Optimizing the cell efficacy of synthetic ribozymes. Site selection and chemical modifications of ribozymes targeting the proto-oncogene c-myb.

Expression of the proto-oncogene c-myb is necessary for proliferation of vascular smooth muscle cells. We have developed synthetic hammerhead ribozymes that recognize and cleave c-myb RNA, thereby inhibiting cell proliferation. Herein, we describe a method for the selection of hammerhead ribozyme cleavage sites and optimization of chemical modifications that maximize cell efficacy. In vitro assays were used to determine the relative accessibility of the ribozyme target sites for binding and cleavage. Several ribozymes thus identified showed efficacy in inhibiting smooth muscle cell proliferation relative to catalytically inactive controls. A combination of modifications including several phosphorothioate linkages at the 5′-end of the ribozyme and an extensively modified catalytic core resulted in substantially increased cell efficacy. A variety of different 2′-modifications at positions U4 and U7 that confer nuclease resistance gave comparable levels of cell efficacy. The lengths of the ribozyme binding arms were varied; optimal cell efficacy was observed with relatively short sequences (13-15 total nucleotides). These synthetic ribozymes have potential as therapeutics for hyperproliferative disorders such as restenosis and cancer. The chemical motifs that give optimal ribozyme activity in smooth muscle cell assays may be applicable to other cell types and other molecular targets.

Since the discovery that certain naturally occurring RNA motifs were capable of catalytically cleaving other RNA molecules in a sequence-specific manner, extensive studies have defined the sequence and structural characteristics that control the in vitro specificity and kinetics of these RNA enzymes or ribozymes (1)(2)(3)(4). Ribozymes have a broad range of potential in vivo applications. These include the use of ribozymes as research tools for probing molecular mechanisms, the use of ribozymes to genetically engineer crops, and the use of ribozymes as therapeutics for human or animal diseases. Each of these applications requires that a ribozyme function efficiently within the intracellular environment. The sequence and structural features that promote optimal intracellular activity of ribozymes are currently under study.
Several factors are likely to contribute to the intracellular efficacy of a ribozyme. A ribozyme must colocalize with its molecular target in the appropriate cellular compartment and must be present at sufficiently high concentration to promote hybridization. In addition, its catalytic cleavage rate must be fast enough, and its half-life must be long enough to allow cleavage of a substantial fraction of the target mRNA population. Finally, the cleavage site in the target mRNA must be accessible to ribozyme binding. When the ribozyme is made synthetically, a variety of modifications can be introduced to increase its half-life within the cell, to change its target sequence binding affinity, and possibly also to alter its intracellular trafficking properties. In this study, we have used chemically synthesized hammerhead ribozymes targeting the protooncogene c-myb to study different chemical modifications and sequence changes that affect cell efficacy. Expression of c-myb is necessary for cell-cycle progression in vascular smooth muscle cells (5,6). Therefore, we have used proliferation of rat aortic smooth muscle cells as a measure of the efficacy of the ribozymes targeting c-myb (7).
Unmodified RNA is subject to rapid nuclease degradation upon exogenous delivery to cells or tissues. For example, the half-life of an all RNA hammerhead ribozyme in human serum is less than 0.1 min (8). In the literature there are several reports of exogenously delivered synthetic ribozymes showing efficacy in cell culture (9 -12). Often, these studies have utilized DNA/RNA chimeric ribozymes to enhance resistance to exonucleases, leaving large regions of unmodified RNA susceptible to endonucleolytic degradation. Extensive modification of the hammerhead ribozyme motif can give dramatic enhancement of the ribozyme half-life in biological fluids (8,13). Modifications of this type have been demonstrated to give efficacy in cell culture (7) and in vivo (13,14).
We have previously described dose-dependent inhibition of smooth muscle cell proliferation by select ribozymes targeting c-myb. Little inhibition was observed with catalytically inactive control RNA molecules. The active ribozymes reduced the level of the target c-myb RNA (7). Thus, the smooth muscle cell assay is a suitable means of measuring ribozyme efficacy in cells. Here, we report on a systematic method for determining accessible ribozyme target sites and for determining the optimal hammerhead ribozyme arm length required for cell efficacy. In addition, we explore the effect of chemical modifications such as those reported by Beigelman et al. (8) on cell efficacy. We have thereby identified chemical motifs that maximize the potency and specificity of synthetic ribozymes delivered exogenously to cells in culture.

MATERIALS AND METHODS
Ribozyme Synthesis and Sequences-Ribozymes were synthesized and purified as described (15)(16)(17). The sequences and modifications of all of the active ribozymes used in this study are shown in Fig. 1. The cleavage site numbering is based on the human DNA sequence numbering (Genbank accession number X52125; transcription starts at * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. nucleotide 198). The inactive ribozymes contain identical binding arms and chemical modifications except that positions G5 and A14 in the catalytic core were changed to 2Ј-O-methyl uridine, thereby eliminating catalytic activity while maintaining nuclease resistance. The catalytic cleavage activity of all of the ribozymes was confirmed on a matched short substrate by standard methods; inactive ribozymes did not show detectable cleavage activity (data not shown).
Template RNA Transcription-A murine c-myb cDNA clone was obtained from Dr. Premkumar Reddy. Full-length c-myb RNA was prepared by T7 transcription. Reactions contained 40 mM Tris, pH 8. Reactions were stopped by addition of formamide gel-loading buffer (95% formamide, 0.1% bromphenol blue, 0.1% xylene cyanol, 20 mM EDTA) and electrophoresed on a 6% denaturing acrylamide gel. Gels were dried and quantified on a phosphorimager.
Proliferation Assay-Cell proliferation assays were performed as described previously (7). Briefly, cells were serum-starved, followed by addition of ribozymes complexed with 7.2 g/ml LipofectAMINE (3.6 M DOSPA 1 ). Following a 3-4-h uptake period, ribozyme-lipid complexes were washed out, and cells were stimulated with fetal bovine serum. The percentage of proliferating cells was measured by bromodeoxyuridine incorporation. Control wells were treated with lipid only and stimulated with growth medium containing either 10% fetal bovine serum or 0% fetal bovine serum. Treatment with lipid alone did not affect proliferation compared to untreated wells (data not shown). All conditions were run in duplicate. Each experiment was performed two to five times, and a representative experiment is presented.

RESULTS
Target Site Selection Strategy-Hammerhead ribozymes can recognize and cleave RNA sequences containing U followed by A, C, or U. This consensus sequence occurs very frequently; the murine c-myb mRNA contains nearly 500 potential hammerhead ribozyme cleavage sites. We wished to develop a systematic method of selecting sites that were amenable to ribozymemediated cleavage in vivo. Several criterion must be considered in selecting sites for optimal cleavage. First, the site must be accessible to ribozyme binding. In the intracellular milieu, different cleavage sites are likely to vary significantly in their accessibility to ribozyme binding. Site accessibility is probably determined by both secondary structure in the mRNA and regions of protein binding. In addition, the ribozyme itself must fold into the correct conformation required for binding and cleaving its substrate. Sequences in the substrate binding arms of the ribozyme can affect its propensity to fold correctly and cleave its target (18).
Hammerhead ribozyme cleavage sites having a high degree of homology between human (19) and murine (20) c-myb se-quences were identified. Using a computer folding algorithm, we eliminated from consideration ribozymes that showed a high probability of forming undesirable intramolecular secondary structure (21). 26 sites were selected for testing in an oligonucleotide binding assay. We designed DNA oligonucleotides spanning the cleavage sites, annealed these to a fulllength in vitro transcript of murine c-myb, and incubated in the presence of RNase H. This enzyme recognizes DNA/RNA hybrids and cleaves the RNA transcript at the site of hybridization. Sites that are single stranded and thus readily accessible to oligonucleotide binding are expected to show high levels of cleavage in this assay. Although the structure of the in vitro transcript may not mimic exactly the structure of the RNA in the cell, we hoped that this approach would at least allow us to identify and avoid regions of prominent local structure. The ribozymes are shown in Fig. 1, and the cleavage results are shown in Table I.
Ribozymes targeting the most accessible sites, as determined by the RNase H assay, were synthesized and tested for their ability to cleave the in vitro transcript under ribozyme excess (single turnover) conditions (Table I). In general, the sites that were accessible to oligonucleotide binding in the RNase H assay were also susceptible to cleavage by ribozymes. There were some exceptions, such as sites 839 and 1017, in which the ribozyme cleavage was relatively inefficient. In addition, there were sites such as 1943 and 575 that showed relatively modest RNase H accessibility yet showed highly efficient cleavage by ribozymes. To fully assess the predictive capacity of the RNase H accessibility screen, one would need to test more sites that were poorly accessible based on the RNase H criterion (e.g. 1334, 1389, etc.) as well as those that scored well in terms of accessibility. In any case, several candidate ribozymes (indicated by asterisks) were selected for testing in cell culture based on ribozyme in vitro cleavage activity.
Two regions of rat c-myb cDNA, encompassing the sites of interest, were sequenced. 2 A single nucleotide difference between rat and murine c-myb was found in the 575 site; the other four sites were conserved. Active and inactive forms of the ribozymes targeting the five sites within rat c-myb were synthesized with an unmodified RNA catalytic core and five 2Ј-O-methyl residues in each binding arm. The inactive ribozymes contained two nucleotide changes in the catalytic core that eliminate cleavage activity. The effect of the ribozymes on proliferation of rat aortic smooth muscle cells was assessed (data not shown). Although four of the five active ribozymes did show statistically significant inhibition of proliferation, the degree of inhibition was relatively low (approximately 15-30% inhibition by the active ribozyme compared to the inactive ribozyme control). Poor performance in cell culture could be the result of rapid intracellular degradation. Beigelman et al. (8) have reported extensive modifications designed to enhance the nuclease resistance of hammerhead ribozymes while retaining catalytic activity. We decided to test such modifications, focusing on the ribozyme targeting site 575.
Effect of Backbone and 2Ј-Sugar Modifications on Cell Culture Efficacy- Fig. 2 shows the site 575 with and without a nuclease-stable core and with and without phosphorothioate linkages in the binding arms. All of the modifications showed some efficacy by active ribozyme versus inactive control, suggesting that the inhibitory effect was mediated by ribozyme cleavage of c-myb RNA. For both the U4 2Ј-C-allyl and the U4,U7 2Ј-amino variants, the ribozymes containing both a "stabilized" core and phosphorothioate linkages gave the most enhanced inhibition of smooth muscle cell proliferation. The in-active U4 2Ј-C-allyl phosphorothioate ribozyme control showed no inhibition. The inactive U4,U7 2Ј-amino phosphorothioate ribozyme did show some inhibition, although less than its active counterpart.
Comparison of 5Ј-End versus 3Ј-End Modifications-The re-sults in Fig. 2 indicated that both a nuclease-resistant core and phosphorothioate linkages in the binding arms were advantageous for obtaining maximum cell culture efficacy. Since phosphorothioate linkages are associated with some degree of cytotoxicity and nonspecific effects (22,23), we wished to determine the minimum number of phosphorothioates sufficient for cell efficacy. Fig. 3A shows a comparison of ribozymes containing 5 phosphorothioate linkages at the 5Ј-end, 5 phosphorothioate linkages at the 3Ј-end, or 5 phosphorothioate linkages at both the 5Ј-and 3Ј-ends. The ribozyme containing phosphorothioates only at the 3Ј-end showed almost no efficacy, while the ribozyme containing phosphorothioates at the 5Ј-end showed equivalent efficacy to that containing phosphorothioates at both the 5Ј-and 3Ј-ends. In this experiment, the inactive ribozyme showed some inhibition relative to the vehicle-treated control. A ribozyme with scrambled sequence binding arms exhibited an equivalent degree of inhibition, indicating that this effect was not mediated by ribozyme binding but was truly a "nonspecific" effect on proliferation. Next, we compared ribozymes with varying numbers of phosphorothioates at the 5Ј-end (Fig. 3, B and C). The degree of efficacy gradually decreased as the number of phosphorothioate linkages was reduced. From these experiments, we concluded that four to five phosphorothioate linkages at the 5Ј-end gives optimal efficacy in this cell culture system. The ribozymes used in this study contained either 3Ј-phosphorothioate linkages or a 3Ј-3Ј "inverted thymidine" modification (24) to protect against 3Ј-exonuclease activity (8). We have subsequently shown that the outcome of this assay is not particularly sensitive to the presence or absence of this 3Ј-protect-  ing group. Anti-c-myb ribozymes containing various protecting groups including a 3Ј-3Ј inverted thymidine, a 3Ј-3Ј inverted abasic residue, a 3Ј-butanediol or no 3Ј protecting group at all showed equivalent efficacy in inhibiting smooth muscle cell proliferation (data not shown). This may indicate that over the timecourse of this assay, the additional stability conferred by these modifications is not significant.
Optimization of Binding Arm Length-Ribozymes targeting c-myb site 575 were synthesized with arm lengths ranging from 5 to 12 nucleotides. The effects of these ribozymes on cell proliferation are shown in Fig. 4. The data are presented as specific inhibition by active versus inactive ribozyme. The optimal arm length was 6 to 7 nucleotides. We confirmed in five separate experiments that there was no significant difference in efficacy between the 6/6 (StemI/StemIII) and 7/7 arm ribozymes (data not shown). We also tested ribozymes with asymmetric arm lengths (StemI/StemIII with 5/10 or 10/5 nucleotides). Both asymmetric variants performed similarly to the 7/7 symmetric ribozyme (data not shown). The symmetric ribozyme containing seven nucleotide binding arms (no. 2972) was used as a standard for comparison in the experiment that follows. We have shown previously that this ribozyme inhibits rat, pig, and human vascular smooth muscle cells in a dose-dependent fashion and that the inhibition of proliferation correlates with a reduction in c-myb RNA levels (7).
Effect of 2Ј-Sugar and Base Modifications-Beigelman et al. (8,16,17) have developed a broad spectrum of different modifications in the catalytic core of hammerhead ribozymes that enhance resistance to nuclease degradation while preserving significant catalytic activity. Differences in intracellular stabil-ity, cleavage rate, uptake, and localization properties conferred by these modifications could alter the cell culture efficacy of the ribozyme. We tested several of these modifications, as shown in Fig. 5. A number of modifications showed equivalent cell efficacy to that exhibited by the U4 2Ј-C-allyl modified ribozyme. Others, such as U7 2Ј-C-allyl and U4 2Ј-fluoro, showed a somewhat lower magnitude of inhibition at the 100 nM dose. Experiments performed at lower doses supported the conclusion that none of the variants differed by more than 2-fold in the dose required to achieve 50% inhibition (data not shown). The inactive ribozyme containing the U4,U7 2Ј-amino modification . U4 C-allyl and U4 C-allyl p ϭ S indicate U4 2Ј-C-allyl "stabilized" cores without and with phosphorothioate linkages at the 5Ј-and 3Ј-ends (2321 and 2550, respectively). U4,7 NH 2 and U4,7 NH 2 p ϭ S indicate U4 and U7 2Ј-amino "stabilized" cores without and with phosphorothioate linkages at the 5Ј-and 3Ј-ends (2320 and 2547, respectively). Data are expressed as proliferation relative to the serum-stimulated control. Relative proliferation is calculated as follows: (% proliferation with ribozyme Ϫ % basal proliferation)/(% proliferation with serum Ϫ % basal proliferation) ϫ 100; error bars represent the range of duplicate wells.

FIG. 3. Optimization of phosphorothioate content.
Ribozymes targeting c-myb site 575 were complexed with LipofectAMINE and delivered to rat aortic smooth muscle cells at a 50 nM dose. Each of the ribozymes contained the U4 2Ј-C-allyl "stabilized" core and varying amounts of phosphorothioate. Active and inactive versions of the ribozymes were tested. A, 10 p ϭ S has five phosphorothioates at both the 5Ј-and 3Ј-ends (2550); 5Ј p ϭ S has five phosphorothioates at the 5Ј-end only (2826); 3Ј p ϭ S has five phosphorothioates at the 3Ј-end only (RPI 2828). B, a series of ribozymes containing a decreasing number of 5Ј-phosphorothioate linkages were compared. 5 p ϭ S (2826), 4 p ϭ S (2972), 3 p ϭ S (2974), 2 p ϭ S (2976), or 1 p ϭ S (2978). C, another experiment comparing 10 p ϭ S (2550), 5 p ϭ S (2826), or 4 p ϭ S (2972). The results are expressed as proliferation relative to the serum-stimulated control as described in Fig. 2; error bars represent the range of duplicate wells. showed greater inhibition than inactive ribozymes containing any of the other modifications. 3 This was seen with the related ribozyme in Fig. 2 as well. We believe that this inhibition represents a truly nonspecific effect because controls using a ribozyme containing the U4,U7 2Ј-amino modified chemistry and scrambled binding arm sequences showed similar levels of inhibition in side-by-side comparisons (data not shown). The results in Fig. 5 indicate that none of the modified chemistries results in an improvement in cell efficacy in the smooth muscle cell proliferation assay compared to the U4 2Ј-C-allyl modification. DISCUSSION We have used a systematic method to identify c-myb ribozymes that inhibit cell proliferation when delivered exogenously to cultured vascular smooth muscle cells. The inhibition is mediated by ribozymes containing a catalytically active core, while inactive controls fail to inhibit. The degree of inhibition observed can be affected profoundly by both backbone and 2Ј-sugar modifications. The optimal ribozyme configuration for inhibition of cell proliferation in this system consists of four phosphorothioate linkages at the 5Ј-end, six or seven nucleotide binding arms, 30 2Ј-O-methyl residues, and any of a variety of 2Ј-sugar or base modifications at positions U4 and U7.
The results in Figs. 2 and 3 demonstrate that both a "stabilized" core and several 5Ј-phosphorothioate linkages are advantageous for achieving substantial ribozyme-mediated inhibition of smooth muscle cell proliferation. Beigelman et al. (8) have shown that the U4-C-allyl and the U4,U7-amino ribozyme motifs increase the serum half-life of the ribozyme Ͼ16,000fold relative to an all RNA ribozyme. Addition of 5Ј-phosphorothioate linkages increase the resistance to 5Ј-exonuclease activity. 4 Thus, the enhanced cell efficacy observed with the combination of these two types of modifications could be attributed solely to enhanced resistance to nuclease. However, we cannot rule out the possibility that the phosphorothioate moieties are advantageous for some additional reason, such as conferring altered intracellular localization or trafficking properties. 5 The optimal hammerhead ribozyme binding arm length for intracellular activity may be a function of the thermodynamic stability of the duplex formed upon substrate binding, the kinetics of substrate binding and release, or of competing intramolecular ribozyme structures that can compromise substrate binding or conformation of the catalytic core. We have found that there is a distinct arm length optimum of six or seven nucleotides in each arm for the ribozyme targeting c-myb site 575. Depending on the sequence of the binding arms, the optimal arm length could vary for ribozymes targeting different sites.
We have shown that a variety of different chemical modifi- 3 Although the majority of experiments performed with this ribozyme chemistry showed greater inhibition by the active version compared to the inactive, we have occasionally observed that inhibition by the inactive ribozyme is equal to that of the active, especially at higher doses Thackray, D. T. Stinchcomb, T. C. Jarvis, and N. Usman, manuscript in preparation. 5 LipofectAMINE delivery results in virtually 100% of the cells taking up ribozyme, with a fairly homogeneous distribution within the population, as demonstrated by flow cytometry using a fluorescently labeled ribozyme (26). In addition, uptake studies using radioactive ribozymes show that the sheer number of ribozymes delivered to each cell exceeds the c-myb mRNA copy number by many orders of magnitude (data not shown). Therefore, a very small percentage of the ribozyme that is taken up could represent the "bioactive" fraction responsible for the observed efficacy. Although ribozyme intracellular localization can be studied using confocal microscopy, or by careful fractionation, it is difficult to establish a meaningful correlation between cell efficacy and the observed localization of the bulk population.  Fig. 1). 5/5, 3862; 6/6, 3206; 7/7, 2972; 8/8, 3212; 10/10, 3210; 12/12, 3208. Data are presented as specific inhibition by active ribozyme relative to inactive ribozyme. Specific inhibition is calculated as follows: (% proliferation with active ribozyme Ϫ % basal proliferation)/(% proliferation with inactive ribozyme Ϫ % basal proliferation) ϫ 100. Error bars represent the standard error of three to five separate assays. cations at positions U4 and U7 can be tolerated while preserving a similar level of cell efficacy. The in vitro catalytic cleavage activities of the ribozymes were determined in single turnover assays on short substrates (data not shown). Although the ribozymes shown in Fig. 5 are all catalytically active, the kinetics of substrate cleavage in vitro varies. For example, the cleavage rate for the U4,U7-amino c-myb site 575 ribozyme is at least 10-fold higher than that of either the U4-C-allyl or the U4 6-methyl-U ribozymes. Despite the lower cleavage activity, both the U4-C-allyl and the U4 6-methyl-U ribozymes perform as well or better than the U4,U7-amino ribozyme in inhibiting cell proliferation. Thus, there does not appear to be a strict correlation between the k obs measured in vitro and the level of efficacy measured in cell culture in this system. 6 This may indicate that some other step besides cleavage is rate limiting in the cell culture assay. Alternatively, the buffer conditions used for the in vitro cleavage assays may exacerbate differences between ribozymes that are less significant within the intracellular milieu.
The modified ribozymes used in this study specifically inhibit cell proliferation. The combination of efficacy and resistance to nucleolytic degradation suggests that these ribozymes could have therapeutic utility in treating diseases resulting from inappropriate overexpression of c-myb. There are several disease conditions in which the disease pathology correlates with up-regulation of c-myb. For example, in restenosis, vascular smooth muscle cells hyperproliferate to form a neointima following coronary angioplasty, causing re-occlusion of arteries in 30 -40% of patients (27)(28)(29). In the rat carotid artery model of restenosis, antisense DNA oligonucleotides targeting c-myb inhibit neoinitima formation following balloon injury (30). In addition, c-myb has been implicated in the pathology of various cancers including melanoma (31), leukemia (32), and lymphosarcoma and colon carcinoma (25).
The optimized chemical modifications reported here represent general motifs that can be applied to ribozymes against other molecular targets. For example, synthetic ribozymes targeting the matrix metalloproteinase stromelysin and containing both the U4 2Ј-C-allyl and the U4,U7 2Ј-amino modifications have shown activity in an animal model of arthritis (13). We are continuing to explore new types of ribozyme modifications, as well as a variety of different formulations that can potentially enhance ribozyme delivery, efficacy, and residence time in cells and tissues.