Life without the essential bacterial tRNAIle2–lysidine synthetase TilS: a case of tRNA gene recruitment in Bacillus subtilis


E-mail; Tel. +33 1 3465 2514; Fax +33 1 3465 2521;

E-mail henri.grosjean@igmors.u-psud; Tel. +33 1 6915 4634; Fax +33 1 6915 4629.


In eubacteria, the post-transcriptional modification of the wobble cytidine of the CAU anticodon in a precursor tRNAIle2 to a lysidine residue (2-lysyl-cytidine, abbreviated as L) allows the amino acid specificity to change from methionine to isoleucine and the codon decoding specificity to shift from AUG to AUA. The tilS gene encoding the enzyme that catalyses this modification is widely distributed. However, some microbial species lack a tilS gene, indicating that an alternative strategy exists to accurately translate the AUA codon into Ile. To determine whether a TilS-dependent bacterium, such as Bacillus subtilis, can overcome the absence of lysidine in its tRNAIle2 (CAU), we analysed the suppressor mutants of a tilS-thermosensitive allele. These tilS-suppressor mutants carry a substitution of the wobble guanosine into thymidine in one of the tRNAIle1 genes (the original GAT anticodon is changed to a TAT). In absence of TilS activity, the AUA codons are translated into isoleucine by the suppressor tRNAIle1, although a low level of AUA codons is also mistranslated into methionine. Results are in agreement with rare cases of eubacteria (and archaea), which naturally lack the tilS gene (or tiaS in archaea) but contain a tRNAIle2 gene containing a TAT instead of a CAT anticodon.


Post-transcriptional modifications of selected nucleotides within tRNAs are essential for the translation process (Björk and Hagervall, 2005; Grosjean, 2005).1 In particular, the first base of the anticodon loop of tRNAs (position 34, the so-called wobble base; Crick, 1966) can be changed into a large variety of modified nucleotides, depending on the associated tRNA isoacceptor and the organism in question. Modified nucleotides change the local structural flexibility and rigidity of the tRNA molecule (Motorin and Helm, 2010), which alters recognition of the tRNA by its cognate aminoacyl–tRNA synthetase (reviewed in Giegéet al., 1998; Giegé and Lapointe, 2009). Modified anticodon nucleotides also fine-tune codon-anticodon interactions during mRNA translation in the ribosome, allowing different amino acid residues to be inserted into growing polypeptide chains (reviewed in Gustilo et al., 2008; Grosjean et al., 2010).

Lysidine (2-lysyl-cytidine; L) is a cytidine modification found in the wobble position-34 of a peculiar tRNA from Escherichia coli, tRNAIle2. This L34-modification has a dual role: it causes mature tRNAIle2 to be aminoacylated by isoleucyl–tRNA synthetase and limits the codon-reading specificity to the rare codon, AUA (Muramatsu et al., 1988b). In contrast, the tRNAIle2 lacking the lysyl group on C34 is aminoacylated by methionyl–tRNA synthetase and exclusively reads the methionine codon AUG (Muramatsu et al., 1988a; Soma et al., 2003; Ikeuchi et al., 2005; Nakanishi et al., 2009). A major tRNA species in E. coli, tRNAIle1, harbours a G34AU anticodon and translates the two frequently used isoleucine codons of AUC and AUU. Together, the tRNAIle1 and tRNAIle2 pair is able to decipher efficiently and accurately all isoleucine codons (reviewed in Grosjean and Björk, 2004; Suzuki and Miyauchi, 2010). Since its discovery by Yokoyama's group in 1988, 2-lysyl-cytidine has been identified in the tRNAIle2 (usually a minor species) of a vast majority of eubacterial species.

In Archaea, the isoleucine AUA codon is also translated by a peculiar Ile–tRNAIle2, which harbours a modified cytidine at the wobble position of its CAU anticodon. However, unlike the lysidine of the bacterial Ile–tRNAIle2, the wobble cytidine-34 is post-transcriptionally modified into agmatidine (2-agmatinyl-cytidine, abbreviated as agm2C) (Ikeuchi et al., 2010; Mandal et al., 2010). Similar to the bacterial tRNAIle2, the archaeal tRNAIle2 unmodified at the wobble C34 is recognized by the methionine–tRNA synthetase and charged with methionine. It then specifically decodes methionine AUG codons and not isoleucine AUA codons (Köhrer et al., 2008; Ikeuchi et al., 2010; Mandal et al., 2010).

Formation of lysidine in bacterial tRNAIle2 and agmatidine in archaeal tRNAIle2 are catalysed by the enzymes tRNAIle2–lysidine synthetase (TilS) and tRNAile2–agmatidine synthetase (TiaS) respectively (Nakanishi et al., 2005; Kuratani et al., 2007; Salowe et al., 2009). Although the products of TilS and TiaS activities result in the same decoding function for tRNAIle2, these two enzymes are evolutionarily unrelated (Ikeuchi et al., 2010).

Since the fully matured and modified tRNAIle2 in bacteria and archaea appears to be the only tRNA able to read isoleucine AUA codons, the viability of organisms that utilize tRNAIle2 and TilS or TiaS should depend on the in vivo activity of the respective enzymes. Indeed, partial inactivation of the E. coli tilS gene resulted in an AUA codon-dependent translational defect (Soma et al., 2003), whereas in B. subtilis and the archaea Haloferax volcanii, complete inactivation of tilS/tiaS was lethal (Kobayashi et al., 2003; Blaby et al., 2010). Bacterial TilS was proposed to be one of the 206 essential protein-coding genes required for maintaining bacterial cell life (Gil et al., 2004), and to be essential for reconstructing a minimal synthetic cell (Mushegian and Koonin, 1996; Moya et al., 2009; Gibson et al., 2010). However, the sequenced genomes of a few living organisms appear to lack a tilS homologue and a tDNAIle2 (CAT) (Jaffe et al., 2004; Silva et al., 2006; de Crécy-Lagard et al., 2007; Suzuki and Miyauchi, 2010), which suggests that alternative mechanisms of AUA codon translation exist.

To identify the bacterial strategies used to overcome the need for TilS, we characterized the spontaneous mutations that restored the growth of a conditionally lethal B. subtilis TilS mutant. Survivors were able to recruit one copy of a mutant tRNAIle1 gene, which coded for a uridine in its wobble base such that the mutant tRNAIle1 could translate AUA codons as isoleucine. However, this strategy did not appear as accurate as the use of a combination of TilS/TiaS and tRNAIle2 (anticodon CAU). We analysed several eubacterial and archaeal species, which naturally lack a tilS or tiaS gene, and found that these organisms have an alternative system for reading the isoleucine AUA codon, which is similar to that found in the B. subtilis suppressor strains.


Construction of a thermosensitive tilS allele

To search for mutations which circumvent the need for TilS activity, we constructed a conditionally lethal tilS mutant. Thermosensitive (Ts) tilS mutants were obtained following random mutagenesis of the tilS gene and isolation of mutations conferring Ts growth. Because tilS forms an operon with the downstream essential hprT gene (hypoxanthine-guanine phosphoribosyltransferase), a phleomycin-resistant gene (phleo) was introduced downstream of hprT to provide a selectable marker without disrupting the structure of the operon. We generated random mutations of the tilShprT region using PCR and integrated them into the B. subtilis 168 chromosome by homologous replacement (Fabret et al., 2002), which resulted in expression of the phleo marker. From 302 phleomycin-resistant transformants, we isolated 14 Ts mutants (4.6%). Complete sequencing of the PCR-amplified regions identified two different Ts strains (tilS1 and tilS12), which carried mutations exclusively in the tilS gene. The corresponding amino acid changes in the TilS mutant proteins were: isoleucine 245 to serine (ATC to AGC), phenylalanine 287 to leucine (TTT to CTT) and glutamine 340 to proline (CAA to CCA) for the tilS1 allele; and isoleucine 341 to methionine (ATT to ATG) and aspartate 394 to glycine (GAC to GGC) for the tilS12 allele.

Both Ts strains grew at near wild-type rates in rich Luria–Bertani (LB) medium at 30°C, but their doubling times on minimal medium (tilS1 strain, 135 min; tilS12 strain, 115 min) were longer than the wild-type doubling time (90 min). The Ts phenotype was sharpest for the tilS1 allele, with 0.02% survival at 47°C and no detectable survival at 50°C, whereas it was less severe for the tilS12 allele, with 90% survival at 47°C and 0.002% survival at 50°C.

All mutations were located in the C-terminal part of B. subtilis TilS protein. Based on a mutation study of E. coli TilS (Ikeuchi et al., 2005), I245S is within a long α-helix that forms the hinge between the catalytic N-terminal domain and the CTD1 subdomain, which carries the F287L mutation, and CTD2 subdomain, which carries the Q340P, I341M and D394G mutations. In the case of E. coli TilS, the mutations that gave rise to a Ts phenotype occurred in CTD1 (L262P, which corresponds to E281 in B. subtilis TilS) and CTD2 (W323R and F425S, which correspond to S342 and I444, respectively, in B. subtilis). These mutations were shown to destabilize the correct folding of TilS. Specifically, the CTD1 mutations affected tRNA recognition and the CTD2 mutations caused TilS aggregation (i.e. insolubility) (Ikeuchi et al., 2005). Thus, as in E. coli, the Ts phenotype of the B. subtilis TilS mutants is likely due to instability and/or misfolding.

Suppression of TilS thermosensitivity by a change in the anticodon of tRNAIle1

To isolate spontaneous suppressor mutants capable of restoring growth at non-permissive temperature, tilS1 cells were plated and incubated at 50°C for 4 days. Under these conditions, seven independent suppressor strains (supTr1–7) emerged. The genetic linkage between each suppressor mutation and the phleomycin-resistant marker was measured by a back cross into wild-type B. subtilis. The linkages for supTr1, 2, 5, 6 and 7 were between 6% and 25%, which indicated that the supressors were extragenic. In contrast, the supTr3 and supTr4 mutations were tightly linked (> 95%) to the phleo marker and corresponded to intragenic suppressors. Indeed, the tilS sequence in the supTr4 mutant revealed a true reversion of the I245S mutation back to isoleucine and the continued presence of the other two tilS1 mutations (F287L and Q340P). This suggested that the thermosensitivity of the tilS1 strain was caused by the I245S mutation alone and that the I245 residue likely plays an important role in the correct folding and activity of TilS. The tilS gene in the supTr3 mutant contained the three original tilS1 mutations along with two compensatory mutations: L416V (TTG to GTG) and a +1 frameshift caused by the insertion of a G residue, which changed the last four residues of the protein from QAKS to ASKI and produced an in-frame fusion with the downstream HprT protein.

The extragenic suppressors were mapped using a derivative of the Tn10 transposon carrying a chloramphenicol resistance gene. Briefly, we produced a library of random Tn10 insertions in the chromosome of each suppressor strain and used it to search for a genetic linkage between the Tn10 derivative and the suppressor mutation, which conferred the thermoresistant phenotype to the tilS1 strain. We isolated Tn10 insertions in the vicinity of the supTr1 and supTr6 mutations, but could detect no significant genetic linkage for any of the five other suppressors. In the tilS1 supTr1::Tn10 and tilS1 supTr6::Tn10 strains, each Tn10 insertion was located in a 23S ribosomal RNA gene. As the B. subtilis chromosome contains 10 repeated ribosomal operons organized in seven clusters, we used pMutin vectors that had been inserted in a gene flanking each of the ribosomal operons to identify the precise position of each Tn10. The strains were transformed by several different chromosomal DNA regions, each carrying a pMutin insertion, and we assayed for the simultaneous loss of Tn10 chloramphenicol resistance and gain of erythromycin resistance from the pMutin vector that resulted from homologous chromosomal DNA replacement. As a result, supTr1::Tn10 and supTr6::Tn10 were mapped to the rrnB-23S and rrnO-23S genes respectively. Complete resequencing of the rrnB and rrnO loci in both suppressor strains and in the parent tilS1 strain allowed the supTr1 and supTr6 mutations to be unambiguously identified. Strikingly, both mutations corresponded to a single base change (position 34) of the wobble G of tDNAIle1 (GAT) into a wobble T, which gave rise to a new TAT anticodon (see detailed sequences in Fig. S2). The genes modified by the supTr1 and supTr6 mutations were trnB–ile2 (position 3172 kb on the genome) and trnO–ile (position 11 kb on the genome) respectively. These suppressor mutant tRNAs are thereafter referred as tRNAIle1>3. In both cases, no additional mutations were detected elsewhere in the tRNAIle1 genes or in the rrnB and rrnO operons. As B. subtilis encodes a third tRNAIle1 gene with a GAT anticodon (trnA–ile at position 31.9 kb), we sequenced the trnA–ile gene in the five uncharacterized suppressor strains but found no mutations.

Suppressor mutations in the tRNAIle1 anticodon render tilS dispensable for cell survival

The capacity of the suppressed tilS1 strain to grow at a non-permissive temperature suggested that TilS activity was dispensable for cell survival in this suppressed background. To test this hypothesis, we examined whether the tilS gene could be deleted from the supTr strains. To this end, the essential hprT gene located downstream of tilS1 in the operon was placed under the control of the IPTG-inducible PSpac promoter by single cross-over integration of a plasmid (Fig. 1A, see Experimental procedures). The resulting tilS1 PSpac–hprT strain, which required IPTG for growth, was used to transfer the tilS1 PSpac–hprT construct into the tilS1 supTr6 suppressor background to generate a tilS1 PSpac–hprT supTr6 strain. Deletion of tilS in the sup+ and supTr6 backgrounds was attempted by replacing it with a spectinomycin resistant gene (Spc). SpcR transformants were obtained for both strains, albeit at a very low frequency in the sup+ background. Analysis of the transformants by Southern blot revealed that the expected chromosome structure of the corresponding tilS deletion was obtained in the supTr6 but not in the sup+ background. The sup+ transformants retained the tilS gene and were likely spontaneous SpcR colonies. The ΔtilS PSpac–hprT supTr6 strain grew well at non-permissive temperature (Fig. 1B), indicating that the tilS gene was dispensable in the supTr6 background. However, its growth at the permissive temperature appeared to be somewhat impaired since these colonies were smaller than those of wild-type B. subtilis.

Figure 1.

Strain genotypes and growth phenotypes. Schematic representation of the tilS chromosomal locus of different strains (A), and their growth at 30°C and 51°C on LB plates (B).

The deletion of tilS was also attempted in the suppressed strains supTr1, supTr2, supTr5 and supTr7, by transforming them, and the sup+ parental strain, with ΔtilS PSpac–hprT supTr6 chromosomal DNA. Genuine deletion of tilS was obtained in the supTr1 strain only. In the other suppressed backgrounds, the few SpcR transformants had all acquired the supTr6 mutation (trnO–ile encoding tRNAIle1>3 with a TAT anticodon), indicating that tilS remained essential in these backgrounds. No mutations were found in the tRNAMet and tRNAIle genes of the supTr2, supTr5 and supTr7 strains, or in the tRNAVal genes, which are the closest homologues of the tRNAIle genes in B. subtilis (Staves et al., 1987; Widmann et al., 2010). This analysis suggests that tilS1 suppression in these strains takes place independently of tRNA mutations, and occurs possibly through stabilization of the TilS1 enzyme at high temperature.

In summary, mutations which changed the isoleucine anticodon GAT to TAT in one of the three redundant tRNAIle1 genes (tRNAIle1>3) allowed the thermosensitive B. subtilis tilS1 strain to grow at non-permissive temperature. That the suppressed tRNAIle1>3 mutants withstood deletion of tilS indicated that a complete lack of TilS function can be tolerated in this background. The most probable tRNA decoding specificities for the Ile/Met decoding box in the parental and suppressed tRNAIle1>3 backgrounds are summarized in Fig. 2.

Figure 2.

Schematic representation of the Ile/Met tRNA decoding specificity in tilS1 parental and suppressed strains. IleRS and MetRS stand for Isoleucyl–tRNA synthetase and Methionyl–tRNA synthetase respectively. TilS1 corresponds to the thermosensitive lysidine–tRNAIle2 synthetase and the acronym ‘Enz?’ stands for an unidentified tRNA modification enzyme. Temporal order of phenomena are the following: first, because the TilS1 enzyme does not function at 50°C (Step 1 in the grey box corresponding to decoding route #5), tRNAIle2, the product of the unique gene trnB–Met is aminoacylated by MetRS instead of IleRS, and the resulting Met–tRNAIle2 will incorporate Met at AUG codons in mRNA (dashed arrow towards eAUG). However, as discussed in text, depending on the degree of competition with other cellular tRNAIle, it may also infrequently incorporate Met at Ile-AUA codons (dashed arrow – ambiguous decoding of AUA). Second, due to a G-to-T mutation in the wobble position of one of the three genes coding tRNAIle1 (trnB–Ile2 or trnO–Ile; Step 2 in the grey box corresponding to decoding route #6), the wobble U34 of the corresponding gene transcript will be eventually modified (see text). The mature tRNAIle1>3 will then be aminoacylated by IleRS before it decodes AUA codons (plain arrow), and the AUU and AUG codons (dashed arrows, see text). However, depending on the degree of competition with the genuine AUG-decoder, Met–tRNAeMet, a small fraction of Ile may be incorporated in place of Met at AUG codons (dashed arrow towards eAUG). The decoding properties of tRNAIle1 (GAU) and tRNAeMet (CAU) or tRNAiMet (CAU) are indicated by plain arrows (decoding routes #1, #3 and #4 respectively).

Decoding of AUA and AUG codons is less accurate in the tilS1 suppressor strain than in the wild-type strain

The tilS1 supTr1 and tilS1 supTr6 strains lacked the Ile–tRNAIle2 (LAU) due to the inactivation of TilS1 at the non-permissive temperature and had a new tRNAIle1>3 (UAU) for which the in vivo aminoacylation status and decoding properties were unknown (Fig. 2, dashed arrows). To determine whether tRNAIle1>3 (UAU) was charged by Ile and which codon it translated in vivo, we investigated how AUA and AUG codons were deciphered in cells grown at 50°C by determining the nature of the incorporated amino acids. The abundant RpoB protein (β subunit of RNA polymerase) contains three isoleucine AUA codons corresponding to residues 52, 211 and 1152 of the protein and 30 methionine AUG codons. An RpoB reporter gene containing an affinity tag was expressed in wild-type and suppressor strains, and the full-length RpoB protein was purified using the affinity tag. RpoB samples were treated with trypsin and the resulting peptides were analysed by liquid chromatography-mass spectrometry (LC-MS). No substitutions between Ile and Met were detected in any of the many precisely identified RpoB peptides from the wild-type strain (see Fig. S1). This indicates that AUA and AUG codons are correctly translated into isoleucine and methionine by Ile–tRNAIle2 (LAU) and Met–tRNAeMet respectively. In RpoB peptides from the tilS1 supTr6 strain, isoleucine was incorporated at positions 211 and 1152 in most peptides but a minor fraction of the same peptides contained methionine at these positions. This indicates that the AUA codon is mistranslated at a low frequency by a tRNA carrying methionine, possibly Met–tRNAIle2 (CAU). Similar results were obtained for the tilS1 supTr1 strain (see Fig. S1). In both suppressor strains, methionine was faithfully incorporated at 27 out of 30 positions corresponding to AUG codons, whereas for three positions (471, 611 and 1043), a minor fraction of isoleucine misincorporation was detected (see Fig. S1). We performed the same analysis on the tilS1supTR2 strain, which carried an uncharacterized suppressor mutation not localized in any tRNAIle gene, and detected no isoleucine or methionine misincorporation (data not shown). This result indicates that misincorporation is not linked to the tilS1 background but to suppressor mutations in the tRNAIle1>3 genes.

The extent of the AUA mistranslation defect in suppressor strains was further assessed using single reaction monitoring (SRM) quantification. We focused on a 12-amino-acid RpoB peptide ILSGDEEEIEMR (amino acids 1144–1155) (actually ILSGDEEEIEMoxR/theoretical m/z = 1436; for details see Experimental procedures and Supporting information), which contained an isoleucine and a methionine residue (bold underlined). The Ile1152 and Met1154 amino acids are encoded by Ile-AUA and Met-AUG codons respectively. In wild-type B. subtilis, we detected a specific ion corresponding to this peptide but no specific ion that corresponded to the ILSGDEEEMoxEMoxR peptide (m/z = 1470) resulting from incorporation of a methionine in place of Ile1152. In the tilS1 supTr1 strain, although the ILSGDEEEIEMoxR peptide ion was the major component, a minor fraction of the ILSGDEEEMoxEMoxR peptide ion was also detected (Fig. 3). Methionine was detected in only 0.18% of these peptides from the supTr1 strain, indicating that the great majority of AUA codons are decoded as isoleucine by the mutant Ile–tRNAIle1>3 (UAU). A similar assay was performed using the RNA polymerase β′ subunit as a reporter in strain supTr1, and methionine misincorporation was observed in 0.06% of the indicator peptide KPETINYR. This value corresponds to a misincorporation frequency lower than 10−3 per AUA codon, which is clearly compatible with the viability of the tilS-deficient B. subtilis mutant.

Figure 3.

Single Reaction Monitoring (SRM) quantification. The specific signal of y6+ ion fragment for two RpoB-SPA peptides ILSGDEEEIEMoxR (top) and ILSGDEEEMoxEMoxR (bottom) is shown for the 15–22 min retention time. Proteins were purified from the 168 wild-type (WT) and the tilS1 supTr1 (supTr1) strains grown at 50°C and analysed as described in Experimental procedures. Mox means oxidized methionine. Correspondences between codon and amino acid are indicated on the figure.

tDNAIle3 (TAT) is naturally present in some prokaryotes (bacteria and archaea)

The genomes of several organisms were reported to contain no tilS gene and no tDNAIle2 (CAT), but do contain a tDNAIle3 (anticodon TAT) (Jaffe et al., 2004; Silva et al., 2006; de Crécy-Lagard et al., 2007; Suzuki and Miyauchi, 2010). These include three bacteria: Mycoplasma mobile, Bifidobacterium adolescentis and Neorickettsia sennetsu (two Gram-positive and one Gram-negative eubacteria respectively); and two archaea: Nanoarchaeum equitans and Korarchaeum cryptofilum. Remarkably, the bacteria Lactobacillus casei BL23 and Rhodopirellula baltica SH1 contain a tDNAIle3 (TAT) gene as well as the tDNAIle2 (CAT) and tilS gene pair (Table 1, see accessions in Table S1). To seek more organisms of this kind, we queried the tDNA database tRNA DB-CE, which presently includes 880 bacterial genomes (Abe et al., 2009) and found one additional prokaryotic genome with a tDNAIle3 (TAT) gene (Cyanothece sp. 7424). Two other genomes, Planctomyces limnophilus and Ktedonobacter racemifer, were also identified through blast using the tDNAIle3 sequence from Rhodopirellula baltica (Table 1). Next, we explored the genomes of bacteria closely related to Bifidobacterium adolescentis (15 genomes) and Neorickettsia sennetsu (9 genomes, available in ‘BLAST with Microbial Genomes’ (, and found that Bifidobacterium angulatum (a human gut symbiont similar to B. adolescentis) and Neorickettsia risticii (the causal agent of Potomac horse fever (Lin et al., 2009) also carry a tDNAIle3 (TAT) gene in place of the tilS and tDNAIle2 gene pair (Table 1).

Table 1.  Ile/Met tDNA and isoleucine codon usages in various bacteria, archaea and eukaryotes.
AnticodonstDNA usage (number of genes)Codon usageReference of tDNAIle3 (TAT) first report
Ile1Ile3Ile2iMeteMetIle codons (%)
Codons readATT       
  • Information for this table is described in the text.

  • a. 

    tDNAIle1 (GAT) trnB–Ile2 mutated into a tDNAIle3 (TAT) trnB–Ile2Tr1.

  • b. 

    tDNAIle1 (GAT) trnO–Ile mutated into a tDNAIle3 (TAT) trnO–IleTr6.

  • c. 

    in the absence of TilS, tRNAIle2 (CAT) reads AUG elongator Met codons.

  • d. 

    two identical tDNA copies differing from the third tDNA in one base-pair.

  • e. 

    two tDNA genes differing at seven positions.

  • f. 

    archaeal functional equivalent of bacterial TilS enzyme.

  • g. 

    value is indicated in bold if ATA is the major codon.

Reference bacterial genomesTilS         
 Bacillus subtilis ssp. subtilis str. 168+3013249.736.813.5 
 Escherichia coli str. K-12 substr. MG1655+3024248.842.68.6 
 Haemophilus influenzae Rd KW20+3012171.420.08.6 
Genomes with experimentally created tDNAIle3 (TAT)          
 Bacillus subtilis mutant supTr121a1c3249.736.813.5This work
 Bacillus subtilis mutant supTr621b1c3249.736.813.5This work
Bacterial genomes with tDNAIle3 (TAT)          
 Bifidobacterium adolescentis ATCC 157031101121.967.710.4Suzuki and Miyauchi (2010)
 Bifidobacterium angulatum DSM 200981101121.566.911.6This work
 Mycoplasma mobile 163 K1101171.47.621.0Silva et al. (2006)
 Neorickettsia risticii str. Illinois1101138.322.739.0This work
 Neorickettsia sennetsu Miyayama1101138.622.638.8Suzuki and Miyauchi (2010)
 Lactobacillus casei BL23+3d112157.037.65.4Suzuki and Miyauchi (2010)
 Rhodopirellula baltica SH 1+12e11230.965.63.5Suzuki and Miyauchi (2010)
 Planctomyces limnophilus DSM 3776+1111145.749.05.3This work
 Ktedonobacter racemifer DSM 4496+2112135.751.313.0This work
 Cyanothece sp. 7424+2111163.821.614.6This work
Reference archaeal genomesTiaSf         
 Sulfolobus acidocaldarius DSM 639+1011132.612.455.0 
 Thermococcus kodakaraensis KOD1+1011119.935.344.8 
 Methanococcus maripaludis S2+1012151.817.530.7 
Archaeal genomes with tDNAIle3 (TAT)          
 Nanoarchaea equitans1101111.824.963.3Randau et al. (2005)
 Candidatus Korarchaeum cryptofilum1101129.27.263.6Elkins et al. (2008)

The codons ATT and ATC are generally considered to be more frequently used than the minor ATA codon. We examined the genomes of bacteria and archaea that harbour a tDNAIle3 (TAT) gene to determine whether this pattern of codon usage holds true for these organisms. Surprisingly, the ATA codon is greatly preferred (> 60%) over the ATT and ATC codons in N. equitans and K. cryptofilum, which lack TiaS and tDNAIle2 (CAT), and also in a few other species (in bold in Table 1).

Possible origin of bacterial and archaeal tDNAIle3 (TAT)

To explore the possible origin of tDNAIle3 (TAT) in bacteria and archaea, in a context where most prokaryotes have instead evolved the AUA decoding system based on the functional pair tilS and tDNAIle2 (CAT), we used each of the identified tDNAIle3 (TAT) gene sequences to query (Blast) against the widest possible database (all GenBank + EMBL + DDBJ + PDB sequences at: The most closely related tDNA genes are listed in Fig. S3, together with their number of differences between tDNAIle1 (GAT) and the query tDNAIle3 (TAT) from the same organism (see Table S2 for accessions). First, the best hits were naturally occurring tRNA gene sequences. Two distinct situations were encountered. (i) The best hit was the tDNAIle1 (GAT) from the same species. This suggests that the tDNAIle3 (TAT) gene arose (a) from one of the redundant tDNAIle1 genes (including a G-to-T transversion in the first base of the anticodon, reminiscent of the B. subtilis mutations studied here) or (b) by direct mutation of an ancient tRNAIle2 (CAT) gene, followed by subsequent loss of the tilS/tiaS gene. The latter might be the case for the two archaea, N. equitans and K. cryptopfilum, where only 8 and 7 differences, respectively, exist between the tDNAIle3 and tDNAIle1 sequences of the same organism. (ii) The best hit for the tDNAIle3 (TAT) sequence corresponded to a tDNA (not necessarily a tDNAIle) sequence from another organism. This is likely the case for most bacteria listed in Fig. S3. In some cases, a tDNA of another species was closer in sequence to a naturally occurring tDNAIle3 (TAT) than to the sequence of any other tDNA of the same species. This observation suggests that a novel tDNA was acquired by horizontal gene transfer (HGT) and subsequently evolved into a functional tDNAIle3 (TAT) (Wang and Lavrov, 2010). Such foreign tDNA recruitment may have occurred via mobile elements, such as tDNA genes carried by prophages (Canchaya et al., 2003; Widmann et al., 2010).


B. subtilis can bypass the lack of the essential tilS gene

In E. coli, disruption of tilS, a quasi-ubiquitous bacterial gene, was shown to be lethal, whereas partial inactivation of tilS resulted in an AUA codon-dependent translational defect (Soma et al., 2003). Indeed, in all eubacteria and archaea, the pyrimidine nucleotide-ending codons AUC and AUU are read exclusively by G34-containing tRNAIle1 (Fig. 2, decoding route #1). So, in tilS mutants, AUA becomes an unassigned codon without a cognate decoding tRNA, which causes a lethal defect in AUA translation.

In the present work, we developed a strategy to identify suppressor mutations that would enable B. subtilis cells to survive the complete loss of TilS enzyme activity. At the non-permissive temperature, the cells of a B. subtilis strain carrying the thermosensitive allele tilS1 were depleted of tRNAIle2–lysidine synthetase activity. We found that cells able to survive without lysidine-containing tRNAIle2 had gained a new function, which enabled AUA codons to be decoded as isoleucine. Specifically, a single point mutation in the anticodon of one of the three copies of the tRNAIle1 (GAU) gene (G34 to U34) created a tRNAIle1>3 (UAU) mutant able to translate the AUA codon as Ile. The rescue of the translation defect caused by the lack of TilS in bacteria, and probably of TiaS in archaea, depends on the set of tRNAs (identity and relative abundance) able to decode the degenerate codons within the Ile/Met decoding family box, with the AUA/AUG codons being among the most difficult to discriminate accurately. Indeed, in the TilS-suppressed strains, translation of AUA codons as Ile and AUG as Met did not appear as accurate as in the wild-type strain. Low levels of miscoding for both codons were detected with an estimation of 10−3 Ile/Met misincorporation per translated AUA codon. Such levels of decoding ambiguity for the AUA and AUG codons in the absence of TilS activity are clearly well tolerated in B. subtilis cells.

Wild-type B. subtilis produce a minor tRNAIle2 (CAU) species in which the wobble C34 is quantitatively modified into L34 in an early step of maturation (Nakanishi et al., 2009). The fully mature L34-containing tRNAIle2 is recognized and charged exclusively by the IleRS isoleucyl–tRNA synthetase (Fig. 2, decoding route #2). The resulting Ile–tRNAIle2 (LAU) reads only the rare Ile AUA codon. The Met AUG codon is read by either of two distinct and abundant species, tRNAeMet or tRNAiMet, depending on whether the codon is located within the coding region (tRNAeMet) or at the translation initiation site (tRNAiMet) (Fig. 2, decoding routes #3 and #4; (reviewed in Mayer et al., 2001).

In the absence of TilS (Step 1 + 2 in Fig. 2), the unmodified C34-containing tRNAIle2 is likely to be fully charged with Met and able to read Met-AUG codons (Fig. 2, decoding route #5, dashed arrow towards cognate eAUG). However, in B. subtilis cells, this decoding route might not be efficient because of the competition with the genuine and abundant AUG-decoder Met–tRNAeMet (Fig. 2, decoding route #3). More likely, the Met–tRNAIle2 (CAU) might effectively outcompete the vanishing Ile–tRNAIle2 (LAU), and introduce Met at the Ile AUA codons of newly transcribed mRNA (Fig. 2, decoding route #5, dashed arrow towards AUA-codon ambiguity), thereby producing aberrant proteins. In other words, the degree of Met misincorporation at AUA codons should depend upon the ratio of competitor Met–tRNAIle2 (CAU) and the vanishing Ile–tRNAIle2 (LAU). This hypothesis has been proposed previously to explain the misincorporation of a given amino acid at codons recognized by several tRNA species (Kramer and Farabaugh, 2007). If this hypothesis is correct, the deletion of the tRNAIle2 (CAU) gene in a suppressed tilS1 strain should abolish the misincorporation of Met at AUA codons (Fig. 2, decoding route #5). Unfortunately, our attempts to delete this gene were unsuccessful, most likely because the flanking ribosomal operons in B. subtilis are highly redundant.

At 50°C, the major cause of B. subtilis tilS1 lethality might be a high frequency of Met misincorporation at AUA codons during translation. For the B. subtilis tilS1 suppressor strain at 50°C, the tRNAIle1>3 (UAU) mutant derived from a genuine Ile-specific tRNA, which is charged with Ile by the IleRS Ile–tRNA synthetase (Nureki et al., 1994), likely conferred survival by ultimately outcompeting the poisonous Met–tRNAIle2 (CAU) at Ile AUA codons (Fig. 2, decoding route #6, heavy arrow towards cognate AUA codon). The observed low frequency of Met incorporation at AUA codons in the tilS1 supTr1/6 suppressor strains probably reflected the more efficient decoding of AUA codons by the suppressor Ile–tRNAIle1>3 (UAU) than by the competing Met–tRNAIle2 (CAU).

Even though efficient AUA codon decoding by the tRNAIle1>3 (UAU) mutant allowed B. subtilis to survive in absence of TilS function, we still do not know whether its wobble U34 nucleotide is modified. Some bacteria (e.g. Mycoplasma species) and organelles (e.g. the mitochondria of budding yeasts) read all four codons of a codon family with a single tRNA harbouring an unmodified wobble U34 in absence of any other competitor isoacceptor tRNAs (reviewed in Grosjean et al., 2010). However, all four degenerate codons of a family are not read with the same efficiency, as this highly depends on the tRNA and the set of codons to be read (reviewed in Björk and Hagervall, 2005). In B. subtilis, a post-transcriptional modification of the wobble U34 in tRNAIle1>3 (Fig. 2, decoding route #6, annotated U?) could possibly improve the reading selectivity of the different degenerate codon(s) of the Ile/Met codon family. Efforts to identify the putative modification(s) of the wobble U34 of tRNAIle1>3 in the B. subtilis suppressor strains supTr1 and supTr6 are in progress.

Evolutionary implications of tilS suppression

The rare prokaryotic organisms that naturally lack the TilS/TiaS activity and the corresponding C34-containing tRNAIle2 substrate possess a U34-containing tRNAIle3 (Table 1). As noted previously, this strategy may have evolved from mutations similar to those of our B. subtilis mutants, or by the subsequent mutagenesis of tDNA genes acquired from other species. The five bacteria (Lactobacillus casei BL23, Rhodopirellula baltica SH1, Planctomyces limnophilus DSM 3776, Ktedonobacter racemifer DSM 4496 and Cyanothece sp. 7424), which contain a U34-harbouring tDNAIle3 and a tilS and C34-containing-tDNAIle2 gene pair in their genomes (Table 1), are particularly interesting because they may represent evolutionary intermediates between the predominant group of TilS-containing organisms and the rare group of organisms lacking the functional pair of the tilS/tiaS and tDNAIle2(CAT) genes. A similar ‘gain-loss’ evolutionary scenario, which was based mainly on theoretical considerations and included a temporary ambiguous decoding system, was initially proposed by Schultz and Yarus (the ‘ambiguous intermediate’ theory; Schultz and Yarus, 1994; 1996) and further developed by others to explain the emergence of TilS in bacteria (Silva et al., 2006; Sengupta et al., 2007).

The TilS and TiaS enzymes, which act exclusively on C34-containing tRNAIle2, belong to completely different clusters of orthologous genes (COG0037 and COG1571 respectively). These different types of enzymes, together with their corresponding substrates, emerged independently after the evolutionary split between Bacteria and Archaea. They evolved separately to enable efficient and accurate decoding of the single AUA-codon as Ile and not Met (i.e. convergent evolution; Grosjean et al., 2010). However, the selective advantage(s) of splitting the degenerate codons of the Ile/Met codon family into a 3:1 partition mode and not the 2:2 (duet) mode of other mixed amino codon families remains to be discovered.

Experimental procedures

Strains and media

Bacillus subtilis strains were constructed from trpC2 strain 168 (Table S3). They were propagated in LB with phleomycin 1 µg ml−1, chloramphenicol 6 µg ml−1 or erythromycin 0.3 µg ml−1. Mutant strains containing the pMutin vector were constructed during the functional analysis project (Ogasawara, 2000).

Strain construction

Two DNA fragments were PCR-amplified from the B. subtilis 168 chromosome. The first fragment (2.2 kbp) encompassed the essential tilS and hprT genes and was amplified under mutagenic conditions (40 µM dATP and 200 µM dCTP, dGTP, dTTP with 1.5 mM MgCl2) using the TilS1 (5′-CATCACCGCAGCCTGCCACACG-3′) and TilS2 (5′-TTCTCACGCACCTCCTTCCCTTTCTCAGCTTTCATAAACTGCCGGTTT-3′) oligonucleotides (from −283 bp to 1957 bp relative to the tilS ATG). The second fragment (1.7 kbp), downstream of the first and corresponding to the ftsH gene, was amplified in accurate conditions (200 µM dNTP and 1.5 mM MgCl2 with the High Fidelity polymerase, Boehringer) using the TilS3 (5′-TTAGCTTTTTTCAACAAATAAAAAGCTAATCGGCAGCCTGCTTCCGAGATGG-3′) and TilS4 (5′-TGTTCGTTGTTGAAATCACGGCC-3′) oligonucleotides (from 1958 bp to 3682 bp relative to the tilS ATG). The underlined TilS2 and TilS3 sequences were complementary to the 5′ and 3′ ends of a phleomycin resistance open reading frame respectively. Both fragments were PCR-joined to the cassette amplified from the pUC19-upp-phleoR plasmid DNA, as described in Fabret et al. (2002). B. subtilis 168 competent cells were transformed with the resulting 4.3 kbp fragment for phleomycin resistance at 30°C. A total of 302 colonies were screened for thermosensitivity at 51°C, and 14 Ts candidates were isolated. The genetic linkage (80 to 92%) between the Ts phenotype and phleomycin resistance was determined by transforming the wild-type 168 strain with chromosomal DNA from the 14 Ts candidates. The PCR-amplified region was sequenced and most strains contained mutations in the tilS and hprT genes. Two strains (tilS1 and tilS12) harboured mutations in the tilS sequence only. For plating efficiency, strains were cultivated in LB medium at 30°C and dilutions were plated on LB plates incubated at 30°C, 37°C, 42°C, 47°C and 50°C. The proportion of surviving cells was estimated at the different temperatures relative to 30°C. Experiments were reproduced three times independently.

Selection of Ts mutant suppressors was done by spreading 10 independent overnight cultures in phleomycin-containing LB medium at 30°C onto LB plates, which were incubated at 50°C for 4 days. The suppression frequency was ∼ 2.6 × 10−8 and yielded seven independent suppressors (supTr1 to supTr7). The genetic linkage between the suppressor phenotypes and phleomycin resistance was measured by transforming the wild-type 168 strain with chromosomal DNA from each of the seven suppressor strains. Transformants were first selected on phleomycin-containing plates at 30°C, and were then streaked onto LB plates at 47°C, 50°C and 30°C to score the growing colonies and the thermosensitive ones.

To place the hprT gene under the control of the IPTG-inducible PSpac promoter, a fragment from position −251 to +239 relative to the ATG of hprT was PCR-amplified from the 168 chromosomal DNA using the BD75 (CCGGAATTCACGAGAAAAGCTGGAGACCG) and BD76 (CGCGGATCCTACTTCTCCAGAAGAAACCG) oligonucleotides. The PCR product was then ligated into the pMutin2 plasmid, using EcoRI and BamHI restriction sites (underlined oligonucleotide sequences). Plasmid DNA was prepared from an ampicillin-resistant (100 µg ml−1) E. coli TG1 transformant and sequenced. The B. subtilis tilS1 strain was then transformed for erythromycin resistance in presence of 1 mM IPTG to induce expression from the PSpac promoter. Integration of the plasmid by single crossover was checked by Southern blotting using the pMutin2 plasmid as probe as previously described (Vagner et al., 1998). The construction was then transferred into the tilS1 supTr6::Tn10 strain by transformation to produce the tilS1 PSpac–hprT supTr6::Tn10 strain.

To delete the tilS gene, two 1.5 kbp DNA fragments were PCR-amplified from the chromosomal DNA of the strain: one 20 bp upstream of the tilS gene using the BD79 (AAGAACTGAAACAGATTCTCGG) and BD77bis (GTCGACCTGCAGGCATGCAAGCTATGTCCTCCTCACAATGAGC) oligonucleotides and the other downstream of tilS, in the lacZ gene of the pMutin, using the BD80 (TACATCGGGCAAATAATATCGG) and BD78 (CCCCGGGTACCGAGCTCGAATTCCCAGCTTGTTGATACACTAA) oligonucleotides. Underlined sequences of BD77bis and BD78 are complementary to the spc gene from the pIC156 vector and allowed all three DNA fragments to be joined by PCR as previously described (Fabret et al., 2002). The resulting 4.4 kpb fragment was transformed into strains carrying the PSpac::hprT fusion. Spectinomycin-resistant clones were then analysed for antibiotic resistance markers (EryR, CmR, SpcR, PhleoR), and the chromosomal structure of the final ΔtilS PSpac–hprT supTr6::Tn10 was checked by Southern blotting using the pMutin2 plasmid as probe. The tilS deletion was transferred into the tilS1 supTr1::Tn10 strain by transformation with ΔtilS PSpac–hprT supTr6::Tn10 chromosomal DNA. Transformants were first selected on erythromycin or spectinomycin plates supplemented with IPTG, then they were purified on selective plates containing both antibiotics. Twelve [EryR SpcR] transformants were PCR-analysed for the tilS deletion, and the trnO–ile locus of at least four deleted transformants was sequenced to verify that the supTr6 mutation had not been introduced at the same time.

Mapping of the suppressors

The suppressor strains supTR1 to supTr6 were transformed with pHV1248 for erythromycin resistance at 30°C. Tn10 transposition was performed in chloramphenicol (Cm)-containing LB medium (Cm is used to select the transposition event) for 4 h at 30°C (to allow the transposition) followed by 4 h at 51°C (to abolish plasmid replication and lose it). To estimate the fraction of cells in which transposition occurred, colonies were counted on LB with or without Cm at 51°C. These fractions were estimated to 2 to 4 × 10−3 in supTr1, supTr2, supTr6 and supTr7. No transposition event was obtained in the supTr5 strain. Libraries of Tn10 transposons were prepared for each suppressor by extracting the chromosomal DNA from all pooled cells, and using it to transform the tilS1 strain for Cm resistance at 30°C. Fifty colonies were streaked onto Cm-containing plates at 30°C and 51°C to isolate thermoresistant clones. The fraction of colonies that grew at 51°C was 32% for supTr1, 2% for supTr2, 12% for supTr6 and 4% for supTr7. The genetic linkage between the Cm resistance and the thermosensitive suppressor phenotype was then measured by transforming the tilS1 strain with chromosomal DNA extracted from each of the thermoresistant clones. Of the 16 thermoresistant clones obtained from the supTr1 Tn10 library, only one clone (supTr1.12 hereafter named supTr1::Tn10) displayed a 94% genetic linkage between Cm resistance and temperature resistance. Other clones demonstrated linkage values of 2–30%. Six thermoresistant clones were obtained with the supTr6 Tn10 library, and one clone (supTr6.4 hereafter named supTr6::Tn10) displayed a 96% linkage value, whereas it was between 2% and 10% for the others. No significant linkage was found with the supTr2 and supTr7 Tn10 libraries (< 8% linkage for the isolated thermoresistant clones). The chromosomal DNA of tilS1 supTr1::Tn10 and tilS1 supTr6::Tn10 clones was digested by TaqI for 2 h at 65°C, ligated into the compatible AccI-digested pBluescript KS+ vector, and then used to transform the E. coli DH10B strain and select for ampicillin- and Cm-resistant (20 µg ml−1) clones. The plasmid inserts were then sequenced using the universal and reverse primers. The DNA sequences were aligned with the genome sequence to locate the Tn10 insertion.

As there are 10 rDNA loci in the B. subtilis genome, the localization of the rDNA loci linked with Tn10 was performed by genetic means. The chromosomal DNA samples of B. subtilis strains containing a pMutin vector (erythromycin resistance) integrated in a gene proximal to a rDNA locus were used to transform the tilS1 supTr1::Tn10 and tilS1 supTr6::Tn10 strains in which Cm and temperature resistance was tightly linked. pMutin insertion mutants targeted yaaC for the rrnO operon (position 12 kb), yaaN for the rrnA operon (32 kb), yacD for the rrnJ–rrnW operons (92–98 kb), ybaR for the rrnI–rrnH–rrnG operons (163–168–173 kb), ydhU for the rrnE operon (637 kb), sspE for the rrnD operon (948 kb) and yuaJ for the rrnB operon (3176 kb). Transformants were selected on erythromycin-containing plates, streaked on LB plates and LB plates supplemented with Cm, and then incubated at 37°C to estimate the frequency of loss of Cm resistance. For the tilS1 supTr6::Tn10 strain, the integration of yaaC–pMutin (position 16 kb) DNA resulted in a 78% loss of Tn10, and integration of yaaN–pMutin (36 kb) DNA resulted in a 5% loss of Tn10. No effect above background was detected for the other integrations, suggesting that the Tn10 was in the rrnO-23S gene. The Tn10 chromosomal position was confirmed using pMutin insertions upstream of the rrnO operon, with gidB– (4208 kb), yaaA– (3 kb), and yaaB– (5 kb) pMutin DNAs producing Cm-resistant loss frequencies of 32%, 66% and 72% respectively. For the tilS1 supTr1::Tn10 strain, the integration of a yuaJ–pMutin (position 3178 kb) DNA established the Tn10 insertion position to be in the rrnB-23S gene. For both Tn10 insertions, exact positions were confirmed by PCR amplification of the insertion sites using a primer in the Cm-resistant gene and one in the chromosomal DNA surrounding the rrn-23S candidate genes, and DNA sequencing.

Protein purification and analysis

The RpoB-SPA tagged proteins (the Sequential Peptide Affinity tag sequence that contains the 3XFLAG epitope and the calmodulin-binding peptide separated by a TEV protease cleavage site; Supporting information) were purified from 50 ml LB cultures at 50°C (OD600 = 0.6) by a two-step affinity purification procedure (Zeghouf et al., 2004; Lecointe et al., 2007) from the following isogenic strains: wild-type 168, supTr1 and supTr6. After electrophoresis on 10% SDS-polyacrylamide gels, protein bands corresponding to the RpoB-SPA or RpoC proteins were excised and subjected to in-gel trypsin digestion with the Progest system (Genomic Solutions, see Supporting information). After drying, peptide extracts were resuspended in 25 µl 0.08% TFA, 2% ACN prior to LC-MS/MS analysis on an Ultimate 3000 LC system (Dionex, Voisins le Bretonneux, France) connected to LTQ (Linear Trap Quadrupole) Orbitrap mass spectrometer (Thermo Fisher, USA). Peptide ions were automatically analysed using a Bacillus subtilis database (see Supporting information) and the spectra corresponding to the relevant point of substitution were manually verified (see Fig. S1). This allowed us to detect the replacement of Ile by any other amino acid as well as the Met/Ile misincorporation. Finally, SRM experiments were conducted by specifically monitoring the signal of one ion fragment in the LTQ mass analyser for peptides with and without point substitutions.

In silico analysis

Searches for bacterial genomes harbouring the anomalous tDNAIle TAT were first performed by querying the tRNADB-CE database (Abe et al., 2009). This tRNA gene database presently holds 880 bacterial genomes. Codon usage values in individual genomes were uniformly estimated from an automatic attribution of coding sequences greater than 100 codons carried over whole genomes.


The authors would like to thank Jean-Pierre Rousset for interest in the work and for providing laboratory facilities to C.F. and H.G. We are also grateful to Marie-Françoise Noirot-Gros, Olivier Delumeau, François Lecointe and Veronique Monnet for stimulating discussions. This work was supported by the EU-funded BaSysBio project LSHG-CT-2006–037469 (to P.N.), and by INRA (to A.G and P.N.). H.G. holds a CNRS position of Emeritus scientist at the University Paris-Sud in Orsay-France.


  • 1

    For clarity/internal consistency, we defined tRNAIle1 as any naturally occurring tRNAIle harbouring a GAU or IAU anticodon (where I represents inosine), tRNAIle2 as tRNAIle having a CAU or LAU anticodon (where L represents lysidine), and tRNAIle3 as tRNAIle having a UAU or U?AU anticodon (where U? represents a yet unidentified putative wobble modified U). tRNAIle1 in which the (GAU) anticodon is mutated into (UAU) are referred to as ‘tRNAIle1>3’. These designations were irrespective of other previously used conventions in the scientific literature. The acronym tRNAAA(XYZ) corresponds to a cellular tRNA coding for amino acid AA and harbouring an XYZ anticodon (with wobble base at position 34 in bold and underlined), whereas ‘tDNAAA (XYZ)’ and ‘tRNAAA (XYZ) gene’, in which X, Y and Z represent deoxynucleotides, are used to define the chromosomal gene coding for the corresponding cellular tRNAAA (XYZ). mRNA codons are indicated as Z′Y′X′ with the third base X′ being the one that pairs (or wobbles) with the first base X of the anticodon in the tRNA.