Correct expression of the genetic code at translation is directly correlated with tRNA identity. This survey describes the molecular signals in tRNAs that trigger specific aminoacylations. For most tRNAs, determinants are located at the two distal extremities: the anticodon loop and the amino acid accepting stem. In a few tRNAs, however, major identity signals are found in the core of the molecule. Identity elements have different strengths, often depend more on kcat effects than on Km effects and exhibit additive, cooperative or anticooperative interplay. Most determinants are in direct contact with cognate synthetases, and chemical groups on bases or ribose moieties that make functional interactions have been identified in several systems. Major determinants are conserved in evolution; however, the mechanisms by which they are expressed are species dependent. Recent studies show that alternate identity sets can be recognized by a single synthetase, and emphasize the importance of tRNA architecture and anti-determinants preventing false recognition. Identity rules apply to tRNA-like molecules and to minimalist tRNAs. Knowledge of these rules allows the manipulation of identity elements and engineering of tRNAs with switched, altered or multiple specificities.
Transfer RNAs are the interface between DNA encoded genetic information and its expression in proteins. Once Crick proposed the ‘adapter hypothesis’ (1), tRNAs were discovered 3 years later (2). This led to a desire to understand how the amino acid activating enzymes, the aminoacyl-tRNA synthetases (aaRS), developed (3) and how they recognize tRNAs (4). The challenge was to decipher the recognition sites on the tRNAs which allow unique reading of the genetic code. This pivotal problem in biology was long referred to as a recognition problem and is now known as the tRNA identity problem. The term ‘identity’ dates back to 1965 (5), and identities were tackled by a variety of methods. These include enzymatic dissection and chemical modification of tRNAs, sequence comparison of isoacceptor tRNAs including mischarged tRNAs and the genetics of suppressor tRNAs (6–8). No unified picture emerged from the early work and tRNA aminoacylation systems appeared idiosyncratic. Nonetheless, attempts were made to find links between them as in the proposal of a discriminator site at N73 (9) and its conservation among molecules charged or mischarged by a same synthetase (7). The idea of a second genetic code was also discussed (4,10,11); it implies conservation of recognition signals in evolution and was rejuvenated (12–14) after the explicit deciphering of the Ala identity (15,16).
Because anticodon nucleotides specify the relationship between an amino acid and a trinucleotide of the genetic code, they were considered privileged candidates to specify recognition by the synthetases. The first experimental fact supporting this view dates to 1964. Kisselev's group showed inactivation of yeast tRNAVal upon chemical modification of its anticodon, an assumption confirmed in Bayev's laboratory (reviewed in 17) using dissected molecules. Evidence for a role of the anticodon in the acceptor function of other tRNAs was obtained for Gly, Met, Phe and Trp specificities. For several others (Glu, Gln, Tyr) the likelihood of anticodon participation in synthetase recognition was high (17). On the other hand, the region near the amino acid −CCAOH accepting end was another obvious recognition candidate. Support for this expectation came from Chambers' laboratory as they used a dissected tRNA to demonstrate a recognition signal for yeast AlaRS embedded in the tRNAAla acceptor stem, likely within its three first base pairs (6).
An important aspect in all discussions on tRNA identity is the link between specific tRNA-aaRS complex formation and catalysis leading to tRNA aminoacylation. Studies of unspecific tRNA-aaRS interactions and comparisons of cognate tRNA charging with tRNA mischarging led to the concept of kinetic specificity. Correct tRNA aminoacylation is governed more by kcat effects than by Km effects (7). Specific complex formation is accompanied by conformational changes of the interacting partners (18,19). In a broader perspective, specificity of tRNA aminoacylation relies also on that of the amino acid activation step, on correction mechanisms and on the balance between the concentrations of tRNAs and synthetases that favor cognate aminoacylation and disfavor mischarging (20,21). Discovery in the early 1970s, of tRNA-like molecules (22) that differ in sequence from canonical tRNAs but are substrates of synthetases was puzzling. These molecules include structures present at the 3′ extremities of plant viral RNAs (23) as well as other mimics more recently discovered (24–27). A global description of tRNA identity has to integrate all these facts. Additional information on early literature has been reviewed (6,17,28–31).
Definitions, General Concepts and Prerequisites
It is a general belief that tRNA identity is governed by positive (determinants) and negative (anti-determinants) elements that respectively trigger specific aminoacylation and prevent false charging. In each tRNA, positive elements are limited in number and constitute the identity sets. They include isolated nucleotides in single-stranded regions, base pairs in helices, and can also be structural motifs. A priori, conserved and semi-conserved residues responsible for tRNA architecture are not part of identity sets. This is, however, not a general rule. In the literature the term ‘identity element’ often designates elements found in vivo, while the term ‘recognition element’ refers to elements determined in vitro. For simplicity we utilize the terminology identity element for both in vivo and in vitro data. Also, one has to distinguish between ‘major’ and ‘minor’ elements. The distinction is not easy and is often subjective. In this survey we consider as major elements those which, upon mutation, lead to the strongest functional effects, either when tested in vitro (strongest decrease in aminoacylation efficiency) or in vivo (strongest suppression effects). In contrast, minor elements have moderate effects and tune specificity. These elements can act by indirect mechanisms. Altogether, tRNA structure can be considered as a scaffold that permits proper presentation of identity signals to synthetases. After amino acid activation by a synthetase, the catalytic site of the enzyme will be completely switched on by the correct interaction with the tRNA, and tRNA charging can proceed. The implication, a priori, is that different types of RNA scaffolds can achieve functional activation of synthetases.
Although this classical view on identity accounts for many of the observed effects, it was recently disturbed with unexpected discoveries such as the existence of alternate identity sets, permissive elements and cryptic determinants (see below). In fact, expression of identities is more subtle than was anticipated and appears to be an intricate interplay of molecular events where mutual adaptation of tRNA and synthetase play a crucial role.
Several technical breakthroughs were essential for deciphering tRNA identities. Methods using RNA polymerases for the in vitro transcription of tRNA genes permit preparation of almost all kinds of mutants. The SP6 RNA polymerase was first used for preparation of functional tRNA-like domains from plant viral RNAs (32) and soon after, Uhlenbeck and colleagues popularized the use of T7 RNA polymerase for the systematic synthesis of tRNA variants (33,34). The transcriptional methods yield tRNAs deprived of modified nucleotides, which can be a drawback if epigenetic modifications on native tRNAs play a role in identity. In that case, chemical synthesis of RNA can provide molecules with modified residues. The search for identity elements in vivo took advantage of suppressor genetics. A reporter system utilizes a dihydrofolate reductase (DHFR) gene with an amber mutation at position 10 that can be read by any engineered suppressor tRNA provided it is recognized by one or several synthetases (35). Characterization of the amino acids incorporated in DHFR at position 10 gives then direct information on the identity(ies) of the suppressor tRNA. The drawback is that most anticodon identity elements cannot be checked by this methodology. An alternate system using mutants of initiator tRNAMet with unusual anticodon sequences removes this drawback (36,37). A difficulty is the choice of the mutations to introduce in a tRNA. For experimental simplicity and because of the great combinatorial diversity of possible variants, workers in the field have essentially studied single mutants. Doing so, possible combinations of nucleotides that could have functional significance escape detection. Recent studies have shown that such combinations actually exist (see below).
Depending on how identity is achieved, it follows that specificity of a tRNA can be changed by transplantation of its identity set into another tRNA background. Transplantations help to verify completion of identity sets. Sets are considered as complete if their transplantation yields molecules as active as the wild-type tRNA. But caution is required in interpretation, since determinants can escape detection if present in the host tRNA in which functionality of the putative identity set is assayed. In vivo, one might consider that mutations in a suppressor tRNA that yield 100% incorporation of a given amino acid in the reporter protein, fully define the identity set for this amino acid specificity. Identity is not defined by physical-chemical properties, but by functional criteria where the balance between correct recognition by one synthetase and non-recognition by the other synthetases play key roles. Also, the in vivo approach does not allow screening of all possible tRNA variants with the potential to interact with a synthetase. A mutation introduced in a suppressor tRNA can affect the recognition of this tRNA by other partners of the translational machinery. Identity sets defined according to in vitro or in vivo criteria, therefore, are not necessarily identical.
Different aspects of tRNA identity have been reviewed in literature (31,38–47). Here we discuss the body of available data in an attempt to uncover universal rules and to understand idiosyncratic features in aminoacylation systems. Deciphering identities relies on functional studies of engineered tRNAs and their understanding is dependent upon advances in structural biology. Thus, the accumulation of tRNA sequences (48) and a better knowledge of the structure of tRNAs (31,49,50), synthetases and their complexes with tRNA (51–60) were pivotal in perceiving interrelations between aminoacylation systems within a given phylum and across phylogenetic lines as well as to understand the origin of tRNA identities.
A compilation of identity elements found to date in cytosolic tRNAs, arranged according to the classification of synthetases in two classes (61,62) and their division into subgroups (55,56), is given in Table 1. Identity elements have been determined for the 20 aminoacylation systems from Escherichia coli, 14 systems from Saccharomyces cerevisiae, four systems from Thermus thermophilus and a few systems from other organisms, often higher eukaryotes including humans. They are in most cases standard nucleotides. Modified nucleotides are identity elements in tRNAs specific for Ile, Glu and Lys in E.coli and Ile in yeast. These modifications are exclusively located in the anticodon loop. The scarce involvement of modified nucleotides in tRNA aminoacylation is fortunate and explains why the wide use of unmodified transcripts for deciphering identities by the in vitro approach was successful.
The distribution of identity nucleotides in tRNA is shown in Figure 1 for the 20 E.coli aminoacylation systems, with the distinction between class I and class II synthetases. In both cases, identity elements lie predominantly at the two distal ends of the tRNA, with a strong participation of anticodon residues and discriminator nucleotide N73 together with the most distal base pairs of the amino acid accepting stem. Position 37 in the anticodon loop is only involved in identities of tRNAs charged by class I synthetases. The most distal residues (N73 and the three anticodon nucleotides) are identity elements for most tRNAs. Identity determinants in the tRNA core (nt 8–31 and 39–65) are more system-dependent and are scattered over 21 positions (see the small size of the spheres in Fig. 1A at positions 8, 10–15, 20, 20a, 22–24, 29, 41, 46 and 48, and in B at positions 10, 11, 15, 20, 24, 25, 44, 45, 48, 59 and 60). Such determinants participate in six class I (Ile, Leu, Cys, Glu, Gln and Arg) and three class II (Ser, Pro and Phe) identities. Identity base pairs in anticodon stems were found in four systems (Ile, Ser, Pro and Phe). A few tRNAs with sequence peculiarities utilize them as identity elements: this is the case of the large variable loop in tRNASer, of residue N-1 in tRNAHis and of the atypical G15·G48 Levitt pair in E.coli tRNACys. Altogether, 40 positions have been detected as sites for identity signals, including the seven positions of the anticodon loop, N73 and the last 5 bp of the acceptor stem. The remaining base pairs 6–12 of the acceptor branch of tRNA (the continuous acceptor and T-stem helix) are never used in identity, what can be explained a posteriori, knowing the contact patterns of tRNAs with synthetases (Fig. 2). In conclusion, classification of identity sets according to the class of the corresponding synthetases is not easily rationalized. It is, however, noticeable that use of position 37 in identity is synthetase class dependent. Also noticeable is the structural rational that emerges when comparing the binding of identity residues to synthetases, which is subclass dependent (see below).
The large occurrence of anticodon and of discriminator residues in identity is emphasized in Table 2. In the 16 E.coli systems where anticodon residues contribute to identity, the middle position 35 is always used; positions 36 and 34 are used less often (12- and 11-fold, respectively) with no strict conservation of the identity nucleotide (for Ile, Gln and Arg). Other anticodon loop positions are only sparsely used in identity. Leu, Ser and Ala identities do not rely on anticodon positions. The discriminator position N73 participates in 18 E.coli identities, the Glu and Thr identities being the exceptions. Noticeable are the Asp, Lys and Asn identities that specify tRNAs charged by synthetases of subclass IIb; here the discriminator position and the three anticodon nucleotides with conserved U35 are used in the three systems. The similar folding of the anticodon and catalytic domains of the corresponding synthetases accounts for these similarities (55–57). For the known yeast identities, involvement of anticodon and discriminator nucleotides is essentially the same as in E.coli, except for the Leu, Arg, His and Ala identities. Opposite to E.coli, Leu identity is dependent on residue 35, and Arg and Ala identities do not require the discriminator base. Altogether, out of the 14 yeast identities, 12 (except Ser and Ala) are dependent on the anticodon, and 11 (except Thr) on the discriminator base. A remarkable feature is the phylum-dependent conservation of the discriminator residue in prokaryotic and eukaryotic tRNAGly and tRNAHis species (48) and their probable conserved role in Gly and His identities, as demonstrated in E.coli and yeast.
Efficient identity switches of tRNAs with transplanted identity sets are indications for the completion of such sets. Table 3 gives examples where transplantations lead to full activity of the host molecules. In some cases, transplantation of a unique element makes the switch efficient. However, in general, full switches require transplantation of several elements and sometimes engineering of the tRNA scaffold. This is true for switches from tRNALeu to tRNASer and tRNAPhe to tRNAAsp (see below).
Phenomenology of Identity Expression
Strength of identity determinants in vivo
The strength of identity determinants expressed in vivo is correlated with suppression of a stop codon in a reporter gene. Strength is estimated by the frequency of amino acid incorporation at the suppressed position. Examples are given in Table 4. The amplitude of the effects ranges from 0 to 100% as measured by incorporation of the appropriate amino acid into the reporter protein. Strong effects (>90% non-cognate amino acid incorporation) concern, for example, the A20U mutation in suppressor tRNAArg which leads to a complete loss of Arg incorporation (thus, A20 is a major Arg identity element). Weaker effects are found, e.g. for the wild-type amber tRNAArg which has its Arg identity reduced to 37%. This decrease is explained by the presence in tRNAArg of the amber CUA anticodon which changes the C35 Arg determinant to U35 (Table 1). Interestingly, this tRNA acquires Lys identity (55%) because U35 is a Lys determinant. Mischarging of mutated suppressor tRNAs is common (Table 4). It is explained by the dual effect of the mutations and/or the properties of the suppressor anticodons, which concomitantly withdraw recognition elements for one identity while introducing elements for other identities. Thus, wild-type and mutated suppressor tRNAs often possess multiple identities.
Strength of identity determinants in vitro: kcat effects versus Km effects
The strategy to measure strength of determinants in vitro is completely different to making in vivo estimates. Here, one determines intrinsic physical-chemical effects of mutations on aminoacylation reactions instead of evaluating their ultimate fate in protein synthesis. Measurements of loss of kinetic specificities (L) is easy within a wide range, from moderate (L < 10) to important (L > 1000) effects (Table 5). Moreover, the individual contributions of kcat and Km to L values can be obtained (Table 6). Noticeably, in vitro studies seldom report possible mischarging of tRNA variants that are detected de facto by the in vivo approach.
Contribution of individual nucleotides to identity differs from one tRNA to another and the strength of determinants is phylumdependent (Table 5). So, the discriminator base is as important (Val), more important (Met and Asp) or less important (Gly and Leu) in prokaryotic than in eukaryotic systems. Furthermore, nucleotides at position 35 are highly represented as identity elements and their alteration produces the same range of effects (Leu, Tyr, Arg, Val and Gly), higher (Trp and Asp) or lower (Phe) effects in prokaryotes than in eukaryotes. No simple picture emerges when comparing prokaryotic and eukaryotic or class I and class II systems. However, the large strength often encountered for N73 and anticodon residues, especially for N35, has to be noticed. Also, in some systems with few identity nucleotides, the relatively low contribution of each residue is noticeable, as in the yeast Phe and Asp systems in which the strongest effects are <1000-fold. In contrast, in the E.coli Ala and His systems, also specified by a reduced number of identity residues, effect of their mutation can be >1000-fold.
The relative contributions of kcat and Km to identity are variable (Table 6). When normalized values are compared (calculated according to a formalism defined in 162) it appears that identity elements produce a continuum of effects ranging from mostly kcat [(kcat)N>(Km)N] to mostly Km [(kcat)N<(Km)N] through mixed kcat and Km effects [(kcat)N≈(Km)N]. Mutations at positions in the vicinity of, or in close contact with synthetases mainly alter kcat suggesting a direct effect on the catalysis of aminoacylation. In contrast, mutation of identity elements involved in tRNA architecture often produce Km effects. This suggests that these changes modify the functional binding of the tRNA on the enzyme.
Assay conditions influence identity expression. Note the effects of the G73→U73 mutation in E.coli tRNAGln under subsaturating (108) or steady-state kinetic (110) conditions. The influence of Mg2+ concentration was mentioned for Phe identity (163). Mg2+ concentration modifies the relative strength of identity elements. Also, mutation of the discriminator residue in yeast tRNAHis infers moderate (137) or dramatic (138) charging depending on the MgCl2 and ATP concentrations (Table 5).
Additivity, cooperativity and anti-cooperativity between identity elements
Kinetic data expressed as free energy variation at transition state of multiple mutants and comparison of experimental values with those calculated from data on single mutants define three types of relationships between identity nucleotides. Since such nucleotides are often scattered over three tRNA domains (the anticodon arm, the acceptor arm and the core of the tRNA), the question is how they act together to achieve global specificity. The question is of particular importance if the individual identity residues have moderate strength. Studies of double mutants established that three determinants (anticodon, A73 and G20a) act independently in yeast Phe identity (149). The functional relationship between Asp determinants in yeast tRNAAsp (G73, G34, U35, C36 and G10·U25) was studied with transcripts mutated at two or more identity positions (164). Multiple mutations affect activity of AspRS mainly at the level of kcat. Nucleotides located far apart in the 3D structure of the tRNA act cooperatively whereas those of the anticodon triplet act either additively or anti-cooperatively. In the latter case, the effect is lower than the sum of individual effects. These properties are correlated with specific interactions of chemical groups on identity nucleotides with amino acids in the protein as revealed by crystallography (165). Similar effects were found between Asp identity residues in the T.thermophilus system (143). Analysis of multiple tRNA mutants in this system revealed cooperativity between determinants of the anticodon loop and G10 and G73. The cooperative effects in the Asp system suggest that conformational changes trigger formation of a functional tRNA-aaRS complex. In yeast, the strongly anti-cooperative triple anticodon mutant indicates that this molecule behaves like an Asp minihelix because it has lost all contacts with the anticodon domain of AspRS. Strikingly, its overall kinetic specificity is identical to that of an Asp minihelix (166).
Aminoacylation of minihelices and partial identity sets
The L-shaped architecture of tRNA with a two-fold symmetry, allows its dissection into two components corresponding to double-stranded helical structures closed by T or anticodon loops. Acceptor branch minihelices are thus constituted by acceptor and T-stems and in many cases were shown to be charged by the homologous synthetases (167,168). Prediction of function of minisubstrates containing identity determinants was first verified in the Ala system (169) (Fig. 3). Since then, aminoacylation of many other minisubstrates (Table 7) including microhelices with short helical domains or helices closed by tetraloops instead of 7 nt canonical T loops has been observed. Escherichia coli minihelixAla and minihelixHis with major Ala and His identity elements are good substrates for their corresponding synthetases. In contrast, minihelixSer is a poor but specific substrate for SerRS, although identity nucleotides of E.coli tRNASer are within the acceptor arm, and no Ser determinant is found in the anticodon domain (an anticodon binding domain is missing in SerRSs). The low activity of minihelixSer is probably due to the absence of the long variable region, a particular feature of tRNASer involved in recognition by SerRS (59). The efficient serylation of human tRNASer deprived of an anticodon arm but possessing its extra arm agrees with this view (176).
Minihelices recapitulating acceptor branches of tRNAs in which the major identity elements are found in the anticodon loop should a priori be inactive. But such minihelices were charged, although to low levels in most cases. Addition of an anticodon helix enhances charging of minihelices derived from E.coli tRNAIle (70) and yeast tRNAVal (174). Asp minihelices behave like strongly anti-cooperative mutants of tRNAAsp and charge well (166). In contrast to a Val anticodon hairpin that stimulates valylation of the Val minihelix (174), an Asp anticodon hairpin has no effect on minihelix aspartylation. Mini- and microhelices derived from E.coli tRNAIle are also efficiently aminoacylated by IleRS although they contain only partial recognition sets, and the strongest Ile identity elements are within the anticodon branch (Table 5).
Mutation of identity positions in minihelices impairs aminoacylation and, thus, minihelix charging follows the same identity rules as those of canonical tRNAs. However, aminoacylation does not necessarily require the presence of major identity nucleotides. In the Asp system, impairment is equivalent in minihelix and full-length tRNA (166), whereas there are differences in the Gln and Ser systems (119,171). While the first 3 bp of the acceptor stem contribute largely to recognition of full F1 length tRNAGln by GlnRS, base pairs 2–71 and 3–70 moderately affect minihelix charging. G73 keeps its role in both contexts (171). In the Ser system, contribution of G2:C71 is consistent both in full-length tRNA and in minihelicies, whereas effect of base pairs 3–70 and 4–69 is substantial in minihelices but negligible in the tRNA. Thus, investigation of minihelices reveals ‘cryptic’ identity elements (119).
Minihelices expressed in E.coli, as shown for homologous Ala and Gly (177) and E.coli and yeast Asp minihelices (J.Rudinger-Thirion and R.Giegé, unpublished), affect cell growth. Inhibition in Ala and Gly minihelices is correlated with the presence of identity elements in the RNA (177).
Finally, it is noticeable that minimalist RNAs corresponding to class II synthetases are generally aminoacylated better than those of class I synthetases. This may reflect a more ancient origin of class II synthetases (see below). On the other hand, minihelices constitute models for the structural understanding of identities. So, NMR studies of microhelices derived from human tRNALeu suggest a decreased stability of the G1-C72 base pair when mutating the discriminator base A73 to G73 (178). This destabilization may explain the Leu→Ser switch upon mutation of the discriminator base (75,76). Similarly, distortion of the phosphodiester backbone of Ala minihelices around the G3·U70 pair revealed by NMR, may explain their Ala identity (179,180).
Each tRNA has, by definition, a single identity. However, tRNAs with multiple aminoacylation identities have been discovered in nature and could be created by engineering. A tRNA with two identities occurs naturally in certain Candida species (181). In these organisms, a leucine codon is to some extent translated into serine, thanks to serine-specific tRNAs [tRNASer(CGA)]. These tRNAs are aminoacylated in vitro and in vivo by SerRS but also by LeuRS. Thus, two distinct amino acids are assigned by a single ‘polysemous’ codon (181). Interestingly, SerRS and LeuRS belong to two different classes of synthetase.
A second family of tRNAs with multiple identities is illustrated by E.coli tRNAIle and yeast tRNAAsp. In both cases, the naturally occurring post-transcriptional modifications hide a second, underlying identity. A minor Ile tRNA from E.coli (tRNAIle2) is an efficient substrate for IleRS thanks to the presence of a Lysidine residue at position 34 (69). However, when this modification is removed, converting L34 to C34, the tRNA undergoes a 10-fold loss in isoleucylation efficiency and becomes a substrate for another class I enzyme, MetRS (69). A similar situation occurs in the yeast Asp system, where the in vitro transcript deprived of modified nucleotides has a double identity (182). It is an efficient substrate for both class II AspRS and class I ArgRS. Many other tRNAs have the potential to be recognized by non-cognate synthetases because they contain partial identity sets. This is observed for tRNAs mischarged by the same synthetase. They often have the same discriminator base and/or common anticodon residues (31).
Engineered suppressors as well as rationally designed in vitro transcribed chimeric tRNAs typically have several aminoacylation identities (Table 8). Suppressor tRNAs often acquire Gln or Lys identities in addition to their primary identity (46) because the corresponding identity sets include U34 and U35 present in opal (UCA) and amber (CUA) anticodons. Likewise, simultaneous introduction into yeast tRNAAsp of the non-overlapping Phe and Ala identity sets converts the chimeric molecule into a triple aminoacylatable substrate (150). Progressive introduction of Gln identity elements into yeast tRNAAsp, or Asp identity elements into E.coli tRNAGln, leads to intermediate engineered tRNAs with dual specificity (142). This potential of tRNAs to display several specificities indicates a role, if not a need, of anti-determinants to restrict these possibilities (see below).
Structural View of Identity Expression
Contacts between tRNAs and synthetases
Six crystallographic structures of tRNA-aaRS complexes are known: the E.coli Gln (187), yeast Asp (188) and T.thermophilus Lys (189), Phe (190), Pro (191) and Ser (192) complexes. Footprinting data are available for yeast Asp (193–195), Arg (196,197), Phe (198) and Val (198,199), E.coli Ile (71) and Thr (200) and bean Leu (201) complexes. These data are critical to the understanding of the structural basis of identity expression. As examples, Figure 2 compares the 3D structures of the Gln and Asp complexes, representative of class I and class II systems. It emphasizes the tRNA residues in contact or close proximity with the synthetases: those are non-identity and identity elements highlighted in yellow and purple colored spheres, respectively. Contacts mainly occur at the two distal ends of the tRNA with the clear distinction that class I GlnRS interacts essentially with residues from the 5′-terminus of the tRNA and class II AspRS from the 3′-terminus. Among them, only a few are identity elements. The situation is the inverse for the anticodon stem-loop residues, especially in the Gln system, where many residues in contact with the synthetase are identity elements.
Direct recognition of identity determinants by synthetases, as first shown in the Gln and Asp complexes, is a general theme. It is suggested in other systems by footprinting (71,193–197) and shown more directly by crystallography for anticodon identity residues contacting the synthetases in the Lys (189), Phe (190) and Pro (191) complexes. So, G35 and G36, the Pro determinants in E.coli (126,127) conserved in all tRNAPro species, are contacted by thermophilic ProRS (191). Most contacts involve H-bonds between identity nucleotides on tRNA and identity amino acids on the synthetases, as seen in the Asp and Gln systems (165,187). These contacts are needed for activity since the decreased activity of tRNAAsp variants mutated at identity positions is roughly proportional to the number of H-bonds lost (164). Beside direct recognition modes, synthetases can recognize structural features in tRNA, like in the Ser complex from T.thermophilus where SerRS recognizes the tRNA shape through backbone interactions and without any contact with the anticodon loop (192). Recognition of specific architectural features likely accounts for the identity of many mitochondrial tRNAs with bizarre structures (202).
It is possible that identity elements act indirectly in promoting specific RNA conformations that are recognized. This could explain why mutation of G10·U25 in tRNAAsp, which is not in direct contact with AspRS (165) leads to a Km-dependent decrease in activity (141), probably because of a perturbed (N10–N25)-45 triple interaction in the mutants. The situation is more complex for Ala identity and recognition of the G3·U70 wobble pair is not definitively understood. Some data support the view of a direct recognition of this pair by AlaRS (203–206) while other data are better interpreted in the light of an indirect involvement (155,183,207,208) where the synthetase recognizes an RNA conformation induced by the wobble pair. A more definitive answer awaits crystallographic data.
The interaction mode of tRNA on multimeric synthetases has been questioned. A combined structural and genetic approach proved tRNATyr interacts with both subunits of TyRS from Bacillus stearothermophilus. Docking suggests that the acceptor arm interacts with the N-terminal domain of one TyrRS subunit and the anticodon arm with the disordered C-terminal domain of the other subunit (209). Similarly, crystallography shows cross-contacts of tRNAPro and tRNASer over the two subunits of T.thermophilus ProRS (191) and SerRS (192). In the case of tetrameric PheRS, the tRNA binds across the four subunits (190). The situation is quite different in the Asp system, where each tRNAAsp binds to one subunit of AspRS with the exception of one H-bond between a phosphate oxygen of U1 and a lysine residue (K293) from the other subunit (165).
Conformational changes in complexed tRNAs
Interaction with synthetases modifies the conformation of tRNA (210). This was demonstrated directly in the Gln and Asp systems by crystallography (187,188). In tRNAAsp the prominent deformation occurs in the anticodon loop; in contrast, in tRNAGln it concerns the acceptor end with the −CCAOH terminus kinked and the first base pair of the stem disrupted (Fig. 2). The kink of the free −N73CCAOH in class I systems seems to be a necessity for geometrical reasons. This is not the case for disruption of the first base pair which is maintained in complexed tRNAVal interacting with class I ValRS as shown by 19F NMR (161).
Occurrence of conformational changes in tRNA is a common characteristic and was seen by solution analysis in the Asp (195), Arg (197), Gln (211), Met (212), Ile (71) or Val (65,213) systems. The changes are likely enzyme induced as shown for the in vitro transcribed tRNAAsp. It is bent over AspRS by its variable region side and over ArgRS by the opposite D-loop side (197). The deformation of the anticodon loop of tRNAAsp is altered if anticodon Asp identity residues are mutated and their contacts with AspRS lost. Both contacts and conformational changes are recovered if the Asp identity set is transplanted into tRNAPhe (195).
Importance of tRNA architecture
Specific architectural characteristics in tRNAs can be identity elements. This is so in the variable region of tRNASer (93,117,120,121), in the atypical G15·G48 Levitt pair in E.coli tRNACys (86), in the G10·U25 pair in yeast tRNAAsp that makes a triple interaction with G45 (141), and most probably in determinants that act by Km effects (Table 6). In the E.coli Cys system the functional effect of G15·G48 is determined by the A13·A22 mismatch in the D-stem of tRNACys, since its introduction in tRNAGly confers Cys identity (89). Similarly, in the E.coli Pro system, the integrity of the G15·C48 pair is important for aminoacylation (127). The case of the E.coli Ala system in which a structural distortion induced by the G3·U70 pair may trigger identity is discussed above.
Optimal aminoacylation efficiency often relies on conservation of specific architectural features in the core of the tRNA (142,150,214–218). For recognition by E.coli AlaRS (150) or GlnRS (142) and yeast PheRS (214), variable region and D-loop length is important. In the E.coli Pro system a relationship between ribose of U8 and G46 makes a contribution to aminoacylation (215). In yeast initiator tRNAMet, the interplay between D- and T-loops is crucial. Interestingly, mutation of A20 and A60 into U20 and G60 abolishes methionylation (83) but not binding to MetRS (217). This is consistent with the increased stability of the mutant and with the idea that the inactive molecule is not able to adopt the correct conformation required at the transition state of the reaction.
On the other hand, synthetases can tolerate architectural variations in their tRNA substrates. Such variations can originate from the absence of base modifications (219,220) or by the replacement of individual (215,216) or families (217,218) of ribonucleotides by their deoxy counterparts. In most systems these variations are accompanied by moderate losses of charging efficiency. Note that ribonucleotides in the anticodon loop are necessary for binding of tRNA to yeast MetRS, as shown in competition experiments with anticodon hairpin DNA-RNA hybrids (217). Modifying tertiary interactions is another way to vary tRNA conformations. A systematic study on the effects of changing the sequence of the tertiary 15·48 Levitt pair in the framework of E.coli tRNAAla found that all but an A15·A48 variant were functional in vivo (221). The A15·A48 mutation in the defective tRNAAla variant corresponds to the Levitt pair found in human tRNAAla. Interestingly, this molecule has an altered conformation in the anticodon region, and introduction of the D-loop sequence of human tRNAAla in the variants restores Ala activity and anticodon conformation (221).
More drastic architectural changes in tRNA compatible with specific aminoacylation have been described. The absence of Tor D-arms leads to mimics of mitochondrial tRNAs (222,223). Thus, in E.coli tRNAIle (71) and yeast initiator tRNAMet (83), removal of the T-arm only slightly impairs aminoacylation. This capability of cytosolic synthetases to recognize mitochondrial-like tRNAs is supported by in vivo expression in E.coli of such molecules (224,225). Other peculiar architectural features (extended D-stem and unusual D- and T-loop interactions) exist in the core of selenocysteine inserting tRNAs and support recognition by SerRS (226). For E.coli tRNASec this leads to a 100-fold impaired serylation efficiency (227). Similar losses in aspartylation are found in yeast tRNAAsp variants with extended D-stems (228).
The most radical deviations from the canonical tRNA architecture are found in viral tRNA-like domains and in several other tRNA mimics (229–233). In the models of these mimics the identity elements are located at positions spatially similar to those in canonical tRNA. This allows similar interactions with the identity counterparts on the synthetases (Fig. 3). This explains the alanylation capacity of 10Sa tmRNA (27,232) and of minihelices (169). The histidinylation capacity of the viral tRNA-like domains is accounted for as well because the pseudoknot fold proximal to the acceptor end provides a position in loop L2 that mimics identity determinant N-1 in tRNAHis (234,235). Likewise, the Val→Met identity switch in TYMV RNA is explained by a mutation in the anticodon and a loosening of the pseudoknot facilitating adaptation to MetRS (236,237). Furthermore, a chargeable polyU construct mimics a tRNALys with its Lys identity determinants U73 and UUU anticodon triplet (238). The binding of polyU and tRNALys to an isolated N-terminal anticodon binding domain of E.coli LysRS, as shown by NMR, is in line with these functional data (239). Thus, the same identity rules account for the aminoacylation capacity of all RNAs with the alternate scaffolds displayed in Figure 3. This also applies for the recognition of the thr mRNA regulatory structure (Fig. 3, panel 5) by synthetases, as shown by mutations in the anticodon loop mimic of this structure, making regulation of the messenger by MetRS rather than by ThrRS (240).
Thus, there is an apparent contradiction. Synthetases sense subtle architectural details in tRNAs but they can also recognize a wealth of different RNA frameworks. The flexibility of RNA allows mutual tRNA-aaRS adaptation and resolves this contradiction. Alternate scaffolds carrying identity signals can sustain synthetase recognition. Following this rationale, specific structural features in the core of tRNA may be considered as additional identity signals that enhance specificity (and in some cases even as major determinants, as for Cys identity). Another conclusion is that restricting flexibility of an RNA scaffold limits its adaptability to synthetases and favors specific recognitions. Conversely, increasing its plasticity decreases its specificity.
A Refined View of Trna Identity
Bases, riboses and chemical groups
Crystallography has revealed the existence of contacts between chemical groups in nucleotide bases and the ribose phosphate backbone of tRNAs with amino acids of synthetases. However, crystallography does not indicate whether these contacts have direct consequences on the aminoacylation capacities of the interacting tRNAs, in other words if they are the actual identity determinants or are features contributing to binding in a neutral way. Functional studies on appropriate tRNA mutants are required to distinguish between the two possibilities. Several strategies can be employed for this aim, like the so-called ‘atomic mutagenesis’, crystallography combined with mutagenesis approaches, comparison of mutational effects for a single position, or the use of a T7 RNA polymerase mutant for the incorporation of deoxyriboses into RNA transcripts. A few examples of chemical groups identified by these methods as essential for aminoacylation are given in Table 9.
Atomic mutagenesis was pioneered to decipher the functional role of chemical groups in the G3·U70 Ala identity pair (203–206). To this end, minimalist RNA substrates of AlaRS were chemically synthesized with G3 or U70 replaced by base analogs. Experiments led to the conclusion that the exocyclic 2-NH2 group of G3 is essential for alanylation. For Gln identity, full-length tRNAGln variants were synthesized with G2, G3 or G10 identity residues replaced by inosine. Impairment of activity indicated the importance of the 2-NH2 groups of these G residues for tRNA discrimination by GlnRS (109). Likewise, for Pro identity, the importance of the 2′-OH in ribose from U8 was highlighted (216).
Yeast tRNAAsp and tRNAMet transcripts with deoxyriboses were prepared with the mutant polymerase (218). These tRNAs fold in conformations close to those of natural tRNAs, but their in vitro activities indicate striking differential effects depending upon the nature of the substituted ribonucleotides. The strongest decrease in charging occurs for initiator tRNAMet with dG or dC and for tRNAAsp with dU or dG. In the Asp system, the impaired aminoacylation is correlated with the substitution of the ribose moieties of U11 and G27 that disrupt two H-bond contacts with AspRS. This suggests that specific 2′-OH groups in tRNAAsp are identity determinants (218).
Prediction of the chemical groups on identity bases that are H-bonded to synthetases were made on the basis of mutational analyses (31,72,244; Table 9). The approach is useful in the absence of crystallographic data. Its validity has been tested in the yeast Asp system and predictions were found in good agreement with crystallography (244).
Investigations on Ile systems led to the proposal that two functional groups (an NHCO structure) shared by identity nucleotides at the wobble position N34 in tRNAIle species are the actual determinants recognized by yeast and E.coli IleRS. Interestingly, beside on G34 (at N1 and O6), these groups can also be brought by modified nucleotides (inosine, pseudouridine or lysidine) present at position 34 in the anticodon of tRNAIle species (69,72). From another point of view, the role in aminoacylation of the hypermodified wybutine residue at position 37 in yeast tRNAPhe is striking. Although this residue is not a Phe determinant it plays a crucial role in the activation of the catalytic site of PheRS since tRNAPhe lacking this modification and the 3′-terminal A is unable to trigger aminoacylation of free A while the modified tRNA does (246).
Identity of tRNAs is not only dictated by the presence of sets of positive identity elements allowing recognition by cognate synthetases, but also by negative signals, the anti-determinants, which hinder interaction of a tRNA with non-cognate synthetases. The first anti-determinant was found in E.coli tRNAIle where a Lysidine residue (a modified C) at position 34 of the anticodon hinders misaminoacylation by MetRS (69). Another modified nucleotide, m1G37, plays a similar role in yeast tRNAAsp where it prohibits aminoacylation by ArgRS (182,247). Other examples of anti-determinants are given in Table 10. They are unmodified single or base-paired nucleotides located in any structural domain of the tRNA. So, anticodon residue A36 in E.coli tRNAArg prevents recognition by TrpRS (114) and base pair U30·G40 in the anticodon stem of tRNAIle prevents that by GlnRS and LysRS (248). Interestingly, protection of tRNAs from non-cognate synthetases is not restricted to synthetases from the same class. Examples in Table 10 show indeed that tRNAs which are specific substrates for class I synthetases contain anti-determinants against class II enzymes (tRNALeu/SerRS) and vice versa (tRNAAsp/ArgRS). The Leu/Ser systems are of special interest since cross-protections are governed by nucleotides located at opposite positions in the tRNAs. Thus, A73 protects tRNALeu against human SerRS (class II) whereas G37 protects tRNASer against yeast LeuRS (class I) (75,77).
Although only a few anti-determinants in tRNA are known, it is likely that their occurrence is not restricted to a few systems. We believe that each tRNA contains anti-determinants against several synthetases. In addition to well-defined nucleotides or to chemical groups, anti-determinants could be of a structural nature. Structural features may restrict a given tRNA from aminoacylation by a non-cognate synthetase. PheRS efficiently recognizes a transcribed tRNAAsp with the five major Phe identity elements, only if the structural background has been changed from that of tRNAAsp to that of tRNAPhe (214). Another example is the so-called ‘orthogonal’ suppressor tRNA derived from E.coli tRNAGln. Appropriate mutations completely abolish recognition by GlnRS. These mutations introduce anti-determinants, in fact structural ‘knobs’, that prevent contacts with GlnRS (186).
Analysis of in vitro transcripts derived from yeast tRNAPhe and E.coli tRNAAla bearing the complete Phe identity set (A73, G20 and the three anticodon nucleotides; Table 1) showed that PheRS is sensitive to additional nucleotides within the acceptor stem. Insertion of G2:C71 has dramatic negative effects in both tRNA frameworks. These effects are compensated by the insertion of the wobble G3·U70 pair, which by itself has no effect on phenylalanylation. From a mechanistic point of view the G2:C71 pair is not a ‘classical’ recognition element since its anti-determinant effect can be compensated by a second-site mutation (162). Thus, tRNA identity is not only the outcome of a combination of positive and negative signals forming the so-called identity sets, but is also based on the presence of non-random combinations of sequences elsewhere in tRNA. These sequences are ‘permissive elements’ and were retained by evolution so that they do not hinder aminoacylation. It is likely that no nucleotide within a tRNA is of random nature but has been selected so that a tRNA can fulfill all its functions efficiently.
Alternate identity sets
Conceptually, each aminoacylation system was expected to have a unique set of identity elements. In other words, they should be shared by all isoaccepting tRNAs. The available data largely supports this view. However, hints suggesting that this is not completely general were obtained. Microhelix aminoacylation by a class I synthetase showed that non-conserved base pairs are required for specificity (249). In the special case of the yeast arginylation system, it has been shown that ArgRS requires two different sets of nucleotides for the specific aminoacylation of its substrates (115). Thus, arginylation of yeast tRNAArg is governed by C35 and G/U36, whereas efficient arginylation of the in vitro transcribed version of yeast tRNAAsp (not protected against mischarging due to the absence of the modified bases) is based on the presence of C36 and G37 (115). Footprinting (193) and a kinetic analysis of mutants (250) are in favor of the existence of alternate mechanistic routes by which each identity set triggers the activation of the same synthetase. These unexpected facts provide new conceptual routes towards a deeper understanding of aminoacylation specificity.
The manifold role of synthetases
Synthetases are pivotal for correct expression of tRNA identities. Physiological tRNA identities are determined by competitions among synthetases (93), reaction conditions (137,138,163) and compromises between recognition specificities and steps in the charging pathways (60,91,251–253). Furthermore, the specificity of amino acid activations and the stability of the transition state are governed by the class defining catalytic cores of the synthetases (53–57,60). On the other hand, the mutual adaptation of cognate tRNA-aaRS couples requires matching of tRNA identity determinants by protein counterparts. In contrast to the great body of data about tRNA determinants, only limited knowledge on synthetase determinants is available. No complete amino acid identity set has yet been precisely characterized. Such sets may comprise amino acids revealed by crystallography to be near to the interacting tRNAs (165,187,190).
As an illustration of the critical role of particular amino acids in the recognition of tRNAs, we give four examples. In the Gln complex, a leucine from the acceptor binding domain of GlnRS (L136) stabilizes the disruption of the U1:A72 pair in tRNAGln by stacking between A72 and G2. Substituting this leucine by amino acids that would favor or disfavor stacking, modulates the discrimination potential of GlnRS against non-cognate tRNAs. Remarkably, the evolutionary solution retained has optimized both cognate tRNA recognition and discrimination against non-cognate tRNAs (253). In the Asp system, a stacking interaction of a phenylalanine (F127) between U35 and C36 in tRNAAsp likewise stabilizes the conformational change of the anticodon loop and thus favors matching of the anticodon identity nucleotides on AspRS (165). It is notable that this residue, as well as two other amino acids recognizing the anticodon (R119 and Q138), are conserved in the AspRS, AsnRS, LysRS class II subgroup. While these stacking interactions in the Gln and Asp systems act indirectly in facilitating other interactions, in the E.coli Ile and Met systems a single amino acid, namely R734 in IleRS and W461 in MetRS, is sufficient to provide recognition specificity. Indeed, the R·W swap within homologous helix loop peptides from the anticodon binding domain of the two structurally related class I synthetases switch tRNA recognition (254). This switch is the converse of that between tRNAIle and tRNAMet when swapping the (G/LAU)Ile and (CAU)Met anticodons (69).
Architectural motifs in synthetases are essential for tRNA recognition. One has been identified in TyrRS within its class I characteristic CP1 (connective peptide 1) sequence (173). As in tRNAs, anti-determinants exist on the surface of synthetases to repel non-cognate tRNAs (38). Only a few have been identified, such as two acidic amino acids in E.coli MetRS that prevent binding on non-cognate tRNA anticodons (255). Negative determinants for rejection of non-cognate tRNAs were also found in bacterial TyrRSs (209,256).
Conservation and Divergence in Evolution
The following is based on theoretical considerations and the principle of an early molecular evolution of tRNA aminoacylation systems (167,237,257–261). Thus, contemporary tRNAs and synthetases would be derived from simplified versions restricted to the minimal structural elements needed for function. The two domains of the L-shaped tRNA would have arisen independently, with the acceptor branch appearing first. In a later stage in history, the catalytic cores of synthetases emerged independently in their class I and class II versions. We conjecture that class II cores had a primordial importance, still reflected in their modern progeny. The core is best at aminoacylating minimalist tRNAs and for two of its members (SerRS and AlaRS) does not recognize anticodon in full tRNA. Co-evolution of catalytic cores of synthetases and accepting RNA hairpins led to an operational RNA code that associated specific amino acids with RNA hairpin structures (167). Much experimental evidence supports this scenario, in particular functionality of minimalist tRNAs and conservation in evolution of catalytic cores of synthetases, e.g. in AlaRSs (262), and of major identity nucleotides in acceptor branches of tRNA (Table 11).
The anticodon domain of tRNA and the additional domains of synthetases appeared later in evolution. Anticodon domains brought the link between the RNA operational code, and the correlated tRNA recognition by synthetases, with the anticodondependent recognition by mRNA. The additional synthetase domains introduced the source of structural diversity that led to the present idiosyncrasies in the expression of tRNA identities. As a result, for several identities (e.g. Gly, Ile, Met) the strength of identity elements in the anticodon region of tRNA overtook that of elements in the acceptor branch (Table 5).
Additional evolutionary events led to idiosyncrasies within individual identities in different species. They were often of divergent nature, but convergence also occurred. Convergent processes may account for the intriguing discovery of the first class-switch of a synthetase, namely that of an archaeal LysRS that is class I-like instead of class II (263) as well as for the origin of tRNA mimics (233). Divergent evolution may explain species differences in the functional expression of the same identity. Such differences were quantitated for Met identity in E.coli and yeast (80,82) and Asp identity in E.coli, T.thermophilus and yeast (143). Functional idiosyncrasies became so important for certain identities that species barriers to tRNA aminoacylation occurred, especially between prokaryotes and eukaryotes (96,264–267). For example, GlyRSs from E.coli and human cannot cross-charge human and E.coli tRNAGly (265). Species barriers have been correlated with non-conservation of the discriminator nucleotide in tRNAGly (265,267) or of the first base pair in the acceptor stem of tRNATyr which is G1:C72 in prokaryotes and C1:G72 in eukaryotes (96). They are compensated by changes either in tRNAs (97,148,268) or synthetases (173), probably in domains at the interface with tRNA. In line with this view is the transplantation of a 39 amino acid peptide from human to E.coli TyrRS that enables the bacterial enzyme to aminoacylate eukaryotic yeast tRNATyr (173). Similarly, discriminator N73 residue is A in prokaryotes and often G in eukaryotes, in the Lys identity set. Human LysRS aminoacylation is relatively insensitive to the nature of N73, a fact accounted by the peculiar sequence of ‘motif 2’ in the human synthetase that can accommodate degeneracy at N73 (269). Another notable example is E.coli tRNATyr that is a Leu acceptor in S.cerevisiae (270). Its identity switch is correlated with the structural similarity of the E.coli tRNA with yeast tRNALeu, both with a large variable region, while in yeast and other eukaryotes tRNATyr has a small variable region (48). Finally, domains not directly involved in the aminoacylation function of synthetases, like the N-terminal sequences that tag eukaryotic synthetases (52), may be essential to rescue prokaryotic enzymes inactive for charging of eukaryotic tRNAs. This possibility has been demonstrated in Glu systems where tagging of E.coli GluRS provides eukaryotic functionality to the bacterial enzyme (271). In general, species idiosyncrasies in identity are restricted to subtle sequence features in tRNAs correlated to evolutionary variations in synthetases.
From what precedes, it is likely that conservation of many major identity nucleotides reflects a common evolutionary origin (Table 11). But what about the origin of the different isoaccepting families of tRNAs? Do they each originate from a common ancestor or are other mechanisms possible? Recent genetic experiments by Saks et al. (272) cast doubt on the paradigm of a common ancestor. According to their ‘tRNA gene recruitment’ model, a tRNA gene can be recruited from one isoaccepting group to another by a point mutation in its anticodon sequence that leads to an identity switch accompanied by a change in codon recognition. The model was tested by the recruitment of a tRNAArg gene that codes for a tRNA in which the Arg anticodon (UCU) is changed into a Thr anticodon (UGU) and replaces an essential tRNAThr(UGU) gene that was inactivated (272). Related to such a mechanism, and based on the functional relationship in yeast of the Arg and Asp identities (115,250), is the proposal of the ‘capture’ of a tRNAAsp species by ArgRS that became a tRNAArg (250).
Spectacular idiosyncrasies exist in mitochondrial systems due to the unusual structure of their tRNAs (48,222), but also to peculiarities of their genomes. Thus, a resourceful mechanism occurs in a mitochondrial gene from marsupials (273). By partial editing in a primary transcript of a tRNAAsp variant with Gly identity, a Gly anticodon (GCC) is changed into an Asp anticodon (GUC), so that a fraction of both tRNAAsp and tRNAGly are produced (273). Thus, an epigenetic phenomenon compensates for the parsimony of the compact mitochondrial genome of the marsupials.
Two classes of conclusions arise from this survey on tRNA identity: first, general rules, and second, idiosyncrasies that distinguish individual or groups of identities. We list those we consider the most representative.
(i) General rules
Similar principles govern tRNA-synthetase recognition, namely that RNA scaffolds carry a limited number of signals (the determinants) that confer recognition specificities.
Most identities rely on determinants present at both extremities of the tRNA (exceptions are the Ala and Ser identities that are anticodon-independent).
Primordial identity features are contained in the tRNA accepting branches (they may be hidden by more recent evolutionary changes in tRNA, but can be deciphered in peculiar systems, e.g. in minimalist tRNAs).
H-bonds often couple identity bases to individual amino acids of synthetases. In some cases, however, the same amino acid can H-bond with two adjacent identity bases (165).
Specific recognitions are accompanied by major conformational changes of the tRNA.
Identity sets determined in vivo correlate well with those determined in vitro. However, in vivo expression of identity appears less sensitive to intrinsic changes in aminoacylation kinetic efficiency than in vitro expression. Activity changes of variants mutated at identity positions are better revealed in isolated systems than in the cellular context where competitions due to relative concentration levels of tRNA and synthetases modulate activities.
Anti-determinants (present on either tRNAs or synthetases) restrict expression of multiple identities. Anti-determinants can be modified nucleotides.
(ii) Idiosyncratic features
The precise interaction modes of tRNAs with the class-defining catalytic cores of synthetases and especially recognition of anticodon binding domains (if present) display great diversity.
Differences in identity expression correlated with sequence changes in tRNAs or synthetases exist between prokaryotic and eukaryotic systems; they can be strong enough to provoke species barriers.
Usage of modified nucleotides as recognition elements occur in a few identities (Glu, Ile, Lys).
Absence of certain synthetases characterizes some organisms (274).
As a final remark, it is fair to say that a number of features are not well understood in tRNA identity and that the recent discoveries of cryptic determinants, of permissive elements and of alternate identity sets have added new intricacies to the simplified classical view. It is likely that these novel features revealed in a few systems are not isolated examples. Among others, we believe that differences in identity elements may exist within isoacceptor tRNA families (e.g. in Table 1, the sequence differences at identity positions in the yeast tRNAIle and tRNAGly families, but which could also imply recognition of similar chemical groups). Also, relationships between identity nucleotides and more generally between non-identity nucleotides are believed to play crucial roles in identity expression. They, as well as the nature of the amino acid identity determinants on the synthetases, remain to be deciphered. The complexity in the tRNA-synthetase interrelations was well illustrated in a recent work aimed to engineer, starting from the E.coli Gln system, a synthetase that would charge an unnatural amino acid on an ‘orthogonal’ tRNA that has lost its ability to be recognized by natural synthetases (275). As noted in the accompanying commentary (276), the difficulty in finding a novel 21st tRNA-synthetase pair lies in the multifactorial nature of the RNA-protein interactions that concern not only specificity elements but also parts in tRNA and synthetase not involved in aminoacylation. It can be anticipated that such specificity manipulations will provide new openings in tRNA research. As in the above example, engineering can be based on structural knowledge combined with advanced RNA sequence analysis (277,278), but exploration of the complexity of aminoacylation systems will also benefit from combinatorial approaches in both the RNA (279) and protein (280) worlds.
This work was partly supported by grants from the Centre National de la Recherche Scientifique (CNRS) and Université Louis Pasteur (Strasbourg). We thank Nancy Martin for advice. M.S. was supported by grants from Ministère de l'Enseignement et de la Recherche (MER) and Association pour la Recherche contre le Cancer (ARC). C.F. was supported by a NATO grant during her sabbatical at California Institute of Technology, Pasadena, USA.