Research Article

Evolution of a Transfer RNA Gene Through a Point Mutation in the Anticodon

See allHide authors and affiliations

Science  13 Mar 1998:
Vol. 279, Issue 5357, pp. 1665-1670
DOI: 10.1126/science.279.5357.1665


The transfer RNA (tRNA) multigene family comprises 20 amino acid–accepting groups, many of which contain isoacceptors. The addition of isoacceptors to the tRNA repertoire was critical to establishing the genetic code, yet the origin of isoacceptors remains largely unexplored. A model of tRNA evolution, termed “tRNA gene recruitment,” was formulated. It proposes that a tRNA gene can be recruited from one isoaccepting group to another by a point mutation that concurrently changes tRNA amino acid identity and messenger RNA coupling capacity. A test of the model showed that an Escherichia coli strain, in which the essential tRNAUGU Thrgene was inactivated, was rendered viable when a tRNAArgwith a point mutation that changed its anticodon from UCU to UGU (threonine) was expressed. Insertion of threonine at threonine codons by the “recruited” tRNAArg was corroborated by in vitro aminoacylation assays showing that its specificity had been changed from arginine to threonine. Therefore, the recruitment model may account for the evolution of some tRNA genes.

Transfer RNA participates in two functionally related but physically decoupled biochemical events that are crucial to protein synthesis. The tRNA amino acid–accepting activity (identity) is determined by the aminoacyl-tRNA synthetases, which specifically recognize and aminoacylate the cognate tRNA or tRNAs on the basis of sequence and structural elements (1-5). The tRNA amino acid–donating activity occurs at the ribosome after the three nucleotides of the tRNA anticodon engage in complementary base pairing with messenger RNA (mRNA) codons. Cumulatively, these two biochemical events determine the sequence of each protein. Thus, to ensure the fidelity of translation, the correspondence between a tRNA's amino acid identity and its anticodon must recapitulate the genetic code.

Given the central role of tRNA in translation and the near universality of the genetic code, it seems likely that tRNA genes were among the earliest genes to arise (6). Efforts to elucidate the events leading to the evolution of tRNA have focused on deducing (i) the sequence and structure of the primordial RNAs that may have originally carried amino acids (7, 8), (ii) the molecular events that account for the transition of these RNAs into an RNA having the prototypical tRNA structure (8, 9), and (iii) the sequence of the common ancestor of present-day tRNAs (6, 10).

In extant organisms, tRNA comprises a multigene family that can be partitioned using the criteria of amino acid specificity, such that each tRNA falls into only 1 of 20 possible amino acid–accepting groups. In turn, several of these groups comprise isoacceptors (tRNAs that accept the same amino acid but have different mRNA codon selectivities). A reasonable a priori assumption is that isoacceptors within each group arose by gene duplication from a common ancestor having the same amino acid identity. However, although this seems to be true in some cases, patterns of sequence similarity that might reflect other evolutionary scenarios have also been observed (10,11). None of these possible scenarios has yet been examined empirically.

One powerful approach to molecular evolution is the use of phylogenetic analysis to deduce the sequence of ancestral genes coupled with the use of biochemical techniques to test their function (12). We developed a different but related approach to investigate the molecular mechanisms that account for the evolution of isoacceptors. In this case, sequence (13) and functional (1-5) information on tRNA was used to formulate a hypothesis about previous historical events, and the hypothesis was tested in vivo using E. coli as a model system. The results show that isoacceptors can evolve by gene recruitment events whereby a point mutation in the anticodon trimer recruits a tRNA from one isoaccepting group to another, provided that the synthetase corresponding to the new amino acid already exists.

Models of tRNA gene evolution. Relations among tRNA genes can be explored now that the complete sets of E. coli andHaemophilus influenzae tRNA gene sequences are available (14-16). The similarities among type I tRNA genes fromE. coli and H. influenzae are shown in Fig.1 (17). Of the eight isoaccepting groups in E. coli with multiple isoacceptors, three of these groups (Ala, Gln, and Pro) form discrete clusters within the dendrogram. In contrast, the isoacceptors within the five other groups (Gly, Arg, Ile, Thr, and Val) are dispersed throughout the dendrogram. This dispersion contradicts a classical model of tRNA gene evolution in which isoacceptors are derived from a single ancestral tRNA having the corresponding amino acid identity. Thus, the mechanisms that account for it are of particular interest. One possibility is that neutral mutations have accumulated since these isoaccepting groups were established. However, this does not seem to be the case because there is relatively little divergence between the E. coli andH. influenzae genes that encode the same isoacceptor, even though these two species must have diverged more than 160 million years ago (18). Alternatively, tRNA genes with the same amino acid identity may be derived from different ancestors. This explanation is consistent with our sequence analyses and has been proposed by others (10, 11) to explain unexpected relations among tRNA genes.

Figure 1

Dendrogram showing similarities among tRNA gene sequences from E. coli (13, 14) and H. influenzae (15). Genes are named by anticodon and species (E, E. coli; H, H. influenzae). Isoaccepting groups discussed in the text are indicated by the amino acid they accept. The bracket encompasses a cluster that primarily consists of tRNAs having a uridine at position 36 of the anticodon. The dendrogram was generated from aligned sequences (16) with the Pileup algorithm [Genetics Computer Group package (44)]. Before the analysis, the three anticodon nucleotides were removed from each tRNA so that the resulting clusters would not be influenced by the current decoding capacity of each tRNA.

A possible mechanism whereby tRNAs with the same amino acid identity could evolve from different ancestral genes is suggested by relations among tRNAs (Fig. 1). One of the deeply branched clusters in the dendrogram primarily comprises tRNAs that have different amino acid identities and differ at the middle position of the anticodon (position 35), but share a uridine at the 3′ position of the anticodon (U36). Although tRNA affinities in the deeply branched clusters could not be established with a high degree of certainty (19), similar patterns have been observed in phylogenetic analyses of archaeal tRNAs (20) and in other types of sequence comparisons (10,11). Together, these observations raise the possibility that amino acid assignments evolved by anticodon mutations.

X-ray diffraction analyses of tRNA-synthetase complexes reveal that the molecular interface at the anticodon can involve a network of interrelated contacts (21). That these contacts are sufficient to redirect the amino acid specificity of full-length tRNAs is evidenced by in vitro studies involving anticodon “transplants” (22-24) as well as by in vivo studies showing that the amino acid inserted at a single engineered codon in a reporter gene can be dictated by the tRNA anticodon sequence (5, 25-27). We reasoned that because the anticodon can dictate tRNA amino acid specificity as well as mRNA coupling, a tRNA gene could be recruited from one isoaccepting group to another by an anticodon mutation that decreased interactions with the cognate synthetase and promoted interactions with the synthetase corresponding to the new anticodon trimer. As a consequence of concurrently changing tRNA identity and mRNA coupling capacity, a single mutation could potentially result in a tRNA that would be competent to correctly translate a new set of codons in all of the essential endogenous mRNAs. Thus, an anticodon mutation might recruit a tRNA gene from one isoaccepting group to another. Because such a phenomenon offers a possible explanation for the patterns observed in sequence analyses, we explicitly tested the recruitment hypothesis in vivo, using E. coli as a model system.

Test of the recruitment hypothesis. The original approach for studying tRNA in vivo identity was not adequate for testing the recruitment hypothesis because it required the tRNA anticodon to correspond to an engineered nonsense (termination) codon in a reporter gene (28). An alternative method that could accommodate tRNAs with normal anticodon trimer sequences was subsequently developed (27). However, in this method as well as in the original method, tRNA function is assessed with respect to the translation of only a single codon in a reporter gene. Moreover, reporter genes are typically not essential, and the sites used to screen for tRNA amino acid identity usually accommodate a variety of amino acids. Thus, in these reporter systems, there is no inherent cost to cell growth when amino acid misincorporation occurs.

To test the recruitment hypothesis, we used chromosomal manipulation (29, 30) to develop a system for stringently evaluating the in vivo function of a “recruited” tRNA (Fig.2). We met this criterion by using homologous recombination to create an E. coli strain in which the essential, single-copy chromosomal gene for tRNAUGU Thr (31) was inactivated and in which the wild-type copy of the gene was simultaneously transferred to the plasmid pMAK705 having a temperature-sensitive replicon (29). The chromosomal tRNAUGU Thr gene was inactivated by replacing its anticodon stem and loop with a sequence corresponding to a Drd I restriction endonuclease site (32). This altered the normal anticodon stem-loop structure and created a stem consisting of six base pairs closed by a tetraloop in the RNA product. Consequently, this product (tRNAThr-Drd I) was one nucleotide shorter than the wild-type tRNAUGU Thr. The knockout strain was transformed with the plasmid p177 bearing the test tRNAs (33) and the cells were grown at the nonpermissive temperature to cure them of pMAK705 bearing the wild-type tRNAUGU Thr gene. Thus, after this point, cell viability depended on the activity of the tRNA expressed from p177.

Figure 2

Scheme for testing the recruitment hypothesis. An E. coli knockout strain was created in which the chromosomal copy of tRNAUGU Thr was inactivated and the wild-type copy was expressed from the plasmid pMAK705 having a temperature-sensitive replicon (29, 32). The test tRNAs were cloned into p177 (33) and then transformed into the knockout strain. Subsequent growth at 43°C cured the cells of pMAK705 bearing the wild-type copy of tRNAUGU Thr so that the function of the test tRNAs could be evaluated.

Several factors dictated the choice of tRNAs for testing the recruitment hypothesis (Fig. 3). We chose to knock out the function of the tRNAUGU Thr gene because it is an essential, single-copy gene whose chromosomal location is known (14, 15). Moreover, its manipulation was facilitated by the availability of DNA sequence for the flanking regions (15). The decision to attempt to recruit the tRNAUCU Arg gene to function as a tRNAUGU Thr isoacceptor was based on several observations that suggested that these two genes may have been involved in a previous recruitment event. First, the anticodons of the encoded tRNAs differ by only the single nucleotide at the central position (position 35) of the anticodon. Second, our analyses revealed that the sequence of tRNAUGU Thr was as similar to tRNAUCU Arg (68%) as it was to the two other threonine isoacceptors (58% and 68%, respectively). We also reasoned, on the basis of the results of several in vitro aminoacylation studies, that a UCU → UGU change in the tRNAArg anticodon (a C35G mutation) would decrease aminoacylation by arginyl synthetase [ArgRS (23, 34, 35)] and would promote recognition by threonyl synthetase [ThrRS (24, 36)] while simultaneously enabling the tRNA to decode the corresponding threonine (ACA and perhaps ACG) codons. However, the possibility existed that the tRNA would retain some of its arginine identity because another ArgRS identity element, A20 (26), remained intact. Consequently, we created two versions of the recruited tRNA: tRNAR→T(A20), with only the C35G mutation, and tRNAR→T(U20), which also had an A20U mutation (Fig. 3).

Figure 3

The sequences of tRNAUGU Thr and tRNAUCU Arg (13, 15). Arrows indicate mutations in tRNAUCU Arg that create the two recruited tRNAs. Both tRNAR→T(A20) and tRNAR→T(U20)have the C35G mutation, which creates a threonine anticodon (UGU), but they differ with respect to whether nucleotide 20 is A or U.

Gain of function by a single nucleotide substitution in the tRNA anticodon. Figure 4 shows the growth of the engineered E. coli strains under various conditions. In each strain, the chromosomal copy of the tRNAUGUThrgene is inactivated. At the outset, all strains have the plasmid pMAK705 bearing the wild-type copy of tRNAUGUThr. This plasmid confers chloramphenicol resistance and has a temperature-sensitive replicon. The strains also have the plasmid p177 from which the test tRNAs are expressed and which confers kanamycin resistance. Consequently, all strains can grow at the permissive temperature (30°C) on LB plates containing chloramphenicol or kanamycin. Growth at 43°C blocks replication of pMAK705 and thus causes the associated wild-type copy of the tRNAUGUThrgene to be lost. This is evidenced by the temperature-dependent chloramphenicol susceptibility of all strains. However, strains will grow after being cured of pMAK705-tRNAUGUThr if the tRNA expressed from p177 can compensate for the chromosomal knockout.

Figure 4

Growth of E. coli strains expressing “recruited” tRNAs. The growth of the two strains expressing the recruited tRNAs (tRNAR→T(A20)and tRNAR→T(U20)) is compared to that of strains expressing the positive control (tRNAUGU Thr) and the negative controls (tRNAArg(A20) and tRNAArg(U20)). The diagram illustrates the genotype of each strain before growth at the nonpermissive temperature. The growth of dilute cell suspensions of each strain (45) was tested at 30° and 43°C on LB plates containing either chloramphenicol (Clm) or kanamycin (Kan) at 20 μg/ml.

As expected, strains expressing the two negative control tRNAs (tRNAArg(A20) and tRNAArg(U20)) were inviable when grown at 43°C in the presence of kanamycin. These tRNAs lack the capacity to couple with threonine codons and thus are unable to compensate for the knockout of the chromosomal copy of the tRNAUGU Thr gene. The inviability of the negative control strains also indicates that mutations that could compensate for the knockout were not arising in the chromosome during the course of the experiment. Strains expressing the positive control (tRNAUGU Thr) and either tRNAR→T(A20) or tRNAR→T(U20)were viable at 43°C in the presence of kanamycin, indicating that these tRNAs compensated for the inactivated chromosomal tRNAUGU Thr gene. The strain expressing tRNAUGU Thr grew somewhat better than those expressing the recruited tRNAs, which suggested that both tRNAR→T(A20) and tRNAR→T(U20) are not as efficient as the wild-type tRNAUGU Thr at some step in the translation process. However, growth of tRNAR→T(A20)and tRNAR→T(U20) was comparable, indicating that a single anticodon mutation was sufficient for achieving tRNA recruitment. Hence, these data show that the substitution of a single nucleotide (C35G) in the anticodon of a tRNAArg results in a tRNA that is competent to insert threonine at threonine codons in all essential endogenous mRNAs to the extent necessary for ensuring cell viability. Thus, a point mutation in the anticodon can recruit a tRNA from one isoaccepting group to another.

Although the results for the negative controls indicated that the new in vivo system was stable during the course of the experiments, we also did a molecular check on the genotype of the strains. Polymerase chain reaction (PCR) amplification with primers specific for the sequence flanking the chromosomal copy of the tuf B operon, followed by restriction digests with Drd I as well as DNA sequencing, revealed that the inactivated tRNAThr-Drd I gene had not changed during the course of the experiments. Also, the sequences of tRNAUGU Thr, tRNAR→T(A20), and tRNAR→T(U20) (and of their promoters and terminators) were unchanged after expression from p177 at 43°C in the presence of kanamycin.

Northern blot visualization of tRNA expression (Fig.5) was used to evaluate how tRNA expression correlates with recruitment. The expression of tRNAUCC Gly provided an internal control for these analyses because it is transcribed from the same operon as tRNAUGU Thr (Fig. 4). Moreover, tRNAUCC Gly and tRNAUGU Thrmigrate differently on denaturing polyacrylamide gels and thus can be distinguished in a Northern blot analysis. Lane 1 shows the expression of tRNAUGU Thr and tRNAUCC Gly from the chromosome before engineering the knockout, and lanes 2 to 5 show the appearance of the inactivated tRNAThr-Drd I product. The increased expression of tRNAUGU Thr and tRNAUCC Gly in lane 2 relative to lane 1 results from the expression of both tRNAs from pMAK705 (Fig. 4). Lanes 3 and 4 show tRNA expression when the tRNAR→T(A20) and tRNAR→T(U20) strains were grown at 43°C in the presence of kanamycin. Under these conditions, the products of the chromosomally encoded tRNAThr-Drd I and tRNAUCC Gly genes were detectable, whereas the product of the pMAK705-encoded tRNAUGU Thr was not. This indicates that the strains were cured of pMAK705-tRNAUGU Thr during the actual recruitment test. Lanes 8 and 9 confirm that both tRNAR→T(A20) and tRNAR→T(U20) were expressed from p177.

Figure 5

Northern blots of tRNA expression (46). Before nucleic acid isolation, DH10B (the parental strain) and the knockout strain (having the plasmid pMAK705 but not p177) were grown at 30°C. The remaining strains were grown at 43°C so that the amount of tRNA expression, which correlates with recruitment, could be evaluated. Lanes 1 to 5, tRNA expression when the blot was probed with oligonucleotides complementary to tRNAUCC Gly, tRNAThr-Drd I, and tRNAUGU Thr. Lanes 6 to 10, tRNA expression when the same blot was probed with oligonucleotides complementary to tRNAUCC Gly and tRNAUCU Arg. The probe for tRNAUCU Arg also hybridizes to tRNAR→T(A20) and tRNAR→T(U20). The exposure time for lanes 6 to 10 was about five times that for lanes 1 to 5.

The amount of tRNA expressed from p177 relative to that expressed from the chromosome was estimated from the Northern blot analysis. Using the expression of tRNAUCC Gly as an internal control, a comparison of lanes 1 and 5 indicates that when tRNAUGU Thr is expressed from p177, its expression is only about 2.5 times that normally obtained from the chromosomal tRNAUGU Thr locus. A similar comparison of lanes 6 and 8 and of lanes 6 and 9 indicates that the expression of tRNAR→T(A20) and tRNAR→T(U20) was about 40 to 55 times that of the chromosomally encoded tRNAUCU Arg gene. This difference, which is greater than that estimated for tRNAUGU Thr, could be attributable to natural differences in the amount of expression of tRNAUCU Arg and tRNAUGU Thr from the chromosome rather than to differences in the expression of tRNAs from p177. Indeed, the tRNAUCU Arg isoacceptor is responsible for translating the extremely rare AGA arginine codon (37) and is thus expected to be in low abundance in the cell (38). Thus, it seems likely that with only a modest (2.5-fold) overexpression, both tRNAR→T(A20) and tRNAR→T(U20) can assume the role of the natural tRNAUGU Thr isoacceptor.

Aminoacylation assays were undertaken to determine the suitability of each tRNA as a substrate for ArgRS and ThrRS (Table1). These assays revealed that the efficiency (maximal velocity V max divided by Michaelis constant K m) of ThrRS aminoacylation of a wild-type tRNAArg was lower than that of a T7 transcript of tRNAUGU Thr by a factor of more than 6600. In contrast, the C35G mutation improved aminoacylation by ThrRS, such that the efficiency of ThrRS aminoacylation of tRNAR→T(A20) was lower than that of T7-tRNAUGU Thr by a factor of only 110. The C35G mutation had the opposite effect on aminoacylation by ArgRS; the efficiency of ArgRS aminoacylation of tRNAR→T(A20) was lower than that of a T7 transcript of tRNAUCU Arg by a factor of 2500. The fact that a single mutation had opposing effects on aminoacylation by ArgRS and ThrRS is central to the success of the experiment. Apparently, the window of difference in the aminoacylation specificity of tRNAR→T(A20) for ArgRS and ThrRS was sufficient to yield a tRNA that predominantly inserted threonine at threonine codons. tRNAR→T(A20) would certainly have been lethal had it inserted arginine at threonine codons at rates greater than the normal error rate (39). Thus, although the efficiency of ThrRS aminoacylation of tRNAR→T(A20) is somewhat less than that of a wild-type tRNAThr, its single amino acid identity endows it with the key characteristic of all tRNAs. An amino acid identity corresponding to the anticodon, rather than aminoacylation efficiency, is likely to be the key prerequisite for the appearance of a tRNA variant in a population. Once such a tRNA appeared, its efficiency of aminoacylation could be improved by the combined forces of mutation and natural selection.

Table 1

Aminoacylation of wild-type and recruited tRNAs by ThrRS and ArgRS (see Fig. 3 for tRNA sequences). All tRNAs were prepared by in vitro transcription with T7 RNA polymerase (42). Partially purified E. coli ThrRS and ArgRS were used to determine kinetic parameters for each substrate (43). Reactions with ThrRS were performed at 37°C in a buffer containing 50 mM tris-Cl (pH 7.5), 8.0 mM MgCl2, 4.0 mM adenosine triphosphate (ATP), 0.1 mM EDTA, 1.0 mM dithiothreitol (DTT), and 20 μM [3H]threonine (specific activity, 10 Ci/mmol). Reactions with ArgRS were performed at 37°C in a buffer containing 50 mM tris-Cl (pH 7.5), 7.0 mM MgCl2, 4.0 mM ATP, 0.1 mM EDTA, 1.0 mM DTT, and 20 μM [3H]arginine (specific activity, 10 Ci/mmol). At each time point, tRNA samples were precipitated on Whatman 3MM paper by washing several times in ice-cold 10% and then 5% trichloroacetic acid followed by a final washing in ice-cold 95% ethyl alcohol. Kinetic parameters were determined from replicated assays that typically used five tRNA concentrations from 0.1 to 9.0 μM and synthetase concentrations from 0.14 to 1.94 U/ml. Assays involving tRNAs and a noncognate synthetase used tRNA concentrations from 5 to 17 μM and synthetase concentrations from 1.16 to 1.94 U/ml. Initial rates were calculated from linear least-squares analyses from five time points. Kinetic parameters were calculated with the program ENZFITTER (Biosoft).

View this table:

Synthetases typically recognize sets of elements in their cognate tRNAs, and members of these sets often occur in noncognate tRNAs (3, 40). Both tRNAUCU Arg and tRNAUGU Thr have a D-stem comprising four Watson-Crick base pairs and a variable loop comprising five nucleotides (Fig. 3). In addition, they both have a C2-G71 base pair in their acceptor stems. The existence of specificity determinants in the tRNA acceptor stem is well established for many groups (4). Indeed, this C2-G71 base pair appears to be important to threonine amino acid identity. It is phylogenetically conserved in threonine tRNAs (20), and it comprises one of the tRNAThr recognition elements (36). Thus, it seems likely that sequence overlap is a critical factor that predisposes tRNAs to recruitment into certain isoaccepting groups.

Our results are consistent with the idea that tRNAs within a single isoaccepting group may be derived from different ancestors. As such, they are helpful in interpreting the observed relations among tRNA gene sequences (10, 11, 20). Because recruitment involves a simultaneous change in tRNA identity and mRNA coupling capacity, tRNA gene recruitment also provides a possible molecular mechanism for Jukes' codon capture hypothesis (41).

Recruitment could also play a role in extant organisms once a gene duplication event has occurred. Given the results of the in vitro aminoacylation assays, it seems unlikely that tRNAR T (A20) would replace an active chromosomal copy of tRNAUGU Thr. However, a tRNAR T (A20) gene might replace a functionally debilitated copy of an original tRNAUGU Thr. Moreover, a tRNAR T (A20) gene might be favored in a population that had not yet evolved a threonine isoacceptor having selectivity for ACA codons. Thus, in addition to a possible role of recruitment in the expansion of tRNA amino acid assignments, recruitment events might be favored when codon usage is in flux.

  • * To whom correspondence should be addressed. E-mail: peggy{at}

  • Present address: Department of Biology, University of Oregon, Eugene, OR 97403, USA.

  • Present address: Hewlett-Packard Laboratories, 3500 Deer Creek Road, Mail Stop 25U-5, Palo Alto, CA 94304, USA.


View Abstract

Stay Connected to Science

Navigate This Article