Loss of a Universal tRNA Feature

Bioinformatic analysis.The script tFind.pl was written to manage the tRNA-finding programs tRNAscan-SE and Aragorn (26, 27) and improve their output and was applied to the complete or nearly complete RefSeq genome sequences from NCBI for 41 RRCPP bacteria, 25 other alphaproteobacteria, and two outgroup strains (Escherichia coli K-12 and Geobacter sulfurreducens). Of the 3,181 output tRNA sequences, those with undetermined identity, with introns, flagged as possible pseudogenes (except those called as tRNA^Tyr, since in all of these cases, they were the only tRNA^Tyr species in the genome and defects were not severe) and the small set of selenocysteinyl tRNAs, 92 in all, were rejected. Removing redundant identical copies from each species left 2,818 sequences, which were aligned in isoacceptor groups, and maximum-likelihood trees were generated by using Phyml. Inspection of sequences for outliers on the trees led to the rejection of eight very unlikely tRNA sequences.

HisRS genes and additional tRNA^His genes were found by querying the National Center for Biotechnology Information for “tRNA-His,”“ HisRS,” and “histidyl” (rejecting those not annotated as HisRS); additional eukaryotic sequences were collected by using the BLAST server for eukaryotic genome projects. The results included HisRS for all 303 complete prokaryotic genomes available at the National Center for Biotechnology Information on 12 January 2006, except for those of Brucella melitensis biovar Abortus 2308, Magnetospirillum magneticum AMB-1, and Nanoarchaeum equitans Kin4-M, which were identified by using BLAST. Combining identical sequences, rejecting HisRS paralogs and those not extending through motif 2, and counting multiple instances from the same gene once, there were 521 HisRS sequences (41 from the RRCPP alphaproteobacteria) and 460 tRNA sequences. Motif 2 loop sequences for 1,682 class IIa aaRS (SerRS, ThrRS, ProRS, HisRS, and some GlyRS) were taken from Pfam version 19.0 entry PF00587 after exclusion of incomplete sequences and HisRS paralogs, and the anticodon-binding domain sequences of 1,320 of the same enzymes (less SerRS) were from PF03129 (15). tRNA^His genes were aligned manually, and HisRS sequences were aligned by using MUSCLE (13), adding other aaRS motif 2 sequences as described previously (12).

RNA purification. Sinorhizobium meliloti 1021 was cultured at 30°C in TY medium with shaking to an A ₆₀₀ of 0.6, and cells from 1.8 ml of culture were rapidly harvested for total RNA extraction as described previously (3); the yield was 36μ g. Total RNA from Caulobacter crescentus C15 was a gift from Yves Brun and Ellen Quardokus (Indiana University). Templates for in vitro transcription of S. meliloti tRNA^His with or without G-1 and with the native U59 or the G59 variation were prepared by mutagenic PCR templated with genomic DNA. tRNAs were transcribed with T7 RNA polymerase, including GMP at a 15-fold excess over GTP to produce 5′-monophosphate ends, and purified after single-nucleotide resolution in a 12% polyacrylamide-8 M urea sequencing gel.

RNA circularization.Total bacterial RNA at 80 ng/μl, or a mixture of two in vitro tRNA transcripts (G₋₁/G₅₉ and X₋₁/U₅₉) at 1.5 pM each, was incubated for 60 min at 37°C in 15 μl of RNA ligase buffer (50 mM Tris-HCl [pH 7.8] at 25°C, 10 mM MgCl₂, 1 mM ATP, 10 mM dithiothreitol) with 1.3 U of T4 RNA ligase (New England Biolabs)/μl and then at 70°C for 5 min. In some experiments, RNA ligation was preceded by dephosphorylation with or without rephosphorylation. Total S. meliloti RNA was incubated at 200 ng/μl for 60 min at 37°C in 50 mM Tris-HCl (pH 7.9) at 25°C, 100 mM NaCl, 10 mM MgCl₂, 1 mM dithiothreitol, with 0.4 U of calf intestinal phosphatase (New England Biolabs)/μl, purifying the product RNA by double phenol-CHCl₃ extraction and ethanol precipitation. For rephosphorylation, dephosphorylated RNA was incubated at 125 ng/μl for 50 min at 37°C in RNA ligase buffer with 0.5 U of T4 polynucleotide kinase (New England Biolabs)/μl, purifying the product RNA as described above.

Amplification and sequencing of products.Reverse transcription-PCR (RT-PCR) primers (-R and -F) are indicated in Fig. 2A to C. RT used Superscript II (Invitrogen) according to the manufacturer's directions. PCR used GoTaq green mix (Promega) with 0.8 μM each of the RT and second PCR primers, starting from ice immediately to 3 min at 95°C, and followed by 30 cycles of 45 s at 95°C, 30 s at 50°C, and 30 s at 72°C. PCR products were analyzed in 2.5% agarose-Tris-borate-EDTA gels and in some cases gel-purified or cloned into the pGEM-T Easy vector (Promega) for sequencing by using the BigDye terminator system (Applied Biosystems) with capillary gel electrophoresis.

Primer extension.In vitro transcript standards (100 fmol) of S. meliloti tRNA^His, either with or without an extra 5′ guanylate, or 20 μg of total S. meliloti RNA were hybridized to 5′-³²P-labeled primer His-AC (Fig. 2B) and subjected to RT with Superscript II according to the manufacturer's directions. Products were extracted with phenol, precipitated with ethanol, and resuspended in formamide dye, and portions were subjected to electrophoresis in a 12% polyacrylamide-8 M urea sequencing gel, which was dried for autoradiography.

Unusual tRNA^His gene in an alphaproteobacterial group.The tRNA-finding computer programs tRNAscan-SE and Aragorn are excellent tools for locating tRNA and tmRNA genes in large sequences (26, 27), but they (i) occasionally err on gene endpoints; (ii) do not sort tRNA genes with CAT anticodons into initiator, isoleucine, and elongator methionine classes; and (iii) ignore the −1 position of tRNA^His genes. We wrote a PERL script tRNAfind.pl that solves these problems for combined tRNAscan-SE and Aragorn output, using kingdom-specific rules for distinguishing genes with the CAT anticodon (28).

Examination of the corrected tRNA^His gene sequences showed that a group of alphaproteobacteria fails to encode G at −1 as all other bacteria do; in addition, they have C72T and C73A changes relative to other bacterial genes (Fig. 1). The group, here referred to as RRCPP, includes members with known tRNA^His genes from the orders Rhizobiales, Rhodobacterales, Caulobacterales, Parvularculales, and the species Pelagibacter ubique that has been placed among the Rickettsiales. Although the only tRNA^His gene known for the order Parvularculales has G-1, the presence of T72 and A73 and analysis of its HisRS sequence (below) indicate that it is part of this group. The other alphaproteobacterial orders investigated, Rhodospirillales and Sphingomonadales, and the remaining Rickettsiales, have the canonical bacterial tRNA^His gene with G-1, C72, and C73.

• Open in new tab
• Download powerpoint

FIG. 1.

Unique tRNA^His gene and HisRS sequences in an alphaproteobacterial group. Selected segments of a sequence alignment are shown from eukaryotic cytoplasmic (E), archaeal (A), and bacterial type I (BI) and II (BII) HisRS (5), along with portions of tRNA^His genes from the same strains and from viruses (V). For eukaryotes with multiple tRNA^His genes, only the most abundant is shown. Coordinates of protein sequences are for E. coli HisRS. Abbreviations: Hsa, Homo sapiens; Dme, Drosophila melanogaster; Cel, Caenorhabditis elegans; Sce, Saccharomyces cerevisiae; Ehi, Entamoeba histolytica; Ath, Arabidopsis thaliana; Neq, Nanoarchaeum equitans; Hma, Haloarcula marismortui; Mth, Methanobacterium thermoautotrophicum; Sso, Sulfolobus solfataricus; Ape, Aeropyrum pernix; Pub, Pelagibacter ubique; Pbe, Parvularcula bermudensis; Ccr, Caulobacter crescentus; Rpa, Rhodopseudomonas palustris; Bab, Brucella abortus; Sme, Sinorhizobium meliloti; Rsp, Rhodopseudomonas sphaeroides; Cau, Chloroflexus aurantiacus; Tpa, Treponema pallidum; Rba, Rhodopirellula baltica; Xax, Xanthomona axonopodis; Nos, Nostoc sp. strain PCC 7120; BanII and BanI, Bacillus anthracis; Blo, Bifidobacterium longum; Pgi, Porphyromonas gingivalis; Cac, Clostridum acetobutylicum; Nar, Novosphingobium aromaticivorans; Mma, Magnetospirillum magneticum; Ama, Anaplasma marginale; Pma, Prochlorococcus marinus; Cab, Chlamydophila abortus; Bsu, Bacillus subtilis; Aae, Aquifex aeolicus; Tth, Thermus thermophilus; Eco, Escherichia coli; Rba2, a second tRNA gene from Rba; Mimi, Mimivirus. Species of the RRCPP bacteria are grouped in a thick box, surrounded by their two closest HisRS homologs (2), and other alphaproteobacterial species are grouped in a thin box. Alignable portions of SerRS, ProRS, ThrRS (Eco), and GlyRS (Hsa) are shown.

There is one other bacterial genome project, for Rhodopirellula baltica of the Planctomycetes, with a tRNA^His gene that does not encode G-1; it has T-1 and A73 but not T72. The function of this gene is questionable, since the same genome also contains a conventional bacterial tRNA^His gene.

tRNA^His that lacks the extra 5′ nucleotide.Three main possibilities considered for the 5′ end of tRNA^His in the RRCPP group were that (i) an uncoded guanylate is added enzymatically after standard RNase P cleavage as in eukaryotes and some mitochondria, (ii) an unusual RNase P cleavage as for other bacterial tRNA^His leaves whichever extra 5′ (nonguanylate) nucleotide is encoded in the gene, or (iii) standard RNase P cleavage leaves a mature tRNA^His with no extra nucleotide. Concerning the first possibility, no homolog of the eukaryotic tRNA guanylyltransferase was detected in any of the alphaproteobacteria, although PSI-BLAST did identify a possible homolog in the gram-positive bacteria Bacillus thuringiensis and Nocardia farcinica.

To investigate these possibilities further, we used T4 RNA ligase to join the ends of molecules in a total RNA preparation and then amplified the ligation junction and flanking sequences of circularized tRNAs by using RT-PCR (8, 42). This method has been used to determine terminal sequences of tRNAs and a tmRNA homolog in mitochondria, including a tRNA^His with an extra 5′ nucleotide (23, 42). The approach has several merits. (i) Individual tRNA species need not be purified. (ii) Contaminating genomic DNA should not interfere, since PCR primers designed to converge across the junction of a circularized RNA would have a divergent orientation on genomic DNA. (iii) If products do not contain flanking sequences found in precursor RNAs, they reflect circularization of mature RNAs. (iv) Even with mispriming, primer-internal sequence from either side of the circle junction can unambiguously identify both of the ligated RNA ends.

We first examined tRNAs from Sinorhizobium meliloti of the order Rhizobiales. A test amplification for the circularized tRNA^Trp yielded a ladder of PCR product bands of the sizes expected for a long rolling-circle product of RT. Ensemble sequencing of all products or of the gel-purified smallest band gave the expected end ligation sequence of tRNA^Trp (Fig. 2A).

• Open in new tab
• Download powerpoint

FIG. 2.

tRNA end sequence analysis. A-C. RNA ligation products. Joints formed by T4 RNA ligase (brackets), involving the circled 5′-monophosphate ends, as revealed by RT-PCR with the indicated primers. Sequences of precursor RNA sequences are shown with portions removed by tRNA processing in lowercase. (D) Amplification of tRNA^His circles. S. meliloti total RNA was treated with T4 RNA ligase (L), calf intestinal phosphatase and then RNA ligase (PL), or phosphatase and then T4 polynucleotide kinase then RNA ligase (PKL). These samples, total RNA with none of these treatments (T), and no RNA (−) were subjected to RT with primer SHis-R and then PCR with SHis-R and SHis-F. Dashes on the right mark the positions of the lowest five bands of a 100-bp ladder after electrophoresis in an agarose gel. E. Primer extension. The 5′-³²P-labeled primer His-AC was extended on 50 fmol in vitro transcript S. meliloti tRNA^His with (G) or without (X) the extra 5′ guanylate or on 10 μg of total S. meliloti RNA (S). Portions of the reaction products (7% for G and X and 40% for S) were analyzed by electrophoresis in a denaturing gel, followed by autoradiography.

RT-PCR on untreated S. meliloti RNA using primers for tRNA^His circles yielded a small artifactual product (Fig. 2D, lane T). (The artifact of lane T of Fig. 2D was independent of RNA ligation but did appear to depend on RT, even in PCR with annealing temperatures as low as 42°C, which ruled out simple mispriming on any contaminating genomic DNA. It had the terminal sequence 5′-… CCCAAC CGTAC cagttggttagagcgcaggt[where the sequence from primer SHis-R is in lowercase and the sequence from the 3′ terminus of tRNA^His is underlined], and thus showed effective priming at the penultimate nucleotide of tRNA^His. A single band of the same size was also observed in no-ligation controls with purified in vitro transcript tRNA.) After treatment with RNA ligase to circularize tRNAs, the smallest RT-PCR product was larger than the artifact and was associated with a ladder of bands as for the tRNA^Trp circle (Fig. 2D, lane L). The gel-purified smallest and next-smallest amplification products both had the sequence from end ligation of tRNA^His and showed that no extra nucleotide was present at the 5′ end of the circularized tRNA^His (Fig. 2B). This result was confirmed in both tested clones of the PCR products.

To confirm the correlation of this unusual tRNA structure with the unusual gene, we determined the terminal sequence of tRNA^His from a different order in the RRCPP group, Caulobacterales. As for S. meliloti, no extra nucleotide was found on the tRNA^His of Caulobacter crescentus (Fig. 2C) in six of seven clones of the amplified circularization junction. (The remaining clone appeared to originate from a circularized tRNA^His precursor; its junction sequence contained two additional 3′ nucleotides and ten additional 5′ nucleotides beyond the mature tRNA sequence).

Possible terminus other than 5′-monophosphate.These results were surprising given that all other known tRNA^His have an extra 5′ nucleotide and were subject to an objection. Circularization by T4 RNA ligase requires 5′-monophosphate and 3′-hydroxyl ends. It was possible that the circles we detected were from a minor variant form of tRNA^His and that the major species has a 5′ terminus other than a monophosphate; a precedent comes from a study of chicken mitochondrial tRNA^His, at least a fraction of which was shown to have a 5′-di- or triphosphate (25). To address this issue, we treated S. meliloti RNA first with calf intestinal phosphatase to remove terminal mono-, di-, or triphosphates and produce uniform 5′-hydroxyl ends and then with T4 polynucleotide kinase to produce uniform 5′-monophosphate ends. This was followed by RNA ligation and amplification for tRNA^His circles as before. Dephosphorylation reduced circularization substantially; the main product after RNA ligation and amplification was the same artifact seen when the RNA ligation step was omitted (Fig. 2D, lane PL; compare to lanes T and L). Rephosphorylation was also effective; the artifact was not apparent among the RT-PCR products and only the circularization ladder seen for the original RNA was detected (Fig. 2D, lane PKL [compare to lane L]). Any tRNA^His 5′ ends that were originally hydroxyl, diphosphate, or triphosphate would have been substantially converted to monophosphate by rephosphorylation, and yet these tRNA circles still lacked any −1 nucleotide, as revealed in all of six clones of the PCR products. Thus, the terminal nucleotide sequence of tRNA^His in S. meliloti is uniform and lacks the extra 5′ nucleotide found in all other tRNA^His. Canonical RNase P processing as for non-tRNA^His is indicated by the observations that the mature 5′ sequence ends at the +1 position, and at least a fraction of the tRNA^His had the 5′ monophosphate expected from RNase P cleavage.

Circularizability of tRNA containing G₋₁.Our circularization assay may have failed to detect tRNA with an extra guanylate if that nucleotide specifically prevents circularization, for example, by leaving the 5′ and 3′ ends in such a configuration that they are unable to approach each other. To examine whether tRNA^His with an extra 5′ guanylate can circularize efficiently, we prepared synthetic (unmodified) versions of S. meliloti tRNA^His either with (G₋₁) or without (X₋₁) the extra 5′ guanylate by in vitro transcription. Titration in a circularization assay showed that both tRNAs produced RT-PCR products at concentrations as low as 3 pM (data not shown).

Dilution favors circularization but does not rule out bimolecular ligation. To test whether bimolecular ligation might explain the result with G₋₁ tRNA^His and, if not, to better compare circularization efficiencies, the assay was performed on a dilute mixture of G₋₁ and X₋₁ tRNAs, with one tRNA marked near its 3′ end by an innocuous sequence variation. In this way, circularization and self-ligation products could be distinguished from cross-molecule ligation products. The innocuous variation was placed at position 59, which is the most variable position in the T-loop of tRNAs in general and is unoccupied by secondary or tertiary contacts in tRNA crystal structures. S. meliloti has U₅₉, and among tRNA^His sequences from 41 RRCPP strains, all four bases are found at this position, with G₅₉ in 26 of them. Accordingly a G₋₁G₅₉ version of S. meliloti tRNA^His was prepared and mixed with X_-1U₅₉ tRNA^His at 1.5 pM each, and the mixture was treated with T4 RNA ligase. RT-PCR products were cloned and sequenced to identify ligation partners for individual product molecules (Table 1) . Eighteen of the tested clones had ligation junctions with perfect tRNA sequences, and these were exclusively from the two types of self-ligation (G₋₁/G₅₉ and X₋₁/U₅₉); none had the G₋₁/U₅₉ or X₋₁/G₅₉ ligation junction that would have indicated cross-ligation. If no circularization occurred in the reaction, and all four bimolecular ligations were equiprobable, this result would be expected with a probability of 0.5¹⁸= 4 × 10⁻⁶. Even if the input ratio of molecules was as skewed as the output ratio (2:1) might suggest, the probability that the result was due to bimolecular ligation would still be negligible. Four of these clones were tandem repeats of the same self-ligation junction, further supporting self-ligation (and implying that RT had proceeded twice around a circle). The exclusive detection of self-ligation products indicates that circularization predominated under these dilute conditions. Thus, G₋₁ tRNA^His can circularize, although possibly with a mild (∼2-fold) disadvantage in circularization efficiency relative to X₋₁ tRNA^His. Nine additional clones came from tRNAs with imperfect 3′ ends; despite tRNA purification in a sequencing gel, these clones showed an extra terminal nucleotide, single-nucleotide deletion, or base substitution in the CCA tail. All nine were self-ligation products, which does not provide information on the relative circularization efficiency of the two (perfect) tRNA types but does support the conclusion that cross-molecule ligation was insubstantial under these conditions. We conclude that the failure to detect G₋₁ tRNA^His in total S. meliloti RNA was not because an extra guanylate prevents circularization.

Primer extension.Modifications at the 5′ end of mature tRNA^His, possibly containing an extra nucleotide, may have prevented circularization of the mature tRNA, allowing detection only of immature processing intermediates. To investigate this possibility, the 5′end of tRNA^His was examined by primer extension, using as standards the in vitro transcripts containing or lacking the G-1 nucleotide (Fig. 3). The RT primer used in the previous experiments (Fig. 2B) was designed and demonstrated to perform well in RT-PCR of circularized tRNA but may not have been sufficiently specific in a single round of RT; a new primer was designed to maximize specificity for tRNA^His by positioning the 3′ end in the anticodon sequence. The primer extension products from S. meliloti total RNA (lane S) matched those of the standard lacking the extra nucleotide (lane X) and not those of the standard containing an extra 5′ guanylate (lane G). This result, which did not depend on the use of T4 RNA ligase, is compatible with that from circularization, again suggesting that mature S. meliloti tRNA^His lacks any extra nucleotide.

• Open in new tab
• Download powerpoint

FIG. 3.

Map of changes to histidyl-tRNA synthetase. A complex between HisRS and tRNAHis was modeled from available structures, with five sites (Fig. 1) where RRCPP HisRS differs from canonical bacterial HisRS marked in red. (A) The ribbon is monomer A from a crystal structure (1ADJ) of the Thermus thermophilus HisRS homodimer (1). Its catalytic domain backbone was superposed by using the iterative fit function of the Swiss-Pdb viewer 3.7 (20) (552 atoms, RMSD 1.43Å) with that from the E. coli AspRS-tRNA^Asp crystal structure (1COA) to mark the aspartyl-adenylate (aa-AMP) and the acceptor arm (black) of tRNA^Asp (14). By analogy with AspRS-tRNA^Asp, the insertion domain here in an open conformation would be expected to close toward the tRNA acceptor end, clamping it to the catalytic domain. The anticodon-binding domain backbone was superposed (320 atoms, RMSD 1.39Å) with that from the E. coli ThrRS-tRNA^Thr crystal structure (1QF6) to mark two anticodon bases of tRNA^Thr (34). The elbow of tRNA^His would contact the back of the other HisRS monomer (16). Positions (E. coli numbering) of residues and inserts unique to the RRCPP HisRS and bases of RRCPP tRNA^His are shown. (B) View from the left of panel A.

Sequence covariations between a protein and its RNA substrate.The class II aaRS share sequence motifs and structure in the catalytic domain, and two additional domains are usually present, an insertion domain and an anticodon-binding domain (Fig. 3). A stepwise pathway for tRNA binding has emerged from multiple crystallographic studies of the class II AspRS (7, 35). The tRNA is first recognized by the anticodon-binding domain, which rotates, bending the tRNA elbow and driving the acceptor end toward the catalytic domain. The insertion domain also rotates, beginning to clamp the acceptor end onto the catalytic domain. When the aminoacyl-adenylate substrate is in the active site, the acceptor end can also be properly accommodated in the active site, with increased clamping by the insertion domain. The acceptor end is further anchored by contacts to two loops within the catalytic domain, the flipping loop and the motif 2 loop, after conformational changes in the loops. Analogs or homologs of all of these critical parts of AspRS can be identified in crystal structures of HisRS (16), and even though there is no structure available for a HisRS complex with tRNA, there are indications that lessons from AspRS apply (6).

HisRS sequences from RRCPP bacteria form a discrete cluster, with major differences from other HisRS in all of the regions that might be expected to be involved in tRNA binding. Attention has previously focused on the motif 2 loop of HisRS, where functionally relevant covariation between HisRS and tRNA^His sequences has been noted (21). The tRNA^His discriminator position 73 is a C in bacteria and an A in eukaryotes, while HisRS position 118 (E. coli numbering) is a Gln in bacteria and differs (typically found as an Ala-Met-Thr tripeptide) in eukaryotes. Likely contact between Gln118 and C73, as the corresponding position of AspRS makes with the discriminator base of tRNA^Asp, was demonstrated by directed mutational analysis (21). The RRCPP group shows a different covariation involving the same tRNA and HisRS positions: the tRNA change from C73 to A correlates with a HisRS change from Gln118 to Gly (Fig. 1). The Gly may not directly recognize A73, but there are additional changes in the motif 2 loop not found in other HisRS that could affect tRNA recognition, Asn114 and strikingly Pro119 (not present in the Pelagibacter ubique HisRS, the most divergent of the RRCPP sequences). These changes are unique among HisRS proteins except for a few occurrences of Gly118 in other bacteria but can also be assessed by comparison to the motif 2 loop of other class IIa aaRS. In this broader context, Gly118 is not unusual and is highly conserved among SerRS and ThrRS. Asn114 is likewise not unusual outside of HisRS. Pro119, however, is unique among all class IIb enzymes.

The flipping loop region of the RRCPP also stands out among HisRS; in particular, Pro63 is found uniformly and uniquely in the RRCPP. Unique insertions occur at other sites in the HisRS of RRCPP, in the insertion domain (Fig. 1 and 3, insert 187), in the catalytic domain (insert 273), and in the anticodon-binding domain (insert 405). This latter insert can also be assessed in a broader context because its region is well aligned with ThrRS, ProRS, and GlyRS; it is unique among all class IIa aaRS.

The other alphaproteobacteria, in the Rhodospirillales, Sphingomonadales, and Rickettsiales other than Pelagibacter, do not have the unusual HisRS, except that genes for both types of bacterial HisRS are found in data from the incomplete Magnetospirillum magnetotacticum genome project. However, the source strain for this project was isolated based on its magnetic properties (4) and may be impure, as is also suggested by the unprecedented finding of two dissimilar tmRNA genes among the same genomic sequence; the complete genome of its congener M. magneticum encodes only one of these two HisRS, similar to those of other Rhodospirillales.

Effects on charging systems for other tRNAs.When initially established in an RRCPP ancestor, the unusual tRNA^His and HisRS might have affected global tRNA charging accuracy in the cell, with increases in the histidylation of non-tRNA^His and in the mischarging of tRNA^His. This situation might have selected for compensatory changes in other aaRS or other tRNAs or for further compensation in the HisRS or tRNA^His molecules themselves. As an initial exploration of possible effects on other charging systems, we examined whether any other tRNAs covary with tRNA_His. All unique tRNA sequences were collected from each of 41 RRCPP strains, 25 non-RRCPP alphaproteobacteria, and two outgroup proteobacteria (Escherichia coli K-12 and Geobacter sulfurreducens), for each standard isoacceptor family. The sequences were aligned for each family, and each position in each alignment was tested for whether a single base was present in >90% of the RRCPP sequences and a different base was present in >90% of the other bacteria. Only the three previously noted positions in tRNA^His (−1, 72, and 73) were identified by this test. Reducing the cutoff to 80% added a fourth tRNA^His site (position 70) and two pairing positions (51 and 65) in the tRNA^Glu T-stem. Retaining the 90% cutoff, but allowing any set of two or more of the four bases to occur in one bacterial group, with a mutually exclusive set of bases in the other bacterial group, identified one new position, again in tRNA^Glu, where position 72 is 100% C or G in RRCPP bacteria and 100% A or U in the other bacteria. The pairing partner of position 72 (position 2) in tRNA^Glu was not detected this way because the non-RRCPP alphaproteobacteria frequently have G₂ paired with U₇₂. Thus, a site in tRNA^Glu covaries perfectly with tRNA^His among the tested alphaproteobacteria, producing a three-H-bond acceptor stem pair (C₂:G₇₂ or G₂:C₇₂) in RRPCC bacteria and a two-H-bond pair (A₂:U₇₃, U₂:A₇₃, or G₂:U₇₃) in the other alphaproteobacteria.

Loss of a universal tRNA^His feature.tRNA^His in all previously studied organisms has a regular nucleotide at the extra− 1 position, usually a guanylate with a 5′-monophosphate. If an extra guanylate or other regular nucleotide were present on the RRCPP tRNA^His, with either a hydroxyl or mono-, di-, or tri-phosphate 5′ end, our experiments could have detected it. The primary conclusion from our exclusive detection of tRNA^His species lacking the extra nucleotide is that tRNA^His does not contain a regular nucleotide at the −1 position in RRCPP bacteria, as is otherwise universal.

Remaining possibilities are that (i) the tRNA species that we detected is the mature tRNA which simply has no extra nucleotide or (ii) an immature tRNA species was detected while the mature tRNA has an irregular −1 nucleotide that prevents the action of either RNA ligase or reverse transcriptase. If the primer extension experiment had demonstrated an extra 5′ nucleotide, this would have argued against the first possibility and for the second, but it did not reveal an extra 5′ nucleotide and therefore did not resolve the issue. Definitive resolution will probably await direct analysis of the purified tRNA^His molecule. For now we note that a mature RRCPP tRNA^His lacking an extra nucleotide can be readily explained by standard RNase P cleavage of the precursor and that there is no precedent for 5′ addition of an irregular nucleotide. Thus, the most parsimonious interpretation of the data in hand is that the mature tRNA simply lacks an extra nucleotide.

It was demonstrated here that versions of S. meliloti tRNA^His that contain or lack G₋₁ can both be circularized by using T4 RNA ligase. This contrasts with a report that neither of the corresponding versions of an E. coli tRNA^His microhelix could be circularized (36). Key differences are positions 72 and 73, which are U₇₂A₇₃ in S. meliloti and C₇₂C₇₃ in E. coli.

The anomalous tRNA^His was found in a group of alphaproteobacteria. Another case where tRNA^His may prove to lack an extra 5′ nucleotide is among bacteriophages. Several phages (with hosts as diverse as gammaproteobacteria, gram-positive bacteria, and cyanobacteria) have a tRNA^His gene, but all current examples have the unusual termini T-1/G1/C72/C73. A similar mutant of the E. coli tRNA^His gene generates a tRNA lacking an extra 5′ nucleotide (40), which calls into question the report in an early tRNA sequence compilation (itself re-reported from a meeting abstract and since removed from the compilation) that the tRNA^His of E. coli phage T5 retains U-1 (38). In other cases there is no compelling reason to question the presence of the extra nucleotide. Whereas phage tRNA^His genes differ from those of their hosts, the tRNA^His gene of mimivirus, which infects amoebae, matches those of eukaryotes and so therefore should its product (Fig. 1). Other bacteria, plastids, many mitochondria, and most archaea encode G-1 in their tRNA^His genes. A small number of archaea fail to encode G-1, but these species are all equipped with a Thg1 homolog and so can be expected to add G-1 posttranscriptionally by the eukaryotic mechanism.

The novel tRNA^His might function more efficiently during translation. By occupying its discriminator nucleotide in a base-pair with G-1, typical bacterial tRNA^His has a shorter 3′ tail than does standard tRNA. The potential G-1:A73 noncanonical pair in eukaryotic tRNA^His may likewise limit the flexibility of the tail. These unconformities could impair accommodation in the active site of the ribosome during translation of His codons, an effect that would be relieved with the standard 3′ tail of the novel tRNA^His in the RRCPP group.

Additional novelty of RRCPP tRNA^His.The tRNA^His genes in the RRCPP group resemble those of eukaryotes in encoding A73 at the discriminator position and failing to encode G-1. Recent phylogenetic analyses of tRNA^His gene sequences and HisRS suggest that (i) the novel RRCPP tRNA gene originated by divergence of an ancestral alphaproteobacterial tRNA^His gene, not by lateral transfer from a eukaryotic donor; (ii) that the RRCPP HisRS is of the type found in eukaryotes which has replaced the typical bacterial HisRS in prokaryotes in several cases; and (iii) that a prokaryote is a more likely proximal donor of this HisRS to the RRCPP bacteria (2). The links to the eukaryotic system led to the presumptions by these authors that tRNA^His identity rules would be like those of eukaryotes and that the G-1 would still be present in the mature RRCPP tRNA^His. Instead, the absence of the −1 nucleotide shows that tRNA^His identity rules in the RRCPP group are unlike those in any other system.

An additional novelty in the RRCPP group is that the base pair between positions+ 1 and 72, which is a G:C pair in nearly all other tRNA^His, has become a G:U pair. This serves as the closing base pair for the acceptor stem and is a good candidate for an identity determinant for the novel tRNA (39).

Disjoint ancestry of RRCPP bacteria.If the species of the RRCPP group comprise a single clade, the distribution of the novel HisRS/tRNA^His system could be explained by vertical inheritance after an instance of coevolution of the molecules. If the species are split into multiple clades within the alphaproteobacteria, this might point to lateral transfer of one or both genes. Published genome-based species trees of the alphaproteobacteria have placed Rhizobiales, Rhodobacterales, and Caulobacterales together in a clade, apart from the Sphingomonadales, Rhodospirillales, and Rickettsiales, but do not address the affiliations of the more recently sequenced Parvularcula and Pelagibacter (33). A recent multiprotein study of 72 alphaproteobacterial strains (K. P. Williams, B. W. Sobral, and A. Dickerman, unpublished data) agrees with the above findings and additionally places Parvularcula close to Caulobacter but places Pelagibacter at the base of the Rickettiales branch with high support, apart from the other species with the novel tRNA^His system. This suggests lateral transfer of the novel HisRS gene within the alphaproteobacteria, most likely directed from a member of the Rhizobiales/Rhodobacterales/Caulobacterales/Parvularculales clade into an ancestor of Pelagibacter. The native tRNA^His gene of the recipient could have undergone the two key base substitutions (C72T and C73A) after transfer of the HisRS gene alone, but the juxtaposition of the HisRS and tRNA^His genes within an apparent operon in Pelagibacter suggests instead that the two genes were transferred together.

Correlated variation in HisRS.In vivo and in vitro studies of yeast HisRS and in vitro study of E. coli HisRS show that the presence of G-1 is the major identity element of tRNA^His, whereas an in vivo study in E. coli indicates that the discriminator base C73, unique among E. coli tRNAs, is a more important identity element (18, 22, 29, 40). The RRCPP group has lost both of these key tRNA^His identity elements. This would appear to be detrimental to translational accuracy unless the losses are compensated. Without implying any particular ordered evolutionary sequence, compensation for these tRNA changes might occur at three levels: (i) HisRS must efficiently recognize the novel tRNA^His; (ii) the HisRS changes, allowing this recognition may require further changes in HisRS or in other tRNAs, so that other tRNAs (which the novel tRNA^His resembles more than standard tRNA^His does) are not mischarged with histidine; and (iii) other aaRS-tRNA recognition rules may need to change so that the less distinctive tRNA^His is not mischarged with other amino acids. Regarding the first point, any or all of the unique features of the RRCPP HisRS described in Results may prove to enable recognition. This is especially likely for the motif 2 and flipping loops, and the inserts in the insertion and catalytic domains are roughly positioned such they could also plausibly interact with the acceptor end. It should be noted that very large displacements of the motif 2 and flipping loops are not likely since these same loops also contribute to the activation of histidine.

An insert at position 405 is unique among all homologous anticodon-binding domains of class IIa aaRS. The position of insert 405 is also interesting because an amino acid substitution at this same site was obtained in a selection for E. coli HisRS that could compensate for a mutation at the acceptor end of suppressor tRNA^His (41). In vitro, this HisRS mutant had tighter binding to each form of tRNA tested, although with reduced discrimination between cognate and noncognate tRNAs (6). Insert 405 may function similarly in RRCPP HisRS, compensating for weakened interaction with tRNA in the catalytic domain by promoting tighter binding in the anticodon-binding domain.

Effects on other systems in RRCPP cells.Regarding competition among other tRNAs for charging by HisRS, the RRCPP G₊₁:U₇₂ pairing, although phylogenetically unique among tRNA^His, is not unique among the tRNAs within a cell; for example, an elongator tRNA^Met of S. meliloti has the same closing base pair and discriminator base as tRNA^His.

An initial investigation into whether other tRNAs may have adapted to the unusual His system revealed that a three-H-bond base pair is found in the second position of the acceptor stem in tRNA^Glu of all RRCPP bacteria, where a two-H-bond base pair is found in all non-RRCPP alphaproteobacteria. This may represent an adaptation to prevent mischarging by the unusual HisRS of the RRCPP group, although we do not have a ready structural explanation. Because one RRCPP genus (Pelagibacter) is not of the same ancestry as the others, the phylogenetic distribution of tRNA^Glu is not a simple coincidence of two variants vertically inherited from a single ancestor. The Pelagibacter tRNA^Glu gene may have hitchhiked on the same presumptive transfer event that brought in the tRNA^His-HisRS operon, but these units are currently distant in the genome.

Full compensation for the loss of the extra guanylate may also require changes in other aaRS so that they can discriminate against the less distinctive tRNA^His; it will be interesting to investigate whether non-HisRS aaRS signatures change in the RRCPP group. As an alternative to compensation at the level of global tRNA charging, it is at least possible that the unusual tRNA^His and HisRS have resulted in an uncompensated reduction in tRNA charging accuracy that these bacteria simply abide.

A second arena where there might be a change correlated with the novel tRNA^His is in RNase P structure and function. It will be interesting to investigate whether RNase P in the RRCPP group, like eukaryotic RNase P (9), is unable to perform the unusual cleavage required for the canonical bacterial tRNA^His. If so this could help elucidate features of bacterial RNase P responsible for the unusual cleavage.

Conclusion.While the loss of a universal and apparently ancient tRNA feature was especially surprising, the additional changes to both tRNA^His and HisRS in the alphaproteobacterial group further indicate a radical departure from previously known tRNA^His identity rules. It will certainly be interesting to investigate the structure and in vitro tRNA recognition properties of this novel HisRS.