J. Mol. Biol. (1997) 266, 877±890 Analysis of Interactions Between the Codon ± Anticodon Duplexes within the Ribosome: Their Role in Translation Valery I. Lim Institute of Protein Research Russian Academy of Sciences 142292, Pushchino, Moscow Region, Russia Computer graphics simulation of interactions between the codon ±anticodon duplexes formed by normal elongator tRNAs at the ribosomal A, P and E-sites (the AP and PE interduplex interactions) was made. This demonstrated that only the correct duplexes at the A-site are compatible with the AP interduplex interaction. The selection of synonymous codons and anticodon wobble bases, together with the AP interduplex interaction, prevents frameshifting. In the absence of this interaction the ef®ciency of the selection falls off sharply. This suggests that the AP interduplex interaction should be retained during translocation and in the post-translocation state, i.e. the PE interduplex interaction that is identical with that of AP should exist to avoid frameshifting. In such a model the P-site duplex provides an indirect linkage between the A and E-site duplexes. The indirect linkage prohibits the simultaneous existence of the A and E-site duplexes. The wobble pairs of the P and E-site duplexes can affect the rate of the A-site occupation via the AP interduplex interaction and the AE interduplex indirect linkage. It is demonstrated that frameshifting can occur from the AP or PE codon± anticodon complex destabilization caused, for example, by small mobility of the wobble pairs, misreading of the codon, unmodi®ed adenine and guanine at tRNA positions 34 (wobble) and 37, respectively. The results obtained can be subjected to direct experimental tests. # 1997 Academic Press Limited Keywords: codon-anticodon interaction; codon context effect; translation; misreading; frameshifting Introduction The rate at which a codon is read appears not to be constant but depends on the identity of the codon's neighbours, a phenomenon known as the context effect (for reviews, see Murgola, 1990; Hat®eld & Gutman, 1992; Osawa et al., 1992; Yarus & Curran, 1992; Buckingham, 1994). One possible source of the context effect is interactions between the codon± anticodon duplexes and anticodon loops. Genetic, physical and theoretical studies suggest that the wobble base-pair of the codon ±anticodon duplex located at the ribosomal P (peptidyl-tRNA) site may interact with the A (aminoacyl-tRNA) site duplex (Lim et al., 1992; Yarus & Smith, 1995). Such an interaction (the AP interduplex interaction; Figure 1) can be observed at different mutual orientations of the tRNAs, since signi®cant changes in the mutual orientation 0022±2836/97/100877±14 $25.00/0/mb960802 caused by sterically allowed conformational changes in a tRNA can occur without disturbance of the AP interduplex interaction (see below, property 5 of the AP interduplex interaction). The characteristic feature of the AP interduplex interaction is that the wobble base-pair of the P-site codon ± anticodon duplex is stacked on the sugarphosphate moiety of the A-site codon and that this stacking interaction shields the two A-site codon inter-ribose hydrogen bonds (Figure 1(b)). Direct localisation of tRNAs within the elongating ribosome has been attempted using a newly developed neutron-scattering technique, the nuclear spin contrast variation (Nierhaus et al., 1995). This physical study has revealed that the A and P-site tRNAs are separated by the dihedral angle of 50 formed by their planes, and that the mutual arrangement of these tRNAs does not change during translocation. Very recently a re®ned value (100 ) # 1997 Academic Press Limited 878 of this dihedral angle has been obtained (Dr K. H. Nierhaus, personal communication) using the latest ``Frank model'' of the ribosome (Frank et al., 1995). This value of the dihedral angle is in complete agreement with that found theoretically (100(10) ; Lim & Spirin, 1986; Spirin & Lim, 1986; Lim & Venclovas, 1992). At these values of the dihedral angle, the AP interduplex interaction is observed virtually without distortion of the crystal structure of the tRNAs (Figure 1). The Ap tRNA-mRNA-tRNA complex within the ribosome has also recently been visualized with cryoelectron microscopy by Agrawal et al. (1996) and by Dr R. Brimacombe, (personal communication). A cryoelectron microscopy study of Agrawal et al. (1996) has revealed that the arrangement of the A and P-site tRNAs involves the dihedral angle of ÿ160 (a re®ned value is about ÿ110 ; Dr J. Frank, personal communication). With regard to the cryoelectron microscopy data of van Heel's group, they indicate the dihedral angle of ÿ50 . At these values (ÿ110 and ÿ50 ) of the dihedral angle, the AP interduplex interaction cannot be observed without distortion of the crystal structure of the tRNAs. Analysis of interactions between the codon ± anticodon duplexes should not be limited to a consideration of the AP interduplex interaction alone, since besides the traditional A and P tRNA binding sites, the T (elongation factor Tu) and E (exit) sites, and the intermediary (hybrid) state A/T, A/P and P/E have been de®ned (for reviews, see Noller, 1991; Weijland & Parmeggiany, 1994; Yarus & Smith, 1995). The T-site, or a related formulation called the R-site (Lake, 1977), is traversed before the A and P-sites; it is the location of the newly bound aminoacyl-tRNA when it is still held by elongation factor Tu (EF-Tu). The E-site was proposed by Wettstein & Noll (1965). It has been shown (Rheinberger, 1991; Nierhaus, 1993) that deacylated tRNA does not leave the ribosome immediately from the P-site, but co-translocates from the P-site to the E-site when peptidyl-tRNA translocates from the A-site to the P-site. At the P and E-sites, the tRNAs interact with codons and the deacylated tRNA is expelled from the E-site by the binding of aminoacyl-tRNA to the A-site at the beginning of the subsequent elongation cycle. Here, we analyse the in¯uence of the interduplex interactions on ribosomal translation within the framework of the model proposing the AP and PE interduplex interactions. Results and Discussion Allowed codon ± anticodon duplexes: a wobble in the pairing of the first, second and third codon bases All codon± anticodon duplexes fall into two general groups, one with and the other without disallowed steric overlaps and an uncompensated loss Interactions Between Codon ±Anticodon Duplexes of hydrogen and polar atom ±ion bonds. The duplexes of these groups should be considered as disallowed and allowed, respectively (see Materials and Methods). Our stereochemical analysis (the results will be published elsewhere) has revealed that all non-canonical base-pairs (the pairs UU, UC, UG, CC, CA, AA AG and GG) can be incorporated into codon ± anticodon duplexes of the allowed group. Non-canonical base-pairs can be formed in any duplex position (even simultaneously in all three positions) using the base-pair propeller twist and Hoogsteen-like base-pairing (recently Hoogsteenlike trans UU base-pairs have been found in the crystal structure of an RNA double helix; Wahl et al., 1996). The propeller twist prevents uncompensated losses of hydrogen bonds in non-canonical base-pairs (Lim, 1994, 1995). Hoogsteen-like base-pairing permits incorporation into the codonanticodon mini-helix of such base-pairs that, in the case of Watson-Crick-like base-pairing, are incompatible with a double helix because of uncompensated losses of hydrogen bonds (for example, the CC and GG pairs). Analysis has demonstrated that the allowed group is rather large, 20 to 40% of all conceivable duplexes (642) in accordance with an assortment of anticodon wobble bases, conserved tRNA purine base 37 and their modi®cations, since they affect the formation of the propeller twist in non-canonical base-pairs of codon± anticodon duplexes. The available experimental data support a wobble in the pairing of the ®rst, second and third codon bases. These data show that there is a wide variety of non-canonical base-pairs in different regions of RNA double helices. The non-canonical base-pairs AG, AC and C are placed at the double helix ends in the anticodon stems of crystal structures of yeast tRNAPhe (Kim et al., 1974; Jack et al., 1976; Stout et al., 1978), yeast tRNAAsp (Moras et al., (Woo 1980) and Escherichia coli initiator tRNAMet f et al., 1980). These pairs together with canonical base-pairs of the anticodon stem form a single double helix. Four consecutive non-canonical basepairs are observed in the middle of the crystal structure of RNA double helices: GU, CU, UC, UG (Holbrook et al., 1991; Cruse et al., 1994) and GU, UU, UU, UG (Baeyens et al., 1995). The UU basepair is observed within a conserved ribosomal RNA hairpin (Wang et al., 1996) and in the middle of the anticodon ± anticodon duplexes formed in the crystals of yeast tRNAAsp (Moras et al., 1980). Moreover, the difference in the lifetime of anticodon ±anticodon duplexes of a similar type and of the ``correct'' anticodon ±anticodon duplexes (as determined by the genetic coding rules) is insuf®cient to avoid translation errors (Grosjean et al., 1978). The above-listed experimental and stereochemical evidence indicates that the codon ± anticodon interaction cannot be explained just by the internal stability of the codon ± anticodon duplexes, since noncanonical pairs are observed in any position of al- Interactions Between Codon ±Anticodon Duplexes 879 Figure 1. The AP interduplex interaction. (a) Mutual orientation of the ribosomal A and P-site bound tRNAs. The two CCA ends (black beads) of the tRNAs are drawn together in order to allow formation of the peptide bond, and the two anticodons form codon ± anticodon mini-helices with adjacent codons (curved tube). Black and hatched trapeziums are the codon ± anticodon base-pairs. The horizontally and vertically hatched trapeziums are the wobble pairs of the A and P-site codon± anticodon duplexes, respectively. o is the dihedral angle formed by the planes of tRNA molecules. The long vertical arrow is the axis of the dihedral angle o. The AP interduplex interaction is shown: the interaction between the A-site codon (the 30 portion of the curved tube) and the wobble pair of the P-site duplex (the vertically hatched trapezium). (b) Stereo view of the AP interduplex interaction. The structure is drawn in accordance with the crystallographic coordinates of yeast phenylalanine tRNA at o 100 . In the P-site codon ±anticodn duplex (right) only the wobble base-pair is shown. The A-site codon triplet and the third residue of the P-site codon are shown in bold lines. In the A-site anticodon loop (left) the residues 32 to 37 are shown. The anticodon (residues 34 to 36) is paired with the codon (bold lines). Broken lines show the two A-site codon inter-ribose hydrogen bonds and hydrogen bonding between U33 and phosphate 36 in the A-site anticodon loop. (c) The AP tRNA-mRNA-tRNA complex superimposed on the 70 S ribosome model proposed by Frank et al. (1991). In this model, the CCA ends of the tRNA molecules are ®xed at the base of the central protuberance (the peptidyl transferase center) of the 50 S subunit, and the anticodon loops lie in the neck region (the decoding site) of the 30 S subunit. A detailed description of this Figure see in Lim et al. (1992). 880 Interactions Between Codon ±Anticodon Duplexes lowed duplexes and the differences in the stability of the allowed duplexes cannot play a crucial role in preventing codon misreading. The energetic differences are an order of magnitude less than the energy of a hydrogen bond, which is required to provide the experimentally observed level of misreading of the codon in vivo (see Materials and Methods). All these facts indicate that ``outside'' factors must affect the formation of the ribosomal A-site codon ± anticodon duplex. It is demonstrated below that the AP interduplex interaction can play the role of the main outside factor in binding the A-site tRNA. Properties of the AP interduplex interaction Property 1. The AP interduplex interaction constrains the A-site codon in the rigidly fixed A-form conformation The inter-ribose hydrogen bonds 20 OH ± O40 are observed in RNA double helices (Kim et al., 1974; Robertus et al., 1974; Stout et al., 1978; Moras et al., 1980; Dock-Bregeon et al., 1989). These hydrogen bonds ®x A-helix geometry. The disruption of the A-site codon inter-ribose hydrogen bonds in the AP codon ±anticodon complex should result in a large energetic loss (20 or 40 kJ/mol, the energy of one or two normal hydrogen bonds). This proceeds from shielding of the inter-ribose hydrogen bonds by the P-site wobble base-pair (Figure 1(b)). The shielded ribose 20 OH groups and O40 atoms cannot form polar atom ± ion bonds or new hydrogen bonds with any outside partners, including solvent molecules, in exchange for the disrupted inter-ribose hydrogen bonds. That is why the A-site codon should be ®xed in the A-form by the AP interduplex interaction. Property 2. Only correct duplexes at the A-site are compatible with the AP interduplex interaction Maintenance of the codon in the A-form (property 1) and the structure of the anticodon loop counteract the non-canonical base-pairing in the ®rst and second positions of the A-site codon (Lim & Venclovas, 1992) and enable the formation of only the correct wobble base-pairs at the third position (Lim, 1994, 1995), i.e. only the correct duplexes at the A-site are compatible with the AP interduplex interaction. The formation of the wrong duplexes (including duplexes formed by anticodons and near-cognate codons) in the presence of the AP interduplex interaction is accompanied by an energetic loss of 20 to 40 kJ/mol (an energetic difference between the correct and wrong duplexes that is required to provide the experimentally observed level of misreading of the codon in vivo; see Materials and Methods). Figure 2. The canonical base-pair UA (bold lines), demonstrating the base-pair major and minor edges. These edges in a double helix form its major and minor grooves. Two extrabold lines are glycosyl bonds. Circles are the polar atoms and groups of the codon base that in the minor groove of the A-site codon± anticodon duplex are positioned close to the major edge of the P-site wobble pair. Filled circles are the adenine or guanine nitrogen atom N3 and the oxygen atom O2 of the pyrimidine base U or C (®ne lines). The open circle is the guanine NH2 group. Property 3. In contrast to the A-site duplex, a wide variety of non-canonical base-pairs is admissible in any position of the P-site duplex, even simultaneously in all three positions The AP interduplex interaction does not ®x the codon in the P-site duplex. Therefore the P-site duplex can be considered as an ordinary allowed one. Consequently, a wide variety of non-canonical base-pairs can be observed in any position of the P-site duplex, even simultaneously in all three positions (see above). Property 4. The AP interduplex interaction counteracts the formation of both the correct and the wrong duplexes at the A-site when the wobble pair in the P-site duplex or solvent molecules and massive modifications at its major edge are fixed In the minor groove of any A-site codon ±anticodon duplex the polar atoms O2, N3 and the NH2 groups of the ®rst and second codon bases are positioned close to the major edge of the P-site wobble pair (the polar atoms, groups and the major edge are shown in Figure 2). Therefore, the P-site wobble pair should be mobile within the minihelix. Otherwise, certain of the O2 and N3 atoms, the NH2 groups of the A-site codon and the major edge polar atoms of the P-site wobble pair will form, at best, only severely deformed hydrogen or polar atom ± ion bonds with ordered solvent molecules, a large number of which is observed (Egli et al., 1996; Wahl et al., 1996) in the minor groove of the RNA double helix. The ®xation of solvent molecules (this takes place in adenine; Lim, 1995) or massive modi®cations (anticodon wobble base modi®cations) at the major edge of the P-site wob- 881 Interactions Between Codon ±Anticodon Duplexes ble base-pairs also leads to the formation of severely deformed hydrogen or polar atom ± ion bonds regardless of the mobility of the P-site wobble base-pair in the mini-helix. All this means that ®xation of the P-site wobble pair as well as that of solvent molecules and massive modi®cations at its major edge should counteract the formation of any duplex at the A-site via the AP interduplex interaction. In the case of the P-site wobble pairs UU and UC, besides the atoms O2 and N3, the guanine NH2 groups and the major edge polar atoms, there are water molecules that also lose a hydrogen or a polar atom± ion bond. When the codon residue in the P-site wobble pairs UU and UC is ®xed in the A-form conformation, one base-base hydrogen bond in these pairs should be replaced by a water bridge (Lim & Venclovas, 1992; Lim, 1994). At the P-site a bridging water molecule loses a hydrogen or a polar atom ± ion bond as the result of the Asite codon ± anticodon duplex formation. This proceeds from shielding of a bridging water molecule by the sugar-phosphate backbone of the A-site codon. In the absence of ®xation of the P-site wobble pair, a bridging water molecule can be removed and base-pairing in the P-site wobble pairs UU and UC can occur without a water bridge. There is genetic evidence suggesting that the small mobility of the P-site wobble pair causes an occlusion of the A-site. Smith & Yarus (1989) have shown that the substitution of purines for the conserved pyrimidines at anticodon loop positions 32 and 33 of the P-site tRNA reduces the rate of A-site occupation and that the effect at position 33 is greater than that at position 32. Curran (1995) has examined in vivo decoding properties of Escherichia coli arginine tRNAICG, which is the only tRNA in E. coli that reads the arginine codons CGU, CGC and CGA. It was concluded that of the wobble pairs IU, IC and IA, the P-site purine-purine wobble pair IA destabilizes the A-site codon ± anticodon duplex. The genetic data reported by Smith & Yarus (1989) and Curran (1995) display one structural generality: in both cases a decrease in the mobility of the P-site wobble pair takes place. Bases 32 and 33 in the crystallographic structure of tRNA are located inside the anticodon loop and parallel with its central part (the backbone section between residues 33 and 35 containing the anticodon wobble residue 34; see Figure 1(b)). Therefore a substitution by the larger purine bases (especially at position 33) should result in a stretching of the central part that, in turn, should cause a decrease in the mobility of any wobble pair in the codon± anticodon mini-helix. With regard to the purine-purine pair IA, the distance in this pair between the glycosyl bonds, in contrast to the purine-pyrimidine pairs IU and IC, strongly differs from that in the canonical pair (Crick, 1966). Therefore, the mobility of the wobble pair IA is also small. Thus, the results presented demonstrate that the experimental data reported by Smith & Yarus (1989) and Curran (1995) can actually be explained by a decrease in the mobility of the P-site wobble pair. Property 5. Significant changes in the mutual orientation of the tRNAs can occur without disturbance of the AP interduplex interaction Small, sterically allowed conformation changes in a tRNA (especially in the anticodon stem) can be accompanied by a noticeable rearrangement of the acceptor stem with respect to the codon ±anticodon duplex. Let us consider one such example. It is known (Saenger, 1984) that the RNA double helices display two major, structurally similar conformations, the A-form conformation with 11 base pairs per helix turn (such helices are observed in the tRNA crystal structure; Kim et al., 1974; Robertus et al., 1974; Stout et al., 1978; Moras et al., 1980) and the A0 -form with 12 pairs per turn. Furthermore, X-ray analysis shows that an RNA double helix can be kinked at points dividing the helix into blocks of four to ®ve base-pairs (DockBregeon et al., 1989). The kink angle (divergence from a straight helix) is about 10 . The structural changes within the double helices of the tRNA anticodon and acceptor arms caused by such kinks and the A$A0 transitions do not disturb the ``tertiary'' interactions in the L-shaped tRNA molecules. Even these very small structural variations (Figure 3) allow alteration of ``crystal'' value of 100 (Figure 1) of the angle o by 20 and an increase in the distance between the tRNA CCA Ê , leaving the AP interdutermini by at least 20 A plex interaction unchanged. Besides the kinks and the A$A0 transitions within the double helix regions of tRNAs, structural changes beyond the helices in the vicinity of their ends can occur without disruption of intramolecular hydrogen bonds. For instance, there can be a rotation of the anticodon loop in the anticodon stem around a double helix axis. In this case the sterically allowed rotation of 20 can be performed without disruption of the anticodon conformation. The A$A0 transitions, the structural changes at the helix ends within the anticodon stem and kinks in the acceptor stem of tRNAs can alter the angle o through a wide range (at least 100(50) ) retaining the AP interduplex interaction (Figure 3). It should be pointed out that these intramolecular conformational variations in tRNAs may provide intermediate states A/T, A/P and P/E in the movement of transfer RNA in the ribosome (Moazed & Noller, 1989), since the ®xation of the codon-anticodon duplex on the small subunit (Figure 1(c)) does not counteract the shifts of the acceptor end of tRNA on the large subunit and vice versa the ®xation of the acceptor end does not counteract the shifts of the duplex (for a discussion of intramolecular conformational variations in tRNAs and their functions see also the waggle theory by Yarus & Smith, 1995). 882 Figure 3. Small structural variations in the AP tRNAmRNA-tRNA complex that can occur without disturbance of the AP interduplex interaction (lower part of the drawing) and that lead to marked changes of the angle o and the distance between the tRNA CCA ends (black beads). The curved tube is the sugar-phosphate backbone of two adjacent mRNA codons. A and P are the anticodon loops of the ribosomal A and P-site bound tRNAs. Long ®ne lines are the anticodon and acceptor stem axes of the tRNAs. Rotations around these axes and their kinks shift the acceptor stem axes. A displacement of the CCA ends toward the viewer (increasing the angle o) is shown. The role of the AP interduplex interaction in the prevention of ribosomal frameshifting during a normal elongation cycle Non-slipping and slipping frameshifting All conceivable cases of frameshifting can be separated into two categories: non-slipping and slipping frameshifting (Figure 4). Non-slipping frameshifting can formally be modelled by insertions (frameshifting) and deletions (ÿframeshifting) of mRNA residues within the AP codon± anticodon complex (Figure 4(a)). In the case of slipping frameshifting, a disengagement of the tRNA anticodons from an initial codon interaction allowing the mRNA to slip with respect to the tRNAs ± ribosome complex takes place. The anticodons may then repair with new codons, so that synthesis continues downstream (Figure 4(b)). Neither of these types of frameshifting must take place during a normal elongation cycle. The interduplex interaction in the AP codon ± anticodon complex formed by normal elongator tRNAs should counteract non-slipping frameshifting The obvious assumption to adopt is that the A and P-site codon ± anticodon duplexes formed by normal elongator tRNAs are ®xed at the A and Psites, and are positioned relative to each other in a universal manner that is independent of the tRNA species involved. This should counteract non-slipping frameshifting, since in the absence of the Interactions Between Codon ±Anticodon Duplexes Figure 4. Frameshifting patterns. (a) Non-slipping frameshifting. Acute angles are the tRNA anticodon loops paired with the codons. Bold lines show the anticodon loop of peptidyl-tRNA and the zero frame codon. Fine lines are the anticodon loop of the deacylated tRNA paired with mRNA. Both 1 frameshifting (left, the insertion of the mRNA residue into the AP codon-anticodon complex) and ÿ1 frameshifting (right, the deletion of the mRNA residue) are shown. (b) Slipping frameshifting. Tandem slippage of peptidyl-tRNA and deacylated tRNA (1) as well as single slippage of peptidyl-tRNA (2) are shown. shifts of the duplexes relative to each other the insertions and deletions of mRNA residues within the AP codon ± anticodon complex cannot occur without the disruption of the duplexes (deletions) and without shielding of oxygen atoms in the sugar-phosphate backbone or the disruption of the inter-ribose hydrogen bonds in the A-site codon (insertions). Consequently, non-slipping frameshifting in the AP codon ± anticodon complex should be unfeasible when the complex is formed by normal elongator tRNAs and can be feasible when the shifts of the A and P-site duplexes relative to each other (including the expelling of a duplex from the P-site) are possible. Such shifts can be induced, for example, by non-standard structures of the anticodon loop or by speci®c interactions between the A and P-site tRNAs (for possible mechanisms for non-slipping frameshifting see, e.g. Weiss, 1984; Curran & Yarus, 1987; BjoÈrk et al., 1989; Vimaladithan & Farabaugh, 1994, and references therein). Tandem slippage of the A and P-site tRNAs With regard to slipping frameshifting, only tandem slippage of the A-site tRNA (newly formed peptidyl-tRNA) and P-site tRNA (deacylated tRNA) can occur in the AP codon ± anticodon complex formed by normal elongator tRNAs. Slippage of the A-site tRNA without slippge of the P-site tRNA is impossible because this variant of slipping frameshifting is the equivalent of non-slipping frameshifting (Figure 4(a)), which is prohibited. Therefore we can conclude that only slipping frameshifting resulting from tandem slippage of the A and P-site tRNAs Interactions Between Codon ±Anticodon Duplexes can occur in the AP condon ±anticodon complex formed by normal elongator tRNAs. Tandem slippage rule The A-site tRNA after slippage, according to property 2, should form only correct duplexes, whereas the P-site tRNA is free to interact with any codon if this codon and the anticodon are able to form an allowed duplex (property 3) in which the pair formed by the ®rst anticodon base is not ®xed within the mini-helix (property 4). Elimination of tandem slippage by the selection of synonymous codons and anticodon wobble bases Assume that the A-site anticodon in the AP codon± anticodon complex forms the correct duplex with the ith codon in a coding sequence. According to the tandem slippage rule, the second and third bases of the A-site anticodon after slippage must form canonical base-pairs with new codon partners. In the case of 1 and 2 slippages the third base of the ith codon should be paired with the second and third anticodon base, respectively. After ÿ2 and ÿ1 slippages the third base of the i ÿ 1th codon should be paired with the second and third anticodon base, respectively. In all these four cases the selection of the third base in the i ÿ 1th and ith codons (selection of synonymous codons) can prevent the canonical base-pairing at the second or third anticodon positions, i.e. slippage of the A-site anticodon in the AP codon ± anticodon complex can be prevented by the selection of synonymous codons. In parallel with the selection of synonymous codons, the selection of anticodon wobble bases can make a substantial contribution to the prevention of slippage of the A-site anticodon, since translational speci®city of most of the wobble bases is restricted. Only unmodi®ed A and U in the anticodon wobble position of the correct duplexes can recognize all four bases, U, C A and G, in the third position of codons (Barrell et al., 1980; Heckman et al., 1980; Sibler et al., 1986; Boren et al., 1993; Inagaki et al., 1995). However, A almost never exists in the ®rst anticodon position (Sprinzl et al., 1991). The anticodon wobble U (the wobble base of greatest abundance) is frequently and variously modi®ed and its modi®cations always restrict translational speci®city. The wobble U is modi®ed even in tRNAs that translate the unmixed codon families (where four codons specify the same amino acid) despite the fact that any of these families is potentially translatable by a single anticodon with an unmodi®ed wobble U, while tRNAs with modi®ed wobble U cannot decode all four codons. All known wobble bases except A and U can recognize only 1 to 3 codon bases. For example, 2thiouracil (s2U), unmodi®ed cytosine (C) and 2lysylcytosine (k2C) decode only A, G and A, respectively; G decodes U and C; 5-methyluracil derivatives (xm5U) decode A and G; 5-hydroxyuracil 883 derivatives (xo5U) decode U, A and G; inosine (l) decodes U, C and A (Lim, 1994, 1995; for reviews, see Agris, 1991; BjoÈrk, 1992; Osawa et al., 1992). It is clear that the selection of synonymous codons and anticodon wobble bases is capable of counteracting slippage of the P-site anticodon. According to the tandem slippage rule, the P-site tRNA, besides the correct duplexes, can form any allowed duplex in which the pair formed by the ®rst anticodon base is not ®xed within the mini-helix. The allowed group of codon ± anticodon duplexes, as noted above, is about 30% of all conceivable duplexes. This means that at uniform distribution of codons in coding sequences the probability of the correct duplex ! any allowed duplex transition resulting from slippage of the P-site tRNA should be 0 3 on the average. This probability can be reduced by the selection of synonymous codons and anticodon wobble bases counteracting the formation of the allowed duplexes and mobile wobble base-pairs. However, full elimination of the correct duplex ! any allowed duplex transition at the Psite is impossible because the selection of synonymous codons and anticodon wobble bases can control a wobble in the pairing of only two codon bases, whereas in allowed duplexes a wobble in the base-pairing of all three codon bases is possible when the codon is not ®xed in the A-form by the AP interduplex interaction. Contrary to the correct duplex ! any allowed duplex transition at the P-site, the probability of the correct duplex ! correct duplex transition at the A-site can be reduced practically to zero by the selection of synonymous codons and anticodon wobble bases because a wobble in the base-pairing in the correct duplexes occurs only at the third codon position. Though it should be noted that in rare cases a slippage of the A-site anticodon cannot be prevented by the selection of synonymous codons and anticodon wobble bases. For example, when an amino acid sequence contains Met and Trp (minor amino acid residues). In contrast to the other amino acid residues, each of these residues is speci®ed by a single codon (Met is speci®ed by AUG and Trp by UGG). Therefore, for example, ÿ2 and ÿ1 slippages of the A-site anticodon interacting with the codon GGG in the sequence 50 UGG-GGG-30 that codes Trp-Gly cannot be prohibited by the selection of synonymous codons and anticodon wobble bases. But in such cases tandem slippage can be prohibited by the elimination of the correct duplex ! any allowed duplex transition at the P-site. Thus, we see that in the presence of the AP interduplex interaction, frameshifting can be prevented within coding sequences by the selection of synonymous codons and anticodon wobble bases. It is apparent that within the framework of our model the selection of synonymous codons should occur to prevent frameshifting and it could be one of the reasons why the contexts of both sense codons and stop codons are non-random (Ikemura, 1981; Kohli & Grosjean, 1981; Lipman & Wilbur, 884 1983; Yarus & Folley, 1985; Shpaer, 1986; Gouy, 1987; Hanai & Wada, 1989; Andersson & Kurland, 1990; Hat®eld & Gutman, 1992). The synonymous codon selection allowing avoidance of tandem slippage may partially explain 3-1 and 3-3 doublet bias (the doublets consisting of the third base of one codon and the ®rst or third base, respectively, of the following codon), which has been demonstrated in coding sequences (for reviews, see Yarus & Curran, 1992; Buckingham, 1994). The PE interduplex interaction The AP interduplex interaction should be retained during ribosomal translocation and in the posttranslocation state in the PE codon ± anticodon complex In the absence of the AP interduplex interaction, the correct duplex ! any allowed duplex transition resulting from tRNA slippage can occur at the P-site and at the A-site. As noted in the preceding section, the selection of synonymous codons and anticodon wobble bases counteracts the correct duplex ! any allowed duplex transition, but full elimination of such transitions within coding sequences is impossible. This suggests that the AP interduplex interaction should be retained during ribosomal translocation and in the post-translocation state, since in the absence of a counteractive mechanism, frameshifting can occur in the PE tRNA-mRNA-tRNA complex. Speci®cally, prokaryotic ribosomes recode the HIV-1 gag-pol-1 frameshift sequence by an E/P site post-translocation simultaneous slippage mechanism (Hors®eld et al., 1995). The assumption that the AP interduplex interaction should be retained is strongly supported experimentally by the work of Nierhaus et al. (1995), in which it is shown that the mutual arrangement of the A and P-site tRNAs does not change strongly during translocation. Retention of the AP interduplex interaction means that the PE interduplex interaction that is identical with that of AP should exist. In thise case, the AP interduplex interaction will take place throughout the ribosomal elongation cycle and so prevent slipping frameshifting caused by tandem slippage. Therefore, we hypothesize that the PE interduplex interaction is identical with that of AP. In this case, properties 1 to 5 of the AP interduplex interaction should be true also for the PE interduplex interaction. The AE interduplex indirect linkage The simultaneous existence of the A and E-site codon ± anticodon duplexes should be prohibited by an indirect linkage between them In the APE codon ±anticodon complex the relative position between the A and E-site codon ±anticodon duplexes can be described as a right-handed Interactions Between Codon ±Anticodon Duplexes helical displacement of the A-site duplex to E site: a rotation by 2 100 around the helix axis and a Ê simultaneous shift by approximately 2 8 A (Figure 5). At such a relative position the A-site duplex is not in direct physical contact with the E-site duplex. However, the AP and PE interduplex interactions provide an indirect linkage between these duplexes. The PE interduplex interaction ®xes the P-site codon (property 1) and, consequently, the Psite wobble pair. In turn, a ®xation of the P-site wobble pair counteracts the formation of both the correct and the wrong codon ± anticodon duplexes at the A-site (property 4). Hence, the AP and PE interduplex interactions should destabilize the APE codon ± anticodon complex. The most likely way of eliminating the destabilizing in¯uence of the interduplex interactions is to expel a codon ±anticodon duplex from the A or Esite. With regard to the P-site duplex, its disruption is accompanied by the simultaneous breaking of about six base-base and inter-ribose hydrogen bonds as a result of holding of the codon ends by the A and E-site duplexes. The simultaneous breakdown of six hydrogen bonds creates the barrier, the overcoming time t eE/kT 10ÿ13 second of which at E 6 20 kJ/mol and T 300 K lies between 107and 108 seconds. This time is much greater than that of the elongation cycle (10ÿ2 to 10ÿ1 second; Spirin, 1986). All this means that the simultaneous existence of the A and E-site codon± anticodon duplexes should be prohibited via the AP and PE interduplex interactions (the AE interduplex indirect linkage). The AE interduplex indirect linkage correlates well with experiment. Direct visualization of three deacylated tRNAPhe molecules in the ribosome (Agrawal et al., 1996) has revealed that the anticodons of the A and P-site tRNAs may interact with two adjacent codons, whereas the position of the E-site tRNA appears to exclude codon± anticodon contact. This result supports the main consequence of the AE interduplex indirect linkage, that the simultaneous existence of the A and E-site duplexes should not take place, as well as a conclusion (Robertson et al., 1986) that the E-site is kinetically heterogeneous, exhibiting more than one successive state for the anticodon region of a transiting tRNA molecule. Because of the AE interduplex indirect linkage an occupation of the A-site in the ribosomal pre-translocation state should reduce the af®nity for tRNA of the E-site and vice versa in the post-translocation state. Accordingly, the elongating ribosome should always carry two tRNA: at A and P-sites before translocation and at P and E-sites after translocation. This is entirely consistent with the allosteric three-site model of elongation (Rheinberger & Nierhaus, 1986; Gnirke et al., 1989; Triana-Alonso et al., 1995; for reviews, see Rheinberger, 1991; Nierhaus, 1993), and the AE interduplex indirect linkage can be considered as a molecular mechanism explaining the allosteric interaction between 885 Interactions Between Codon ±Anticodon Duplexes Limitations of the approach and its further elaboration Figure 5. A representation of the APE codon± anticodon complex. The curved tube is mRNA (the A, P and E-site codons). Acute angles are the anticodon loops of the A, P and E-site bound tRNAs forming codon ± anticodon duplexes with mRNA. Broken lines show the wobble base-pairs. The P-site wobble pair interacts with the Asite codon ± anticodon duplex (the AP interduplex interaction), and the E-site wobble pair interacts with the Psite duplex (the PE interduplex interaction). The A, P and E-site codons form a helix. The axis of this helix is the long vertical arrow (the axis of the dihedral angle o, see Figure 1(a)). The relative position between adjacent codon ± anticodon duplexes can be described as a rotation by 100 around the helix axis and a simultaneous Ê. shift by approximately 8 A The main limitation of the approach is that in the proposed model ``outside'' factors affecting the formation and existence of the AP and PE codon ± anticodon complexes are not considered. There is a broad spectrum of such outside factors: spatial structures in a mRNA molecule; the af®nity of tRNA molecule for the ribosomal sites including the af®nity of a nascent peptide; aminoacyl-tRNA imbalance; the selection of the wobble base-pairs; non-standard structures of the anticodon loop; conserved purine base 37 in tRNAs, and so on. The action of these factors leads to non-slipping and slipping frameshifting, ribosome hopping, misreading of the codon, abortive termination of translation (e.g. see Curran & Yarus, 1987; 1989; Jacks et al., 1988; BjoÈrk et al., 1989; Andersson & Kurland, 1990; Jacks, 1990; Chamorro et al., 1992; Gesteland et al., 1992; Morikawa & Bishop, 1992; Osawa et al., 1992; Peter et al., 1992; Tu et al., 1992; Yarus & Curran, 1992; Hagervall et al., 1993; Somogyi et al., 1993; Mottagui-Tabar et al., 1994; Vimaladithan & Farabaugh, 1994; Atkins & Gesteland, 1995; Barak et al., 1996; Chen et al., 1996). In many cases the action of outside factors can be explained within the framework of our model. The action of certain of the factors on the AP and PE codon ± anticodon complexes is analyzed in the following ( a comprehensive analysis is beyond the scope of this work and will be published elsewhere). Outside factors leading to codon misreading and the occurrence of fixed wobble base-pairs can stimulate frameshifting via interduplex interactions the A and E sites on the detailed molecular level in terms of concrete atomic interactions. The wobble pair of the E-site duplex can affect the rate of codon reading at the A-site via the AE interduplex indirect linkage Stabilization or destabilization of the PE codon ± anticodon complex will increase or decrease, respectively, the lifetime of the AE interduplex indirect linkage and so affect codon reading. However, in these cases the in¯uence only on the rate of the A-site occupation can take place. A pronounced effect on the accuracy of codon reading should be absent because the AE interduplex indirect linkage counteracts the formation of both the correct and the wrong codon ± anticodon duplexes. If it is taken into account that the stability of the PE codon ± anticodon complex depends on the mobility of the Esite wobble pairs, solvent molecules and modi®cations at their major edges (property 4), it is quite obvious to think that the wobble pair of the E-ste duplex can affect the rate of codon reading at the A-site via the AE interduplex indirect linkage. There are different outside factors responsible for misreading of a codon. For example, aminoacyltRNA imbalance or an elevation of the aminoacyltRNA af®nity for the A-site caused by aminoacyltRNA mutations outside the anticodon (in this case a lowering of the af®nity caused by the formation of the wrong A-site duplex can be compensated). The ®xed wobble base-pairs can be obtained, for example, by the selection of synonymous codons and the anticodon wobble base or by changing the anticodon loop structure (see property 4). The appearance of the wrong codon± anticodon duplexes at the A-site and the ®xed wobble base-pairs in the P-site duplex should result in destabilization of the AP codon ±anticodon complex in the pre-translocation state and the PE codon ± anticodon complex in the posttranslocation state (see properties 2 and 4). The destabilization of the AP or PE codon ± anticodon complex can be eliminated by tandem slippage or by expelling either of the two duplexes formed by deacylated tRNA and peptidyl-tRNA from the ribosomal A or P or E-site. In the ®rst case, slippage of tRNA molecules should satisfy the tandem slippage rule. In the second case, the exit of the deacylated tRNA anticodon ± codon du- 886 Interactions Between Codon ±Anticodon Duplexes plex from the P-site or E-site permits single slippage of peptidyl-tRNA. Because of the lack of the interduplex interaction, the probability of single slippage should be comparatively high (property 3). Thus, we see that misreading of the codon and the small mobility of the wobble base-pair in the P-site duplex can initiate slipping frameshifting resulting from tandem slippage of deacylated tRNA and peptidyl-tRNA or from single slippage of peptidyl-tRNA. tRNA conserved purine base 37 as an outside factor affecting codon reading and frameshifting In all elongator tRNAs position 37 next to the 30 end of the anticodon contains purine bases (Grosjean et al., 1995). In the crystal structure of tRNA (Jack et al., 1976) the second and third anticodon residues together with residue 37 are in the A-form conformation. Owing to such an orientation, conserved purine base 37 simultaneously stacks on the ®rst codon base and the third anticodon base (Figure 6) and so counteracts the large propeller twist in the base-pair formed by these bases. This, in turn, counteracts the formation of non-canonical base-pairs in the codon ®rst position, since the large propeller twist is required in many non-canonical base-pairs to avoid an uncompensated loss of hydrogen or polar atom ±ion bonds (Lim, 1994, 1995). All this can serve as an explanation for the fact that position 37 in all elongator tRNAs contains a purine base. Apart from the fact that conserved purine base 37 counteracts the formation of non-canonical basepairs in the codon ®rst position, the type of purine base 37 and its modi®cations determine the stability of codon ± anticodon duplexes and so affect codon reading and frameshifting. Let us use unmodi®ed guanine at tRNA position 37 to illustrate evidence supporting this conclusion (a comprehensive analysis of the conserved purine base 37 properties will be published elsewhere). At tRNA position 37 the guanine NH and NH2 groups together with solvent molecules form two hydrogen bonds, which in the codon ±anticodon duplex are stacked on the ®rst codon base when residue 37 is in the A-form conformation (Figure 6). Each of the solvent molecules forming these stacked hydrogen bonds loses one of its 2± 3 hydrogen bonds, which can be formed with other solvent molecules in the absence of stacking. There are three conceivable ways by which a loss of the hydrogen bonds can be precluded: (1) shifts of stacked hydrogen bonds toward the codon base edge; (2) substitution of the ribosomal material for the solvent molecules forming stacked hydrogen bonds; (3) pairing of guanine 37 with a mRNA base, i.e. the formation of the allowed mini-helix extended by the base-pair G37 mRNA base (primarily by GC). The ways (1) and (2) are unlikely. The way (1) is not suited for guanine because two stacked hydrogen bonds of guanine in the vicinity of the A-form Figure 6. Stacking pattern of bases in a double helix demonstrating that a purine base in chain position i (in this case guanine, bold lines) always simultaneously stacks on both bases of the pair formed by the i ÿ 1th base and its partner (the pair UA, ®ne lines positioned under the bold lines). Three extrabold lines are glycosyl bonds. Dotted circles are solvent molecules: water molecules and anions that can be considered as hydrogen bond acceptors. These solvent molecules form hydrogen bonds with the guanine NH and NH2 groups. The hydrogen bonds are stacked on uracil. Guanine can be considered as tRNA purine base 37 that is simulataneously stacked on the ®rst codon base and third anticodon base. conformation cannot be forced out of stacking simultaneously. The way (2) is unlikely because tRNA molecules must move in the ribosome. As for the way (3), the formation of the extended mini-helix at the A-site can occur through expelling the duplex from the P-site. In this case the exit of the duplex from the P-site allows avoidance of steric restrictions within the P/A boundary of mRNA. All this means that guanine at tRNA poition 37 can play the role of an outside factor, the action of which leads to 1 non-slipping frameshifting. Consequently, unmodi®ed guanine at tRNA position 37 should be rare or absent and modi®cations in its derivatives should eliminate one or both stacked hydrogen bonds (a purine base forming one stacked hydrogen bond, for example adenine, is admissible because a single hydrogen bond can be forced out of stacking). These conclusions are entirely consistent with experiment. In contrast to unmodi®ed adenine, unmodi®ed guanine almost never exists at tRNA position 37. There are only two known examples of the presence of unmodi®ed guanine at position 37 and in all guanine derivatives one or both stacked hydrogen bonds are eliminated by nucleoside modi®cations (see the compilation summarising the occurrence of modi®ed nucleosides by Grosjean et al., 1995). As to frameshifting, there is direct evidence that methylation of the NH group of quanine (m1G) at tRNA position 37 prevents frameshifting. In the Salmonella typhimurium trmD mutant, which lacks m1G, the low-level 1 frameshifting caused by the de®ciency has been interpreted to be due to base-pairing between tRNA base 37, G, and the ®rst codon base, C (BjoÈrk et al., 1989; Hagervall et al., 1993). 887 Interactions Between Codon ±Anticodon Duplexes The interduplex interactions and programmed frameshifting The chance of a shift in the reading frame is well below 10ÿ4 (Atkins et al., 1972), but there exist special sites in mRNA at which very high frameshifting frequencies have been detected (programmed frameshifting; for a review see, for example, Atkins & Gesteland, 1995). Besides frameshift sites, programmed frameshifting is controlled by special signals coded in mRNA. These signals ``stimulate'' high levels of shifting at the localized shift site. There is a broad spectrum of stimulatory signals that can be realized at frameshift sites through the interduplex interactions. In fact, in the above sections we have considered outside factors playing the role of stimulatory signals (small mobility of the wobble pairs, misreading of the codon, guanine at tRNA position 37) that can initiate frameshifting. In other words, within the framework of the model of interacting codon± anticodon duplexes it is possible to attempt to design arti®cial systems of programmed frameshifting in terms of concrete atomic interactions and these systems can be subjected to direct experimental tests. Main results Analysis of interactions between the codon ± anticodon duplexes gave the following main results. Firstly, the AP interduplex interaction and not the differences in the stability of the allowed duplexes selects the correct codon ± anticodon duplexes: at the A-site only the allowed duplexes with the Aform conformation of the codon are compatible with the AP interduplex interaction and such duplexes should be considered correct. Secondly, the AP interduplex interaction should be retained during ribosomal translocation and in the post-translocation state in the PE codon± anticodon complex. The retention of the AP interduplex interaction throughout the ribosomal elongation cycle and the selection of synonymous codons and anticodon wobble bases are able to prevent frameshifting within coding sequences. The synonymous codon selection allowing avoidance of frameshifting may partially explain 3-1 and 3-3 doublet bias, which is observed in coding sequences. Thirdly, tandem slippage of peptidyl-tRNA and deacylated tRNA can occur when peptidyl-tRNA after slippage forms only correct duplexes, and deacylated tRNA forms any allowed duplexes in which the pair formed by the ®rst anticodon base is not ®xed within the mini-helix (tandem slippage rule). Fourthly, there is a broad spectrum of factors leading to destabilization of the AP or PE codon± anticodon complex: small mobility of the wobble pairs, misreading of the codon, unmodi®ed adenine at the ®rst anticodon position, unmodi®ed guanine at tRNA position 37, non-standard structures of the anticodon loop, spatial structures in a mRNA molecule, and so on. The destabilization of the AP or PE codon± anticodon complex can lead to frameshifting, since the destabilization of these complexes can be eliminated by tandem slippage or by expelling either of the two duplexes formed by deacylated tRNA and peptidyl-tRNA from the ribosomal A or P or E-site. In the ®rst case, tRNA slippage should satisfy the tandem slippage rule. In the second case, the exit of the deacylated tRNA anticodon ± codon duplex from the P-site or E-site permits single slippage of peptidyl-tRNA. Because of the lack of the interduplex interaction, the probability of single slippage should be comparatively high, as new codon ±anticodon duplexes formed by peptidyl-tRNA need not be correct. Fifthly, the indirect linkage between the codon ±anticodon duplexes bound to the A and E-sites follows from the AP and PE interduplex interactions. This AE interduplex indirect linkage counteracts the simultaneous existence of the A and E-site duplexes: the occupation of the E-site counteracts the formation of any duplex at the A-site and, conversely, the occupation of the A-site is accompanied by expelling the codon ± anticodon duplex from the E-site. Sixthly, the wobble pairs of the P and E-site duplexes can affect the rate of the A-site occupation by aminoacyl-tRNA via the AP interduplex interaction and the AE interduplex indirect linkage, respectively. Seventhly,anticodonwobbleresidue modi®cations can take part in the prevention of frameshifting and in the regulation of the rate of the A-site occupation by aminoacyl-tRNA. Probably, for this reason the anticodon wobble residues are frequently and variously modi®ed. For example, the anticodon wobble U (the wobble base of greatest abundance) is modi®ed, as a rule, even in tRNAs that translate the unmixed codon families (where four codons specify the same amino acid), despite the fact that any of these families is potentially translatable by a single anticodon with an unmodi®ed wobble U, while tRNAs with modi®ed wobble U cannot decode all four codons. Finally, purine base 37 in tRNA molecules simultaneously stacks on the ®rst codon base and the third anticodon base and so counteracts the large propeller twist in the base-pair formed by these bases. This, in turn, counteracts the formation of non-canonical base-pairs in the codon ®rst position, since the large propeller twist is required in many non-canonical base-pairs to avoid an uncompensated loss of hydrogen or polar atom ±ion bonds. All this can serve as an explanation for the fact that position 37 in all elongator tRNAs contains a purine base. Materials and Methods Stereochemical analysis was made using the FRODO program on an Evans & Sutherland PS390 computer graphics system and Corey-Pauling-Koltun atomic models. The experimentally observed level of misreading and frameshifting in vivo is 10ÿ4 to 10ÿ3 (Atkins et al., 888 1972; Spirin, 1986; Parker, 1989; Jùrgensen & Kurland, 1990). This means that if the selection of tRNAs rests only on the difference in the free energy of complex formation, at equilibrium a difference of G 20 kJ/mol between wrong and correct association is required (the probability of misreading or frameshifting exp(ÿG/kT) at G 20 kJ/mol and T 300 K is 5 10ÿ4). Therefore, in considering interactions between the codon ± anticodon duplexes and in the search for allowed structures, a complex was disallowed when an uncompensated loss of hydrogen or polar atom ± ion bonds was observed and/or when at least one interatomic distance was less than the extreme limit (rarely observed steric overlap). Such a disallowed steric overlap and a polar atom ± ion bond are comparable to the hydrogen bond in energy (the energy of a normal hydrogen bond is about 20 kJ/ mol). The contribution of the other interactions (interactions in the sugar-phosphate backbone, base stacking stabilized mainly by dipolar and London dispersion forces and hydrophobic effects) was ignored, since these interactions are an order of magnitude weaker than a hydrogen bond. For details and references, see Materials and Methods (Lim, 1994). Acknowledgements The author is grateful to Drs J. F. Curran, M. B. Garber and K. H. Nierhaus for constructive critical comments on an earlier version of the manuscript and to Drs R. Brimacombe and J. Frank for kindly providing the data on the arrangement of tRNAs within the ribosome. This work was supported by grant 96-04-48774 from the Russian Foundation for Basic Research References Agrawal, R. K., Penczek, P., Grassucci, R. A., Li, Y., Leith, A., Nierhaus, K. H. & Frank, J. (1996). Direct visualization of A-, P-, and E-site transfer RNAs in the Escherichia coli ribosome. Science, 271, 1000± 1002. Agris, P. F. (1991). Wobble position modi®ed nucleosides evolved to select transfer RNA codon recognition: a modi®ed-wobble hypothesis. Biochimie, 73, 1345± 1349. Andersson, S. G. E. & Kurland, C. G. (1990). Codon preferences in free-living microorganisms. Microbiol. Rev. 54, 198± 210. Atkins, J. F. & Gesteland, R. F. (1995). Discontinuous triplet decoding with or without re-pairing by peptidyl tRNA. In tRNA: Structure, Biosynthesis, and Function (SoÈll, D. & RajBhandary, U., eds), pp. 471 ± 490, American Society for Microbiology, Washington, DC. Atkins, J. F., Elseviers, D. & Gorini, L. (1972). Low activity of b-galactosidase in frameshift mutants of Escherichia coli. Proc. Natl Acad. Sci. USA, 69, 1192± 1195. Baeyens, K. J., De Bondt, H. L. & Holbrook, S. R. (1995). Structure of an RNA double helix including uraciluracil base pairs in an internal loop. Struct. Biol. 2, 56 ± 62. Barak, Z., Lindsley, D. & Gallant, J. (1996). On the mechanism of leftward frameshifting at several hungry codons. J. Mol. Biol. 256, 676 ±684. Interactions Between Codon ±Anticodon Duplexes Barrell, B. G., Anderson, S., Bankier, A. T., de Bruijn, M. H. L., Chen, E., Coulson, A. R., Drouin, J., Eperon, I. C., Nierlich, D. P., Roe, B. A., Sanger, F., Schreier, P. H., Smith, A. J. H., Staden, R. & Young, I. G. (1980). Different patterns of codon recognition by mammalian mitochondrial tRNAs. Proc. Natl Acad. Sci. USA, 77, 3164± 3166. BjoÈrk, G. R. (1992). The role of modi®ed nucleosides in tRNA interactions. In Transfer RNA in Protein Synthesis (Hat®eld, D. L., Lee, B. J. & Pirtle, R. M., eds), pp. 23±85. CRC Press, Boca Raton, FL. BjoÈrk, G. R., WikstroÈm, P. M. & BystroÈm, A. S. (1989). Prevention of translational frameshifting by the modi®ed nucleoside 1-methylguanosine. Science, 244, 986 ± 989. Boren, T., Elias, P., Samuelsson, T., Claesson, C., Barciszewska, M., Gehrke, C. W., Kuo, K. C. & Lustig, F. (1993). Undiscriminating codon reading with adenosine in the wobble position. J. Mol. Biol. 230, 739 ± 749. Buckingham, R. H. (1994). Codon contex and protein synthesis: enhancements of the genetic code. Biochimie, 76, 351± 354. Chamorro, M., Parkin, N. & Varmus, H. E. (1992). An RNA pseudoknot and an optimal heptameric shift site are required for highly ef®cient ribosomal frameshifting on a retroviral messenger RNA. Proc. Natl Acad. Sci. USA, 89, 713 ± 717. Chen, X., Kang, H., Chen, L. X., Chamorro, M., Varmus, H. E. & Tinoco, I. Jr (1996). A characteristic bent conformation of RNA pseudoknots promotes ÿ1 frameshifting during translation or retroviral RNA. J. Mol. Biol. 260, 479± 483. Crick, F. H. C. (1966). Codon-anticodon pairing: the wobble hypothesis. J. Mol. Biol. 19, 548± 555. Cruse, W. B. T., Saludjian, P., Biala, E., Strazewski, P., Prange, T. & Kennard, O. (1994). Structure of a misÊ resolution and paired RNA double helix at 1.6 A implications for the prediction of RNA secondary structure. Proc. Natl Acad. Sci. USA, 91, 4160± 4164. Curran, J. F. (1995). Decoding with the A:I wobble pair is inef®cient. Nucl. Acids Res. 23, 683± 688. Curran, J. F. & Yarus, M. (1987). Reading frame selection and transfer RNA anticodon loop stacking. Science, 238, 1545± 1550. Curran, J. F. & Yarus, M. (1989). Rates of aminoacyltRNA selection at 29 sense codons in vivo. J. Mol. Biol. 209, 65 ± 77. Dock-Bregeon, A. C., Chevrier, B., Podjarny, A., Johnson, J., de Bear, J. S., Gough, G., Gilham, P. T. & Moras, D. (1989). Crystallographic structure of an RNA helix: [U(UA)6A]2. J. Mol. Biol. 209, 459 ± 474. Egli, M., Portmann, S. & Usman, N. (1996). RNA hydration: a detailed look. Biochemistry, 35, 8489± 8494. Frank, J., Penczek, P., Grassucci, R. & Srivastava, S. (1991). Three-dimensional reconstruction of the 70 S E. coli ribosomes in ice: the distribution of ribosomal RNA. J. Cell Biol. 115, 597 ± 605. Frank, J., Verschoor, A., Li, Y., Zhu, J., Lata, R. K., Radermacher, M., Penczek, P., Grassucci, R., Agrawal, R. K. & Srivastava, S. (1995). A model of the translational apparatus based on a three-dimensional reconstruction of the Escherichia coli ribosome. Biochem. Cell Biol. 73, 757± 765. Gesteland, R. F., Weiss, R. B. & Atkins, John F. (1992 ). Recoding: reprogrammed genetic decoding. Science, 257, 1640± 1641. Interactions Between Codon ±Anticodon Duplexes Gnirke, A., GeigenmuÈller, U., Rheinberger, H. J. & Nierhaus, K. H. (1989). The allosteric three-site model for the ribosomal elongation cycle: analysis with a heteropolymeric RNA. J. Biol. Chem. 264, 7291± 7301. Gouy, M. (1987). Codon contexts in enterobacterial and coliphage genes. Mol. Biol. Evol. 4, 426 ± 444. Grosjean, H. J., de Henan, S. & Crothers, D. M. (1978). On the physical basis for ambiguity in genetic coding interactions. Proc. Natl Acad. Sci. USA, 75, 610± 614. Grosjean, H., Sprinzl, M. & Steinberg, S. (1995). Posttransciptionally modi®ed nucleosides in transfer RNA: their locations and frequencies. Biochimie, 77, 139± 141. Hagervall, T. G., Tuohy, T. M. F., Atkins, J. F. & BjoÈrk, G. R. (1993). De®ciency of 1-methylguanosine in tRNA from Salmonella typhimurium induces frameshifting by quadruplet translocation. J. Mol. Biol. 232, 756± 765. Hanai, R. & Wada, A. (1989). Novel third-letter bias in Escherichia coli codons revealed by rigourous treatment of coding constraints. J. Mol. Biol. 207, 655± 660. Hat®eld, G. W. & Gutman, G. A. (1992). Codon pair utilisation in bacteria, yeast and mammals. In Transfer RNA in Protein Synthesis (Hat®eld, D. L., Lee, B. J. & Pirtle, R. M., eds), pp. 157±189, CRC Press, Boca Raton, FL. Heckman, J. E., Sarnoff, J., Alzner-DeWeerd, B., Yin, S. & RajBhandary, U. L. (1980). Novel features in the genetic code and codon reading patterns in Neurospora crassa mitochondira based on sequences of six mitochondrial tRNAs. Proc. Natl Acad. Sci. USA, 77, 3159± 3163. Holbrook, S. R., Cheong, C., Tinoco, I. & Kim, S.-H. (1991). Crystal structure of an RNA double helix incorporating a track of non-Watson-Crick basepairs. Nature, 353, 579± 581. Hors®eld, J. A., Wilson, D. N., Mannering, S. A., Adamski, F. M. & Tate, W. P. (1995). Prokaryotic ribosomes recode the HIV-1 gag-pol-1 frameshift sequence by an E/P site post-translocation simultaneous slippage mechanism. Nucl. Acids Res, 23, 1487± 1494. Ikemura, T. (1981). Correlation between abundance of Escherichia coli transfer RNAs and the occurrence of the respective codons in its protein genes. J. Mol. Biol. 146, 1 ±21. Inagaki, Y., Kojima, A., Bessho, Y., Hori, H., Ohama, T. & Osawa, S. (1995). Translation of synonymous codons in family boxes by Mycoplasma capricolum tRNAs with unmodi®ed uridine or adenosine at the ®rst anticodon position. J. Mol. Biol. 251, 486± 492. Jack, A., Ladner, J. E. & Klug, A. (1976). Crystallographic re®nement of yeast phenylalanine transfer Ê resolution. J. Mol. Biol. 108, 619 ± 649. RNA at 2.5 A Jacks, T. (1990). Translational suppression in gene expression in retroviruses and retrotransposons. Curr. Top. Microbiol. Immunol. 157, 93 ± 124. Jacks, T., Madhani, H. D., Masiarz, F. R. & Varmus, H. E. (1988). Signals for ribosomal frameshifting in the Rous sarcoma virus gag-pol region. Cell, 55, 447± 458. Jùrgensen, F. & Kurland, C. G. (1990). Processivity errors of gene expression in Escherichia coli. J. Mol. Biol. 215, 511± 521. Kim, S. H., Suddath, F. L., Quigley, G. J., McPherson, A., Sussman, J. L., Wang, A. H. J., Seeman, N. C. & 889 Rich, A. (1974). Three-dimensional tertiary structure of yeast phenylalanine transfer RNA. Science, 185, 435± 440. Kohli, J. & Grosjean, H. (1981). Usage of the three termination codons: compilation and analysis of the known eukaryotic and prokaryotic translation termination sequences. Mol. Gen. Genet. 182, 430 ± 439. Lake, J. A. (1977). Aminoacyl-tRNA binding at the recognition site is the ®rst step of the elongation cycle of protein synthesis. Proc. Natl Acad. Sci. USA, 74, 1903± 1907. Lim, V. I. (1994). Analysis of action of wobble nucleoside modi®cations on codon-anticodon pairing within the ribosome. J. Mol. Biol. 240, 8 ± 19. Lim, V. I. (1995). Analysis of action of the wobble adenine on codon reading within the ribosome. J. Mol. Biol. 252, 277± 282. Lim, V. I. & Spirin, A. S. (1986). Stereochemical analysis of ribosomal transpeptidation. Conformation of nascent peptide. J. Mol. Biol. 188, 565± 577. Lim, V. I. & Venclovas, C. (1992). Codon-anticodon pairing. A model for interacting codon-anticodon duplexes located at the ribosomal A- and P-sites. FEBS Letters, 313, 133±137. Lim, V. I., Venclovas, C., Spirin, A. S., Brimacombe, R., Mitchell, P. & Muller, F. (1992). How are tRNAs and mRNA arranged in the ribosome? An attempt to correlate the stereochemistry of the tRNA-mRNA interaction with constraints imposed by the ribosomal topography. Nucl. Acids Res. 20, 2627± 2637. Lipman, D. J. & Wilbur, W. J. (1983). Contextual constraints on synonymous codon choice. J. Mol. Biol. 163, 363± 376. Moazed, D. & Noller, H. F. (1989). Intermediate states in the movement of transfer RNA in the ribosome. Nature, 342, 142± 148. Moras, D., Comarmond, M. B., Fischer, J., Weiss, R., Thierry, J. C., Ebel, J. P. & Giege, R. (1980). Crystal structure of yeast tRNAAsp. Nature, 288, 669 ± 674. Morikawa, S. & Bishop, D. H. L. (1992). Identi®cation and analysis of the gag-pol ribosomal frameshift site of feline immunode®ciency virus. Virology, 86, 389± 397. Mottagui-Tabar, S., BjoÈrnsson, A. & Isaksson, L. A. (1994). The second to last amino acid in the nascent peptide as a codon context determinant. EMBO J. 13, 249± 257. Murgola, E. J. (1990). Suppression and the code: beyond codons and anticodons. Experientia, 46, 1134± 1141. Nierhaus, K. H. (1993). Solution of the ribosome riddle: how the ribosome selects the correct aminoacyltRNA out of 41 similar contestants. Mol. Microbiol, 9, 661 ± 669. Nierhaus, K. H., Beyer, D., Dabrowski, M., SchaÈfer, M. A., Spahn, C. M. T., Wadzack, J., Bittner, J.-U., Burkhardt, N., Diedrich, G., JuÈnemann, R., Kamp, D., Voss, H. & Stuhrmann, H. B. (1995). The elongating ribosome: structural and functional aspects. Biochem. Cell Biol, 73, 1011± 1021. Noller, H. F. (1991). Ribosomal RNA and translation. Annu. Rev. Biochem. 60, 191 ± 227. Osawa, S., Jukes, T. H., Watanabe, K. & Muto, A. (1992). Recent evidence for evolution of the genetic code. Microbial. Rev. 56, 229± 264. Parker, J. (1989). Errors and alternatives in reading the universal genetic code. Microbiol. Rev. 53, 273± 298. Peter, K., Lindsley, D., Peng, L. & Gallant, J. A. (1992). Context rules of rightward overlapping reading. New Biol. 4, 520± 526. 890 Interactions Between Codon ±Anticodon Duplexes Rheinberger, H. J. (1991). The function of the translating ribosome: allosteric three-site model of elongation. Biochimie, 73, 1067± 1088. Rheinberger, H. J. & Nierhaus, K. H. (1986). Allosteric interactions between the ribosomal transfer RNAbinding sites A and E. J. Biol. Chem. 261, 9133± 9139. Robertson, J. M., Paulsen, H. & Wintermeyer, W. (1986). Pre-steady-state kinetics of ribosomal translocation. J. Mol. Biol. 192, 351±360. Robertus, J. D., Ladner, J. E., Finch, J. T., Rhodes, D., Brown, R. S., Clark, B. F. C. & Klug, A. (1974). Ê Structure of yeast phenylalanine tRNA at 3 A resolution. Nature, 250, 546± 551. Saenger, W. (1984). In Principles of Nucleic Acid Structure, 1st edit., pp. 242 ± 252, Springer-Verlag, New York. Shpaer, E. G. (1986). Constraints on codon context in E. coli genes: their possible role in modulating the ef®ciency of translation. J. Mol. Biol. 188, 555 ± 564. Sibler, A. P., Dirheimer, G. & Martin, R. P. (1986). Codon reading patterns in Saccharomyces cerevisiae mitochondria based on sequences of mitochondrial tRNAs. FEBS Letters, 194, 131±138. Smith, D. & Yarus, M. (1989). tRNA-tRNA interactions within cellular ribosomes. Proc. Natl Acad. Sci. USA, 86, 4397± 4401. Somogyi, P., Jenner, A. J., Brierley, I. & Inglis, S. C. (1993). Ribosomal pausing during translation of an RNA pseudoknot. Mol. Cell. Biol. 13, 6931± 6940. Spirin, A. S. (1986). Ribosome Structure and Protein Biosynthesis, Benjamin/Cummings Publ. Co., Inc., Menlo Park, CA. Spirin, A. S. & Lim, V. I. (1986). Stereochemical analysis of ribosomal transpeptidation, translocation, and nascent peptide folding. In Structure, Function, and Genetic of Ribosomes (Hardesty, B. & Kramer, G., eds), pp. 556±572, Springer-Verlag, New York. Sprinzl, M., Dank, N., Nock, S. & SchoÈn, A. (1991). Compilation of tRNA sequences and sequences of tRNA genes. Nucl. Acids Res. 19(Suppl.), 2127± 2171. Stout, C. D., Mizuno, H., Rao, S. T., Swaminathan, P., Rubin, J., Brennan, T. & Sundaralingam, M. (1978). Crystal and molecular structure of yeast phenylalanyl transfer RNA. Strucure determination, difference fourier re®nement, molecular conformation, metal and solvent binding. Acta Crystallog. sect. B, 34, 1529± 1544. Triana-Alonso, F. J., Chakraburtty, K. & Nierhaus, K. H. (1995). The elongation factor 3 unique in higher fungi and essential for protein biosynthesis is an E site factor. J. Biol. Chem. 270, 20473± 20478. Tu, C., Tzeng, T.-H. & Bruenn, J. A. (1992). Ribosomal movement impeded at a pseudoknot required for frameshifting. Proc. Natl Acad. Sci. USA, 89, 8636± 8640. Vimaladithan, A. & Farabaugh, P. J. (1994). Special peptidyl-tRNA molecules can promote translational frameshifting without slippage. Mol. Cell. Biol. 14, 8107± 8116. Wahl, M. C., Rao, S. T. & Sundaralingam, M. (1996). The structure of r(UUCGCG) has a 50 -UU-overhang exhibiting Hoogsteen-like trans U U base pairs. Nature Struct. Biol. 3, 24± 31. Wang, Y.-X., Huang, S. & Draper, D. E. (1996). Structure of a U U pair within a conserved ribosomal RNA hairpin. Nucl. Acids Res. 24, 2666± 2672. Weijland, A. & Parmeggiani, A. (1994). Why do two EFTu molecules act in the elongation cycle of protein biosynthesis?. Trends Biochem. Sci. 19, 188 ± 193. Weiss, R. B. (1984). Molecular model of ribosome frameshifting. Proc. Natl Acad. Sci. USA, 81, 5797± 5801. Wettstein, F. O. & Noll, H. (1965). Binding of transfer ribonucleic acid to ribosomes engaged in protein synthesis: number and properties of ribosomal binding sites. J. Mol. Biol. 11, 35± 53. Woo, N. H., Roe, B. A. & Rich, A. (1980). Three-dimen. sional structure of Escherichia coli initiator tRNAMet f Nature, 286, 346 ± 351. Yarus, M. & Curran, J. (1992). The translational context effect. In Transfer RNA in Protein Synthesis (Hat®eld, D. L., Lee, B. J. & Pirtle, R. M., eds), pp. 319± 365, CRC Press, Boca Raton, FL. Yarus, M. & Folley, L. S. (1985). Sense codons are found in speci®c contexts. J. Mol. Biol. 182, 529± 540. Yarus, M. & Smith, D. (1995). tRNA on the ribosome: a waggle theory. In tRNA: Structure, Biosynthesis, and Function (SoÈll, D. & RajBhandary, U., eds), pp. 443 ± 469, American Society for Mibrobiology, Washington, DC. Edited by D. E. Draper (Received 13 May 1996; received in revised form 31 October 1996; accepted 31 October 1996)