I. Introduction

Top Previous Next References

Concepts for a genetic type of therapy were developed almost 20 years ago and have recently been transformed (albeit with limited success) into clinical reality. In 1990, the first gene therapy study for the treatment of ADA deficiency began (Blaese et al., 1995), followed by a rapidly growing number of clinical gene therapy trials, across diseases caused by genetic disorders, viral infections, and malignancies (Miller, 1992; Anderson, 1998). Current strategies are aimed at either the replacement of defective genes or suppression of a pathological gene target. At present, there are several well documented genetic approaches being pursued to affect the ablation of defective gene production. Strategies exemplified by the use of antisense RNA or deoxyoligonucleotides (Milligan et al., 1993; Crook and Bennett, 1996), expression of mutant structural or regulatory genes with dominant repressor activity (Woffendin et al., 1996) and over-expression of competitive RNA sequences (Sullenger et al., 1990; Berkhout and van Wamel, 1995) vary widely in their degree of specificity and serve primarily to block gene expression by interfering with RNA transcription or translation. Unique among these approaches is the use of catalytic nucleic acids, RNA (ribozyme) or DNA (deoxyribozyme, DNA enzyme or DNAzyme) oligonucleotides capable of cleaving a target RNA molecule in a highly sequence-specific manner (Rossi, 1992; Sun et al., 1997, 1999; Welch et al., 1998).

Ribozymes, the most extensively studied of the catalytic nucleic acids, exist in a range of distinct categories of naturally occurring catalytic RNA. These include a series of small ribozymes important for the rolling circle replication of viroid genomes, such as hammerhead and hairpin ribozymes (Haseloff and Gerlach, 1988; Hampel and Tritz, 1989; Rossi, 1992), group I introns (Cech et al., 1981, 1990, 1992), the RNA component of RNase P (Guerrier-Takada et al., 1983; Frank and Pace, 1998), and hepatitis delta virus ribozyme (Branch and Robertson, 1991). In addition to naturally occurring ribozymes, the number of entirely synthetic RNA molecules with identified novel catalytic activities have increased dramatically over the past few years as a result of the development of in vitro selection and evolution techniques. DNAzymes, unlike their naturally occurring counterpart, are a recent development in catalytic nucleic acid technology. Their catalytic activity is exclusively derived from in vitro selection procedures in which deoxynucleotides have been "trained" for activities as diverse as cleavage of RNA or DNA (Carmi et al., 1998), DNA oligonucleotide ligation, and phosphorylation of DNA (Cuenoud and Szostak, 1995; Li and Breaker, 1999a). A new class of DNA enzyme the "10-23" DNAzyme was selected from a combinatorial library of DNA sequences for its ability to cleave a short HIV target RNA (Santoro and Joyce, 1998).

The catalytic activity and specificity of both ribozymes and DNAzymes has been extensively characterized in vitro and in cell culture systems. A wide range of chemical modifications has allowed the synthesis of oligonucleotides with in vivo stability approaching most conventional drugs. Efficient delivery and distribution of oligonucleotide compounds both intracellularly and in vivo remains, however, a critical challenge for a successful transition from the laboratory to the clinic: a limitation common to virtually all aspects of nucleic acid-based therapy. Our discussion will cover a range of issues from structure and catalytic mechanism to delivery and pharmaceutical development. Given the broad scope of this field, we have chosen a limited number of recent references focusing primarily on the investigation of catalytic nucleic acids with potential clinical application.

	II. Ribozymes

Top Previous Next References

Ribozymes are catalytic RNA molecules possessing, at the very least, enzymatic cleavage and ligation activities (Haseloff and Gerlach, 1988; Hampel and Tritz, 1989; Rossi, 1992). In nature, enzymatic RNA molecules catalyze sequence-specific RNA processing. The specificity is determined by Watson-Crick base-paring between ribozymes and nucleotides near the cleavage site of the target RNA. By altering substrate recognition sequences, several intramolecular cis-cleaving ribozymes have been engineered to cleave target RNA in trans (Symons, 1994). Theoretically, these trans-cleaving ribozymes can be designed to cleave any RNA species in a sequence-specific manner. Thus, the mRNA coding of any proteins associated with a disease can be selectively cleaved by ribozymes. Consequently, ribozymes have become potentially valuable tools for the inhibition of virus replication, modulation of tumor progression, and analysis of cellular gene function.

At present, there are five major RNA catalytic motifs that are derived from naturally occurring ribozymes: hairpin, hammerhead, group I intron, ribonuclease P and hepatitis delta virus ribozyme. Among those, two RNA catalytic motifs that originate from plant viriod and virusoids have received much attention for their potential use, due to their inherent simplicity, relatively small size, and the ability to be incorporated into a variety of flanking sequence motifs without changing site-specific cleavage capacities. These are hammerhead and hairpin ribozymes, illustrated in Fig. 1. The hammerhead ribozyme model is based on the satellite RNA strand (+) of tobacco ringspot virus (sTobRV²). It has three basic components: (i) a highly conserved 22 nucleotide-catalytic domain; (ii) base-pairing sequence flanking the susceptible 3',5'-phosphodiester bond; and (iii) a recognition sequence on the target RNA such as GUC. The cleavage reaction occurs 3' to the recognition sequence with formation of a terminus containing a 2',3'-cyclic phosphodiester and a 5'-hydroxyl terminus on the 3'-fragment (Symons, 1994). The () strand of the sTobSV forms a two-dimensional hairpin structure, with four major helical regions possessing catalytic activity. On the basis of this structure, the hairpin ribozyme model consists of four helices and five loop regions formed between a 50-base catalytic RNA and a 14-base substrate RNA. Helices 3 and 4 are within the ribozyme itself, and the ribozyme binds to the target RNA through helix 1 (six base pairs) and helix 2 (four base pairs), separated by a NGUC loop in the substrate strand. The recognition sequence is bNGUC, where b is G, C or U; N is any nucleotide, and cleavage occurs 5' to the G residue (Hampel and Tritz, 1989). Other catalytic motifs, group I introns that are derived from the self-splicing intervening sequence of Tetrahymena thermophila, RNase P, and catalytic domain of human hepatitis delta virus will be discussed in detail in Section IV.

View larger version (20K): [in this window] [in a new window]   Fig. 1.   Structures of hammerhead (A) and hairpin ribozymes (B). Refer to text for details.

A large number of studies have been performed to provide a better understanding of the mechanism of ribozyme-mediated catalysis. In general, a minimal kinetic description for one turnover of hammerhead ribozyme reaction (Fig. 2) involves assembly of ribozyme (E) and substrate (S) into an E · S complex, cleavage of the phosphodiester bond, generating a 5'-product with a 2',3'-cyclic phosphate terminus (P1) and a product with a 5'-hydroxyl terminus (P2) that remains bound to the ribozyme, and release of the products. A Michaelis-Menten mechanism has been established by using multiple- and single-turnover conditions for the formation of the ribozyme-substrate complex and its subsequent conversion to products (Fedor and Uhlenbeck, 1990; Perreault et al., 1990).

View larger version (10K): [in this window] [in a new window]   Fig. 2.   The catalytic cycle of an RNA-cleaving nucleic acid enzyme. This diagram is a proposed schematic representation of the catalytic cycle typical for small RNA-cleaving nucleic acid enzymes such as the hammerhead ribozyme and the 10-23 deoxyribozyme, which bind to their substrate via Watson-Crick interactions. The description is based on the data from kinetic analysis and X-ray crystal structures (Kumar et al., 1996; Scott et al., 1996; Murray et al., 1998). These reactions are initiated when the binding domains or pairing arms of the enzyme specifically engage the single-stranded RNA substrate by forming two double-helical duplexes (enzyme-substrate complex). Between the binding domains, the catalytic domain forms a catalytically active secondary structure that is capable of positioning an attacking divalent cation (such as Mg2+) in proximity to the cleavage site. In this configuration the reaction between metal ions and RNA is enhanced by many orders of magnitude through stabilization of the transition state (transition state complex). After cleavage of the phosphordiester linkage the enzyme and two cleavage products remain associated (enzyme-product complex) before becoming dissociated (product release). The catalytic nucleic acid is then free to bind more substrate in another cycle of the reaction.

Under multiple-turnover conditions, the substrate is in excess of the ribozyme so that the ribozyme can catalyze the cleavage of several substrate molecules. The catalytic rate constant or turnover numbers, k_cat, is a measure of the rate-limiting step, which can be cleavage, conformational transitions of the ribozyme-substrate complex, or product release. When a ribozyme with six bases in each arm was used in kinetic analysis, k_cat and K_M were within typical values of 1-2 min¹ and 20-200 nM, respectively. Extension of the helical arms generally results in an increased stability and markedly decreased k_cat (Hertel et al., 1994). Under single-turnover conditions, the ribozyme is in excess of the substrate, and such conditions are normally used for cleavage of long mRNA substrate. Cleavage rates are generally several orders of magnitude lower, compared with the rates obtained for short substrate (Heidenreich et al., 1994). In the case of hammerhead and hairpin ribozymes, the presence of Mg²⁺ is essential for the cleavage, and the convincing evidence suggested that Mg²⁺ not only assists in RNA folding but also participates directly in the cleavage mechanism (Dahm and Uhlenbeck, 1991).

Hammerhead ribozymes such as those illustrated in Fig. 1, can cleave any 5'-NUH-3' triplets of an RNA, where U is conserved, N is any nucleotide and H can be C, U, A, but not G. This cleavage rule has been further extended from the work of several groups (Perriman et al., 1992; Shimayama et al., 1995; Zoumadakis and Tabler, 1995; Ludwig et al., 1998). Comparative studies revealed that the reaction rate (k_cat) decreases in the following order: AUC, GUC>GUA, AUA, CUC>AUU, UUC, UUA>GUU, CUA>UUU, CUU. Results from recent studies extend the repertoire of cleavage sites for hammerhead ribozyme application to inhibition of gene expression. Kore et al. (1998) report that the NUH rule can be modified to the NHH rule, where H is any nucleotide except G. However, whether this extension can be applied to in vivo applications is yet to be tested in a suitable model. This can be further complicated when the target RNA structure in cellular environment is taken into consideration.

In addition to the NHH triplets for cleavage specificity, the ribozyme arm sequence context can also influence cleavage rate significantly. In a simple term, the longer the binding arms, the lower the turnover in cleavage of short substrates. The length of the 3'-arm is apparently more critical for specificity than that of the 5'-arm (Hertel et al., 1996). For a long substrate RNA, intramolecular structures of the target RNA could interfere with the binding of a ribozyme, and for in vivo application, cellular proteins could also affect ribozyme activity. Results from variations in arm length of ribozymes from 5 to 18 nucleotides (nt) have indicated that the ribozyme activity is closely related to the arm length both in symmetric or asymmetric models, and this depended somewhat on the sequence context (Heidenreich and Eckstein, 1992; Scherr et al., 1997; Sioud et al., 1997; Crisell et al., 1993).

Given that the biophysical principles governing RNA folding in vivo have not been clearly determined, there is no clear means available to determine the accessibility of a potential RNA substrate cleavage site. As a first approximation the gross topography of the substrate RNA can be simulated by analyzing the region surrounding the cleavage site using an RNA secondary structure folding program. In this way, it may be possible to determine whether or not the target site is buried within an obvious thermodynamically stable region of secondary structure. However, there are limitations to the size of RNA that may be analyzed by this method, and this is clearly a gross approximation of RNA secondary structure that does not take into account tertiary structure nor RNA-protein interaction, which may be present in the intracellular context. Other tools for empirical testing of cleavage site accessibility, such as S1 nuclease or RNase mapping, can supply additional information (Pavlakis et al., 1980; Knapp, 1989). As an alternative, an assessment of biological significance may allow one to predict certain features of the RNA that would indicate accessibility. For example, a number of regions in the HIV-1 genome have been targeted by different groups using ribozymes, including the 5'-leader region (Weerasinghe et al., 1991; Ojwang et al., 1992), gag (Sarver et al., 1990) and tat genes (Lo et al., 1992; Sun et al., 1995; Wang et al., 1998), and the  packaging site (Sun et al., 1994). These target sites are generally exposed for potential RNA-RNA or RNA-protein interactions as a result of their biological function, and thus should be potentially accessible in cells.

In addition to enzymatic mapping of accessible sites in an RNA, combinatorial approaches may be used to identify ribozyme accessible sites, in which no assumptions regarding the best target sites and corresponding ribozymes are made. This approach may provide a cost-effective mapping technique and could speed up the process of drug discovery. Lieber and Strauss (1995) first reported the use of combinatorial approach for the selection of ribozyme cleavage sites. A library of ribozyme genes with random sequences of 13 nucleotides on both sides of the hammerhead was generated. Using RACE (rapid amplification of cDNA ends) technique, they successfully identified cleavage sites, and subsequently reamplified and cloned the ribozyme genes. In our laboratory, a similar design was used to construct a synthetic ribozyme library, and sensitivity of detection of cleavage sites was increased significantly by using ligation-mediated polymerase chain reaction (PCR) (Y. Kim and L. Q. Sun, unpublished results). More recently, a simpler approach that requires neither PCR nor cloning was also recently reported by Yu et al. (1998), in which a population of in vitro transcribed hairpin ribozymes with randomized substrate binding arms, was incubated with the substrate RNA, and cleavage positions were determined by primer extension.

Ribozyme-mediated suppression of gene expression has proven to be neither straightforward nor routine partly due to the difficulty in defining the most accessible sites. In vitro screening provides the simplest and most direct approach to identifying the accessible sites although there are some concerns in terms of correlation between in vitro and in vivo results. However, a combined approach of cell-based screening systems for target site selection and in vitro screening using the full-length mRNA will provide valuable information for the design of gene-targeted ribozymes.

The gradual maturation of ribozyme technology from the bench to clinical application involves several major challenges, many of which still need to be resolved. These include extra- and intracellular stability of the ribozyme, delivery of ribozymes to target cells, target accessibility, colocalization of ribozyme and target within cells, and optimal catalytic activity and specificity of the ribozyme. Despite their enzymatic activity, ribozymes must be delivered to target cells in amounts sufficient to both affect a significant proportion of the cell population and to affect a significant proportion of the target mRNA. Although therapeutic application of ribozymes may be achieved by exogenous delivery of chemically produced ribozyme complexed, for example, to a cationic lipid (Section V), for certain applications, such as their use in chronic disease or in an antiviral therapeutic settings, they may require permanent availability to be of therapeutic benefit. This can be achieved by transfer of genes encoding for ribozymes using viral vector systems (for review see Sun and Symonds, 1998).

The preclinical and clinical application of ribozyme-based gene therapy has been primarily focused on AIDS, cancers, and other viral infections. This is summarized in Table 1. Here, as an example, we will discuss some of the aspects of ribozyme-based gene therapy for AIDS.

A simple approach to treating HIV-infected patients is the infusion of transduced and "protected" CD4⁺ peripheral blood lymphocytes. If ribozyme the construct can protect CD4⁺ T cells from HIV-1 infection and its sequelae in patients, then the decline in the numbers of CD4⁺ T cells could be halted or even to some extent reversed, and HIV-infected individuals could benefit clinically. It is relevant to note that the half-life of an HIV-infected CD4⁺ T lymphocyte is of the order of only 2 days. We have initiated two independent Phase I clinical trials to test safety of this approach, the ability to detect ribozyme-containing cells in the bloodstream and the hypothesis that ribozymes can protect CD4⁺ T lymphocytes from rapid HIV-1-mediated destruction within an infected individual. Both trials use LNL6 vector (a Moloney murine leukemia virus-based expression vector) and recombinant LNL6 containing Rz2 (RRz2), and each trial uses a separate target cell populationCD4⁺ peripheral blood lymphocytes and CD34⁺ stem. The first trial involves identical twins, discordant for infection with HIV (Cooper et al., 1999). Healthy CD4⁺ lymphocytes from the uninfected twin are transduced with a retroviral vector containing the ribozyme gene. These transduced cells are cultured and expanded ex vivo before transfusion into the bloodstream of the corresponding HIV-positive twin. The second clinical trial involves the removal, transduction, and transfusion of CD34⁺ stem cells within HIV-positive individuals (Rosenblatt et al., 1999). The rationale being that these transduced stem cells will differentiate and give rise to a variety of lineages that express the ribozyme.

In each trial, separate populations of cells are transduced with the retrovirus vector containing the ribozyme and the vector alone as a control, and equal numbers of these two transduced cell types are then introduced into the recipient patients. This has allowed us to monitor the survival of ribozyme-expressing CD4⁺ lymphocytes relative to a similar population of transduced CD4⁺ lymphocytes that have been processed in an identical fashion but do not contain the ribozyme sequence. Cell survival is being monitored by detecting ribozyme and control vector DNA sequences in peripheral blood using quantitative PCR procedures. Preliminary data indicate that transduction of the ribozyme construct into human cells is safe, and the transduced cells can be detected in all the patients.

In summary, a large body of work on ribozyme catalytic efficiency, mode of action, and basic chemistry have accelerated the possibility of using ribozyme human gene therapy to target specific human diseases. With the development of delivery systems well under way, and with much deeper understanding of molecular genetics of human disease, it is hoped that ribozymes will soon emerge as gene-targeted molecular therapies.

	III. RNA-Cleaving Deoxyribozymes

Top Previous Next References

The therapeutic potential of RNA enzymes or ribozymes with RNA-cleaving activity is now well established (Bramlage et al., 1998). The gene suppression activity of these nucleic acid-based agents is probably mediated by a combination of antisense RNA interactions and catalytic destruction of the mRNA target. Although it's possible to manufacture biologically active RNA, its relative fragility in this environment makes it difficult to administer in a direct delivery mode. For this and other reasons, most ribozyme applications rely on transgenic production of RNA in vivo within the context of a gene therapy. Advances in nucleic acid chemistry, however, have led to progress in ribozyme synthesis, which allow the incorporation of modified ribonucleotide analogues with enhanced nuclease resistance (Scaringe et al., 1990; Wincott et al., 1995). This usually involves substitution of the 2'-hydroxyl moiety with some other chemistry such as 2'-deoxy, 2'-O-methyl, 2'-amino, or 2'-fluoro derivatives. These 2'-modifications, however, reduce or eliminate catalytic activity when made at conserved positions and thus cannot be applied to the entire ribozyme (Perreault et al., 1990; Pieken et al., 1991; Williams et al., 1992; Yang et al., 1992). The most effective synthetic ribozymes have a chimeric composition that enhances nuclease stability while maintaining cleavage activity (Paolella et al., 1992; Beigelman et al., 1995). Substitution of DNA (and other 2'-modified ribonucleotides) into the helix-forming motifs and specific unpaired positions of the catalytic domain has been shown to produce a substantial improvement in the biological stability; however, these chimeric ribozymes are still relatively vulnerable to endoribonucleases. By comparison, antisense DNA oligonucleotides (ODNs) with their uninterrupted DNA composition have a much greater half-life in vivo. The natural biological stability of DNA compared with RNA can also be readily supplemented by chemistry, which provides even greater resistance to nuclease digestion; the more commonly used modification involves the replacement of phosphodiester linkages in the backbone with phosphorothioate or methylphosphonate moieties. However, despite being more suitable for direct delivery, antisense ODNs are devoid of endogenous RNA cleavage activity and thus can only act as passive inhibitors of translation machinery through their sequence-specific binding activity, or mediate destruction of the RNA target component of an RNA-ODN heteroduplex in the nucleus of cells by activating ribonuclease H.

Perhaps an ideal oligonucleotide-based RNA-directed gene inactivation agent is one that could combine the self-sufficient RNA destructive capability of ribozymes, such as the "hammerhead" and the "hairpin", with the biological resilience of the antisense ODN. Although DNA molecules with RNA cleavage activity have not been observed in nature, some are now in existence thanks to an accelerated evolutionary process (in vitro selection) designed specifically to derive DNA sequences with this activity (Breaker and Joyce, 1994, 1995; Santoro and Joyce, 1997). In natural systems the replicability and relative stability of double helical DNA makes it well suited to its role as the custodian of genetic information. In this form DNA secondary and tertiary structure is severely restricted and provides very little opportunity for the exploration of conformations, which might facilitate useful reaction rate enhancement. Despite the dominance of protein catalysts (and a small role for RNA catalysts) in this environment, DNA, which is liberated from its complimentary strand, is also capable of substantial structural diversity and even catalytic activity (Breaker, 1997; Li and Breaker, 1999b).

The possibility of an RNA-cleaving catalytic DNA was initially explored in selection constructs containing a single embedded ribonucleotide substrate component. Downstream from the substrate, a section of randomized DNA sequence (40-50 bp) provided molecular diversity, whereas fixed sequences on each flank provided stable sites for primer binding. A sample population from these combinatorial libraries (complete library = 10¹³-10¹⁴ variants) was immobilized on a solid support containing streptavidin via a biotin tag (Fig. 3). After washing and stripping the complementary template strand in alkali, a buffer containing an appropriate divalent metal ion (Pb²⁺, Zn²⁺, Mn²⁺, Mg²⁺) was introduced to generate conditions conducive to a cis-cleavage reaction (Breaker and Joyce, 1994, 1995). The release of active members in the library by this self-cleavage allowed them to be selectively amplified and regenerated by PCR for subsequent rounds of selection. After numerous cycles of selection and amplification, the self-cleavage activity of the remaining population was examined by electrophoresis of an end-labeled fraction that was allowed to fold and react in solution. When it was established that the selected pool contained a satisfactory level of self-cleavage activity, the PCR products were cloned and sequenced. The cloned sequences were then used to examine the cleavage activity of individual molecules and predict secondary structures. The secondary structure predictions for the intramolecular configuration of enzyme-substrate complex was then used to separate these two components (of the cis arrangement), such that the enzyme and substrate would interact in trans. The trans cleavage capacity of these DNA enzymes or deoxyribozymes was then examined using a separate oligonucleotide enzyme and substrates. Using this selection protocol in the first instance with Pb²⁺ and then with other buffer systems including Mn²⁺, Mg²⁺, and Zn²⁺, different deoxyribozymes were found that could use each of these divalent metal ions to cleave a single embedded ribonucleotide (Breaker and Joyce, 1994, 1995). In the case of the magnesium-dependent deoxyribozyme, this was accomplished with a rate up to 10⁵ times that of the uncatalyzed reaction.

View larger version (11K): [in this window] [in a new window]   Fig. 3.   An in vitro selection scheme for RNA-cleaving catalytic DNA. Adapted from Breaker and Joyce (1994). In this scheme a combinatorial library of different DNA sequences is generated by the synthesis of an oligonucleotide template with a 50 base (N50) stretch of random sequence (hatched lines). To enable amplification and reselection, each molecule from this diverse pool contains a fixed sequence at each end for use as primer binding sites (black). Immobilization is achieved by tethering this assembly of potentially functional sequences through a biotinylated (B) RNA containing substrate (rA) to a solid streptavidin-coated support. After alkali denaturation (to strip away the template strand) the single-stranded residue is allowed to fold around its substrate linkage in the presence of suitable buffer and cofactors. The eluent of this washing (which contains catalytically active molecules) is then used as a PCR template for regenerating the assembly for a second round of selection while the inactive sequences remain tethered to the support. After multiple rounds of selection the relative abundance of each active molecule should be high enough to visualize the activity in a cleavage assay and clone the DNA for sequencing.

After further development of the in vitro selection protocol, Mg²⁺-dependent DNA enzymes capable of cleaving a biologically relevant, all RNA substrate, were evolved (Santoro and Joyce, 1997). From this selection, experiment two prototypes were characterized and denoted the "8-17" and the "10-23" RNA-cleaving DNA enzymes (Fig. 4). Both of these molecules are reminiscent of the hammerhead ribozyme in that they contain conserved catalytic domains flanked by variable binding domains. Each binding domain was originally evolved to specifically complement the substrate sequence used in the selection protocol; however, as this was achieved in both of these deoxyribozymes through Watson-Crick base pairing, the possibility existed that the specificity could be altered to suit any target RNA sequence. This was particularly successful in the 10-23 deoxyribozyme, which was found to have a minimum sequence requirement of a purine-pyrimidine pair at the site of cleavage. Despite being less flexible, the 8-17 deoxyribozyme with its requirement for an AG in the substrate, can also be described as a general-purpose endoribonuclease.

View larger version (21K): [in this window] [in a new window]   Fig. 4.   A schematic representation of the "10-23" and "8-17" general purpose RNA-cleaving deoxyribozymes. Panels A and B contain illustrations of the 10-23 and 8-17 deoxyribozymes, respectively. Watson-Crick interactions for each deoxyribozyme-substrate complex are represented by generic ribonucleotides (N) in the target (top) and the corresponding deoxyribonucleotides (N) in the arms of the deoxyribozyme (bottom) of each. The defined sequences in the loop joining the arms and spanning a single unpaired purine at the RNA target site of each model represent the conserved catalytic motif.

The 10-23 DNA enzyme or deoxyribozyme was named from its origin as the 23rd clone of the 10th cycle of in vitro selection (Santoro and Joyce, 1997). This enzyme has a number of features that endow it with tremendous potential for applications both in vitro and in vivo. These include its ability to cleave almost any RNA sequence with high specificity provided it contains a purine-pyrimidine dinucleotide. This can be accomplished at very high kinetic efficiency with rates approaching and even exceeding those of other nucleic acid and protein endoribonucleases (Santoro and Joyce, 1997). This remarkable activity is all the more spectacular when considering that it is achievable at concentrations of magnesium down in the physiological range. After characterization of the 10-23 catalytic motif, more rounds of reselection and amplification were undertaken with a partially randomized sequence (25% degeneracy) with the aim of further optimizing the catalytic activity (Santoro and Joyce, 1997). Surprisingly, this selection did not yield any refinement of 10-23 deoxyribozyme, indicating that the catalytic motif was highly conserved and had very little sequence redundancy.

1. Kinetic Efficiency. The ability of the 10-23 deoxyribozyme to cleave purine-pyrimidine junctions meant that the AUG start codon of any gene could be used as a target. Early kinetic analysis of the 10-23 deoxyribozyme focused on synthetic substrate sequences derived from the start codons of various HIV genes (Santoro and Joyce, 1997). A key point that emerged from this analysis was that the kinetic efficiency of deoxyribozyme-catalyzed cleavage varied substantially from one substrate sequence to the next. This sequence-dependent variability seemed to be closely associated with the thermodynamic stability of the enzyme-substrate heteroduplex as predicted by the hybridization free energy (Sugimoto et al., 1995). In this relationship, DNA enzymes with the greatest heteroduplex stability indicated by a low free energy of hybridization (calculated using the nearest neighbor method), was often found to have the greatest kinetic activity. The sensitivity to heteroduplex stability in most instances can be counterbalanced to some extent by increasing the arm length until the hybridization free energy decreases to a threshold level. At this point the heteroduplex stability is optimal for catalysis, and the enzyme activity can approach its maximum efficiency. Factors other than length that tend to increase the heteroduplex stability include the general GC content and specific pyrimidine's content of the DNA component (Ratmeyer et al., 1994; Sugimoto et al., 1995; Gyi et al., 1998). The influence of heteroduplex stability on the kinetic efficiency of the deoxyribozyme is probably derived from its effect on the K_M of the reaction. The inverse relationship between K_M and enzyme-substrate complex stability can be observed by increasing the substrate binding domain length such that the heteroduplex stability is increased, which usually causes the K_M to fall toward its minimum. The benefit to the overall kinetic efficiency obtained by increasing binding domain length, however, is limited by the adverse effect it has on catalytic turnover (indicated by the k_cat), which occurs when the enzymes increased affinity for the products slows down the catalytic cycle by reducing the rate of product release. In a more recent investigation of this behavior in reactions where the substrate binding domain length ranged between 4/4 and 13/13, the maximum overall efficiency (k_cat/K_M) under physiological reaction conditions was found with an arm length of between 8 and 9 bp (Santoro and Joyce, 1998).

2. Sequence Specificity. With the potential to bind any RNA sequence and cleave purine-pyrimidine junctions, the 10-23 DNA enzyme has unprecedented target site flexibility. However, despite the enormous capacity to cleave different sequences, the actual substrate specificity of an individual deoxyribozyme with defined RNA binding domains, appears to be very high. This ability to discriminate is particularly important in biological applications where unwanted side reactions between the deoxyribozyme and some closely related or unrelated substrate could be very undesirable. The issue of specificity was central in a study by Taira and coworkers (Kuwabara et al., 1997) comparing the in vitro cleavage activity of deoxyribozymes and hammerhead ribozymes targeting RNA derived from the junction of the bcr-abl fusion. In this system, deoxyribozymes had the advantage for two reasons: one, they were able to get closer to the junction because of their superior target flexibility; and two, their activity was more easily perturbed by mismatch and hence were less reactive with the RNA sequence from the wild type abl gene. The difference between the ribozyme and deoxyribozyme in this study was attributed to the lower stability of the DNA-RNA heteroduplex compared with the RNA-RNA homoduplex (Ota et al., 1998). Indeed when the heteroduplex stability was increased by lengthening the RNA binding domains, the specificity of the bcr-abl (b2a2)-cleaving deoxyribozymes also decreased slightly (Warashina et al., 1999). To achieve the desired specificity in the bcr-abl (b3a2) target, Wu et al. (1999) used deoxyribozyme arm length asymmetry (6/12 bp), such that the short arm is less likely to allow cleavage of bcr transcripts. However, the key point is not so much the difference in duplex stability, but the sensitivity to this difference, which the deoxyribozyme displays because it maintains relatively high activity at comparatively low heteroduplex stability. In our experience the kinetic efficiency of any given 10-23 deoxyribozyme seems to approach a maximum when the heteroduplex stability is just greater than a minimum threshold level. When the stability falls below this threshold by the introduction of a binding domain mismatch or truncation, the catalytic activity of the molecule becomes severely impaired. The influence of single base mismatch was demonstrated empirically at different positions of the substrate by the introduction of point mutations (Santoro and Joyce, 1998). In this analysis, any mismatch with the substrate was detrimental to the catalytic efficiency, although the extent of this effect varied substantially from one position to the next and between different types of mispairing. We have also examined the specificity of the 10-23 deoxyribozyme by observing its ability to discriminate between sequences that differ by as little as a single nucleotide polymorphism (Cairns et al., 2000b). In this experiment, reactions between deoxyribozyme and matching substrate sequences (derived from a polymorphic site in the L1 gene of six different clinically relevant HPV types) were compared with reaction in the unmatched substrates. In each case only the perfectly matched type-specific deoxyribozymes were capable of achieving substantial cleavage of the corresponding substrate despite the similarity between the different sequences. In each of these studies the specificity of cleavage was examined with respect to binding domain-substrate interactions where some mismatches, particularly those producing "wobble" pair, can be tolerated (Santoro and Joyce, 1998; M. J. Cairns et al., 2000a). If however, the difference between the target and nontarget substrate lies at the cleavage site, such that the purine-pyrimidine (R-Y) becomes R-R, Y-Y, or Y-R, then the deoxyribozyme would have no activity on the nontarget substrate.

3. Biological Activity. The ability of the 10-23 deoxyribozyme to specifically cleave RNA with high efficiency under simulated physiological conditions has fueled expectation that this agent may have useful biological application in a gene inactivation strategy. To explore this potential a number of groups (including our own) have attempted to examine the activity of deoxyribozymes in biological systems (summarized in Table 2). In our laboratory, we initiated experiments with deoxyribozymes targeting the viral sequences from the HPV16 E6 and E7 genes; and in smooth muscle cells (SMCs) with deoxyribozymes targeting the c-myc gene (Cairns et al., 1999; Sun et al., 1999). The biological activity of deoxyribozymes that cleave RNA derived from the bcr-abl fusion, Egr-1, huntingtin, the env gene of HIV and CCR5 chemokine receptor have also been examined (Dash et al., 1998; Goila and Banerjea, 1998; Santiago et al., 1999; Warashina et al., 1999; Wu et al., 1999; Zhang et al., 1999).

View larger version (12K): [in this window] [in a new window]   Fig. 5.   The 3'-3'-internucleotide linkage and its effect on the serum nuclease stability of the deoxyribozyme oligonucleotide. Panel A contains the structure of a normal 5'-3'-internucleotide linkage and a 3'-3'-linkage inversion. Panel B shows the stability profile of an unmodified oligonucleotide and its 3'-3'-inversion modified counterpart in human serum.

View larger version (12K): [in this window] [in a new window]   Fig. 6.   Dose response for deoxyribozyme Rs6 mediated suppression of serum-stimulated SV40LT-SMC smooth muscle cell proliferation. Cell growth suppression was determined by cell counting and expressed as percentage of the inactivated control oligonucleotide.

4. Target Site Selection. Underpinning the tremendous versatility of the 10-23 deoxyribozymes is its ability to bind a target RNA sequence via Watson-Crick base pair interactions. This means it has the potential to bind and cleave any RNA molecule both in vitro and in vivo. However, like other agents that function by hybridization with single-stranded RNA, the deoxyribozyme must compete with the targets own stable intramolecular base pairing, which forms its characteristic secondary structure. Fortunately the 10-23 deoxyribozyme cleavage sites are plentiful in most biological substrates and thus provide a host of opportunities to achieve maximum cleavage efficiency. Finding these sites in the target RNA that are amenable to efficient hybridization and cleavage is usually a difficult and time-consuming task (one neglected in most studies to date) involving empirical testing of many deoxyribozyme in long-folded RNA transcript in vitro. In an attempt to streamline this process, we developed a multiplex approach to target site selection, which allowed the simultaneous analysis of many different deoxyribozyme cleavage sites in a single reaction (Cairns et al., 1999). Using this strategy, the cleavage efficiency of 80 different deoxyribozymes targeting the entire E6 component of the full-length E6/E7 transcript from HPV16, was determined in a single experiment. Molecules with both efficient and inefficient RNA cleavage activity were then compared for their ability to inhibit E6 gene expression in a cell-free system. In this assay the most efficient cleavers were also the most efficient inhibitors of E6 gene expression. The deoxyribozyme target site selection strategy was also used to identify the most efficient cleavers of a full-length rat c-myc transcript. Again some of the more active molecules from 60 deoxyribozymes tested under multiplex conditions were compared with less active cleavers in their ability to suppress rat SMC proliferation. As c-myc gene expression is closely associated with proliferation in response to serum stimulation in this model, the level of post-treatment growth could be used as an indicator of anti-c-myc activity. Gene suppression in this system was found to correlate with deoxyribozyme activity in the multiplex cleavage assay, with the most efficient cleavers also having the greatest effect on SMC proliferation. The multiplex assay was found to be an effective screen for deoxyribozyme cleavage sites, because it could efficiently identify the molecules with high activity against long-folded substrate RNA. In addition to this application, the multiplex selection assay may also be useful for identifying target site accessibility for other RNA binding agents such as antisense oligonucleotides and ribozymes.

	IV. Other Catalytic Nucleic Acids

Top Previous Next References

Hammerhead and hairpin motifs have often found application in the design of trans-acting ribozymes. There are, however, several other naturally occurring ribozymes exhibiting useful activities, including those found within group I introns, ribonuclease P (RNase P) and a viroid-like human pathogen, the hepatitis delta virus (HDV).

Introns are noncoding sequences that interrupt the coding sequences of most eukaryotic genes. These must be removed, or "spliced", after transcription to allow expression of functional mRNA, rRNA, or tRNA molecules. In the case of group I introns, excision is mediated by the autocatalytic activity of the intronic sequences themselves. There are several hundred examples of group I introns, including those found in plant and fungal mitochondria, bacteriophage, eubacteria, and chloroplast tRNA. Despite considerable variability in size and sequence, group I introns have phylogenetically conserved secondary structures (Michel and Westhof, 1990) and a common reaction mechanism. The first confirmation of self-splicing activity came from work on ribosomal RNA genes from Tetrahymena thermophila (Cech et al., 1981). In vivo, these reactions occur with the assistance of protein factors that, in some cases (Lambowitz and Perlman, 1990), are encoded within the intron itself. Interestingly, the Neurospora CYT-18 protein, which is known to interact with the Neurospora large mitochondrial ribosomal RNA and ND1 group I introns, can substitute for an RNA domain in the Tetrahymena group I intron (Mohr et al., 1994). The Neurospora introns are able to form most of the RNA secondary structures required for activity but require the CYT-18 protein for in vitro and in vivo splicing activity. This convergence in function between RNA and protein factors perhaps signals a transition from the relatively simple self-splicing group I introns to the more complex splicing pathways seen in higher eukaryotes (Weiner, 1993).

The enzymatic activity of group 1 introns involves a two-step transesterification with a requirement for a divalent cation, such as magnesium, and a guanosine cofactor. The first step involves a nucleophilic attack of the guanosine cofactor 3'-hydroxyl group on the 5'-splice site forming a free 3'-hydroxyl on the 5'-exon. In a second step, this free hydroxyl group makes a nucleophilic attack on the 3'-splice site, releasing the intron as a circular molecule and leaving a ligated exon (Cech, 1990). Substrate specificity is determined by a sequence within the intron, the internal guide sequence (IGS), and there is a requirement for a U at position 1, relative to the cleavage site, paired with a conserved G in the IGS (Cech et al., 1992). Site-specific mutagenesis of the IGS has shown that other RNA sequences can be targeted. Truncated versions of the Tetrahymena group I intron have been shown to catalyze many trans reactions in vitro including endonuclease, ligase, nucleotidyl transferase, phosphatase reactions with RNA (Cech, 1990), and, in some cases, DNA substrates (Herschlag and Cech, 1990; Robertson and Joyce, 1990). In vitro evolution methodologies have been employed to improve or alter the catalytic requirements of the ribozyme (Green et al., 1990; Green and Szostak, 1992; Beaudry and Joyce, 1992; Lehman and Joyce 1993).

The ability of group I ribozymes to perform trans-splicing reactions in vitro suggests the possibility of therapeutic modification of disease-relevant RNA targets in vivo. In a model system, truncated lacZ transcripts have been corrected in Escherichia coli (Sullenger and Cech 1994) and mammalian cells (Jones et al., 1996). Closer to a therapeutic application, the L-21 group I ribozyme has been used to correct mutant ^s-globin mRNA transcripts by trans-splicing with the -globin 3'-exon in sickle cell anemia-derived erythroid lineage precursor cells (Lan et al., 1998). Since the efficiency of the trans-splicing group I ribozymes can approach 25-50% of targeted transcripts (Jones and Sullenger, 1997), this approach has the potential to correct inherited and other diseases caused by expression of mutant mRNAs. Some questions remain regarding the specificity of the reaction as other mRNAs within the cell can be targeted, perhaps because the IGS recognition component is only 6 nucleotides in length (Jones et al., 1996) (see Fig. 7). RNA repair, using group I ribozymes or other methods (Kmiec, 1999; Puttaraju et al., 1999) has significant advantages over gene replacement strategies, and avoiding the need for regulated transgene expression.

View larger version (15K): [in this window] [in a new window]   Fig. 7.   Schematic description of RNA repair mechanism using a group I intron ribozymes (Sullenger and Cech, 1994). A ribozyme with the 3'-end of wild-type transcript was designed based on group I intron model. After cleavage, the ribozyme mediated a transfer of its 3'-exon tag to the 5'-cleavage product; thus RNA repair was accomplished.

RNase P is a ribonucleoprotein involved in processing the 5' termini of tRNA precursors during their maturation. RNA cleavage is via nucleophilic attack on the phosphodiester bond leaving a 5'-phosphate and 3'-hydroxyl at the cleavage site, and there is an absolute requirement for divalent metal ions. The E. coli RNase P comprises an RNA domain of ~400 nucleotides (the M1 subunit) and a protein component of 14,000 Da dubbed C5. In vitro, the M1 RNA component has been shown to possess intrinsic catalytic activity; in vivo, however, there appears to be a requirement for C5 (Guerrier-Takada et al., 1983). The RNase P protein component is believed to facilitate binding between M1 RNA enzyme and tRNA substrate by masking electrostatic repulsion between enzyme and substrate RNAs (Gardiner et al., 1985; Reich et al., 1988). Comparisons between RNase P enzymes from diverse organisms reveals substantial sequence variation; however, there are similar core sequences and secondary structures that likely represent the catalytic domains (Frank and Pace 1998). Eukaryotic RNase P complexes are less well characterized, seem to have a higher protein content, and the RNA component has no apparent in vitro enzymatic activity (Tanner, 1999).

There is no absolute dependence on conserved sequences at the tRNA-precursor cleavage site, the most important domains within the substrate consist of the aminoacyl acceptor stem and T-stem with conserved GUUC sequence (Kahle et al., 1990; Yuan and Altman, 1995). The ubiquitous nature of RNase P, the lack of a requirement for specific nucleotide sequences for cleavage, and the inherent efficiency of utilizing a cellular enzyme, have created interest in directing RNase P-mediated cleavage to therapeutic target mRNAs. Both E. coli and human RNase P were shown to cleave a target RNA that resembles a tRNA substrate in a bimolecular reaction (Yuan et al., 1992). The 3'-proximal sequence of the stem functions to identify the target and is considered to act as an external guide sequence (EGS) that can be targeted to essentially any target RNA. For example, an EGS based on the E. coli tRNA^tyr and modified to hybridize to the bacterial chloramphenicol acetyltransferase mRNA sequence was capable of directing specific human RNase P-mediated cleavage of chloramphenicol acetyltransferase mRNA in vitro and in vivo (Yuan et al., 1992). A variation on this approach is to fuse an EGS to the bacterial M1 RNase P subunit itself, improving substrate binding and cleavage efficiency. A fusion between the M1 RNA and a sequence complementary to the herpes simplex virus I thymidine kinase (TK) mRNA was shown to efficiently cleave TK mRNA in vitro and reduce TK mRNA and protein levels by ~80% in transfected cells (Liu and Altman 1995). Direct expression of EGSs directed against a bacterial resistance gene (Guerrier-Takada et al.,1997) or essential viral genes (Plehn-Dujowich and Altman, 1998) has proven effective in functional cellular assays. Recently, deletion studies have defined that an EGS of only 30 nucleotides is sufficient for efficient cleavage in a biomolecular reaction in vitro (Werner et al., 1999). These results were generated using short, synthetic substrates, and it will be important to demonstrate similar findings when full-length mRNA transcripts are targeted either in vitro or in vivo. As with other approaches to gene suppression mentioned in this review, target selection, stability of synthetic molecules in biological fluids, and delivery of EGS, whether vector-based or as oligonucleotides, remain as challenges. Nevertheless, RNase P-mediated approaches to gene down-regulation offer considerable promise for genomic and therapeutic applications.

The hepatitis delta virus virus (HDV) is a single-stranded circular RNA virus of approximately 1700 nucleotides, which causes significant pathology in man (Lai, 1995). HDV infection requires the presence of the hepatitis B virion either coincident with HDV infection or as a pre-existing infection. Hence, HDV is designated as a satellite virus of hepatitis B. HDV contains two ribozymes, both required for RNA replication (Macnaughton et al., 1993), one on the genomic, infectious RNA strand, the other in the complementary region of the antigenomic strand. The nucleotide sequence and four stem (P1-P4), pseudoknot secondary structure of the HDV ribozymes is different than other ribozymes described so far. The predicted secondary structure has been largely confirmed by X-ray crystallographic studies at a resolution of 2.3 A (Ferré-D'Amaré et al., 1998). The self-cleaving activity of HDV RNA is enhanced in the presence of denaturants (Rosenstein and Been, 1990), which suggests that the active structures are present in nascent transcripts, not the mature, folded, genomic RNA. Substrate recognition requires formation of the P1 stem, comprising a GU wobble pair plus six nucleotides hybridized to the target, with cleavage occurring just 5' to the wobble base pair. The minimal ribozyme sequence is ~85 nucleotides, almost entirely 3' with respect to the cleavage site, useful for the generation of discrete RNA transcript termini or processing ribozyme multimers to monomers in cis-reactions (Tanner, 1999). Trans-cleaving HDV ribozymes have been designed that are active against oligonucleotide substrates; these include a P4-2 ribozyme (Branch and Robertson, 1991; Perrotta et al., 1993; Wu et al., 1993), a P1 ribozyme (Perrotta and Been, 1993), a circular ribozyme (Perrotta et al., 1993; Puttaraju et al., 1993), and a hybrid of both genomic and antigenomic HDV ribozymes (Been et al., 1992). However, the modest activity of trans-cleaving HDV ribozymes against long substrates (Roy et al., 1999a) limits their utility for therapeutic applications.

	V. PharmacologyThe Key to Applications

Top Previous Next References

The most challenging aspect of the use of catalytic nucleic acids in a pharmaceutical role is delivering these molecules to their site of action, the RNA target located within the cytoplasm or nucleus. Delivery requires that the oligonucleotide survive local or systemic administration long enough to bind to the target cells, cross the cytoplasmic membrane or become released from an endosomal/lysosomal vesicle, pass through the nuclear membrane (or cross the endoplasmic reticulum), and be able to functionally hybridize to the target RNA. This represents a long chain of events that are at best unlikely for a typical conventional drug let alone a large, labile, highly charged molecule (Liang et al., 1999). A range of viral vector constructs have been designed to express ribozymes endogenously within target, and potentially successful ex vivo applications have been described (Sun et al., 1995; Bramlage et al., 1998; Wang et al., 1998). The in vivo application DNAzymes and chemically synthesized ribozymes is presently restricted to exogenous delivery although relatively little work in this area has been described. Studies of exogenously delivered oligonucleotides have been dominated for several years in studies of antisense oligonucleotide therapy. The pharmacological issues critical to the delivery of catalytic DNA and RNA can be extrapolated from the wealth of data collected from studies of the pharmacology of antisense drugs (Crooke, 1998; Bennett, 1998; Juliano et al., 1999). By drawing on this wide body of data and including recent studies of ribozyme and DNAzyme delivery, it is possible to discuss how issues such as stability, toxicity, immunology, in vivo and cellular pharmacokinetics, and the employment of delivery agents are challenges common to all forms of oligonucleotide-based therapy.

As discussed in Section III, a wide range of chemical modifications have been employed to enhance the stability of synthetic ribozymes and DNAzymes to facilitate their in vivo application. For antisense oligonucleotides, the first generation phosphorothioate (PS) deoxyoligonucleotides have been studied extensively and are being evaluated in a number of clinical trials against a range of targets (Hogrefe, 1999; Persidis, 1999). However the stability offered by the PS backbone is only partial because oligonucleotides prepared by nonstero-controlled methods contain mixtures of the nuclease-susceptible configuration and the resistant