Importance of the reverse Hoogsteen base pair 54-58 for tRNA function

Abstract

To elucidate the general constraints imposed on the structure of the D- and T-loops in functional tRNAs, active suppressor tRNAs were selected in vivo from a combinatorial tRNA gene library in which several nucleotide positions of these loops were randomized. Analysis of the nucleotide sequences of the selected clones demonstrates that among the randomized nucleotides, the most conservative are nucleotides 54 and 58 in the T-loop. In most cases, they make the combination U54-A58, which allows the formation of the normal reverse Hoogsteen base pair. Surprisingly, other clones have either the combination G54-A58 or G54-G58. However, molecular modeling shows that these purine–purine base pairs can very closely mimic the reverse Hoogsteen base pair U-A and thus can replace it in the T-loop of a functional tRNA. This places the reverse Hoogsteen base pair 54-58 as one of the most important structural aspects of tRNA functionality. We suggest that the major role of this base pair is to preserve the conformation of dinucleotide 59–60 and, through this, to maintain the general architecture of the tRNA L-form.

INTRODUCTION

One of the most conservative elements in the tRNA tertiary structure is the region at the outer corner of the tRNA L-form, where the T-loop interacts with the D-loop. This region, which we will henceforth call the DT region, is comprised of the whole T-loop, the first base pair of the T-stem 53-61 and nucleotides 18 and 19 of the D-loop, which interact, respectively, with nucleotides 55 and 56 of the T-loop (Fig. 1). Out of 11 nt of the DT region, only three, 57, 59 and 60, show a limited variability: 57 is always a purine, while 59 and 60 are pyrimidines in most cases (1). The other 8 nt of this region are invariable in cytosolic tRNAs. The DT region is involved in several important tRNA functions. First, it plays a major role in maintaining the perpendicular arrangement of the two helical domains called the L-form, which provides the proper juxtaposition of the two functional centers, the acceptor terminus and the anticodon. Also, this region is important for correct and efficient maturation of the termini of the molecule (2–4). Finally, it harbors recognition elements for interaction with different tRNA-binding enzymes, including some aminoacyl-tRNA synthetases (5–10).

Figure 1.

The standard tRNA L-form. Rectangles represent individual nucleotides. The DT region at the outer corner of the molecule is boxed. Cross-hatched and filled rectangles represent nucleotides of the D- and T-loop, respectively. Unpaired nucleotides as well as nucleotides at the beginning and the end of the helical regions are numbered in accordance with the standard tRNA nomenclature (1). Nucleotides of the anticodon loop, non-stacked nucleotides of the D-loop and nucleotide 47 are not shown. There are two base pairs, G18-Ψ55 and G19-C56, formed between the D- and T-loops. The reverse Hoogsteen base pair T54-A58, whose structure is seen in Figure 4, is formed within the T-loop. Dinucleotide 59–60 bulges from the double helical stem between base pairs G53-C61 and T54-A58. Nucleotide 59 stacks on the tertiary base pair 15-48, constituting the last layer of the D/anticodon helical domain. This interaction fixes the perpendicular arrangement of the two helical domains called the L-form.

The tertiary structure of the DT region is of special interest and has been the subject of a number of studies (11–14). The presence of such elements as the U-turn between T54 and C55, the unusual non-Watson–Crick base pairs T54-A58 and G18-Ψ55, the mutual intercalation of fragments 57–58 and 18–19, the bulging of nucleotides 59–60 and the interaction of phosphate 60 with the amino group of C61 makes this region one of the most structurally diversified in the whole tRNA. This diversity raises questions concerning the role played by each of these elements in the structure of the DT region and of the whole molecule and the limits within which these elements could be modified without destroying tRNA structure and function. These questions become even more important in view of the recent finding that rRNA also contains motifs resembling the structure of the DT region (15). Thus, elucidation of the role and the sequence requirements for formation of the elements constituting this region in such a relatively small molecule as tRNA would contribute to understanding of structure–function relationships in other RNAs and RNA–protein complexes, including the ribosome. To address this problem, we here undertook an analysis of the general constraints imposed on the structures of the D- and T-loops in a tRNA functioning in vivo. For this, we selected suppressor tRNAs from a specially designed combinatorial tRNA library in which a number of positions in the D- and T-loops were randomized. Analysis of the nucleotide sequences of the successful tRNA clones sheds light on the role of particular elements of the DT region in the global tRNA structure.

MATERIALS AND METHODS

Strains

The Escherichia coli strains TOP10 (Invitrogen) and XAC-1 (F′ lacI₃₇₃lacZ_u118 am proB⁺/F^– Δ(lac-proB)_x111 nalA rif argE_am ara) were used, respectively, for cloning and selection of the suppressor tRNAs. The XAC-1 strain contains amber mutations in the genes lacZ and argE (16).

Construction of the combinatorial library and selection of suppressor tRNAs

The template oligonucleotide coding for the combinatorial tRNA library (Fig. 2) was synthesized at BioCorp Inc. (Montreal, Canada), amplified by PCR to produce the double-stranded DNA using primers 5′-GCGAATTCGGGGCTATA-3′ and 5′-GACTGCAGTGGTGGAGT-3′, and cloned into plasmid pGFIB-1 using EcoRI and PstI restriction sites, as described previously (17). This plasmid provides a constitutive high level expression of a cloned tRNA gene (18). All enzymes were from New England Biolabs. Of 20 µl of the ligation mixture, 5 µl was electroporated into competent TOP10 cells, providing 4.5 × 10⁶ colonies, i.e. about a quarter of the sequence complexity of the library. The plasmid DNA was recovered using the Qiafilter Midiprep kit (Qiagen) and then transformed into competent XAC-1 cells. The positive clones were selected as blue colonies when grown on LB-agar containing ampicillin (100 µg/ml) and X-Gal (200 µl of the 20 mg/ml solution spread on top of each 150 × 15 mm plate). The plasmid DNA of these clones was extracted and retransformed into the XAC-1 cells to confirm the dependence of the phenotype on the presence of the plasmid. The ability of the selected tRNAs to suppress the nonsense mutation in gene argE was checked by plating the retransformed XAC-1 cells on minimal A medium without arginine.

Figure 2.

Construction of the tRNA gene library. In the nucleotide sequence of E.coli tRNA^Ala_UGC, each of the two enclosed regions, 16–19 in the D-loop and 54–58 in the T-loop, was replaced by six fully randomized positions, while nucleotide G20 and the anticodon TGC (boxed) were replaced by T20 and CTA, respectively. Nucleotides 54 and 58, which form the reverse Hoogsteen base pair in the T-loop, are connected by a line. The EcoRI and PstI restriction sites that are seen flanking the 5′ and 3′ termini were used for cloning the library into the pGFIB-1 plasmid.

Sequencing

Sequencing of the selected tRNA genes was performed on the LI-COR DNA sequencing system (Département de Biochimie, Université de Montréal) using primers 5′-GCTTCTTTGAGCGAACGATCAAAAATAAGT-3′ and 5′-GGGTTTTCCCAGTCACGACGTTGTAAAACG-3′ labeled at the 5′-end with IRDye 800 (LI-COR Biosciences).

Measurement of the β-galactosidase activity

β-Galactosidase activity of clones with suppressor tRNA genes was determined as described by Miller (19) using overnight cultures grown in A medium containing 0.4% glucose and 1 mM MgSO₄ to an A₆₀₀ of 0.8–0.9. The values were obtained by averaging the measurements from three independent cultures and calculated as a percentage of the activity of the control tRNA^Ala_su+.

Presence of the suppressor tRNAs in the cytosol and their aminoacylation level

To preserve the aminoacylated form of the tRNAs, the total cellular RNA was extracted under acidic conditions, as described previously (17). To obtain the deacylated tRNA, 4 µg of the total RNA was mixed with 1.5 µl of 0.5 M Tris (pH 9.0), incubated for 30 min at 37°C and deposited on an acid polyacrylamide gel (6.5% polyacrylamide, 8 M urea, 0.1 M sodium acetate) together with the untreated fraction. The gel was run for 24 h at 300 V at 4°C in 0.1 M sodium acetate, after which the part of the gel around the xylene cyanol dye was transferred by electroblotting to a Hybond-N nylon membrane (Amersham). The membrane was hybridized with two radiolabeled DNA probes, one complementary to region 26–44 of the suppressor tRNAs, consisting of the anticodon stem and loop, and the other to region 34–53 of the E.coli 5S rRNA. The 5S rRNA probe was used to monitor the amount of total RNA in each sample. The hybridization was performed overnight at 37°C in 7% SDS, 0.25 M Na₂HPO₄ (pH 7.4), 1 mM EDTA (pH 8.0), 1% BSA using a Robbins hybridization incubator.

Computer modeling

Preliminary modeling was done interactively, using the InsightII/Discover package (Version 2000; Accelrys Inc., San Diego, CA). The X-ray structure of the yeast tRNA^Phe (20) was used as a starting conformation. The presumed structures of RH-GA or RH-GG were appended to the T-stem replacing base pair U54-A58. The other randomized nucleotides were arranged in a way to resemble the structure of the DT region in the normal tRNAs and, at the same time, to provide a reasonable system of hydrogen bonds and base–base stacking interactions. Each model was submitted to unrestrained energy minimization using the AMBER force field (21) until an energy minimum was reached. Visualizations were done in a Silicon Graphics O2 computer.

RESULTS

The library design

The library was built from E.coli tRNA^Ala_UGC as a scaffold (Fig. 2). The choice was determined by the fact that the most important tRNA^Ala identity element for the cognate alanyl-tRNA synthetase, the G3-U70 base pair, was located in the acceptor stem, i.e. neither in the DT region nor in the anticodon, the sites that were modified in this study (22,23). This would minimize the role of interaction with a particular aminoacyl-tRNA synthetase as a factor in the tRNA selection. To enable the tRNAs to recognize the amber stop codon UAG, the anticodon TGC in the gene was replaced by CTA. All five nucleotides of region 54–58 of the T-loop, which were known to be involved in conserved interactions either within the loop or with nucleotides of the D-loop, were fully randomized. Correspondingly, four nucleotides 16–19 of the D-loop, which could be involved in interactions with the T-loop, were also fully randomized. To prevent nucleotide G20 from substituting for either G18 or G19 in their interactions with the nucleotides of the T-loop, it was replaced by T20. To stimulate the formation of alternative structural patterns in the DT region, we added one and two nucleotides to the randomized regions of the T- and D-loops, respectively. Thus, in the design, the T-loop contained 8 nt, one more than in the standard tRNA structure, while the D-loop had 10 nt, which is not unprecedented for the cytosolic tRNAs (1). Each loop had six randomized positions, providing for the total sequence complexity of a library of 1.7 × 10⁷ variants.

Cloning and selection of functional clones

The tRNA gene library was synthesized chemically, amplified by PCR and cloned into the pGFIB-1 plasmid, as described previously (17). The selection of active suppressor tRNA clones was done in the XAC-1 strain of E.coli, which had nonsense amber mutations in genes lacZ and argE. A successful suppression of the first mutation in the presence of 5-bromo-4-chloro-3-indolyl β-d-galactopyranoside (X-Gal) provides blue colonies, which was used for the primary identification of functional tRNA clones. Out of 3 × 10⁴ clones screened, several dozen positive clones were selected, whose suppressor activity was confirmed by a subsequent retransformation and by suppression of the second mutation in gene argE, which converts the arginine-dependent cells into prototrophs. The β-galactosidase activity was evaluated quantitatively for each clone and compared to that of the control tRNA^Ala_su+. The latter tRNA was derived from the normal tRNA^Ala by changing the anticodon from UGC to CUA and cloned in the same plasmid as the other suppressor tRNAs.

The nucleotide sequences of the selected tRNA clones, as deduced from their genes, are presented in Table 1. Only the sequences of those clones whose activity was at least 1% of the control are given. Comparison to the original design revealed six clones with a nucleotide deletion in the T-loop and three clones with a deletion in the D-loop, providing, respectively, for a seven member T-loop or a nine member D-loop. In two clones, K25 and K30, mutations affected the non-randomized part of the T-loop, deleting, respectively, U59 and C60. No other mutation outside the randomized regions was found. For eight clones arbitrarily chosen from Table 1, the in vivo level of the suppressor tRNA and of its aminoacylated form was determined by acid polyacrylamide gel electrophoresis followed by hybridization with a specific probe complementary to the anticodon stem and loop. For all suppressor tRNAs tested, the level in cytosol was detectable, although relatively low compared to that of tRNA^Ala_su+ (Fig. 3). For each clone, most of the tRNA was found in the aminoacylated form.

Table 1. The nucleotide sequences and the β-galactosidase activity of the selected tRNA clones.

Clone	D-loop	T-loop	β-Galactosidase activity (%)
T7-tRNAs
	1619	5458
	||	||
K30	AGUGAGGAUA	UCCAAAU	10.8 ± 1.1
K25	AGGAACGCUA	UGAAAAC	17.7 ± 6.2
K3	AGAACGAAUA	UGAAAUC	4.1 ± 0.3
K15	AGGCAUAUUA	UGAAAUC	11.0 ± 1.4
K29	AGGAAAAAUA	UGGGAUC	5.1 ± 1.0
K6	AGAGGGAGUA	GCACAUC	25.0 ± 5.0
T8-tRNAs
	1619	5458
	||	||
K26	AGAACGACUA	UAAACAUC	3.9 ± 0.8
K18	AGAACAAAUA	UAAACAUC	2.5 ± 0.6
K1	AGGAGAACUA	UAACCAUC	1.3 ± 0.1
K7	AGGACAAAUA	UAACCAUC	1.3 ± 0.2
K24	AGAAAAACUA	UAGCCAUC	6.0 ± 0.8
K5	AGCGAAGAUA	UAGCCAUC	1.7 ± 0.3
K20	AGGAGAUCUA	UAGCCAUC	3.2 ± 0.2
K27	AGUGAAAUUA	UAGCCAUC	9.9 ± 2.0
K19	AGA-CAACUA	UAUACAUC	2.0 ± 0.4
K2	AGAAAGACUA	UGACGAUC	7.9 ± 1.7
K23	AGUAAGGUUA	UGCCAAUC	5.9 ± 1.1
K9	AGAGCGAAUA	GACGCAUC	1.3 ± 0.3
K17	AGAGGCCAUA	GAGCCAUC	6.5 ± 1.8
K4	AGA-CGGGUA	GCACAAUC	1.1 ± 0.1
K28	AGGGCAAAUA	GCAGCAUC	2.9 ± 0.6
K21	AGUGAAAGUA	GCCACAUC	2.8 ± 0.5
K31	AGAGAGGGUA	GCCCAAUC	6.3 ± 0.8
K10	AGA-AGGAUA	GUACCAUC	5.9 ± 2.2
K14	AGGAGGGAUA	GGACCGUC	4.1 ± 0.5
K13	AGAGGAAAUA	GUACCGUC	1.3 ± 0.2
K16	AGGGGGAUUA	GUCAAGUC	4.8 ± 1.2
K32	AGUCGGUAUA	GUCGAGUC	38.5 ± 1.8

The sequences are deduced from the genes. Only the D- and T-loops, where the sequences differ one from the other, are given. Positions of nucleotides 16–19 in the D-loop and 54–58 in the T-loop mark the beginning and the end of the each randomized region. Nucleotides forming the RH base pair in the T-loop are underlined. The activity of tRNA^Ala_su+ is taken as 100%.

Figure 3.

Northern blot showing the presence in the cytosol and the level of aminoacylation of some suppressor tRNAs. For each clone the – and + lanes correspond to the samples not treated and treated with Tris. In the – lanes the aminoacylated and deacylated forms of the suppressor tRNA move as individual bands, while in the + lanes the total tRNA is deacylated and the suppressor tRNA moves as one band. In all – lanes the bands corresponding to the aminoacylated form of the tRNA are much larger than those corresponding to the deacylated form and are comparable to the bands in the + lanes, representing the total amount of the suppressor tRNA. This indicates that in all clones most of the tRNA is present in the aminoacylated form. A smaller size of the bands of the suppressor tRNAs compared to tRNA^Ala_su+ indicates a notably lower presence of the selected tRNAs in the cytosol. 5S rRNA was visualized to monitor the amount of total RNA in each sample. Because the signal from 5S rRNA was much stronger than that from suppressor tRNAs, the upper and lower parts of the same membrane have been exposed, respectively, for 4 h and overnight. The nucleotide sequence of clone K8, due to its low β-galactosidase activity, is not included in Table 1, but is available upon request.

Analysis of the nucleotide sequences

In the experiments described above, on average, only one in every 1000 clones showed a detectable level of nonsense suppression activity. This indicated that the nucleotide sequence space available for the DT region was rather small. A systematic analysis of the sequences of the selected clones could help to reveal the rules imposed on the structure of this region in functional tRNAs.

We started the analysis with the ‘quasi-normal’ clones, those that had the normal 7 nt in the T-loop. Henceforth, we will call such molecules T7-tRNAs, in contrast to T8-tRNAs which have 8 nt in this loop. Analysis showed that in T7-tRNAs, the fifth position of the T-loop was always occupied by A and was the only invariable position in both randomized regions (Table 1). The second most conservative nucleotide was the first one of the T-loop, which in all sequences except K6 was U. The presence of U and A, respectively, in the first and the fifth position of the T-loop would allow the formation of the reverse Hoogsteen base pair U54-A58 (RH-UA), as in the normal tRNAs. Although in the normal tRNAs U54 is always modified to T, it is not yet known whether it is the case in the selected tRNAs. On the other hand, because this modification does not interfere with the ability of the base to form hydrogen bonds, its absence would not affect the formation of base pair U54-A58. Another conservative feature consistent with wild-type tRNA is the presence of a purine in position 57 of all but one T7-tRNA. Other randomized nucleotides, including all 6 nt in the D-loop, were essentially more diversified and did not seem to provide for a common structural pattern.

Among T8-tRNAs, half of the sequences (11 out of 22) also contained U in the first position of the T-loop (Table 1). If this U plays the same role as it does in T7-tRNAs, there should be an A a few nucleotides later that is able to form RH-UA with this U. Generally, this A could occupy either the fifth or the sixth position of the T-loop, depending on the position of the additional eighth nucleotide. Analysis showed that in those T8-tRNAs whose T-loop started with U the only other conservative nucleotide was the A occupying the sixth position of the same loop. Thus, the formation of RH-UA by these two nucleotides would place the additional nucleotide in the region between them.

In all other T8-tRNAs, the first position of the T-loop was occupied by G (Table 1). If this G plays a structural role analogous to that played by U, its possible partner would occupy either the fifth or the sixth position of the T-loop. Neither of the two positions was conserved in these sequences: the fifth nucleotide was allowed to be either C or A, while the sixth was either A or G. To explore the abilities of both the fifth and the sixth nucleotides to pair with the first G, we looked for possible arrangements of three different dinucleotide combinations, GC, GA and GG, that would be close to the arrangement of U and A in RH-UA. For GC, we did not find any satisfactory arrangement. However, for both combinations GA and GG we found arrangements that are presented in Figure 4. In these arrangements, the G that is equivalent to U in RH-UA donates two hydrogen atoms for formation of hydrogen bonds with atom N7 of the other purine. This purine can be either A or G. In the latter case, an additional hydrogen bond can be formed between N2-H and O6 of the first and second G residues, respectively. The two arrangements GA and GG are superimposable in the sense that if one overlaps the positions of the glycosidic bonds of the first nucleotides, the glycosidic bonds of the second nucleotides in both arrangements would occupy about the same position. In the same sense, these two arrangements are fairly close to RH-UA. Accommodation of any of these arrangements based on the standard RH-UA would require a shift and rotation of one of the bases by only 1.5 Å and 20°, respectively. Therefore, a replacement of RH-UA in the T-loop by either GA or GG would require only relatively minor changes in the positions of the neighboring nucleotides. To reflect the closeness of these GA and GG arrangements to RH-UA, we will call them RH-GA and RH-GG, respectively.

Figure 4.

Juxtaposition of the bases in RH-GA, RH-GG and other alternative base pair candidates for replacement of RH-UA. (A) Positions of the glycosidic bonds in the alternative base pairs compared to that in RH-UA. In each base pair the position of the glycosidic bond corresponding to the base on the right is superimposed on that of U in RH-UA. The glycosidic bond of the other nucleotide will thus occupy a particular place depending on the structure of the base pair. The numbers indicating particular positions of the glycosidic bonds correspond to the base pairs in (B).

Further analysis revealed a few additional nucleotide combinations like CA and AA seen in Figure 4 that could also be arranged relatively closely to RH-UA while having a reasonable system of hydrogen bonds. Still, all these additional combinations were more distant from RH-UA than RH-GA or RH-GG and, therefore, their incorporation into the T-loop instead of RH-UA would cause greater changes in the conformation of the whole DT region. This latter aspect was expected to render these combinations less preferable in this place than RH-GA or RH-GG.

The fact that GG and GA can be accommodated close to RH-UA, while GC cannot, makes the sixth rather than the fifth nucleotide of the T-loop in T8-tRNAs the most probable partner to form a base pair with the G occupying the first position of this loop. As to the other randomized nucleotides in both loops of T8-tRNAs, they, as in T7-tRNAs, were much more diversified and did not seem to provide for a common structural pattern. Finally, we can consider K6, the only T7-tRNA clone whose T-loop starts with G rather than with U. This clone also has A in the fifth position of the T-loop, which would allow these G and A to form RH-GA, analogous to RH-UA existing in all other T7-tRNAs.

From this analysis, a clear picture emerges: in all selected tRNAs the first and the last randomized positions of the T-loop are always able to form a RH base pair, i.e. either RH-UA, RH-GA or RH-GG. The last randomized position is either the fifth in the T7-tRNAs or the sixth in the T8-tRNAs. The region between the first and the last position varies in length and sequence and does not seem to have a common pattern.

Modeling of the tRNA structures

To confirm that the exchange of RH-UA for either RH-GA or RH-GG in the T-loop did not cause any steric problem, we modeled the structure of the DT region for several clones having either RH-GA or RH-GG. After unrestrained energy minimization, the bases constituting the RH base pair always retained their juxtapositions and the inter-base hydrogen bonds, as one can see in the example of the model for clone K31 (Fig. 5). Comparison of the models with the structure of the yeast tRNA^Phe (20) showed that the whole region that included the T-stem and the RH base pair (RH-GA or RH-GG in the models and RH-UA in tRNA^Phe), as well as nucleotides 59 and 60, was superimposable in all structures.

Figure 5.

The model of the structure of the DT region for clone K31 (red) superimposed on the corresponding region in yeast tRNA^Phe (green). The figure also includes the T-stem and the tertiary base pair 15-48. For both tRNAs, the ribbon follows the sugar–phosphate backbone. Explicitly shown are base pairs 15-48 and 54-58 and nucleotide 59 in tRNA^Phe, as well as all nucleotides of the DT region and pair 15-48 in K31. Comparison of the modeled structure with tRNA^Phe demonstrates a good superposition of the T-stem and base pairs RH and 15-48, as well as nucleotide 59. The proper arrangements of the nucleotides in the RH base pair thus guarantees the proper position of nucleotide 59, whose stacking to base pair 15-48 would fix the juxtaposition of the two helical domains known as the L-form. Still, one can notice a difference in the conformation of the backbone in the two structures, which is highest between nucleotides 58 and 59. Such a difference makes a universal interaction of this region with a particular protein factor unlikely.

DISCUSSION

The results presented here show that a tRNA could be functional even if the structure of its DT region is substantially modified compared to the standard. Although for all selected clones the efficiency of the nonsense codon suppression was lower than for tRNA^Ala_su+, it was strong enough to provide a level of β-galactosidase synthesis sufficient to change the color of the colonies in the presence of X-Gal and to allow cell growth without external arginine. Additional examination of several clones showed that suppressor tRNAs had a detectable in vivo level and existed mainly in the aminoacylated form.

The nucleotide sequences of the selected suppressor tRNAs demonstrated a range of diversity never seen in the natural cytosolic tRNAs. In spite of this, the selected tRNAs constituted only a tiny fraction of the whole tRNA gene library, which implied the existence of strong constraints imposed on the structure of functional tRNAs. To elucidate these constraints, we undertook a comparative analysis of the nucleotide sequences of the selected tRNAs. It may be a little surprising that among the selected clones there were no clones having the wild-type sequence pattern. On the other hand, the wild-type sequence G18-G19-…-U54-U55-C56-R57-A58 is expected to appear on average only once in 8000 clones. Moreover, this probability can easily get beyond the technically detectable level if some additional requirements are imposed on the identity of the nucleotides flanking the conservative dinucleotide G18–G19 in the D-loop and on the additional eighth nucleotide in the T-loop. For most of the randomized nucleotides, our analysis did not find any obvious common pattern. The only exception consisted of the first and the last randomized positions in the T-loop, which were always able either to form RH-UA, analogous to base pair U54-A58 in the normal tRNAs, or to mimic it closely via formation of RH-GA or RH-GG. Modeling experiments showed that a replacement of RH-UA with either RH-GA or RH-GG did not cause any major rearrangement in the conformation of the DT region and provided for stable, sterically reasonable tRNA structures. Because such an RH base pair can be formed in all selected tRNAs, its existence is judged to be one of the most important requirements imposed on the structure of the DT region in a functional tRNA. In fact, this requirement has been the only one satisfied in all selected tRNAs, which allows us to conclude that the preservation of a RH base pair in the T-loop is more important for tRNA function than that of other universal elements, including inter-loop base pairs G18-Ψ55 and G19-C56.

Different explanations of the importance of the RH base pair for tRNA function can be suggested. For example, this base pair could be involved in a specific, vitally important interaction with a protein or other factor and thus should be preserved as such. However, a specific interaction like this would probably not tolerate an exchange of RH-UA for either RH-GA or RH-GG, because the juxtaposition of the glycosidic bonds and, therefore, the conformation of the backbone in the two latter base pairs, however close it is to that in RH-UA, is still notably different. Moreover, the three base pairs have different chemical groups exposed on the surface and thus are unlikely to be recognized by the same factor. In another, more probable explanation, an RH base pair is needed to stabilize a particular conformation of a neighboring region in the tRNA structure and thus to enable this region to serve its function. We do not expect this region to include the top of the T-loop closed by the RH base pair or the randomized part of the D-loop. Indeed, in the selected tRNAs, these regions vary in length and in nucleotide sequence and do not seem to have a universal structure. There is, however, another region, dinucleotide 59–60 at the end of the T-loop, whose position may need a particular structure of the RH base pair. This dinucleotide bulges from the double helical stem between base pairs 54-58 and 53-61. Therefore, its conformation is to a great extent determined by the positions of the flanking nucleotides 58 and 61, which, in turn, depend on the structures of base pairs 54-58 and 53-61, respectively. Thus, the presence of base pair 54-58 with the correct juxtaposition of the bases is necessary for the proper positioning of dinucleotide 59–60. The reverse Hoogsteen base pair U54-A58 perfectly suits this purpose. However, as our results show, combinations GA and GG can also be arranged in an appropriate way. In the modeled tRNAs containing these combinations, the position of dinucleotide 59–60 remains virtually the same as in the standard tRNA structure (Fig. 5).

We have already suggested (24) that the bulged dinucleotide 59–60 plays a crucial role in maintenance of the general shape of tRNA. Indeed, in the normal tRNAs, nucleotide 59 stacks on the tertiary base pair 15-48, which constitutes the last stacked layer of the D/anticodon helical domain (Figs 1 and 5). This interaction determines the juxtaposition of the two domains, the D/anticodon and the acceptor/T, i.e. the general geometry of the tRNA L-form, and is invariable in all normal tRNAs (24,25). Thus, the importance of maintaining the standard juxtaposition of the helical domains within the tRNA L-form would justify the necessity to preserve the conformation of dinucleotide 59–60 via formation of the reverse Hoogsteen base pair 54-58. Because the bases of nucleotides 59 and 60 are mainly involved in stacking interactions between themselves and with nucleotides 15 and 48, and not in hydrogen bonding, their identity is not that important. This reflects the partial variability of nucleotides 59 and 60 in the normal tRNAs (1), and also fits the fact that replacements U59A in clones K25 and K30 and C60U in clone K30 did not impair the tRNA function (Table 1).

The importance of this base pair for tRNA function, especially compared to other randomized elements of the DT region, correlates well with the data of Nazarenko et al. (26) on the efficiencies of mutants of yeast Phe-tRNA^Phe in different partial reactions of the tRNA functional cycle. According to these data, the efficiencies of the mutants in tertiary complex formation with factor Tu and GTP, in binding to the A and P site of the poly(U)-programmed ribosome and in peptide formation, are generally more sensitive to nucleotide replacements in pair 54-58 than in pairs 18-55 and 19-56. There have also been reports of using tRNA libraries with randomized positions in the DT region for selection of clones in vitro by affinity to either the phenylalanyl- or glutamyl-tRNA synthetase and to the EF-Tu factor (27,28). Interestingly, in none of these studies was the necessity for a RH base pair in the T-loop detected. However, because different steps of the tRNA functional cycle are probably not equally dependent on the proper position of the two helical domains, concentration on only some steps of this cycle would not necessarily reveal all sequence requirements for a fully functional tRNA. Moreover, the mutations in the tRNA clones selected by affinity for a particular protein may be detrimental not only for other steps of the functional cycle not involved in this selection, but even for this very step if, for example, they hinder the dissociation of the complex (29,30). We here used an alternative approach of tRNA selection in vivo based on its suppressor activity. This guarantees that the selected clones are correctly transcribed, processed and folded, that they are able to interact productively with all necessary factors of the protein biosynthesis machinery, including the aminoacyl-tRNA synthetase and EF-Tu, as well as the UAG-charged ribosome. The sequence requirements revealed in this way have been balanced between all these steps.

Another essential aspect of our approach relates to the original design of the library. As one can see in Table 1, among T7-tRNAs only clone K6 does not have RH-UA, while among T8-tRNAs there are more than half of such clones. It may reflect the existence of certain constraints on fitting either RH-GA or RH-GG into the seven member T-loop, which would relax when an additional eighth nucleotide is added. Therefore, if in the T-loop of the original design the normal 7 nt were preserved, most probably we would have seen an overwhelming majority of the selected clones having RH-UA, only sporadically intermingling with those having RH-GA or RH-GG. This would have made it more difficult to understand that it is the conformation of the sugar–phosphate backbone rather than base pair U54-A58 per se which is crucial for tRNA function. Although the introduction of an additional base into the T-loop probably has some negative effect on the suppressor activity, most importantly it increases the chance of selecting alternative structures. The elucidation of the common structural pattern in the selected tRNAs designed in this way has allowed us to recognize the RH base pair as the most important aspect that must be preserved, even when the structure substantially deviates from the standard.

In contrast to the nucleotides composing the RH base pair, other randomized nucleotides do not seem to have a common structural pattern. This means that the existence of a particular universal structure of this region is not required for tRNA function. On the other hand, it does not mean that this region is not structured or does not play any functional role. Instead, there are indications that it has a particular structure and that this structure is important for tRNA function, even if it is not the same in all selected tRNAs. First, as discussed above, the seven member T-loop may limit the use of RH-GA and RH-GG, so that these latter base pairs are able to replace RH-UA in only a fraction of all successful T7-tRNAs. This ability will thus depend on the identity of other randomized nucleotides. Also, the activities among the selected clones differ by almost 40 times, despite the presence of a RH base pair in all of them. This indicates the existence of other factors within the tRNA structure that affect the activity. Because these clones differ one from another only in the randomized regions, we have to conclude that other randomized nucleotides not involved in the RH base pair play a role in tRNA function. Still, an overall inspection of the nucleotide sequences shown in Table 1 has provided no obvious consensus pattern able to explain the differences in the activity. In normal tRNAs, the randomized nucleotides not involved in the RH base pair take part in the interactions between the D- and T-loops. These interactions affect the overall stability of the tRNA tertiary structure, but may also be directly involved in a particular tRNA-related process. Similar interactions could exist in the selected suppressor tRNAs as well, although they would be different in different clones. An example of such interactions is seen in the model of K31 in Figure 5. A systematic analysis of the possibility of forming these interactions in the selected tRNA clones is now in progress and will be published elsewhere. One should admit, however, that this analysis may not be able to explain the existing differences in tRNA activity. Indeed, there are many reasons that could negatively affect tRNA activity, such as formation of an alternative secondary structure, a higher susceptibility to a nuclease activity, a too low or too high affinity for the cognate aminoacyl-tRNA synthetase or for the elongation factor Tu, etc., and for each clone, the real reason can be different. A complete understanding of this phenomenon needs an analysis of the behavior of individual tRNA clones at each step of their functional cycle.