RNA aptamers that bind to and inhibit the ribosome-inactivating protein, pepocin.

Pepocin, isolated from Cucurbita pepo, is a ribosome-inactivating protein (RIP). RIPs site-specifically recognize and depurinate an adenosine at position 4324 in rat 28 S rRNA, rendering the ribosome incapable of interacting with essential elongation factors. Aptamers that target pepocin were isolated from a degenerate RNA pool by in vitro selection. A conserved hairpin motif, quite different from the sequence of the toxin-substrate domain in rat 28 S rRNA, was identified in the aptamer sequences. The aptamers selectively bind to pepocin with dissociation constants between 20 and 30 nM and inhibit the N-glycosidase activity of pepocin on rat liver 28 S rRNA. Competitive binding experiments using aptamer variants suggest that the conserved hairpin region in the anti-pepocin aptamer binds near the catalytic site on pepocin and prevents the interaction of pepocin and 28 S rRNA. Anti-RIP aptamers have potential use in diagnostic systems for the detection of pepocin or could be used as therapy to prevent the action of pepocin in mammalian cells.

Ribosome-inactivating proteins (RIPs) 1 are widely distributed throughout nature (for review, see Ref. 1). They inhibit protein synthesis in mammalian cells by disabling the ribosome. Type II RIPs (e.g. ricin) are heterodimeric proteins; an A-chain catalyzes the depurination of a specific adenosine in rRNA, and a disulfide linked B-chain binds cell surface galactosides (for review, see Ref. 2). Binding of the B-chain to the cell surface results in translocation of the A-chain to the cytosol, where it hydrolyzes the N-glycosidic bond of a universally conserved adenosine in the large subunit of rRNA. Cleavage of the adenine residue disrupts the binding of elongation factors, inactivating the ribosome. Type I RIPs, of which pepocin is a member, are considerably more numerous in nature but much less cytotoxic because they lack the B-chain required for cell entry (3).
The depurination site in the large subunit of rRNA is highly conserved in many eukaryotes and is located in a hairpin loop of 17-mer containing a GAGA tetraloop, in which the second A is removed (4,5). Recent studies using RNA variants of the toxin-substrate domain have shown that small hairpin fragments containing a GAGA tetraloop are also able to act as substrates for ricin (6). The cleavage reaction is remarkably specific, and the GAGA sequence is essential for toxin recognition. For instance, the A-chain of ricin does not recognize hairpin variants with all of the possible transitions and transversions of each nucleotide in the GAGA tetraloop sequence (6). NMR studies indicate that RNA fragments corresponding to the toxin-substrate domain of rRNAs form the GAGA tetraloop hairpin (7)(8)(9)(10). The structure of the GAGA tetraloop is similar to those of stable GAAA and GCAA (GNRA) loops, whereas these are not recognized by the toxins (6). These results illustrate the importance of both sequence and structure for recognition in RNA-protein complexes; studies of the contacts between RNA hairpins and RIPs are informative as a model system for scrutinizing RNA-protein interactions.
In vitro selection can be used to generate nucleic acid ligands (aptamers) that bind to target proteins and thus can answer questions regarding which components of an RNA molecule are essential for binding to its target protein (11)(12)(13). For instance, crystal structures of RNA aptamers that bind MS2 coat protein shed light on the nature of specific interactions between coat protein and operator (14,15). Despite the structural differences between the aptamers and the wild type operator, the aptamers bind in the same location on the coat protein as the wild type RNA and maintain many of the same contacts with the protein compared with the wild type RNA. In a similar example, RNA aptamers that bind elongation factor Tu (EF-Tu) were isolated by in vitro selection (16). EF-Tu is known to interact with an RIP recognition domain in rRNA. A consensus sequence identified within the aptamers was found in the toxinsubstrate domain of Thermus thermophilus 23 S rRNA.
To identify novel anti-RIP compounds, we used in vitro selection to produce RNA ligands that bind the toxin, pepocin, which is a type I RIP isolated from the sarcocarp of Cuburbita pepo (17). Because the single A-chain of pepocin contains Nglycosidic cleavage activity against the large subunit of rRNA but is not highly toxic, it was considered to be a suitable model system for the study of RIP recognition and inhibition.

EXPERIMENTAL PROCEDURES
Materials-Pepocin and gypsophilin were isolated from the sarcocarp of C. pepo. (17) and from Gypsophila elegans (18), respectively. The purity of proteins was checked by SDS-polyacrylamide gel electrophoresis with silver staining. Oligodeoxyribonucleotides were synthesized using standard phosphoramidite chemistry on a DNA synthesizer (392, Perkin-Elmer). [␣-32 P]UTP and [␥-32 P]ATP were purchased from Amersham Pharmacia Biotech (Uppsala, Sweden). RNase T1 and RNase V1 were purchased from Amersham Pharmacia Biotech, nuclease S1 was from Takara (Tokyo, Japan), and RNase I was from Promega (Madison, WI).
In Vitro Selection-In vitro selection was carried out using an RNA pool that contained 30 randomized nucleotide positions (19). The RNA pool was transcribed from the DNA template using an Ampliscribe T7 * 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.
in vitro transcription kit (Epicenter Technologies, Madison, WI). RNA was dissolved in 10 mM Tris-HCl (pH 7.6), 50 mM NaCl, and 10 mM MgCl 2 and then heated at 75°C for 3 min and cooled to room temperature. To exclude filter-binding RNA sequences from the pool, the RNA was passed over a 0.45-m HAWP filter (Millipore, Bedford, MA) two or three times before incubation with target. The RNA was mixed with a pepocin preparation (10 mM sodium phosphate (pH 6.5), 100 mM NaCl, 0.5 mM dithiothreitol, and 40% glycerol). The final solution contained 3 mM sodium phosphate, 3 mM MgCl 2 , 45 mM NaCl, 2.5 mM EDTA, 3 mM Tris-HCl (pH 7.5), 0.15 mM dithiothreitol, and 12% glycerol. Table I summarizes the conditions used in each round of the selection. The binding was carried out for 1 h at 37°C. After 1 h, the solution was vacuum filtered over a HAWP filter at 5 p.s.i. and washed five times with 0.2 ml of the selection buffer. The RNA on the filter was eluted twice with 0.2 ml of 7 M urea, 100 mM sodium citrate (pH 5.0), and 3 mM EDTA for 5 min at 100°C, and the eluted RNAs were precipitated with isopropyl alcohol. In the fifth through eighth rounds, an additional filtration step was carried out on the eluted RNA pool to further exclude filter-binding species (20 Cloning and Sequencing-The polymerase chain reaction products of the eighth round pool were ligated into a TA cloning vector (Invitrogen, Carlsbad, CA) and cloned into Escherichia coli Inv␣FЈ (Invitrogen). Plasmid DNAs were isolated and sequenced using an AutoRead sequencing kit and a DNA sequencer (ALF, Amersham Pharmacia Biotech).
Filter Binding Assays-The RNA pools from each round and individual clones from the eighth round were internally labeled with [␣-32 P]UTP (3000 Ci/mmol) during T7 transcription. The labeled RNA (0.3 M final concentration) was mixed with pepocin (0.3 M final concentration) in 0.1 ml of the selection buffer for 1 h at 37°C. The solution was then filtered and washed three times with selection buffer. The filter was exposed to a PhosphorImager plate, and the amount of retained radioactivity was determined using a PhosphorImager (Fuji Photo Film, Kanagawa, Japan). The fraction of bound RNA was calculated by comparing the radioactivity on the filter with the total radioactivity of the input RNA.
For the determination of dissociation constants, individual RNA clones were end labeled with [␥-32 P]ATP after 5Ј-dephosphorylation of   (19,21). The radioactivity on both filters was quantified, and dissociation constants (K d ) were determined by curve fitting with the program Kaleidagraph (Abelbeck Software, Reading, PA).
Binding Competition Assays-The binding activity of each aptamer variant was determined by competition experiments between a 5Јlabeled aptamer variant and 5Ј-labeled clone 8-09. Both RNAs (each at 0.45 M final concentration) were mixed with pepocin (0.3 M final concentration) in 100 l of the selection buffer for 1 h at 37°C. The mixture was passed over a HAWP filter and washed three times with 0.2 ml of the selection buffer. The RNAs on the filter were eluted with 0.1 ml of 0.025% of bromphenol blue, 7 M urea, 100 mM sodium citrate (pH 5.0), and 3 mM EDTA for 5 min at 100°C. The solution was applied directly to a gel containing 7 M urea. The radioactivity in each band on the gel was quantified using a PhosphorImager. Binding ratios were calculated by comparing the counts in the retained and input RNA.
T m Measurements-For T m measurements, each aptamer (0.7 A 260 final concentration) was dissolved in 0.25 ml of a buffer containing 10 mM sodium phosphate (pH 7.0), 1 M NaCl, and 1 mM EDTA. Before measurement, the aptamer solution was heated at 75°C for 3 min and then cooled to room temperature. Melting profiles were obtained at 260 nm using a spectrophotometer (DU650, Beckman) at a heating rate of 0.5°C/min with 1-cm cuvettes. T m values were calculated from the first derivatives of the melting curves.

RESULTS AND DISCUSSION
Selection of Anti-pepocin Aptamers-We carried out selection experiments using an RNA pool that contained a region of 30 randomized nucleotides flanked by constant regions. The relatively short length of the randomized region was chosen because several plant toxins recognize and hydrolyze a specific adenine in small synthetic hairpin fragments (4,22). In addition, RNA pools containing relatively short randomized regions have frequently proven sufficient to identify aptamers that bind tightly to a target protein. For example, the same 21residue motif was found in anti-tyrosine phosphatase aptamers isolated from two pools that contained random regions that were either 30 or 71 nucleotides in length (23).
Selection experiments were initiated with an RNA pool containing 2 ϫ 10 13 different sequences. Table I  increased from 0.03 to 24%. Although some filter-binding species accumulated during the middle rounds of the selection, postfiltration of selected RNAs successfully removed the filterbinding species.
Sequence variations observed among the 23 similar clones are consistent with the formation of a stem-loop structure. For example, although 7 ligands have a CAG/CUG sequence in stem 2, 12 ligands change this stem sequence to GGA/UUC, 2 ligands to AGG/CUU, and one ligand to AAG/CUU. Therefore, most ligands have an NUAU loop and two bulged cytidines, one between stems 1L and 2L, and one between stems 2L and 3L (Fig. 2). Interestingly, the sequence of the anti-pepocin aptamer hairpin motif is quite different from the toxin-substrate domain GAGA in rRNAs.
Based on sequence differences in stem 1L, the anti-pepocin aptamers can be classified into two groups, class 1 and class 2. Class 1 ligands have a UAG or a UCG sequence and a mismatched base pair in stem 1L. In contrast, class 2 ligands have a UUG sequence in this region and form an uninterrupted stem 1 structure. Thus, class 1 aptamers may form a similar but slightly different hairpin structure than class 2 aptamers. Fig.  2 summarizes the structural variations in each class and the results of binding assays with pepocin. It was observed that a G-U base pair in stem 2 of both class 1 and class 2 aptamers produced better binding than canonical base pairs. The fact that class 1 and class 2 aptamers, and individual aptamers within class 1 and class 2, were likely derived from different initial molecules further emphasizes the importance of the non-canonical loop structure for pepocin binding.
Minimal Anti-pepocin Aptamers-To determine the minimal sequence necessary for anti-pepocin aptamer binding, we prepared various mutant fragments of the class 1 aptamer 8-09, which bound the best in our initial assays. To quantify the binding abilities of each fragment better, competition experiments were carried out. Selected small RNA fragments were competed with 8-18, an aptamer that bound specifically but poorly to pepocin. A 45-mer fragment of aptamer 8-09 (termed 9-45), which contained the AUAU loop, stems 1-4, and the two bulged cytidines competed roughly twice as well with 8-18 for binding to pepocin (Fig. 3a). The binding activity of aptamer 9-45 was improved slightly by introducing substitutions (UAA to AAG) that closed the apparent internal loop of aptamer 8-09, generating aptamer 9-45b. The fully paired stem was observed in many other class 1 aptamers such as 8-08 and 8-01. A 41-mer fragment, 9-41, had binding activity similar to that of 9-45b, but binding activity decreased after additional truncation (9-37 (37-mer) and 9-33 (33-mer)) (Fig. 3b). These results are consistent with the predicted secondary structure, which shows 10 base pairs between stems 3 and 4. Most of the aptamers isolated from the eighth round pool were also predicted to contain approximately 10 base pairs in stems 3 and 4. It should be noted that all of the aptamers bound much better to pepocin than the isolated toxin-substrate domain of rRNA; for example, aptamer 9-45b binds roughly 100-fold better than the isolated toxin-substrate domain.
There is some sequence degeneracy at specific positions in the loop and stem regions of the aptamers. Four different bases appeared at the first position of the loop, although the UAU motif was highly conserved at the remaining three positions in the loop. We examined the binding abilities of these variants by introducing substitutions into the minimal aptamer 9-41. The binding activities of these variants are shown in Fig. 4. As has been observed in other selection experiments (11), sequence variations between the aptamers provided valuable information about the function. The order of the binding activity of the class 1 loop sequence variants was UUAU Ͼ CUAU Ͼ AUAU Ͼ GUAU. Similarly, 6ix of 13 class 1 NUAU loops are UUAU, 3 are CUAU, 4 are AUAU, and no GUAU loops appear. The tightest binding aptamer was the class 1 aptamer (9-41U22) containing a UUAU sequence in the loop. The K d of 9-41U22 was 17.9 Ϯ 2.2 nM, which was similar to that of the full-length aptamers 8-09 (21.0 Ϯ 2.0 nM) and 8-14 (23.4 Ϯ 2.3 nM) (Fig. 5a).
Changes outside the loop region also affected activity. The FIG. 5. Binding curves for anti-pepocin aptamers with pepocin or gypsophilin. Different concentrations of protein were mixed with a fixed concentration (0.6 nM) of aptamer at 37°C for 1 h, and the percent binding was determined using a filter binding assay. a, binding curves for aptamers with pepocin. The calculated dissociation constants for aptamers 8-14, 8-09, and 9-41U22 were 23 Ϯ 2, 21 Ϯ 2, and 18 Ϯ 2 nM, respectively. b, binding curves for aptamer 8-09 with pepocin or gypsophilin. The pepocin binding curve is the same as in a.
second base pair in stem 1L was predicted to be A-A or C-A in class 1 aptamers but U-A in class 2 aptamers. The substitution of the class 1 C-A base pair to a class 2 U-A (9-41U20) decreased binding activity slightly. Interestingly, the class 1 aptamer, 9-41 (T m ϭ 82.5°C) was found to be more thermally stable than the class 2 aptamer, 9-41U20 (T m ϭ 80.0°C). Finally, both bulged cytidines located in the stem region appear to be important for binding. Substitution of the bulged C to either U or A reduced the binding activity significantly. Overall, it appeared that the sequences in and around stems 1 and 2 contributed to interactions with pepocin.
Mapping the Secondary Structure of the Aptamers-Although all sequence and binding data were consistent with the predicted secondary structures, we enzymatically probed the structures of both class 1 (9-41) and class 2 (9-41U20) aptamers. Labeled 9-41 and 9-41U20 were treated with RNase V1, S1 nuclease, and RNase I (Fig. 6). RNase V1 digests doublestranded regions in the RNA specifically. S1 nuclease and small amounts of RNase I preferentially digest single-stranded regions in RNAs.
Both 9-41 and 9-41U20 gave similar digestion patterns, with the exception of RNase V1 digestion in the stem 1 region (positions 18 -20 and 26 -28 in Fig. 6) and RNase I digestion throughout stem 1 and the loop regions (position [17][18][19][20][21][22][23][24][25][26][27]. The digestion pattern of 9-41U20 was consistent with the predicted tetra loop (the NUAU loop) hairpin structure. However, 9-41 may have a structure different from that of 9-41U20 at positions 18 -28, including the loop and stems 1 and 2. RNase V1 digestion of 9-41 did not give strong bands at position 18 -20 in the stem 1L region compared with 9-41U20. In contrast, RNase I extensively digested 9-41 throughout the loop and the stem 1 regions, although S1 nuclease treatment gave similar digestion patterns for 9-41 and 9-41U20. Under the conditions that we used, the RNase I activity was expected to be slightly stronger than nuclease S1 activity. Therefore, although enzymatic probing confirms that class 1 and class 2 aptamers form secondary structures similar to those predicted, the class 1 aptamers seem to have more flexible stem 1 and loop regions.
Inhibition of Pepocin Activity-To determine whether the anti-pepocin aptamers could inhibit ribosome inactivation by pepocin, we examined the depurination of the specific adenosine in rat liver 28 S rRNA. Rat liver ribosomes and pepocin were incubated together in the presence of increasing concentrations of aptamers. The 28 S rRNA was extracted from the ribosome and treated with aniline to cleave the rRNA at the depurination site. The aniline-treated rRNA was analyzed on a denaturing gel (Fig. 7), and the expected cleavage product of 28 S rRNA was observed (Fig. 7, lane 2).
The addition of a class 1 aptamer, 8-09, in 10,000-fold excess over pepocin inhibited the depurination of 28 S rRNA (Fig. 7a,  lane 9). The inhibitory effect of a class 2 aptamer, 8-14, was weaker than that of class 1 and was observed at around a 25,000-fold excess (Fig. 7a, lane 6). Although the small fragments 9-41U22 and 9-41U20U22 bound tightly to pepocin, they showed no inhibitory effect when present in ten-thousand fold excess (Fig. 7b).
Mechanism of Toxin Inhibition-The simplest explanation of anti-pepocin aptamer inhibition of rRNA depurination is that the aptamer competes with rRNA for binding to pepocin. However, to account for the large excess of aptamer needed for inhibition, we considered the possibility that the aptamer might also be a substrate for pepocin. To determine if pepocin FIG. 6. Enzymatic probing of 9-41 and 9-41U20 using RNase V1, S1 nuclease, and RNase I. a, limited digestion patterns of 5Ј-labeled 9-41 (lanes 1-7) and 9-41U20 (lanes 8 -14). Lanes 1 and 8, intact RNAs; lanes 2 and 9, alkaline digestion; lanes 3 and 10, digestion with RNase T1; lanes 4 and 11, with RNase V1 (0.05 unit); lanes 5 and 12, with RNase V1 (0.1 unit); lanes 6 and 13, with S1 nuclease (1 unit); lanes 7 and 14, with RNase I (0.3 unit). b, digestion patterns on the secondary structure of 9-41. c, digestion patterns mapped onto the secondary structure of 9-41U20. depurinated one or more positions in the aptamers, resulting in a loss of activity, the aptamers were incubated with pepocin for 1 day at 37°C followed by treatment with aniline. No depurination or cleavage products were observed (data not shown).
Because the sequence of the toxin-substrate domain in 28 S rRNA is highly conserved in many eukaryotes, if the antipepocin aptamers mimicked the rRNA domain then it seemed possible that the anti-pepocin aptamers might bind and inhibit other RIPs. We examined the ability of these aptamers to bind to and inhibit another RIP, gypsophilin, isolated from G. elegans (18). Aptamers 8-09, 8-14, and 9-41U20 did not bind to gypsophilin (Fig. 5b), and gypsophilin activity was not inhibited by the addition of the aptamers in up to a 25,000-fold excess over protein (data not shown). Thus, anti-pepocin aptamers are remarkably selective for pepocin.
Thus, although the anti-pepocin aptamers compete with rRNA for binding to pepocin, they do not appear to do so by mimicking universal features of rRNAs. In concert with our other results, two additional features of the mechanism of binding and inhibition are suggested. First, the fact that the class 1 aptamers form a semistable stem-loop structure that nonetheless binds tightly and specifically to pepocin suggests that the conformation of the aptamers may change on binding pepocin. This hypothesis is also consistent with the observed specificity of the aptamers for pepocin.
Second, it is possible that the anti-pepocin aptamers bind adjacent to but do not necessarily fill the pepocin active site. This assumption is supported by the fact that the full-length class 1 aptamer, 8-09, inhibits pepocin activity, whereas the small fragment, 9-41U22, of the class 1 aptamer is not effective even though its K d is smaller than that of 8-09. Partial occlusion of a protein active site by an aptamer has been observed previously for protein tyrosine phosphatases (23).
Along these lines, it is also possible that two RNA recognition sites exist in pepocin; one site could bind efficiently to a portion of a RNA substrate, and the other is the catalytic site that hydrolyzes the adenine residue in rRNA. This interpretation is consistent with the fact that small synthetic RNA hairpins bind poorly and are not hydrolyzed by pepocin. This model would be similar to the two-site recognition known to occur with other ribosome-associating proteins such as elongation factors EF-G and EF-Tu (24). EF-Tu and EF-G interact with both the thiostrepton binding site in the 1067 region and the toxin recognition site in the 2660 region of 23 S rRNA. From this vantage, the selected aptamers might block the hypothesized substrate binding site but not necessarily the catalytic site.
Applications-The identification of anti-RIP aptamers may suggest new approaches to the development of therapeutics or diagnostics for toxic agents. To the extent that the anti-pepocin aptamer has been shown to inhibit RIP activity, it provides proof of principle for the development of prophylactic or therapeutic aptamers to other cytotoxins, such as the Vero toxin produced by E. coli O157 (25)(26)(27). Such aptamers could be introduced as expression constructs into exposed or at-risk individuals to minimize toxin pathology. Moreover, anti-toxin diagnostics using aptamers might be powerful tools for detecting trace amounts of toxins. For example, anti-pepocin aptamers might be used in biosensors that would detect the presence of pepocin in foods made from pumpkin.