Nucleic Acids Research

The E.coli tRNA^Ser isoacceptor set: recognition and identity

Escherichia coli SerRS recognizes five tRNA isoacceptors plus tRNA^Sec, none of whose anticodons are involved in recognition. Instead, the acceptor stem, D-arm and extra stem/loop nucleotides most strongly contribute to serine tRNA identity (50,65,66). Comparison of the sequences of the five E.coli serine tRNAs revealed that they share the absolutely conserved G1:C72 and G2:C71 base pairs, the conserved chemistry of purine-pyrimidine base pairs at positions 4:69, 5:68 and 7:66 and the AU and UA alternatives at position 3:70. The only bases that are absolutely conserved among the five tRNA^Ser isoacceptors, but never present in other two tRNAs with a long extra arm, which are presumably the best candidates for being misacylated by serine, are D20, G20B, G47C and G73. The in vivo identity switch experiments of Normanly et al. (50) have shown the importance of the discriminator base and bases from the first three pairs of the acceptor stem for serine identity. This was further emphasized by recent work on on tRNA^Ser minihelices (53) which revealed that the relationship between acceptor stem recognition and aminoacylation specificity strongly depends on the content and accessibility of information presented by the base pairs of A-form RNA helices (67). Based on previous tRNA^Ser footprinting (51) and X-ray crystallographic studies of the tRNA^Ser:SerRS complex (33) it was clear that SerRS primarily binds tRNA from its variable loop side. This orients SerRS toward the major rather than the minor groove of the acceptor stem helix. Regardless of the fact that the long variable arm of tRNA^Ser shows variation in both length and sequence within isoacceptors, it makes the largest contribution to the specificity of serylation, as noted by different experimental approaches (52,64,68,69). The most interesting observation is that the mechanism of recognition of this tRNA domain variesin different species. In Bacteria, it is recognized stem-length specifically, but not sequence specifically, which is the consequence of the conserved stem pairing pattern of the variable arm within the tRNA^Ser isoacceptors. Each member of the class 2 tRNA family has an unique orientation of the long variable arm. There are a few unpaired nucleotides between the possible stem of the variable arm and the base at position 48 (before the T[psi]C stem starts): none in tRNA^Ser, one in tRNA^Leu and two in tRNA^Tyr. While in the Leu system deletion of 1 bp caused only a small decrease in aminoacylation efficiency (70), the interaction with SerRS is impaired by shortening of the extra arm stem length in the cognate tRNA. There is direct interaction between the lower part of the stem and SerRS, which can be indirectly disturbed by a conformational change upon base pair deletion (64). Thus, regardless of the fact that the long variable arms of different tRNA^Ser isoacceptors in Bacteria are not similar to each other, their orientation is preserved due to conservation of other structural features, as are some of the D-arm nucleotides. Each of the three type 2 tRNA species has a characteristic D-loop sequence following the invariant bases G18 and G19. Bases 20A and 20B are inserted into the D-loop in prokaryotic tRNA^Ser and both play novel roles in tertiary interactions in the core of the tRNA. In particular, base 20B (which is absent in all eukaryotic cytoplasmic tRNAs^Ser, but present in archaeal serine isoacceptors) is stacked against the first base pair of the long variable arm and thus defines the direction of the latter. As shown by Himeno et al. (48), a base change at position 20B had a large effect on the specificity conversion from Tyr to Ser. Since neither the sequence in the D-loop, the stem-pairing pattern of the variable arm, the tertiary base pair 15:48, nor the nucleotide at position 59 in the T[psi]C-loop are involved in base-specific recognition by the synthetase, it was concluded that SerRS selectively recognizes tRNA^Ser on the basis of its characteristic tertiary structure (64). Thus, the tertiary structural variations among the type 2 tRNAs enable precise discrimination by their respective synthetases.

Recognition of tRNA^Ser in eukaryotes reflects changes in the type 2 tRNA subset

The structure of tRNA^Ser appears to vary extensively from species to species (21) which raises the question as to how the mode of tRNA discrimination has changed during evolution. There are significant differences between tRNA^Ser from yeast and from E.coli, notably in the acceptor stem sequences and the determinants that affect the orientation of the long variable arm. This could explain the low cross-acylation of yeast tRNA^Ser with E.coli SerRS. On the other hand, it has been demonstrated that yeast SerRS expressed in E.coli recognizes bacterial serine-specific tRNAs in vivo (55). Footprinting experiments (54) showed that, as in E.coli, the long variable arm of yeast tRNA^Ser makes contact with the synthetase, though protection of the acceptor stem was not observed. This is in agreement with in vitro experiments performed with yeast tRNA^Ser transcripts (57,58) which found the discriminator base to be unimportant for SerRS recognition; instead, its role in discrimination against misacylation by other synthetases, predominantly LeuRS, has been proposed. In contrast to the G2:C71 base pair which is conserved in E.coli, most Saccharomyces cerevisiae tRNA^Ser isoacceptors contain a G:U wobble pair at position 2:71, which could be involved in different types of interactions with the cognate synthetase than observed in bacterial systems. In Schizosaccharomyces pombe tRNA^Ser there is either a U:A or A:U base pair at the same position. The S.pombe tRNA^Ser isoacceptor with A2:U71 is efficiently aminoacylated by S.cerevisiae SerRS, while the E.coli enzyme failed to recognize it in vivo (56). Footprinting experiments also revealed a strong protection of the upper part of the anticodon stem (54), which together with the absolute conservation of the A27:U43 base pair in S.cerevisiae tRNA^Ser isoacceptors, suggests its involvement in the recognition process. It has not yet been experimentally proved that the interaction with the cognate synthetase requires the accessibility of particular chemical groups of the A27:U43 base pair, but we have previously shown that all the heterologous but cognate tRNAs that were efficiently recognized by S.cerevisiae SerRS, both in vivo and in vitro, possess this particular base pair. Another conserved nucleotide found in a pool of yeast and other eukaryotic serine specific tRNAs is U44. It is unpaired, located at the beginning of the extra arm and may influence the geometry of this domain and the interaction with cognate synthetases. On the other hand, different base pairing in the variable arm of prokaryotic and eukaryotic tRNAs^Ser, together with the absence of the nucleotide at position 20B in eukaryotes, may cause the formation of sufficiently different tertiary structures in eukaryotic serine-specific tRNAs, which cannot be recognized by bacterial SerRS enzymes. Recent work (48,57) suggests that in contrast to E.coli, where tertiary structural elements play a key role in discriminating from other class II tRNAs, such discrimination in yeast is more sequence dependent and less tertiary structure dependent. While in E.coli, every type 2 tRNA has a different number of unpaired nucleotides between the T[psi]C-stem and the first base pair of the variable arm, together with a different tertiary 15:48 base pair at the base of this arm in tRNA^Ser and tRNA^Leu, in S.cerevisiae tRNA^Leu and tRNA^Ser share the same number of unpaired nucleotides at the base of the long variable arm and the G15:C48 tertiary base pair. The simpler, i.e. less exclusive, recognition by yeast SerRS could be due to less constrained recognition by eukaryotic enzymes specific for type 2 tRNAs; they must reject only one kind of long-variable-arm tRNA (tRNA^Leu), while their bacterial counterparts have to reject two (tRNA^Leu and tRNA^Tyr). Thus, evolutionary adaptation toward less stringent recognition of cognate tRNAs may be the consequence of tRNA type switching by tRNAs^Tyr. Comparison of higher eukaryotic cytoplasmic serine isoaccepting tRNAs, both from animals and plants, revealed the conservation of certain structural elements that might influence the orientation of the long variable arm (20). This domain in human tRNA^Ser and tRNA^Sec functions as the major identity element in an orientation-dependent but not sequence-specific manner (60). The orientation can also affect the interaction of the bound synthetase with other identity elements. Some of the cytoplasmic leucine tRNAs in eukaryotes also contain a long extra arm in an orientation similar to that of tRNA^Ser, although their sequences are different from those of tRNA^Sec and tRNA^Ser. It seems plausible that an interaction can occur between the extra arm of leucine tRNA and seryl-tRNA synthetase, since there is no requirement of sequence specificity in the extra arm for SerRS recognition. Therefore it seems that at least some elements of tRNA^Ser tertiary structure, possibly different to those employed in eubacteria, are also important for recognition in eukaryotes. In agreement with this suggestion it has been observed that human amber suppressor tRNA^Ser functions in vivo with the cognate enzyme from yeast, but not with that from E.coli (56). In contrast, recent experiments showed that supF, an E.coli amber supressor tRNA^Tyr carrying the G73 mutation, fails to be misacylated with serine by yeast SerRS, but is a substrate for the E.coli enzyme, probably due to the greater similarity between tRNA determinants in the two bacterial tRNAs. In vitro experiments (59,71), revealed that exchange of the discriminator base A73 for G is alone sufficient to convert human tRNA^Leu into a serine acceptor in vitro. Since G73 is the only requirement for the identity swap between leucine and serine accepting tRNAs, it is probably involved in a direct interaction with SerRS. In Bacteria (48,70) and lower eukaryotes (48,57) this base serves only as an antideterminant. Taken together, the body of work on tRNA^Ser identity suggests that the mechanisms by which seryl-tRNA synthetases recognize the orientation of the variable arm as well as the acceptor stem have diverged during evolution. Such differences in the contributions of the various recognition elements to the specificity of aminoacylation and the mechanism of tRNA interaction with the cognate synthetases in various organisms have also been documented for several type 1 tRNAs (37-39). Perhaps the most notable gap in this area is the lack of any comparable data for archaeal tRNA^Ser recognition.

Recognition of selenocysteine tRNAs

Selenocysteine-inserting tRNA species can be found in all three kingdoms: they were originally discovered in Bacteria, but during more recent surveys of the various branches of the Archaea and Eukarya, tRNA^Sec genes have been detected in virtually all organisms examined (47). The discovery of selenocysteine tRNAs in the archaeal lineage (72) indicates that the principles of selenocysteine biosynthesis and co-translational insertion may have been established at a time before the divergence of the three lineages. The selenocysteine-inserting tRNAs are the longest tRNAs known to date, due to the length of their extra arms and their extended 8 bp acceptor stems. In addition, their primary structure deviates from the canonical tRNA consensus at several positions. Despite these differences, the three-dimensional model of the solution structure of tRNA^Sec is similar to that of canonical tRNAs, with tertiary structure stabilization achieved by a set of novel tertiary interactions (73). These structural peculiarities provide the basis for the specialized biochemical functions of tRNA^Sec. Although the tRNA^Sec species are charged with serine by seryl-tRNA synthetase (74) charging efficiencies were found to be 100-fold reduced compared with that of a canonical serine isoacceptor. The factors that may contribute to the low charging efficiency in E.coli are: the unusually long acceptor stem (8 bp), the deviation from the identity set of serine isoacceptors at position 11:24, and the special conformational features imposed by the unconventional mode of stacking the D-T loop region.

The primary structures of tRNAs^Sec from Bacteria, Archaea and Eukarya bear little sequence similarity, but the crucial features of their structure necessary for function are highly conserved. Thus, the divergence of the selenocysteine systems within the two lineages is not greater than that observed for other components of the translational apparatus. Although the long variable arm of tRNA^Sec reiterates the major recognition site for SerRS (60), the length of the acceptor stem is a major determinant for discrimination between tRNA^Ser isoacceptors and tRNA^Sec with respect to several other proteins which interact specifically with Sec-tRNA^Sec such as selenocysteine synthase, and the selenocysteine-specific elongation factor SELB.

EVIDENCE FOR PARAPHYLETIC ORIGIN OF SerRS FROM THE PRESENT-DAY ARCHAEA

The first archaeal SerRS sequenced, isolated and characterized was that of the extreme halophile Haloarcula marismortui (81). As noted earlier (78) preliminary phylogenetic analyses showed this enzyme to be most similar to its counterparts from Gram-positive Bacteria, and not to eukaryotic SerRSs as expected from the universal tree of life (12). Taupin et al. (81) also showed that H.marismortui SerRS serylates tRNA^Ser from E.coli, but not from yeast, which is in accord with both the structure of archaeal tRNA^Ser and the observed phylogenetic position of the enzyme.

The appearance of five other archaeal SerRSs in sequence data banks, from Archaeoglobus fulgidus, Methanococcus jannaschii, Methanococcus maripaludis, Methanobacterium thermoautotrophicum and, most recently, Pyrococcus furiosus, has now allowed a more stringent and reliable phylogenetic analysis of SerRS evolution. Inspection of preliminary multiple sequence alignments led to the conclusion that (i) parts of the alignment corresponding to the most variable parts of the sequence, the N-terminal coiled coil and the C-terminal extensions, are too ambiguous and should be excluded from the alignment; and (ii) due to the ambiguous positions of gaps in the alignment which are the result of intervening stretches of amino acids found in some sequences (especially those from the methanogens M.jannaschii, M.maripaludis and M.thermoautotrophicum) the positions with gaps were also excluded from the analysis. Under either maximum parsimony (Fig. 1A), neighbor-joining with correction for multiple substitutions invoked (Fig. 1B), and maximum likehood (Fig. 1C) methods the SerRSs from A.fulgidus and P.furiosus formed a clade with their eukaryotic counterparts, as expected. However, the SerRSs of H.marismortui and the three methanogens clustered with the bacterial enzymes 100% of the bootstrap replicates (82) (at least using the programs of the PHYLIP package, where this information is available), even though mere inspection of the sequence reveals that, while the SerRSs of methanogens have numerous `loops' not found in their other counterparts, the H.marismortui SerRS has a `normal' appearance. These sequences clustered most closely with SerRS of Gram-positive B.subtilis and the cyanobacterium Synechocystis sp. (from independent sources, Gram-positive bacteria and cyanobacteria are known to be sister taxa) although a reliable inference of their mutual positions in that part of the tree could not have been established by either method. These results indicate the paraphyletic origin of serS genes in today's Archaea. According to the present-day classification of Archaea, all six organisms (A.fulgidus, P.furiosus, H.marismortui and the three methanogens) belong to the group Euryarchaeota; the halophiles and the methanogens share a more recent common ancestor than the remaining two organisms. What seems most plausible is to postulate that this common ancestor of H.marismortui and the methanogens acquired its serS gene by horizontal (lateral) transfer from an eubacterial donor. It is possible that the original archaeal and the laterally transferred serS gene co-existed for a certain time in the course of evolution before the original gene was lost. The newly acquired enzyme was probably instantly functional, i.e. it serylated the host's tRNA^Ser immediately after its acquisition, and after the transfer it evolved so as to optimize its function under the extreme conditions, the most obvious feature being an increased proportion of aspartate and glutamate residues in the sequence of the extreme halophile H.marismortui SerRS (for explanation see e.g. 81). The function of numerous `appendages' acquired by the SerRS from methanogens is unknown.

A    B    C

Figure 1. Results of phylogenetic inferences based on the sequences of seryl-tRNA synthetases from different organisms. The alignment included 23 different sequences over their globular domains performed with Clustal_X (94). The gaps were excluded from the analysis in order to be able to include the SerRS sequences of methanogens, which contain numerous extra loops in comparison with other sequences. The organisms, with total length (in amino acids) of SerRS given in parentheses, are as follows: Arabidopsis thaliana (451), A.fulgidus (446), Bacillus subtilis (425), Caenorhabditis elegans (487), Coxiella burnetii (423), E.coli (430), Haemophilus influenzae (429), H.marismortui (460), Homo sapiens (514), M.thermoautotrophicum (513), M.jannaschii (521), M.maripaludis (514), Mycobacterium tuberculosis (419), Mycoplasma genitalium (417), Mycoplasma pneumoniae (420), P.furiosus (455), S.cerevisiae (462), S.cerevisiae, putative mitochondrial (446) S.pombe (450), Staphylococcus aureus (428), Synechocystis sp. (430), T.thermophilus (421), Zea mays, organellar (489). (A) Most parsimonious unrooted tree constructed from a bootstrap analysis of the SerRS alignment (100 data sets). The programs used were SEQBOOT, PROTPARS and CONSENSE of the PHYLIP package (95). Numbers at the branches correspond to percentage bootstrap frequencies for each branch (82). Only values >50 are shown. (B) Unrooted neighbor-joining tree constructed from a bootstrap analysis of the SerRS alignment using Clustal_X with correction for multiple substitution (1000 replicates; percentages >50% are shown in the figure). (C) Maximum likehood tree constructed using PROTML (96); the tree was obtained by star decomposition followed by evaluation of 16 user trees constructed by permuting sequence positions at the most uncertain nodes.

The proposed evolutionary event for the archaeal SerRS genes is not the sole example of such an event among genes coding for aaRSs. Other scenarios concerning the `late' evolutionary acquisition of aaRSs have been proposed for GlnRS and GluRS, where horizontal genetic transfer of the GlnRS gene from eukaryotes to Bacteria has been postulated (12,83).

SERYLATION IN ORGANELLES

In agreement with the bacterial origin of organelles, components of the organellar translational apparatus are structurally most like their prokaryotic counterparts. All serine-specific tRNAs, except those in animal mitochondria, possess a long variable arm comprising usually more than 14 nucleotides, with no unpaired nucleotides at the base of the extra arm. Since these tRNAs normally contain both 20A and 20B nucleotides, these positions may influence the orientation of the variable arm as in eubacteria. The results of phylogenetic analysis of tRNA^Ser sequences (B.Lenhard, unpublished results) suggest an especially interesting branching pattern for the organellar tRNAs^Ser: while mitochondrial tRNA^Ser of fungi and animals cluster together, plant organellar tRNAs^Ser form a separate clade that includes both mitochondrial and chloroplast sequences, as if there was selection in favor of common identitiy elements. It will be of great interest to find out the interrelation of the corresponding seryl-tRNA synthetases, or even to find out whether two separate organellar enzymes exist.

The primary structures of only two organellar seryl-tRNA synthetases are known thus far: one is from yeast mitochondria (P38705) and the other belongs to maize organelles, and is probably also mitochondrial (84). Phylogenetic analyses revealed that these organellar synthetases cluster together, and are phylogenetically closer to the bacterial SerRS enzymes. The structural resemblance between organellar seryl-tRNA synthetases and their bacterial counterparts is also evident from the lack of any C-terminal extension, which characterizes all cytoplasmic eukaryotic seryl-tRNA synthetases. We have recently shown the complementation of an E.coli serS mutant strain with the gene encoding maize organellar SerRS. Furthermore, its mature protein product overexpressed in E.coli efficiently aminoacylated bacterial tRNA^Ser in vitro, while yeast tRNA was a poor substrate (84).

Serylation in yeast mitochondria is especially intriguing, since the organelle contains three tRNA^Ser isoacceptors, which differ considerably in primary structure (85,86). tRNA₁^Ser has only 39 bp in common with tRNA₂^Ser. tRNA₂^Ser and tRNA₃^Ser are encoded by the same gene, but the mature tRNAs differ in their modification pattern. Phylogenetic analyses of the relation of tRNA₁^Ser to the other two isoacceptors tends to yield ambiguous results (B.Lenhard, unpublished observation), although both genes are also known to exist in some other fungi (based upon inspection of tRNA sequences). Both tRNA₂^Ser and tRNA₃^Ser should recognize all four UCN codons. As demonstrated earlier (87), only these two isoacceptors (UCN) were aminoacylated by E.coli SerRS, while tRNA₁^Ser (AGY) was not. Both types of mitochondrial tRNA^Ser isoacceptors contain a long variable arm, whose base pairing pattern and orientation probably differ due to a very different primary structure in the arm per se and in the D-loop of the tRNAs. It would be interesting to learn what the mechanism is of recognition of such diverse tRNA substrates with mitochondrial SerRS, which is, according to sequence alignment and modeling studies, rather similar to its bacterial counterparts. Such a unilateral aminoacylation specificity also exists between bovine mitochondria and eubacteria. It has been shown that e.g. bovine mitochondrial SerRS charges cognate E.coli tRNA species and misacylates non-cognate E.coli species, whereas their bacterial counterparts do not efficiently charge cognate mitochondrial tRNAs (88). The latter is probably due to a very specific tRNA^Ser structure in these organelles: unlike all other serine-specific tRNAs, only those from animal mitochondria do not possess a long variable arm, and in some cases (those for the AGY codons) the D-stem/loop is missing (22). The T-loop of this truncated tRNA^Ser is the main recognition site for the mitochondrial SerRS (89). Serylation in animal mitochondria has recently gained much attention, since the necessity for evolutionary adaptation may accordingly produce seryl-tRNA synthetases which differ substantially from all other SerRS enzymes, especially in the N-terminal domain.

IDENTITY SWITCHES AND TYPE-CONSTRAINED MISACYLATION AMONG TYPE 2 tRNAs

Since the long variable arm is an important recognition element for three aminoacyl-tRNA synthetases in E.coli (LeuRS, SerRS and TyrRS), it is apparent that these enzymes have to recognize particular sequences and/or orientations of this domain, as well as the additional determinants in their cognate tRNAs in order to avoid misacylation. The complete specificity change of tRNA^Tyr to tRNA^Ser in vitro was facilitated by insertion of three nucleotides into the variable arm (stem) of tRNA^Tyr plus two nucleotide changes at positions 9 and 73. Both G73 in tRNA^Ser and A73 in tRNA^Leu and tRNA^Tyr are phylogenetically conserved, implying an important role for this base during aminoacylation of all three tRNAs (70). In this context it is not surprising that a G73 mutant in the E.coli tRNA^Tyr amber suppressor, required no other structural changes in order to be misacylated with serine in vivo after increasing the level of bacterial SerRS (M.Nalaskowska and I.Weygand-Durasevic, unpublished results). Interestingly, similar misacylation was not observed upon overexpression of yeast SerRS in E.coli. While bacterial suppressor tRNA^Tyr inserts only tyrosine at amber codons in E.coli, this tRNA inserts leucine in S.cerevisiae (insertion of serine is probably prevented by an A73 antideterminant). Thus, the E.coli tyrosine tRNA is functionally a leucine tRNA in yeast cytoplasm, indicating that in evolution tRNA^Tyr is more closely related to a tRNA of different acceptor specificity, but of the same class type, than to one with the same amino acid specificity, but of a different class type. An obvious explanation would be that, after tRNA^Tyr switched from type 2 to type 1, eukaryotic (and possibly archaeal) LeuRS lost its ability to discriminate against tRNA^Tyr of type 2. However, since the determinants for identity between the E.coli and yeast mitochondrial tyrosine tRNAs are conserved (i.e. yeast mitochondrial tRNAs^Tyr are bacteria-like, comprising a long extra arm), E.coli tyrosyl-tRNA synthetase can substitute for yeast mitochondrial enzyme function in vivo (90). The ability of overexpressed GlnRS to misacylate variants of both E.coli tRNA^Tyr and tRNA^Ser amber suppressors (91) suggests that the long variable arm of type 2 tRNAs does not necessarily prevent misrecognition by synthetases specific for type 1 tRNAs, in the presence of other overlapping determinants (amber anticodon and/or altered acceptor stem).

CONCLUSIONS

The tRNAs with long extra arms are likely to have originated with the closure of the genetic code. One explanation for their role was that they allowed for specific recognition by the corresponding aaRSs without resorting to the anticodon (i.e. Ser and Leu have six anticodons each). However, the ancient existence of a long extra arm in bacterial tRNA^Tyr requires another explanation. One imaginable scenario might reside in the fact that TyrRS and TrpRS are structural isomers which separated much later than other aaRSs (92,93), and that tRNA might have played a more active role in establishing the amino acid specificity of these enzymes. The later evolutionary divergence of the two systems would then allow the loss of the extra arm in tRNA^Tyr of Eukarya.

The archaeal systems are especially interesting, with tRNAs in general being more bacteria-like than their corresponding aaRS. The case of SerRS in methanogens and halophiles mentioned above is a noteworthy exception, suggestive of early events in evolution. Furthermore, it is likely that co-evolution of aaRSs and their cognate tRNAs might account for some of their idiosyncratic features that cannot be otherwise explained.

ACKNOWLEDGEMENTS

This work was supported by grants from National Institutes of Health, NIH/FIRCA, International Centre for Genetic Engineering and Biotechnology (Trieste, Italy), Fondecyt (Chile) and Ministry of Science and Technology (Croatia).