RNA-Protein Interaction AN ANALYSIS WITH RNA OLIGONUCLEOTIDES OF THE RECOGNITION BY a-SARCIN OF A RIBOSOMAL DOMAIN CRITICAL FOR FUNCTION*

alpha-Sarcin is a cytotoxic protein that inactivates ribosomes by hydrolyzing a single phosphodiester bond on the 3' side of G-4325 in eukaryotic 28 S rRNA. We have examined the requirements for the recognition by alpha-sarcin of this domain using a synthetic oligoribonucleotide (35-mer) that reproduces the sequence and, we presume, the secondary structure (a stem, a bulged nucleotide, and a loop) at the site of modification. The wild type structure and a large number of variants were transcribed in vitro from synthetic DNA templates with phage T7 RNA polymerase. Recognition of the substrate is strongly favored by a G at the position that corresponds to 4325. There is an absolute requirement for a helical stem; however, it can be reduced from the 7 base pairs in the natural structure to 3 without loss of specificity. The nature of the base pairs in the stem modifies but does not abolish recognition; whereas, the bulged nucleotide does not contribute to identification. Cleavage is materially affected by altering the nucleotides in the universal sequence surrounding G-4325 and changing the position in the loop of the tetranucleotide GAG(sarcin)A leads to loss of recognition by the toxin. We propose that the alpha-sarcin domain RNA participates in elongation factor catalyzed binding of aminoacyl-tRNA and of translocation; that translocation is driven by transitions in the structure of the alpha-sarcin domain RNA initiated by the binding of the factors, or the hydrolysis of GTP, or both; and that to toxin inactivates the ribosomes by preventing this transition.

a-Sarcin is a cytotoxic protein that inactivates ribosomes by hydrolyzing a single phosphodiester bond on the 3' side of G-4325 in eukaryotic 28 S rRNA. We have examined the requirements for the recognition by cY-sarcin of this domain using a synthetic oligoribonucleotide (35-mer) that reproduces the sequence and, we presume, the secondary structure (a stem, a bulged nucleotide, and a loop) at the site of modification. The wild type structure and a large number of variants were transcribed in vitro from synthetic DNA templates with phage T7 RNA polymerase.
Recognition of the substrate is strongly favored by a G at the position that corresponds to 4325. There is an absolute requirement for a helical stem; however, it can be reduced from the 7 base pairs in the natural structure to 3 without loss of specificity. The nature of the base pairs in the stem modifies but does not abolish recognition; whereas, the bulged nucleotide does not contribute to identification.
Cleavage is materially affected by altering the nucleotides in the universal sequence surrounding G-4325 and changing the position in the loop of the tetranucleotide GAG(sarcin)A leads to loss of recognition by the toxin. We propose that the Lu-sarcin domain RNA participates in elongation factor catalyzed binding of aminoacyl-tRNA and of translocation; that translocation is driven by transitions in the structure of the a-sarcin domain RNA initiated by the binding of the factors, or the hydrolysis of GTP, or both; and that the toxin inactivates the ribosomes by preventing this transition.
cu-Sarcin is a small, basic, cytotoxic protein produced by the mold Aspergillus giganteus (1,2) that inhibits protein synthesis by inactivating ribosomes (3)(4)(5). The molecular basis of the inhibition is the hydrolysis of a phosphodiester bond (6)  In the original sequence determination it was G-4325 (33); in the corrected seqience itwas G-4327 (8); in the drawings of the secondary structure of 28 S rRNA it is G-4326 (8,34). For simplicity we retain the designation G-4325.
a single-stranded loop 459 residues from the 3' end of 28 S rRNA (7,8). The cleavage site is embedded, in a purine-rich single-stranded segment of 14 nucleotides that is near universal (8,9). This is one of the most strongly conserved regions of rRNA and, indeed, the ribosomes of all the organisms that have been tested, including the producing fungus, are sensitive to the toxin (10). cr-Sarcin catalyzes the hydrolysis of only the one phosphodiester bond and this single break accounts entirely for its cytotoxicity (10). This remarkable specificity is peculiar to cY-sarcin; treatment of ribosomes with other ribonucleases causes extensive digestion of rRNA. The finding that cleavage of a single phosphodiester bond in the ol-sarcin domain inactivates the ribosome implies that this sequence is crucial for function since ribosomes ordinarily survive treatment with nucleases despite many nicks in their RNA (11); indeed, some organisms physiologically divide their 2% rRNA into domains; in Trypanosomes, for example, the fragmentation is into two large and four small molecules (12). Thus, intact rRNAper se is not essential for protein synthesis. The presumption that the a-sarcin region of 28 S rRNA is critical for ribosome function has gained considerable reinforcement from the elucidation of the mechanism of action of ricin (13-16). Ricin, which is among the most toxic substances known, is a RNA N-glycosidase and the single base that is depurinated is A-4324 in 28 S rRNA, i.e. the nucleotide adjacent to the oi-sarcin cut site (13-16).
Until recently little was known of the function of individual ribosomal components or even of ribosomal domains. None of the ribosomal proteins or nucleic acids have activity when separated from the particle. In the beginning it was thought that the rRNAs only provided a scaffolding to support the ribosomal proteins which catalyzed the partial reactions of protein synthesis. However, the pendulum has swung in the other direction. Now the rRNAs are envisioned as being responsible for the basic biochemistry of protein synthesis, for the binding of aminoacyl-tRNA, mRNA, and the initiation, elongation, and termination factors; for peptide bond formation; and for translocation. The ribosomal proteins, which are presumed to be a later evolutionary embellishment, are thought to facilitate the folding and the maintenance of an optimal configuration of the rRNA (17), perhaps, in this way conferring on protein synthesis speed and accuracy.
The value that derives from an analysis of the mechanism of action of antibiotics and of toxins that affect ribosomes is in concentrating attention on regions where efforts to comprehend functional correlates of structure are likely to be rewarded. There are good reasons to suspect that the cu-sarcin/ ricin domain is involved in EF-l*-dependent binding of ami- The Substrate Specificity of cu-Sarcin noacyl-tRNA to ribosomes and EF-2 catalyzed GTP hydrolysis and translocation.
These are the partial reactions most adversely affected by cu-sarcin (3) and by ricin (18), respectively, and cleavage at the cu-sarcin site in Escherichia coli 23 S rRNA interferes solely with the binding of EF-Tu and . Moreover, EF-Tu and EF-G footprint in the cu-sarcin/ ricin domain (20). EF-Tu protects only four nucleotides in prokaryotic 23 S rRNA against chemical modification and these correspond in eukaryotic 28 S rRNA to A-4324 (ricin), G-4325 (Lu-sarcin), and G-4319 and A-4329 which latter are also in the universal sequence (20). EF-G also protects only four nucleotides and three are the same as the ones protected by EF-Tu; they are the bases that correspond to G-4319, A-4324, and G-4325 (20).
We have undertaken to determine how ol-sarcin recognizes a single phosphodiester bond in a particular domain in rRNA. This is part of an effort, unfortunately but necessarily oblique, to understand how ribosomal proteins which like cu-sarcin are small and basic recognize specific sites in rRNA. An RNA oligonucleotide, a 35-mer, was prepared using a synthetic DNA template and phage T7 RNA polymerase (21); the oligoribonucleotide has the sequence and should have the secondary structure (a stem, a bulged nucleotide, and a loop) of the cu-sarcin region of 28 S rRNA. This synthetic oligoribonucleotide is cleaved by oc-sarcin at a position that corresponds to G-4325, precisely where the toxin hydrolyzes the nucleic acid in intact ribosomes (16).
We have now systematically altered nucleotides in the (Ysarcin domain RNA. The aim is a more precise definition of the sequence of nucleotides and of the higher order structure that prescribes the binding of the protein to the RNA and allows the catalysis of hydrolysis. The determinants of recognition that we have examined are: 1) nucleotide specificity at the site of covalent modification; 2) the necessity for a bulged nucleotide in the helical stem; 3) the requirement for the stem itself; 4) the influence of context, i.e. the importance of the nucleotides in the universal sequence in the loop in which the cY-sarcin guanosine is located; and, finally, 5) the effect of the position in the loop of the tetranucleotide GAG((rsarcin)A.

EXPERIMENTAL PROCEDURES
General-The preparation of the RNA oligonucleotides using synthetic DNA templates, phage T7 RNA polymerase, 5'-[w3*P]ATP, and unlabeled GTP, UTP, and CTP, and the purification by polyacrylamide gel electrophoresis of the radioactive oligoribonucleotides were described before (16,21).

Assay of the Effect of wSarcin on Synthetic Oligoribonucleotides-
The RNA oligonucleotides (in concentrations ranging from 1.08 to 2.26 pM) were heated at 90 "C for 2 min in 10 mM Tris-HCl, pH 7.6, 50 mM NaCl, and 10 mM MgCI;! and allowed to renature at 0 "C. The concentrations of the substrate oligoribonucleotides varied with the number of labeled adenosines since the amount of radioactivity was kept constant. After renaturation appropriate amounts of cu-sarcin, water, buffer (50 mM Tris-HCl, pH 7.6; 250 mM KCl), and EDTA (in an amount sufficient to chelate the magnesium ions in the renaturation buffer) were added to give a final concentration of 15 mM Tris-HCl, pH 7.6, 15 mM NaCI, 50 mM KCl. The concentrations of the substrate RNAs and of cr-sarcin (specified in the figure legends) were chosen so as to approximate those that attend on cleavage of 28 S rRNA in intact ribosomes (22). Incubation was for 20 min at 20 "C. The reaction was stopped by addition of an equal volume of 178 mM Tris-HCI, pH 8.3, 178 mM boric acid, 5 mM EDTA, containing 0.05% bromphenol blue and saturated with urea; an aliquot was analyzed by electrophoresis for 3 h at 1.5 kV in 20% polyacrylamide gels containing 7 M urea (16); the oligonucleotides were visualized by autoradiography.

RESULTS
The Nucleotide Specificity at the Site of Covalent Modification of RNA by a-Sarcin-A radioactive oligoribonucleotide that reproduces the sequence and the secondary structure of the Lu-sarcin domain in 28 S rRNA was prepared using a synthetic DNA template and phage T7 RNA polymerase (Fig.  IA). The three guanosines at the 5' end of the 35-mer derive from the T7 promoter; they are not easy to be rid of, although they make no effective contribution to the identification of the substrate by the toxin. This has been established by chemical synthesis of an oligoribonucleotide that lacks only these three nucleotides (a 32-mer) and the demonstration that it is cleaved specifically by cY-sarcin with normal efficiency (data not shown).
We refer to the oligoribonucleotide (the 35-mer in Fig. LA) that reproduces the ol-sarcin domain in 28 S rRNA as wild type. Treatment of the oligomer with cu-sarcin in concentrations ranging from 2.94 x lOma to 2.94 x 10d6 M led to the formation of two fragments; one of 21 nucleotides derived from the 5' side of the substrate and the other of 14 nucleotides from the 3' end (Fig. 1B and Ref. 16). This is the same specificity as the toxin has with ribosomes (16) The estimate of sensitivity is an approximation since the reaction rate in these experiments is pseudo first-order with respect to a-sarcin concentration (22). The symbols are: C-O, wild type 35mer from A; l ---0, 34-mer lacking the wild type bulged nucleotide at position 6 (cf. Fig. 3 about 85% of the substrate is cleaved (Fig. lB, lane 4, and Fig. 1C). At higher concentrations of a-sarcin, 2.94 x 10m6 M or greater, cleavages occur at appreciably all the purines and it is no longer possible to distinguish between specific and nonspecific hydrolysis (Fig. lB, lane 5). The extent of the nonspecific cleavage cannot be accurately evaluated from these gels since only the adenosines in the substrate are radioactive. We do not know what it is that distinguishes the specific from the nonspecific action of cu-sarcin. Position 21 in the wild type substrate (corresponding to G-4325 in 28 S rRNA) was systematically varied. An oligoribonucleotide with a transition of the G to an A was a less suitable substrate for cY-sarcin; cleavage was decreased to 35% of the control (Fig. 2B). Transversions of the wild type G to U (Fig. 2C) or to C (Fig. 20) reduces hydrolysis to approximately 17 and to l%, respectively. (The comparison, by densitometry of the radioautograph, was of the results in the experiments with the lowest concentrations of cy-sarcin (5.88 x 1Om8 M), i.e. of lanes 2 in A and B and of lanes 2 and M in C and D (Fig. 2)). Thus, there is strong but by no means absolute dependence on preservation of the G at the site of covalent modification; the preference is G > A > U >> C. What is notable is that cu-sarcin which has been presumed to be a purine-specific nuclease (22) has activity with pyrimidines. The results prejudice one to consider structure rather than setiuence as the more important determinant of specificity.
Implicit in the interpretation of these and subsequent experiments is the assumption that the decrease in sensitivity to the hydrolytic effect of ol-sarcin is predominantly if not exclusively on K,,, rather than KC,,. We are testing the assumption.
The Requirement for a Bulged Nuckotide in the Stem of the a-Sarcin Domain-The cu-sarcin domain RNA has a canonical protein binding structure: a stem, a loop, and a bulged nucleotide. The last occurs in a number of ribosomal protein binding sites, albeit the bulged nucleotide is usually an A rather than the U that is found here (23). We suspected that the bulged U at position 6 in the substrate (position 4310 in 28 S rRNA) is not necessary for cY-sarcin action since it does not occur at the comparable site in E. coli 23 S rRNA (24), although the bacteria's ribosomes are sensitive to the toxin (6,25). Still we tested the possibility directly by synthesizing a variant oligoribonucleotide (a 34-mer) that lacked the bulged U in the stem. Just as we suspected the variant is as sensitive to the toxin as the wild type substrate (Figs. 3 and 1C). Thus, a bulged nucleotide is not required for recognition by a-sarcin. Witherell and Uhlenbeck (26) found that a bulged adenosine in the bacteriophage Qp genomic RNA was not necessary either for association with the coat protein.

Sarcin-To
test the importance of the stem for cu-sarcin action we constructed a linear molecule (35-mer) that retained the 17 nucleotides in the loop, including the universal purine-rich sequence of 14 bases, but with 5' (11 nucleotides) and 3' (7 nucleotides) ends altered so that they would not pair. The The Substrate Specificity of a-Sarcin linear molecule migrated slower during electrophoresis in 20% polyacrylamide-urea gels than the wild type oligomer (Fig. 4, compare lanes M and l), which is consistent with the former lacking the organized secondary structure of the latter. a-Sarcin did not cleave the linear substrate specifically; there was no formation of the 21 and 14 nucleotide fragments that mark specific hydrolysis by the toxin (Fig. 4). However, there was nonspecific digestion as is apparent from the progressive disappearance of the radioactive oligoribonucleotide with increasing concentrations of a-sarcin (Fig. 4); indeed, the linear RNA may be more sensitive to this nonspecific effect of the toxin than the one with higher order structure.
Having established that the stem is required, we inquired as to the number of base pairs in the helix that are needed. Successive additional base pairs and the bulged uridine were deleted from the wild type oligoribonucleotide (Fig. 5A). Omission of nucleotides 4, 6, and 34, leaving 6 pairs (Fig. 5B); 4-6 and 34, 35, leaving 5 pairs (Fig. 5C); 4-7 and 33-35, leaving 4 pairs (Fig. 50); and 4-8 and 32-35, leaving 3 pairs (Fig. 5E), yielded variant oligoribonucleotides all of which are recognized specifically by a-sarcin. It is possible that the helix is necessary only to tether the ends of the loop and in nature may be longer and hence more stable than is required for recognition by a-sarcin; we presume it is neither longer nor more stable than is required for its contribution to the function of this ribosomal domain in protein synthesis.
There is evidence that the nature of the base pairs in a RNA helix, i.e. GC as opposed to AU, can affect the structure of an included loop (27). For this reason we have begun to systematically change those in our wild type substrate (Fig.  6A). The variants we have assessed so far are: 1) CG to AU at positions 11 and 29 at the base of the loop (Fig. 6B); 2) GC to UA at positions 10 and 30 (Fig. 6C); 3) UA to GC at positions 9 and 31 (Fig. 6D); and 4) CG to GU at positions 11 and 29 (Fig. 6E), this last is a change to a noncanonical pair. All are recognized specifically by a-sarcin; however, the change in the second base pair (Fig. 6C) or the substitution of a GU for a GC pair at the base of the loop (Fig. 6E) decreases sensitivity to the toxin. The alterations may affect recognition by modifying the conformation of the loop.
The Requirement for Ribonucleotides for Recognition by a-Sarcin-Synthetic oligodeoxynucleotides corresponding to E. The Substrate Specificity of a-Sarcin coli tRNAPh' or tRNAL"" are recognized by their cognate synthetases (28). This led us to synthesize a DNA oligomer (32-mer) that corresponds to the sequence of the wild type asarcin RNA; in this instance the three 5' guanosines were omitted and deoxythymidine replaced ribouridine. This synthetic oligodeoxynucleotide was not cleaved specifically by asarcin (data not shown). However, there was nonspecific digestion and, as we had observed before (22), the hydrolysis of the DNA required magnesium; nonspecific cleavage of RNA by cY-sarcin on the other hand is inhibited by magnesium (22).
The Effect of the Context of the a-Sarcin Site Guanosine on Recognition-The purpose was to evaluate the contribution to recognition of the nucleotides in the single-stranded region surrounding the cY-sarcin site guanosine (G-21 in the synthetic oligonucleotide).
It was anticipated that alterations in the 5' adjacent adenosine would not have an appreciable effect since depurination of A-4324 by pretreatment with ricin did not affect subsequent cleavage by a-sarcin at G-4325 in the same ribosomes (29). Nonetheless, a series of variants were constructed with alterations of the ricin site A to G, U, or C (Fig.  7). As we had expected all were recognized by a-sarcin with specificity and normal efficiency (Fig. 7).
The context was changed in another way: the tetranucleotide GAG(sarcin)A was left intact for reasons that will be apparent shortly and the remainder of the universal portion of the lbop sequence was engineered so it was entirely uridines (Fig. 8A)  The wild type (1.58 pM) and the mutant (1.75 pM) oligoribonucleotides with alterations at A-20 (the structure of each is given) were incubated for 20 min at 20 "C without (lanes I) or with a-sarcin: 5.88 SB). Neither oligonucleotide is a competent substrate for asarcin (Fig. 8). Thus, this part of the context is an essential feature of the recognition of the substrate by cu-sarcin. The Contribution of the Position of the Loop Tetranucleotide GAGA to Recognition-E.
coli ribosomes are not sensitive to ricin (30); the ribosomes are not inactivated and A-2660 in the a-sarcin/ricin domain is not depurinated (15). However, naked E. coli rRNA is a substrate for the N-glycosidase activity of the toxin (14). The two sites of covalent modification, one each in 16 and 23 S rRNAs, have stems with 7 base pairs and loops that have the sequences GAGA. There are sites with this structure, however, that are not modified by ricin. A comparison of the two subsets, modified and unmodified by ricin, indicated that the tetranucleotide GA(ricin)G(cY-sarcin)A in the loop had to have a particular position with respect to the stem. In 28 S rRNA the ricinmodified adenosine in the loop has an equal number of nucleotides on either side of it and, hence, in two dimensions is centered over the stem.
In agreement with a prediction that comes from this observation, if the tetranucleotide GAGA is moved either four (Fig.  9B) or two (Fig. 9C) positions closer to the 5' end or two positions (Fig. 9D) closer to the 3' end the specific response to cu-sarcin is entirely lost; only the nonspecific hydrolytic effect on purines is seen. The results with the variant in which the GAGA is moved two positions closer to the 5' end ( Fig.  9C) are instructive.
Recognition of the original guanosine would have yielded fragments of 19 and 16 nucleotides from the 5' and 3' sides of the substrate. Oligonucleotides of this length were not found (Fig. 9C). In this variant there is still a guanosine at position 21 as in the wild type substrate (cf. in the sequence. We cannot be sure from these experiments that ol-sarcin is appreciating the position or the geometry of the tetranucleotide, since changing the location of the GAGA in the loop also alters the surrounding sequence and we have already seen that that context affects identification (Fig. 8).
Lu-Sarcin appears to appreciate the structure of the loop and this structure is no doubt affected by the nucleotide sequence; nonetheless, the position of the tetranucleotide GAGA may have a special influence on loop structure.

DISCUSSION
The conclusion from these experiments is that specific recognition of rRNA by a-sarcin requires, in the first instance, a stem and a loop but that a bulged nucleotide is not necessary. There is a strong preference for a guanosine at the site of covalent modification in the loop and the context surrounding this nucleotide, i.e. the 14-base universal sequence, affects binding of cy-sarcin and catalysis. The exception is the immediate 5' adenosine which has no influence on identification of the RNA. The stem is absolutely essential; however, only 3 of the 7 base pairs found in 28 S rRNA are needed; the identity of the pairs modifies but does not abolish recognition. It is possible that the helical stem contributes to recognition only indirectly by tethering the ends of the loop and conferring on the latter a specific conformation. Finally, the position of the tetranucleotide GAG(sarcin)A in the loop conditions rec-ognition perhaps directly, perhaps indirectly by altering the context.
Although, the problem has not been addressed directly in these experiments the results have bearing on the important question of the function of the cu-sarcin domain in 28 S rRNA. There are earlier observations that are consistent with the proposition that the structure of this region is complex and capable of reversible alterations (13,16,31). We cite now others: oligodeoxynucleotides complimentary to the universal sequence in the loop of the cu-sarcin domain will not bind to either E. coli (32) or rat ribosomes3 suspended in buffer, suggesting, but by no means proving, that the structure is not simply single-stranded. Occlusion by ribosomal proteins could account for the failure of the cDNA to bind to the site; however, this seems less likely since oc-sarcin and ricin have access to the domain. It is most important in this regard that if ribosomes are catalyzing protein synthesis they will bind the complimentary oligodeoxynucleotide (manifest as sensitivity to ribonuclease H) suggesting a reversible change in the structure of the domain.4 A similar observation has been made with E. coli ribosomes.5 Further support for the proposal comes from experiments with inhibitors of protein synthesis (cycloheximide, GMP-PNP, and sparsomycin) in which (Ysarcin is used as a probe of structure." These indicate that the oc-sarcin domain RNA alternates between open and closed states during translation since ribosomes are sensitive to (Ysarcin only when peptidyl-tRNA is in the A-site prior to translocation. The interpretation we favor is that during translocation the loop goes from a more complex structure, possibly one involving a tertiary interaction, to a simple single-stranded one. The process, if it occurs, could provide the motive force for transitions in the structure of ribosomal 60 S subunits or in the relationship of the large to the small subunit and might underlie the movement required for the translocation of peptidyl-tRNA from the A-to the P-site and for the movement of mRNA one codon after each of the reiterative rounds in translation. In this paradigm it is the elongation factors EF-1 and EF-2 which, directly or indirectly (i.e. indirectly through the binding or the hydrolysis of GTP), initiate the reversible transition or switch in rRNA structure that propels translocation. Cleavage at G-4325 in 28 S rRNA by cu-sarcin or depurination at A-4324 by ricin might abolish the capacity to reversibly switch structures and account in this way for the catastrophic effect of the toxins on ribosome function.