNA substrate, a new set of chemical questions arise. The first question is to identify the nucleophile on the left-hand RNA substrate. Three types of nucleophile are possible: the 3'-hydroxyl group on the 3'-terminal nucleotide; the 2'-hydroxyl group on the 3'-terminal nucleotide; and any of the internal 2'-hydroxyl groups within the substrate strand. The first two nucleophiles lead to linear RNA (non-native 2'–5' and native 3'–5' linkages, respectively), and the third type of nucleophile leads to 2',5'-branched RNA.

Our first selection effort with a 5'-triphosphate RNA substrate identified deoxyribozymes that create 2',5'-branched RNA using an internal 2'-hydroxyl group on the left-hand substrate as the nucleophile (see Wang & Silverman, J. Am. Chem. Soc. 2003, 125, 6880-6881 [#22]). The branched RNA product is biochemically interesting because 2',5'-branched RNA is an intermediate in the important process of RNA splicing, during which the non-coding "introns" are removed from messenger RNA and other RNA molecules, leaving just the exons (see more about this below). Our original branch-forming deoxyribozymes have impressive catalytic abilities for nucleic acid enzymes. For example, some of these deoxyribozymes ligate RNA with >90% yield and k_obs = 2 min^–1, which rivals the rate constants for many natural ribozymes. Despite such favorable features, these particular RNA ligase deoxyribozymes have fairly restrictive sequence requirements for the RNAs that may be used as substrates (see Wang & Silverman, Biochemistry 2003, 42, 15252-15263 [#24]), so they are not very practical branch-forming reagents. Nevertheless, branched RNA is quite challenging to synthesize using conventional organic chemistry approaches, and this first report established that deoxyribozymes can provide synthetic access to interesting products—here, branched RNAs—that are not readily prepared by other means.

Although branched RNA is a biochemically intriguing ligation product, an important goal in pursuing RNA ligation is to synthesize linear RNA linkages, which will enable synthesis of site-specifically modified RNAs for structure-function studies. Towards this goal, we re-evaluated our selection strategy to determine how to obtain linear rather than branched RNA products. One revised approach was to locate the incipient ligation junction within an RNA:DNA duplex region, such that branched RNA formation would be disfavored geometrically (the RNA termini to be joined would be held closely together, favoring linear ligation). An important benefit of this approach was anticipated to be formation of 3'–5' and not 2'–5' linkages. This was expected because 3'–5' linkages are known to be more stable than 2'–5' in an RNA:DNA duplex; the 2'–5' linkage changes the local helical geometry and is more susceptible to intramolecular transesterification at phosphorus. To the extent that the relative stability of 3'–5' linkages in the ground state is represented in the transition state for ligation, formation of 3'–5' linkages should occur with a higher rate constant than for 2'–5' linkages. When this approach was tested experimentally, indeed native 3'–5' linear RNA linkages were synthesized by the new deoxyribozymes (see Coppins & Silverman, J. Am. Chem. Soc. 2004, 126, 16426-16432 [#31]). However, once again there were fairly restrictive sequence requirements imposed upon the RNA substrates, such that the new deoxyribozymes have limited practical utility. In retrospect, these sequence requirements for the RNA substrates are not surprising, because the characteristic functional groups of the RNA nucleobases are exposed in the minor groove of an A-form RNA:DNA duplex. It is reasonable to hypothesize that some of the DNA enzyme nucleotides interact with the RNA substrates at these locations, thus enforcing the observed (and unwanted) RNA substrate sequence requirements.

From these results, it became clear that we needed a more general approach to obtain 3'–5' RNA ligase deoxyribozymes. During the selection strategy, it is difficult to impose control over whether the 3'-hydroxyl group (desired) or 2'-hydroxyl group (undesired) of the L substrate attacks the 5'-triphosphate of the R substrate. The experiments described above demonstrate that this control can be achieved geometrically by arranging the RNA and DNA appropriately in a duplex, albeit at the sacrifice of substrate generality for the resulting deoxyribozymes. An alternative approach to 3'–5' linkages is addition of a step to the selection strategy that stringently requires only 3'–5' RNA ligation to permit DNA molecules to pass through to the next selection round, regardless of the geometrical relationship between the RNA and DNA. This was accomplished by adding a suitable step between step B and step C of the selection scheme that is shown at the top of this page. In the new step, the previously reported 8–17 DNA enzyme was used to cleave at the RNA ligation junction that was just created by the deoxyribozyme in step B. Because 8–17 selectively cleaves 3'–5' RNA linkages, only if the deoxyribozyme just created a 3'–5' linkage will 8–17 then cleave that linkage, thus providing stringent 3'–5' selection pressure. When this strategy was applied, the resulting deoxyribozymes almost universally create 3'–5' RNA linkages (see Wang & Silverman, Biochemistry 2005, 44, 3017-3023 [#32]).

We have combined the selection approaches described above to identify 3'–5' ligase deoxyribozymes that satisfy all of our key design criteria: they synthesize native 3'–5' RNA linkages rapidly, in high yield, and with useful generality for their RNA substrates (see Purtha, Coppins, Smalley & Silverman, J. Am. Chem. Soc. 2005, 127, 13124-13125 [#41]). For example, the Mg²⁺-dependent 9DB1 deoxyribozyme and the Zn²⁺-dependent 7DE5 deoxyribozyme have sequence requirements of merely D¯RA and A¯R, respectively (D = A, G, or U; R = A or G; the arrowhead denotes the RNA ligation site between the L and R substrates). These sequence requirements for RNA ligation are comparable to those of useful RNA-cleaving deoxyribozymes like 10–23 (R¯Y) and 8–17 (A¯G), which have been used in practice for several years as biochemical reagents in many laboratories. We are currently applying 9DB1, 7DE5, and other deoxyribozymes for RNA ligation in the context of projects that aim to understand RNA structure-function relationships. In ongoing efforts, we are also generalizing our selection approach to obtain a larger family of RNA ligase deoxyribozymes that will collectively be able to achieve RNA ligation for almost any combination of RNA substrates, regardless of their sequences.

Click here to return to the top of this webpage

Branched RNA synthesized by deoxyribozymes: Implications for natural RNA splicing

As described above, our first experiment with RNA ligation using a 5'-triphosphate electrophile led to biochemically relevant branched RNA. Subsequently we again obtained branched RNA, but in a rather surprising fashion (see Coppins & Silverman, Nature Struct. Mol. Biol. 2004, 11, 270-274 [#26]). A selection experiment was devised that we originally thought would favor formation of linear RNA, because the nucleotides that provided the 2'-hydroxyl nucleophiles in the first experiment were constrained by RNA:DNA base pairing. However, the resulting deoxyribozymes again created branched RNA. As exemplified by the 7S11 deoxyribozyme, DNA nucleotides within the originally random N₄₀ region were found to be base-paired with RNA nucleotides of the left-hand substrate, such that a single RNA adenosine nucleotide remained unpaired (bulged). The 2'-hydroxyl group of this bulged adenosine served as the nucleophile to attack the 5'-triphosphate, forming branched RNA. The 7S11 deoxyribozyme is particularly interesting because the first step of natural RNA splicing pathways also uses a branch-site adenosine nucleotide that is flanked by Watson-Crick duplex regions (e.g., group II introns and the spliceosome). This close structural similarity between artificial branch formation and natural RNA splicing leads to the hypothesis that natural branch formation may inherently be mechanistically favored. Additional experiments are required to address this hypothesis.

Our interest in the 7S11 deoxyribozyme grew with two additional discoveries about its applicability and structure (see Coppins & Silverman, J. Am. Chem. Soc. 2005, 127, 2900-2907 [#34]). First, by testing a comprehensive set of RNA substrates, we showed that 7S11 is highly general for forming 2',5'-branched RNAs of wide sequence composition. Strict sequence requirements for the RNA substrates are found only at the two nucleotides directly involved in the ligation reaction, and other branch-forming deoxyribozymes that we have independently identified have different requirements (see Pratico, Wang & Silverman, Nucleic Acids Res. 2005, 33, 3503-3512 [#37]). Second, we demonstrated that 7S11 adopts a three-helix-junction structure, in which the ligation reaction occurs in a complex that has three RNA:DNA helical arms extending outward from a common junction. Because many natural ribozymes adopt multi-helix-junction structures with two to four helices, 7S11 is an interesting model system to understand multi-helix junctions. We are currently pursuing both X-ray crystallography and NMR spectroscopy of the 7S11 deoxyribozyme.

An intriguing facet of the 7S11 deoxyribozyme was revealed by examining its leaving group dependence. When a 5'-triphosphate reacts as an electrophile, the leaving group is pyrophosphate (PP_i), which is a stabilized anion and therefore a very good leaving group. In contrast, during the first step of natural RNA splicing the leaving group is the 5'-exon, which is an oligonucleotide and thus not a particularly good leaving group. We demonstrated that 7S11 can function with an oligonucleotide leaving group (see Coppins & Silverman, Biochemistry 2005, 44, 13439-13446 [#40]). The ligation rate and yield are reduced by the change in leaving group from PP_i to an oligonucleotide, but this was expected based on the comparison of their leaving group abilities (indeed, a Brønsted plot of log k_obs versus pK_a of leaving group was linear). These findings further strengthen the parallel between the 7S11-catalyzed ligation reaction and natural RNA splicing. It is an interesting and unanswered question whether or not a single deoxyribozyme that performs both the first and second steps of splicing can be identified by in vitro selection.

We have also obtained data indicating that the branch-site adenosine nucleophile used by 7S11 is preferred over the other nucleotides, at least in the 7S11 structural context that mimics natural splicing (see Zelin, Wang & Silverman, Biochemistry 2006, 45, 2761-2771 [#47]). We changed the branch-site nucleotide to uridine, randomized the two DNA enzyme loops, and repeated the selection efforts. Despite never having been provided with branch-site adenosine, the resulting deoxyribozymes preferred branch-site A over U, with k_A/k_U between 20–1000. It should be possible to dissect chemically the basis for the adenosine preference via atomic-scale changes to the adenine nucleobase, such as replacement with purine, 2,6-diaminopurine, and inosine.

Click here to return to the top of this webpage

Lariat RNA synthesis by deoxyribozymes: Exquisite site-selectivities for reactions that are impossible to achieve by traditional organic synthesis approaches

The deoxyribozymes such as 7S11 that create branched RNA allow the synthesis of  RNA products that are difficult to prepare by conventional methods of organic synthesis or solid-phase synthesis. The actual intermediates in biological RNA splicing are a subclass of 2',5'-branched RNAs called lariat RNAs. A lariat RNA has a 2',5'-branched core but additionally has a closed loop formed by two of the three single-stranded "arms" that emerge from the branch-site nucleotide. A lariat is formed during the first step of RNA splicing, when the branch-site adenosine nucleophile attacks the 5'-splice site and displaces the 5'-exon. Due to their special topology, lariat RNAs are almost impossible to synthesize by traditional organic synthesis or solid-phase synthesis methods. In contrast, we have recently shown that deoxyribozymes can be used for efficient synthesis of biologically related lariat RNAs (see Wang & Silverman, Angew. Chem. Int. Ed. 2005, 44, 5863-5866 [#38]). A specific deoxyribozyme, 6BX22, was identified that tolerates almost any set of RNA substrate sequences and also is not inhibited by the incipient lariat loop during branch formation. This reaction is formally a macrocyclization, and in one case 6BX22 was used to synthesize a 1597-membered ring (266 nucleotides) in 72% yield with <0.1% side products—and with no protecting groups on any of the 308 other 2'-hydroxyl groups that do not react during the lariat formation. Such exquisite site-selectivity without the use of protecting groups is essentially impossible by any conventional organic chemistry approach. Nevertheless, deoxyribozymes catalyze the desired reaction efficiently.

In some cases, one-step deoxyribozyme-catalyzed lariat synthesis does not work efficiently. We have developed an alternative two-step strategy in which a 2',5'-branch is formed first, and the lariat loop is then closed (see Wang & Silverman, RNA 2006, 12, 313-321 [#44]).

Click here to return to the top of this webpage

Using the RNA ligation products of deoxyribozymes to test specific biochemical hypotheses

Because the branched and lariat RNA products formed by deoxyribozymes are directly relevant to biochemistry, we have focused substantial effort on using these RNA products to test specific biochemical hypotheses. As one example of these efforts, we have recently described the study of two potential proofreading mechanisms for 5'-splice site selection by group II introns (see Wang & Silverman, ACS Chem. Biol. 2006, 1, 316-324 [#50]). Over a decade ago, the reversibility of the first step of RNA splicing was noted, and it was hypothesized that this first-step reversibility may serve a proofreading function for mis-splicing at the 5'-splice site. However, until now this hypothesis could not be tested because the requisite mis-spliced RNAs could not be synthesized. Using the 7S11 deoxyribozyme, we synthesized mis-spliced 2',5'-branched RNAs derived from the ai5g group II intron and tested directly their ability to undergo the reverse of the first step of splicing. This reversibility was not observed, indicating that the hypothesis of a proofreading role for such reversibility is probably incorrect. In parallel, we identified an alternative proofreading mechanism for mis-spliced RNAs that involves hydrolysis of the incorrect 5'-exon at the proper 5'-splice site and subsequent ligation to the 3'-exon, all catalyzed by the mis-spliced intron.

We have also applied deoxyribozymes to evaluate the proposed role of branched RNA in retrotransposition of the yeast Ty1 element. Retrotransposition is a process by which mobile DNA genetic elements such as Ty1 move around the genome via RNA intermediates. Others recently proposed that 2',5'-branched RNA is an obligatory Ty1 retrotransposition intermediate for which cleavage by the lariat debranching enzyme, Dbr1p, enables reverse transcription to continue synthesizing the complete Ty1 complementary DNA (cDNA). Cells lacking Dbr1p can still produce substantial Ty1 cDNA, and therefore the obligatory  intermediacy of branched RNA would require that the Ty1 reverse transcriptae (RT) can read through the proposed branch site with considerable efficiency. The proposed branched RNA has a branch-site uridine. We used a deoxyribozyme that can synthesize branched RNA with any branch-site nucleotide (see Pratico, Wang & Silverman, Nucleic Acids Res. 2005, 33, 3503-3512 [#37]) to prepare branched RNA that corresponds directly to the proposed Ty1 branch site for a direct test of this read-through ability. We found that Ty1 RT cannot efficiently read through the proposed branch site (see Pratico & Silverman, RNA 2007, 13, 1528-1536 [#57]), which is inconsistent with the hypothesis that branched RNA is an obligatory Ty1 retrotransposition intermediate. Our study is notable because we used a deoxyribozyme product to enable a direct test of a biochemical hypothesis, in a fashion that was essentially impossible without the deoxyribozyme.

Click here to return to the top of this webpage

DECAL: Deoxyribozyme-catalyzed labeling of RNA

The ability to attach an RNA strand to a specific 2'-hydroxyl group of an RNA target can be used as an RNA labeling strategy (see Baum & Silverman, Angew. Chem. Int. Ed. 2007, 46, 3502-3504 [#56]). In this approach, which we term DECAL (for "DEoxyribozyme-CAtalyzed Labeling"), we first synthesize a short "tagging RNA" by in vitro transcription using T7 RNA polymerase. The tagging RNA incorporates a 5-aminoallylcytidine (or uridine) nucleotide at its second position. The aminoallyl-modified transcript is coupled with the amine-reactive form of a desired biophysical probe to form the labeled tagging RNA. The tagging RNA is then attached by the deoxyribozyme to a specific internal 2'-hydroxyl group of the RNA. Because the intact target RNA is derivatized directly with the tag, splint ligation is avoided, and no nucleotide mutations are required in the target RNA.

We have performed DECAL with useful biophysical probes such as fluorescein, tetramethylrhodamine, or biotin. Labeling RNA with fluorescein and tetramethylrhodamine at two separate sites allows fluorescence resonance energy transfer (FRET) experiments to be performed. Using this approach, we efficiently generated dual-labeled P4-P6 RNA, allowing FRET experiments to be performed with only a small fraction of the time and effort that is necessary using the conventional splint ligation approach.

In ongoing work, we are applying DECAL to enable folding studies for large RNAs such as group II introns, which have >800 nucleotides. We are also using DECAL with biotinylated tagging RNAs to enable sequence-specific biotinylation of RNA targets in complex mixtures of RNA. We anticipate that this methodology will be useful in biologically relevant experiments such as investigation of in vivo RNA-protein interactions.

Click here to return to the top of this webpage

Branched RNA-DNA conjugates applied to studies of DNA as a constraint

We have used in vitro selection to identify deoxyribozymes that attach DNA strands to a specific 2'-hydroxyl groups of large RNA targets. These deoxyribozymes enable our efforts with DNA as a constraint to study RNA folding and catalysis. See our webpage on DNA as a constraint for more information.

Click here to return to the top of this webpage

Engineering a three-helix-junction deoxyribozyme to function with a small-molecule substrate

One of the main challenges for deoxyribozymes is their application to small molecules as substrates. Upon inspecting the three-helix-junction structure of 7S11 and related deoxyribozymes, we recognized that we should be able to disconnect the RNA nucleotide that provides the electrophilic 5'-triphosphate from the remainder of the corresponding RNA oligonucleotide substrate. We hoped that the deoxyribozyme would still function, but now with GTP as the separate small-molecule substrate (and the remainder of the original oligonucleotide substrate as an obligatory cofactor). Our efforts were successful: we engineered the 10DM24 deoxyribozyme to mediate the multiple-turnover ligation of GTP (see Höbartner & Silverman, Angew. Chem. Int. Ed. 2007, 46, 7420-7424 [#59]). We also examined in detail the requirements for productive substrate binding. We found that hydrogen bonding contributes substantially to selective binding of GTP (or other NTP substrates), and structural preorganization within the substrate is important for its efficient utilization.

We envision that the use of GTP analogs or other small molecules as deoxyribozyme substrates can be combined with the DECAL approach described above. This will allow RNA labeling with tags that are small molecules rather than oligonucleotides, which will have practical advantages.

Click here to return to the top of this webpage

Beyond RNA ligation: Deoxyribozymes that catalyze a variety of reactions

There is no chemical reason why DNA-catalyzed synthetic reactions should be limited merely to RNA ligation. We have initiated experiments to identify deoxyribozymes that catalyze reactions of even wider interest and utility. Two of our research directions are described briefly here.

First, we have used the three-helix-junction deoxyribozyme platform to explore DNA-catalyzed reactions of amino acid sidechains, which are previously unknown. This was done by replacing the branch-site adenosine nucleophile with a single amino acid (one of tyrosine, serine, or lysine) and performing new in vitro selection experiments. We identified a new deoxyribozyme, Tyr1, that efficiently creates a Tyr-RNA nucleopeptide linkage (see Pradeepkumar, Höbartner, Baum & Silverman, Angew. Chem. Int. Ed. 2008, 47, 1753-1757 [#63]). In contrast, deoxyribozymes that function with Ser or Lys were not found. These studies demonstrate that DNA has the catalytic ability to create nucleopeptide linkages, thereby further expanding the scope of DNA catalysis beyond oligonucleotide substrates. Our current efforts focus on improving the DNA-catalyzed reactivity of many amino acid sidechains and widening the scope to use polypeptide and protein substrates.

Second, we have used in vitro selection to identify deoxyribozymes that catalyze the Diels-Alder reaction (see Chandra & Silverman, J. Am. Chem. Soc. 2008, 130, 2936-2937 [#64]). This represents the first effort from any laboratory toward the use of catalytic DNA with solely small-molecule substrates (in our efforts above with the 10DM24 deoxyribozyme and GTP substrate, the other substrate is still an oligonucleotide). One particular deoxyribozyme was effective at catalyzing the Diels-Alder reaction between two small-molecule substrates (anthracene and maleimide derivatives) in multiple-turnover fashion with a separate deoxyribozyme as catalyst. These findings with the Diels-Alder reaction demonstrate that DNA is as catalytically efficient as RNA, at least for this important C–C bond-forming reaction of organic chemistry. Therefore, the lack of 2'-hydroxyl groups does not impede DNA catalysis relative to RNA catalysis.

We hope that these two examples are the first steps on the eventual pathway to using deoxyribozymes in broad fashion for bioorganic chemistry. Despite our hope, an honest assessment is that deoxyribozymes will almost certainly not replace more traditional catalysts for most synthetic chemistry applications. Nonetheless,  in cases where the high selectivity of an "enzyme" is required along with a modest reaction scale (e.g., milligrams or less), the use of deoxyribozymes is plausible. At present, the two principal factors that limit this use are (1) the scope of reactions known to be catalyzed by DNA, combined with (2) the ability to obtain DNA catalysts that function with non-oligonucleotide substrates, especially small molecules and proteins. We hope to make advances on both frontiers in the coming years.

Click here to return to the top of this webpage

Elucidating deoxyribozyme structures and reaction mechanisms

One of the main challenges for all of our new DNA enzymes is to understand the chemical mechanisms by which they achieve catalysis. Conventional biochemical experiments involving mutagenesis (i.e., changes of one nucleotide to another) will be useful in this regard, as will high-resolution information from structural biology methods. For the latter, we are pursuing both X-ray crystallography and NMR spectroscopy analysis of some of our DNA enzymes, such as the three-helix-junction 7S11 deoxyribozyme that is described above. In parallel, we are interested to use lower-resolution techniques such as comparative gel electrophoresis (CGE) and fluorescence resonance energy transfer (FRET) to obtain information on deoxyribozyme structure and folding. In collaboration with Prof. Taekjip Ha in our Department of Physics, we are additionally interested in pursuing single-molecule studies of deoxyribozyme folding and catalysis.

Miscellaneous issues related to deoxyribozymes

Please click here for a discussion of the relationship of DNA catalysis to DNA-templated synthesis (DTS).
Please click here for a discussion of the term "catalysis" as applied to deoxyribozymes.

Click here to return to the top of this webpage

Click here to return to the Silverman lab overview of research

Click here for a full publications list

Research description last updated May 4, 2008