key: cord-0001390-amtrihpb
authors: Moukadiri, Ismaïl; Garzón, M.-José; Björk, Glenn R.; Armengod, M.-Eugenia
title: The output of the tRNA modification pathways controlled by the Escherichia coli MnmEG and MnmC enzymes depends on the growth conditions and the tRNA species
date: 2013-11-26
journal: Nucleic Acids Res
DOI: 10.1093/nar/gkt1228
sha: 12590435a221d10824ab106e66d0f9fec397db40
doc_id: 1390
cord_uid: amtrihpb

In Escherichia coli, the MnmEG complex modifies transfer RNAs (tRNAs) decoding NNA/NNG codons. MnmEG catalyzes two different modification reactions, which add an aminomethyl (nm) or carboxymethylaminomethyl (cmnm) group to position 5 of the anticodon wobble uridine using ammonium or glycine, respectively. In [Image: see text] and [Image: see text], however, cmnm(5) appears as the final modification, whereas in the remaining tRNAs, the MnmEG products are converted into 5-methylaminomethyl (mnm(5)) through the two-domain, bi-functional enzyme MnmC. MnmC(o) transforms cmnm(5) into nm(5), whereas MnmC(m) converts nm(5) into mnm(5), thus producing an atypical network of modification pathways. We investigate the activities and tRNA specificity of MnmEG and the MnmC domains, the ability of tRNAs to follow the ammonium or glycine pathway and the effect of mnmC mutations on growth. We demonstrate that the two MnmC domains function independently of each other and that [Image: see text] and [Image: see text] are substrates for MnmC(m), but not MnmC(o). Synthesis of mnm(5)s(2)U by MnmEG-MnmC in vivo avoids build-up of intermediates in [Image: see text]. We also show that MnmEG can modify all the tRNAs via the ammonium pathway. Strikingly, the net output of the MnmEG pathways in vivo depends on growth conditions and tRNA species. Loss of any MnmC activity has a biological cost under specific conditions.

Transfer RNAs (tRNAs) are highly and diversely modified, and each has a unique pattern of modification (1) (2) (3) . These modifications are post-transcriptionally introduced at precise positions by specific enzymes and play important roles in folding, stability, identity and in the functions of tRNAs.

Modifications at the wobble position of the anticodon contribute to the accuracy and efficiency of protein synthesis (1, 3, 4) . Changes in the modification levels of the wobble position affect the synthesis of specific proteins and also lead to complex phenotypes through unknown mechanisms (3, (5) (6) (7) (8) (9) (10) .

Wobble modifications are often complex and require for their synthesis the participation of several enzymes (3) . Even though all of the enzymes involved in a specific modification pathway are known, it is unclear how their actions are modulated and coordinated to produce the final modification. Progress in this area may lead to a better understanding of tRNA biology.

Proteins MnmE and MnmG (formerly TrmE and GidA, respectively) are evolutionary conserved from bacteria to eukaryotic organelles. In Escherichia coli, they are involved in the modification of the wobble uridine of tRNA Lys mnm5s2UUU , tRNA Glu mnm5s2UUC , tRNA Gln cmnm5s2UUG , tRNA Leu cmnm5UmAA , tRNA Arg mnm5UCU and tRNA Gly mnm5UCC , all of which read NNA/G codons of the split codon boxes, with the exception of tRNA Gly mnm5UCC , which reads codons of the glycine family box (11) . MnmE and MnmG are dimeric and form a functional a2b2 heterotetrameric complex (MnmEG) in which both proteins are interdependent (12, 13) . MnmE is a GTP-and tetrahydrofolate (THF)-binding protein, whereas MnmG is a FAD-and NADH-binding protein (14) (15) (16) (17) (18) . The MnmEG complex catalyzes the addition of the aminomethyl (nm) and carboxymethylaminomethyl (cmnm) groups to position 5 of the wobble uridine using ammonium and glycine, respectively [ Figure 1 ; (13) ]. Both reactions require GTP and FAD as well as NADH if FAD is limiting in the in vitro reaction. Moreover, a THF derivative, likely methylene-THF, serves as the donor of the methylene carbon that is directly bonded to the C5 atom of U34. However, the detailed mechanism of the basic reaction has not yet been demonstrated (11, 13) . According to current models, GTP hydrolysis by the MnmE G-domain, which is located distant from the MnmEG active center, leads to structural rearrangements in the MnmEG complex, which are essential for tRNA modification (11, 19) .

The wobble uridine (U34) of tRNAs modified by MnmEG may be further modified by other enzymes at position 5 or other positions. In tRNA Lys mnm5s2UUU , tRNA Glu mnm5s2UUC and tRNA Gln cmnm5s2UUG , thiolation at position 2 of the U34 is performed by MnmA (20) , whereas in tRNA Leu cmnmUmAA , TrmL methylates the 2 0 -OH group of the U-ribose (21) . Therefore, some tRNA substrates for the MnmEG complex are also substrates for MnmA or TrmL.

Moreover, in tRNA Lys mnm5s2UUU , tRNA Glu mnm5s2UUC , tRNA Arg mnm5UCU , and tRNA Gly mnm5UCC , the resulting products of MnmEG activity are not the final modifications because these tRNAs carry the methylaminomethyl (mnm) group at position 5 of U34. Formation of the mnm 5 -final group is mediated by the action of the twodomain, bi-functional enzyme MnmC (22) (23) (24) (25) (26) . The oxidoreductase activity of the C-terminal domain, here designated as MnmC(o), is responsible for the FADdependent deacetylation that transforms cmnm 5 U into nm 5 U, whereas the methyltransferase activity of the N-terminal domain, designated as MnmC(m), catalyzes the SAM-dependent methylation that transforms nm 5 U to mnm 5 U (Figure 1 ).

The crystal structure of MnmC consists of two globular domains interacting with each other (27) . The structure also reveals that the two catalytic centers of MnmC(o) and MnmC(m) face opposite sides of the protein, thus favoring a model in which the two domains could function in a relatively independent manner. In fact, when two MnmC mutant proteins possessing a catalytically dead MnmC(o) or MnmC(m) domain were mixed, partial recovery of mnm 5 -group synthesis was observed (25) . However, considering the relatively hydrophilic nature of the domain interface, the possibility that conformational changes within the entire MnmC protein may occur in vivo and affect its functioning, cannot be excluded. A previous attempt to separately express both MnmC domains revealed that MnmC(m) was soluble, whereas MnmC(o) produced inclusion bodies, suggesting that MnmC(m) is required for the correct folding or structural stability of MnmC(o) (25) .

How the activities of the MnmEG and MnmC enzymes are organized and modulated remains unclear. Whether tRNAs preferentially utilize one of the two MnmEG pathways, the glycine or the ammonium pathway (Figure 1 ), in vivo depending on metabolic circumstances has not been investigated. The fact that only cmnm 5 , but not nm 5 or mnm 5 has been detected to date at position 5 of U34 in tRNA Gln cmnm5s2UUG and tRNA Leu cmnm5UmAA suggests that both tRNAs do not use the ammonium pathway and are not substrates for MnmC(o). Accumulation of the nm 5 s 2 U nucleoside which may have a dual origin [ Figure 1 ; (13) ], has not been observed in total tRNA purified from wild-type E. coli cells (23) . Therefore, the mnm 5 s 2 U synthesis in tRNA Lys mnm5s2UUU and tRNA Glu mnm5s2UUC appears to be organized to prevent nm 5 s 2 U accumulation. However, the presence or absence of cmnm 5 s 2 U as an intermediate in mnm 5 s 2 U biosynthesis is difficult to determine from nucleoside analysis of bulk tRNA due to the natural occurrence of cmnm 5 s 2 U in tRNA Gln cmnm5s2UUG . Biosynthetic tuning of mnm 5 s 2 U to avoid the accumulation of intermediates could be achieved by either kinetic tuning of the activities of MnmEG and MnmC, which requires that the sequential modifications are performed at similar or increasing rates or selective degradation of partially modified tRNAs. A previous steady-state kinetic analysis of the activities of the full MnmC protein indicated that the MnmC(m)dependent reaction (nm 5 !mnm 5 ) occurs faster than the MnmC(o)-dependent reaction (cmnm 5 !nm 5 ) or at a similar rate at very high substrate (tRNA) concentrations (26) . However, this study did not take into account the efficiency of the reactions performed by MnmEG (Figure 1 ), which is crucial to understand the organization of the mnm 5 s 2 U biosynthetic process.

Modifications located within or adjacent to the anticodon are important for stabilization of codon-anticodon pairing as well as for restricting the dynamics of the anticodon domain and shaping its architecture (1, 4) . Considering the network of pathways leading to mnm 5 s 2 U production ( Figure 1 ), it is important to determine whether mutations affecting the different activities of MnmC [MnmC(o) or MnmC(m)] have distinct biological consequences and to what extent an accumulation of modification intermediates affects bacterial biology.

In this study, we investigated the activities and the tRNA specificity of the MnmEG and MnmC enzymes in vivo and in vitro, analyzed how the U34 modification status is influenced by genetic and physiological conditions in bulk and specific tRNAs and assessed the effects of mnmC mutations on bacterial growth. Our study was facilitated by the cloning and separate expression of the two MnmC domains, MnmC(o) and MnmC(m). In contrast to a previous report (25) to those of the full protein. We also show that tRNA Gln cmnm5s2UUG and tRNA Leu cmnm5UmAA are substrates for MnmC(m), but not for MnmC(o). Our data suggest that MnmEG and MnmC are kinetically tuned to produce only the fully modified nucleoside mnm 5 U in tRNA Lys mnm5s2UUU . We demonstrate that all the tRNA substrates of MnmEG are modified in vitro through the ammonium pathway. However, the net output of the ammonium and glycine pathways of MnmEG in vivo depends on growth conditions and tRNA species. Finally, we demonstrate that the loss of any MnmC activity has a biological cost under specific conditions.

Escherichia coli strains and plasmids are shown in Table 1 . A list of the oligonucleotides and primers used in this work is provided in Supplementary Table S1 . Transduction with phage P1 was performed as previously described (28) . Deletion of the mnmC(o) coding region was performed by targeted homologous recombination (29) 

To determine bacterial growth rates, overnight cultures were diluted 1/100 in fresh LBT medium and incubated at 37 C with shaking. The doubling time was measured by monitoring the optical density of the culture at 600 nm. Samples were taken from exponentially growing cultures after at least 10 generations of steady-state growth. The growth rate was calculated as the doubling time of each culture in the steady-state log phase by linear regression. Competition experiments were performed as previously reported (21) . Briefly, the reference strain (BW25113) and the strain to be tested (TH48, TH49 or TH69) were grown separately to stationary phase by incubation at 37 C. Equal volumes of both strains were mixed and a sample was immediately taken to count viable cells on LAT plates with and without the antibiotic required to estimate the content of each strain (tetracycline to identify strains of the TH48 background). Six cycles of 24-h growth at 37 C (with shaking) were performed by diluting mixed cultures of 1/1000 in LBT medium. After the sixth cycle, the mixed culture was analyzed for its strain content as before. The final ratio was calculated as the ratio of the number of colony forming units (CFU) per milliliter recovered on LAT supplemented with tetracycline versus CFU per milliliter on LAT.

Routinely, a single colony was inoculated into 5 ml LBT supplemented with adequate antibiotics and grown overnight at 37 C. The pre-culture was then inoculated (1:100 dilution) into 1000 ml LBT plus antibiotic and grown until the OD 600 reached $0.5. Protein expression was induced by the addition of 0.2% arabinose (Flag-tagged proteins). The cultures were grown for an additional 3-4 h at 30 C with gentle shaking, harvested by centrifugation (4500g for 15 min at 4 C), washed with TBS buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl and 5 mM MgCl 2 ) and stored at À20 C. For enzyme purification, frozen cells were thawed on ice, resuspended in 10 ml of lysis buffer (TBS containing 1 mM PMSF and 1 mM EDTA) and disrupted by sonication. The lysate was centrifuged at 10 000g for 45 min at 4 C and the supernatant was mixed with 0.3-0.6 ml of pre-equilibrated anti-Flag agarose resin (ANTI-FLAG M2 Affinity Gel, A2220-Sigma) and incubated at 4 C for 1 h with mild shaking. Instructions from the manufacturer were followed for washing and eluting Flag-tagged proteins. The fractions containing the proteins were pooled, concentrated with Amicon Ultra-15 30k devices in TBS buffer and stored in 50-ml fractions with 15% glycerol at À20 C. The His-MnmC(m) domain was purified using Clontech's TALON Metal Affinity Resin according to the manufacturer's instructions. The MnmE and MnmG proteins were purified as previously described (14, 17) . Protein concentrations were determined with a NanoDrop spectrophotometer at 280 nm. The purity of all enzymes was >95% as estimated by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and coomassie blue staining.

FAD was released from recombinant proteins by heating at 75 C for 5 min in the dark and analyzed by highperformance liquid chromatography (HPLC). Briefly, HPLC separation was achieved with a Synergi 4u Fusion column (25 cm  4.6 mm, 5 mm) using a gradient of 5 mM ammonium acetate, pH 6.5, to acetonitrile 50% and water 50% over 35 min at a flow rate of 0.6 ml/min. Flavins were detected by fluorescence emission (525 nm) using a Waters 474 scanning fluorescence detector set at 450 nm for excitation.

Strain IC6010 carrying pIC1253, pIC1339 or pIC1340 was grown overnight in LBT with ampicillin at 37 C. Overnight cultures were diluted 1/100 in the same medium supplemented with 0.1% L-arabinose (inducer of the AraC-P BAD system) and incubated at 37 C for 2 h. To halt the expression of the MnmC recombinant proteins, the cells were recovered by centrifugation at 3000g for 10 min, washed once with fresh LBT medium, resuspended in the same volume of pre-warmed LBT containing ampicillin and 1% glucose (repressor of the AraC-P BAD system) and incubated at 37 C. Samples were taken several times after the addition of glucose. The cells were lysed by brief ultrasonication and the lysate was centrifuged at 10 000g for 10 min at 4 C. The soluble fractions (50-150 mg of total proteins) were analyzed by western blotting using anti-Flag and anti-GroEL antibodies. Isolation of bulk tRNA from E. coli and reverse-phase HPLC analysis of nucleosides Total tRNA purification and analysis of nucleosides by reverse-phase HPLC were performed as described previously (13, 35, 36) . HPLC analysis was monitored at appropriate wavelengths to achieve optimal adsorption of the target nucleosides, 314 nm for thiolated nucleosides and 254 nm for non-thiolated nucleosides. The nucleosides were identified according to their UV spectra (35) and by comparison with appropriate controls.

The E. coli tRNA Lys mnm5s2UUU , tRNA Gln cmnm5s2UUG and tRNA Leu cmnm5UmAA genes were cloned into pBSKrna (37) digested with either EcoRV (to produce chimeric tRNAs in which the E. coli tRNA is inserted into a human cytosolic tRNA Lys 3 sequence that is used as a scaffold) or EcoRI and PstI to produce a tRNA without the scaffold (overexpressed 'native' tRNAs). The overproduction of tRNAs in strains transformed with pBSKrna derivative plasmids was performed as described previously (37) . Specific chimeric tRNAs were purified from bulk tRNA by the Chaplet Column Chromatography method (38) using a biotinylated DNA probe that is complementary to the scaffold human cytosolic tRNA Lys 3 moiety (common to all chimeric tRNAs). The probe was immobilized on a HiTrap Streptavidin HP column. The same approach was used to purify overexpressed 'native' and true native tRNAs using biotinylated DNA probes that are complementary to the specific sequence of each E. coli tRNA (Supplementary Table S1 ).

The E. coli genes encoding tRNA Cys GCA , tRNA Arg mnm5UCU , tRNA Glu mnm5s2UUC , tRNA Gly mnm5UCC and tRNA Leu cmnm5UmAA were PCR-amplified from genomic DNA using 'vent' polymerase and the primers Cys-F/Cys-R, Arg-F/Arg-R, Glu-F/Glu-R, Gly-F/Gly-R and Leu-F/Leu-R, respectively. The amplicons were cloned into a SmaI-linearized pUC19 plasmid to produce pIC1394, pIC1395, pIC1536, pIC1550 and pIC1537. The E. coli gene encoding tRNA Gln cmnm5s2UUG was PCR-amplified from genomic DNA using expand-long polymerase and the primers Gln-F and Gln-R and cloned into pBAD-TOPO. The plasmid pIC1083 (containing the tRNA Lys mnm5s2UUU gene), a derivative of pUC19, was a gift from Dr Tamura (34) . Unmodified E. coli tRNAs were prepared by in vitro transcription from BstNI-digested plasmids (pIC1083, pIC1394 and pIC1395) and HindIII-digested plasmids (pIC1535, pIC1536, pIC1550 and pIC1537) using the Riboprobe T7 transcription kit (PROMEGA) and 2-5 mg of each digested plasmid as a DNA template in a 50-ml reaction mix.

To analyze the in vitro activity of the recombinant proteins, total tRNA (40 mg) from a mnmC-W131stop or mnmC(m)-G68D mutant was incubated in a 200-ml reaction mixture containing 50 mM Tris-Cl (pH 8.0), 50 mM ammonium acetate, 0.5 mM FAD or 0.5 mM S-Adenosyl-L-Methionine (SAM), 5% glycerol and 2 mM of the purified protein [Flag-MnmC, Flag-MnmC(o), or Flag-MnmC(m)] for 40 min at 37 C. tRNA was recovered by phenol extraction and ethanol precipitation and subsequently treated with nuclease P1 and E. coli alkaline phosphatase. The resulting nucleosides were analyzed by reverse-phase HPLC as described (13, 30) . For in vivo complementation studies, overnight cultures of strains mnmC-W131stop or mnmC(m)-G68D containing pBAD-TOPO or derivative plasmids (pIC1340 and pIC1339) were diluted 1:100 in 100 ml LBT containing 0.2% arabinose and grown to an OD 600 of 0.5. The cells were recovered by centrifugation at 4500g for 15 min at 4 C. Bulk tRNA was obtained and analyzed by HPLC as described above.

To assay MnmC(o) or MnmC(m) activity, the reaction mixture (100 ml) contained 50 mM Tris-HCl (pH 8), 50 mM ammonium acetate, 3% glycerol, 2 mM NaCl, 73 mM MgCl 2 , tRNA (0.5-5 mM), Flag-tagged protein (25 nM) and 100 mM FAD or SAM (depending on the MnmC activity to be determined). The mixtures were pre-incubated at 37 C for 3 min before addition of proteins. All reactions were performed at 37 C. The reactions were halted after 60-90 s of incubation by adding 100 ml of 0.3 M sodium acetate, pH 5.2. To assay the ammonium-dependent MnmEG activity with respect to tRNA concentration, the reaction mixture (100 ml) contained 100 mM Tris-HCl, pH 8, 100 mM ammonium acetate, 5 mM MgCl 2 , 5% glycerol, 5 mM DTT, 0.5 mM FAD, 2 mM GTP, 1 mM methylene-THF, 10 mg bovine serum albumin (BSA) and tRNA (0.1-2 mM). The reaction was initiated by the addition of 0.1 mM MnmEMnmG complex obtained as previously described (13) and stopped after 2 min of incubation at 37 C by adding 100 ml 0.3 M sodium acetate, pH 5.2. In all cases, tRNA was finally recovered by phenol extraction and ethanol precipitation and treated with nuclease P1 and E. coli alkaline phosphatase. The resulting nucleosides were analyzed by reverse-phase HPLC. The area of the synthesized nucleoside was calculated and its amount was extrapolated from standard curves that were prepared using chemically synthesized nucleosides (A. Malkiewicz, University of Lodz, Poland) over a range of 0-250 ng. Experiments were performed in triplicate. To determine the V max and K m values, the data were fitted to the Michaelis-Menten equation using nonlinear regression (GraphPad Prism v4.0).

To analyze the specificity of MnmC(o) and MnmC(m) in vitro, the tRNA (10-15 mg) was first modified using the MnmEG complex through the ammonium and glycine pathways as described previously (13) . The modified tRNA was phenolized, ethanol precipitated and incubated with 2 mM MnmC(o) or MnmC(m) domains in 200 ml (total volume) of MnmC buffer containing 50 mM Tris-HCl (pH 8.0), 50 mM ammonium acetate, 0.5 mM FAD [for the MnmC(o) assay] or 0.5 mM SAM (for the MnmC(m) assay) and 5% glycerol. After incubating for 40 min at 37 C, the tRNA was recovered by phenolization and ethanol precipitation and the tRNA was then treated with nuclease P1 and E. coli alkaline phosphatase. The resulting nucleosides were analyzed by reverse-phase HPLC as described above.

The acid resistance experiments were performed essentially as previously reported (39) . Briefly, strains were grown in LBT containing 0.4% glucose to stationary phase. The cultures were then diluted 1:1000 into EG medium [minimal E medium containing 0.4% glucose; (40) ], pH 2.0, supplemented or not with 0.7 mM glutamate. Samples were obtained at 0, 1, 2, 3 and 4 h post acid challenge and spotted on LAT plates.

In a previous study, we demonstrated that MnmEG catalyzes two different reactions in vitro and produces nm 5 and cmnm 5 using ammonium and glycine, respectively, as substrates (13) . We also reported the presence of cmnm 5 s 2 U and nm 5 s 2 U in total tRNA purified from an mnmC null mutant. However, it has been suggested that the formation of these intermediates is sensitive to growth conditions and specific strains, which might explain why they were not detected in other studies (23, 26, 41) .

To obtain a clear picture of the functional activity of enzymes MnmEG and MnmC during exponential growth (mid-log phase; OD 600 of $0.6) in LBT, we analyzed the HPLC profile of bulk tRNA isolated from several strains and different genetic backgrounds (Table 2 and Supplementary Figure S1 ). In the HPLC analysis, absorbance was monitored at 314 nm to maximize the detection of thiolated nucleosides. We consistently detected a small amount ($10-20%) of cmnm 5 s 2 U in the wild-type strains of three different backgrounds, TH48, BW25113 and MG1655 (Table 2) , which may originate from tRNA Gln cmnm5s2UUG or might be an intermediate of the final product mnm 5 s 2 U that is present in tRNA Lys mnm5s2UUU and tRNA Glu mnm5s2UUC . Notwithstanding, mnm 5 s 2 U was the major final product ($80-90%). We were unable to identify nucleosides mnm 5 U and cmnm 5 Um (which are present in tRNA Arg mnm5UCU , tRNA Gly mnm5UCC and tRNA Leu cmnm5UmAA ) when tRNA hydrolysates were monitored at 254 nm, as a comparison of chromatograms of total tRNA purified from wild-type strains and their mnmE or mnmG derivatives did not allow the detection of any peaks attributable to those nucleosides in the wild-type chromatogram. The percentage of nucleosides represents the distribution of the peak area of each nucleoside compared to the sum of the peak areas of the two nucleosides considered. Each value represents the mean of at least two independent experiments. wt: wild-type.

As expected, the intermediates cmnm 5 s 2 U and nm 5 s 2 U were observed in mnmC null mutants (carrying mnmC-W131stop or DmnmC mutations); these nucleosides were typically distributed in a 70/30 ratio, suggesting that the MnmEG enzyme preferably uses the glycine pathway under the growth conditions used in these experiments ( Table 2 ). This conclusion was consistent with the cmnm 5 s 2 U/mnm 5 s 2 U ratio observed in the DmnmC(o) mutant ($86/14), which clearly exhibited a MnmC(o) À MnmC(m) + phenotype. Moreover, the relative distribution of cmnm 5 s 2 U and nm 5 s 2 U in the mnmC(m)-G68D mutant was $15/85, suggesting that most of the cmnm 5 s 2 U formed by the glycine pathway is converted to nm 5 s 2 U by the MnmC(o) activity present in this strain and that the remaining cmnm 5 s 2 U could proceed, at least partially, from tRNA Gln cmnm5s2UUG . Taken together, these results clearly support the idea that MnmEG uses ammonium or glycine to modify tRNAs in vivo and that Table 2 ] to transform tRNAs modified by MnmEG via the ammonium pathway ( Figure 1 ), which raises the possibility that certain tRNAs may be substrates for MnmC(m), but not for MnmC(o).

In order to study the functional independence of the two MnmC domains, we decided to express them separately. The full mnmC gene and its two domains, mnmC(o) (encoding for amino acids 250-668 of MnmC) and mnmC(m) (encoding for amino acids 1-250 of MnmC), were N-terminal Flag-tagged by PCR and cloned into the pBAD-TOPO expression vector. Recombinant protein expression was induced in the E. coli mutant mnmC::kan (DmnmC) by adding 0.2% arabinose. The Flag-tagged proteins MnmC, MnmC(o) and MnmC(m) were purified close to homogeneity and exhibited apparent molecular masses of 78, 48 and 34 kDa, respectively ( Figure 2A) .

The Flag-MnmC protein and its C-terminal domain [Flag-MnmC(o)] were yellow, and their spectra showed maximum absorption peaks at $375 and 450 nm, indicating the presence of a flavin derivative. The putative cofactor was identified as FAD by its retention time ( Figure 2B) . Therefore, the isolated MnmC(o) domain contains non-covalently bound FAD, which suggests that it is correctly folded. To our knowledge, this is the first report describing the expression and purification of the soluble form of MnmC(o). A previous attempt to purify this domain was unsuccessful because overexpression of a somewhat different, recombinant MnmC(o) produced inclusion bodies, which led to the hypothesis that MnmC(o) is incapable of folding on its own and requires the presence of MnmC(m) (25) . Our data, however, demonstrate that MnmC(o) can fold independently of MnmC(m). Interestingly, the in vivo stability of the separated domains was significantly lower than that of the full protein ( Figure 2C) , suggesting that the physical interaction of both domains in the full protein confers greater stability.

The full MnmC protein and the separate domains MnmC(o) and MnmC(m) behaved as monomeric proteins when subjected to gel filtration chromatography ( Figure 2D ). However, a mixture of MnmC(o) and MnmC(m) displayed the same elution profile as the full protein, indicating that the separate domains interact in vitro ( Figure 2D ). This interaction was further explored by kinetic analysis using surface plasmon resonance (SPR). Flag-MnmC(o) was injected at different concentrations onto the immobilized His-MnmC(m) ligand ( Figure 2E ). The apparent equilibrium constant K D was 87 ± 15 nM, which is indicative of a strong interaction between MnmC(o) and MnmC(m).

Taking into consideration the functional relationships of MnmC with the MnmEG complex, we explored whether MnmC interacts with members of the complex, i.e. MnmE or MnmG. A final concentration of 5 mM Flag-MnmC was mixed with 5 mM His-MnmG or MnmE and the samples were analyzed by gel filtration. As shown in Figure 2F , no interaction between the full MnmC protein and MnmE or MnmG was observed. The same negative result was obtained by SPR (data not shown).

To analyze whether the recombinant MnmC(o) and MnmC(m) proteins are functionally active, we first assessed their tRNA modifying capability in vitro using total tRNA purified from exponentially growing cultures as a substrate. The in vitro activity of the recombinant MnmC(o) protein was investigated by incubating tRNA purified from the null mutant mnmC-W131stop (nm 5 s 2 U/ cmnm 5 s 2 U ratio & 31/69, Table 2 ) with 2 mM Flag-MnmC(o) domain and 0.5 mM FAD. As shown in Figure 3A , a major part of cmnm 5 s 2 U was converted to nm 5 s 2 U, which demonstrates that Flag-MnmC(o) is catalytically active in the in vitro assay, although a small part of cmnm 5 s 2 U, probably proceeding from tRNA Gln cmnm5s2UUG , remained unaltered. When tRNA purified from the mnmC(m)-G68D mutant (nm 5 s 2 U/ cmnm 5 s 2 U ratio & 85/15; Table 2 ) was incubated with 2 mM Flag-MnmC(m) and 0.5 mM SAM, nm 5 s 2 U was modified to mnm 5 s 2 U, demonstrating the capability of the recombinant protein to perform the methylation reaction in vitro ( Figure 3B ). As expected, the small peak corresponding to cmnm 5 s 2 U (according to its retention time and spectrum) remained unchanged after the in vitro reaction mediated by MnmC(m) ( Figure 3B ).

Then, we assessed the capability of the recombinant MnmC(o) and MnmC(m) proteins to modify tRNA in vivo. The strain with the null mutation mnmC-W131stop (IC6019) was transformed with pBAD-TOPO and its derivative pIC1339 expressing MnmC(o), whereas strain mnmC(m)-G68D (IC6018) was transformed with pBAD-TOPO and its derivative pIC1340 expressing MnmC(m). In the resulting strains, which were grown to exponential phase in the presence of the arabinose inducer, the recombinant MnmC(o) protein catalyzed the conversion of cmnm 5 s 2 U into nm 5 s 2 U, although a peak of remnant cmnm 5 s 2 U was observed ( Figure 3C ) and nm 5 s 2 U was fully transformed to mnm 5 s 2 U by MnmC(m) ( Figure 3D ). Similar results were obtained in the absence of arabinose (data not shown). This finding indicates that recombinant MnmC(o) and MnmC(m) are expressed and accumulate to some extent in the absence of the inducer, although it was not possible to detect them by western blotting with an anti-Flag antibody (data not shown). The fact that both proteins were capable of modifying tRNA despite their very low concentrations suggests that these proteins have high activity in vivo. those exhibited by the entire MnmC protein, we conducted steady-state kinetic experiments. The conditions used for each reaction were similar to enable a comparison of the kinetic constants and were based on conditions known to optimize activity (23, 26) . The substrate for the assays was a chimeric version of E. coli tRNA Lys mnm5s2UUU expressed from pBSKrna (37, 42) , which facilitates overproduction and purification of recombinant RNA and has been successfully used by our group in previous studies (21, 33) . Here, the tRNA Lys mnm5s2UUU gene was inserted into the EcoRV site of the region encoding the scaffold tRNA (a human cytosolic tRNA Lys 3 lacking the anticodon region). The resulting chimeric tRNA essentially contained the complete E. coli tRNA Lys mnm5s2UUU sequence fused at its 5 0 -and 3 0 -ends to the truncated anticodon stem of human cytosolic tRNA Lys 3 , producing an RNA of $170 nt (Supplementary Figure S2) .

The FAD-dependent-oxidoreductase activity of the recombinant MnmC(o) and MnmC proteins was compared by determining the conversion of cmnm 5 s 2 U to nm 5 s 2 U on the cmnm 5 s 2 U-containing chimeric tRNA Lys mnm5s2UUU (chi-tRNA Lys mnm5s2UUU ) purified from an mnmC null mutant (IC6010). The SAM-dependent methyltransferase activity of MnmC(m) and MnmC was compared by following the conversion of nm 5 s 2 U to mnm 5 s 2 U on the nm 5 s 2 Ucontaining chi-tRNA Lys mnm5s2U extracted from an mnmC(m)-G68D mutant (IC6018). All proteins displayed Michaelis-Menten kinetics with respect to varying concentrations of chi-tRNA Lys mnm5s2UUU (Supplementary Figure S3 ). As shown in Table 3 , the enzymatic activity of each isolated MnmC domain had kinetic parameters (k cat and K m ) similar to those exhibited by the entire protein, which supports the notion that the MnmC domains function independently.

The k cat constants of the MnmC activities were similar to those of previous studies (23, 26) , although the K m values were higher than those reported by Pearson and Carell (26) , most likely due to the nature of the tRNA substrate used in our experiments. In any case, our data indicated that MnmC(m) displayed a catalytic efficiency (k cat /K m ) 2-fold higher than that of MnmC(o) ($0.1 versus $0.05; Table 3 ), which is in agreement with the proposal that synthesis of the final modification mnm 5 (s 2 )U is favored by kinetic tuning of the MnmC activities, thus avoiding accumulation of the nm 5 (s 2 ) intermediate (26) .

We also analyzed the catalytic parameters of the ammonium-dependent reaction mediated by the MnmEG complex, whose activity precedes that of MnmC(m) in the mnm 5 (s 2 )U synthesis (Figure 1) . We used conditions known to be appropriate for assaying the MnmEG activities in vitro (13) and included the presence of Mg 2+ (5 mM), which is required for GTP hydrolysis by MnmE (43) . In line with this, the assay buffer differed from that of the MnmC reactions where a relatively low concentration of Mg 2+ was used (73 mM), as MnmC activities are known to be severely inhibited at 5 mM of Mg 2+ (23) . Under the proper in vitro conditions for each enzyme, the modification reaction mediated by MnmEG (incorporation of nm 5 into U34) displayed an $5-fold lower catalytic efficiency than that of the MnmC(m)-mediated reaction (Table 3) . However, this comparison should be handled with caution because the conditions used in the MnmEG and MnmC(m) assays were different, and the K m values may change according to the buffer conditions. In addition, the structure of the chimeric tRNA used as a substrate could affect the catalytic cycle of each enzyme in a different manner.

Therefore, in order to gain information on the efficiency of the mnm 5 s 2 U assembly-lines in vivo, we decided to explore the modification status of the native tRNA Lys mnm5s2UUU in several strains. The presence of modification intermediates in the tRNA purified from a wildtype strain would be indicative of the MnmEG and MnmC activities not being kinetically tuned. HPLC analysis of native tRNA Lys mnm5s2UUU purified during the exponential phase revealed no accumulation of the cmnm 5 s 2 U intermediate in the wild-type strain ( Figure 4A ), whereas this nucleoside was predominant in the tRNA extracted from the ÁmnmC strain ( Figure 4B ). Moreover, we observed no nm 5 s 2 U in the native tRNA Lys mnm5s2UUU purified from the wild-type strain ( Figure 4A Figure 4C ). These results suggest that tRNA Lys mnm5s2UUU containing cmnm 5 s 2 U or nm 5 s 2 U at the wobble position is stable, and consequently that the disappearance of these intermediates in the wild-type strain is probably due to their conversion into mnm 5 s 2 U by MnmC activities.

Thus, it appears that MnmEG and MnmC are kinetically tuned to produce only the final mnm 5 group in tRNA Lys mnm5s2UUU . Notably, the intermediate s 2 U, resulting from the activity of MnmA (Figure 1 ), was only observed in a DmnmG strain, which is defective in the MnmEG activity ( Figure 4D ). Moreover, we did not detect the non-thiolated intermediates cmnm 5 U and nm 5 U when tRNA Lys mnm5s2UUU hydrolysates were monitored at 254 nm (data not shown). These data suggest that the coordination of the MnmAand MnmEGC-dependent pathways is tightly coupled to synthesize mnm 5 s 2 U on native tRNA Lys mnm5s2UUU without a build-up of intermediates.

Escherichia coli tRNA Gln cmnm5s2UUG and tRNA Leu cmnm5UmAA contain cmnm 5 at the wobble uridine instead of the mnm 5 found in the remaining MnmEG substrates, (http://modomics.genesilico.pl/sequences/list/tRNA). These data suggest that tRNA Gln cmnm5s2UUG and tRNA Leu cmnm5UmAA are not substrates for MnmC(o). Moreover, they raise the question of what happens with the ammonium pathway of the MnmEG complex, which theoretically produces nm 5 U in all its tRNA substrates ( Figure 1) . Interestingly, the analysis of gln1 tRNA from Salmonella enterica serovar Typhimurium indicated that 80% of the molecules contained cmnm 5 s 2 U, whereas the remaining 20% contained mnm 5 s 2 U (44). In this case, we speculate that mnm 5 s 2 U might be synthesized by MnmC [using the MnmC(m) activity] from the nm 5 s 2 U generated via the ammonium-dependent MnmEG pathway. Therefore, we decided to investigate whether a fraction of tRNA Gln cmnm5s2UUG and tRNA Leu cmnm5UmAA in E. coli contains mnm at position 5 in the wobble uridine, and whether both tRNAs function as substrates for MnmC(m) but not MnmC(o) (Figure 1 ).

To address this question, we constructed a series of tRNA expression plasmids by inserting the coding sequences of tRNA Gln cmm5s2UUG , tRNA Leu cmnm5UmAA and tRNA Lys mnm5s2UUU into EcoRI/PstI-digested pBSKrna (37); thus, we eliminated the tRNA scaffold-encoding region (i.e. the DNA region encoding the human cytoplasmic tRNA Lys 3 ) present in the vector (Supplementary Figure S2 ). Selected strains were transformed with the resulting plasmids. We then examined the HPLC profile of the overexpressed, 'native' tRNAs purified from cultures grown overnight to the stationary phase. It should be noted that the expression of the tRNA genes in these constructs was under the control of the strong constitutive lpp promoter (37) . As mutations leading to reduced expression might be selected over time because overexpression ultimately slows growth, the fresh transformation of cells with the pBSKrna derivatives every time and overnight incubation of the transformed cells prior to the purification of the cloned target (in this case, tRNAs) are highly recommended (37) .

The HPLC analysis of tRNA Lys mnm5s2UUU ( Figure 5A ), used herein as a control, indicated that: (i) mnm 5 s 2 U was predominant in a wild-type strain; (ii) both nm 5 s 2 U and cmnm 5 s 2 U accumulated in the mnmC null strain and (iii) nm 5 s 2 U prevailed in the mnmC(m)-G68D strain. Taken together, these results indicate that both MnmEG-MnmC(o)-MnmC(m) and MnmEG-MnmC(m) pathways operate on tRNA Lys mnm5s2UUU and that they converge to produce the final modification mnm 5 s 2 U in the wild-type strain.

An analysis of overexpressed tRNA Gln cmnm5s2UUG ( Figure 5B ) indicated that cmnm 5 s 2 U and nm 5 s 2 U accumulated at a ratio of $90/10 when tRNA was purified from the mnmC-W131stop mutant. Despite the lower proportion of nm 5 s 2 U, the presence of this nucleotide indicates that the MnmEG complex was indeed able to modify tRNA Gln cmnm5s2UUG through the ammonium pathway. Nucleosides cmnm 5 s 2 U and mnm 5 s 2 U were found at a 90/10 ratio when tRNA Gln cmnm5s2UUG was purified from the wild-type strain. Detection of mnm 5 s 2 U in tRNA Gln cmnm5s2UUG clearly indicates that this tRNA is a substrate for MnmC(m).

It should be noted that both cmnm 5 s 2 U and nm 5 disappeared when tRNA Lys mnm5UUU was purified from the mnmC(m)-G68D strain ( Figure 5A ). This result was expected, considering that cmnm 5 s 2 U was transformed into nm 5 s 2 U by the MnmC(o) activity present in the mnmC(m)-G68D strain. Interestingly, the HPLC profiles of tRNA Gln cmnm5s2UUG purified from the mnmC(m)-G68D and mnmC-W131stop strains were similar ( Figure 5B ). In both cases, cmnm 5 s 2 U was much more abundant than nm 5 s 2 U, suggesting that the MnmC(o) activity present in the mnmC(m)-G68D strain did not function in tRNA Gln cmnm5s2UUG . Overexpression of tRNAs often causes hypomodification. In fact, the hypomodified nucleoside s 2 U, resulting from the action of MnmA, was relatively abundant in the overexpressed tRNA Lys mnm5s2UUU and tRNA Gln cmnm5s2UUG (data not shown). Nevertheless, s 2 U accumulated at a similar proportion in all tested strains so that the ratios of the nucleosides of interest (mnm 5 s 2 U, cmnm 5 s 2 U and nm 5 s 2 U) were not greatly affected. When the tRNA Lys mnm5s2UUU and tRNA Gln cmnm5s2UUG hydrolysates were monitored at 254 nm, the non-thiolated nucleosides nm 5 U, mnm 5 U, cmnm 5 U were hardly detected (data not shown). Therefore, we think that the different HPLC pattern of tRNA Lys mnm5s2UUU and tRNA Gln cmnm5s2UUG in the mnmC(m)-G68D strain cannot be attributed to the presence of hypomodified nucleosides. Rather, it reveals that tRNA Gln cmnm5s2UUG is not a substrate for MnmC(o). We also analyzed the HPLC profile of tRNA Leu cmnm5UmAA purified from a trmL strain in which the methylation of the ribose in the wobble uridine is impaired due to the ÁtrmL mutation (21) . We adopted this approach to examine the modification of tRNA Leu cmnm5UmAA because we were unable to distinguish nucleosides cmnm 5 U m , nm 5 U m and mnm 5 U m in our HPLC chromatograms. However, we were able to identify the peaks corresponding to cmnm 5 U, nm 5 U and mnm 5 U using proper markers ( Figure 5C ). As shown in Figure 5D , overexpressed tRNA Leu cmnm5UmAA purified from a ÁtrmL strain contained cmnm 5 U. However, no traces of mnm 5 U or nm 5 U were detected in this tRNA. Similar results were obtained when analyzing native, non-overexpressed tRNA Leu cmnm5UmAA purified in the exponential phase ( Figure 5E ). Altogether, these data suggest that tRNA Leu cmnm5UmAA is not modified by the ammonium-dependent MnmEG pathway and that it is not a substrate for MnmC(o).

To further explore the specificity of MnmC(o) and MnmC(m) for tRNA Gln cmnm5s2UUG , tRNA Leu cmnm5UmAA and tRNA Lys mnm5s2UUU , we performed in vitro modification reactions using in vitro synthesized tRNAs as substrates. These modification reactions included two steps. First, substrate tRNA was modified by the MnmEG complex through the ammonium ( Figure 6A , B, E and F) or the glycine ( Figure 6C and D) pathways. The resulting tRNA carrying nm 5 U or cmnm 5 U (solid black lines) was subsequently used as a substrate to characterize the activity of the MnmC(m) or the MnmC(o) domains, respectively.

The data in Figure 6 indicate that MnmC(m) catalyzed the conversion of nm 5 U into mnm 5 U in both tRNA Lys mnm5s2UUU (panel A) and tRNA Gln mnm5s2UUG (panel B), whereas MnmC(o) catalyzed the conversion of cmnm 5 into nm 5 in tRNA Lys mnm5s2UUU (panel C), but not in tRNA Gln cmnm5s2UUG (panel D). These results support the idea that tRNA Gln cmnm5s2UUG does not function as a substrate for MnmC (o) .

Surprisingly, we observed that MnmEG catalyzed the formation of nm 5 U from in vitro synthesized tRNA Leu cmnm5UmAA and, in turn, nm 5 U was converted into mnm 5 U through MnmC(m) ( Figure 6E ). This result contradicts the data provided in Figure 5D and E, where no traces of nm 5 U or mnm 5 U were observed in the HPLC analysis of the tRNA Leu cmnm5UmAA purified from a trmL strain. Considering that the experiments shown in Figure 6E were performed using an in vitro synthesized tRNA Leu cmnm5UmAA (therefore lacking modifications), we thought that the presence of some modified nucleoside(s) in the in vivo synthesized tRNA Leu cmnm5UmAA ( Figure 5D and E) could hinder the recognition of this substrate by MnmEG. However, when we used overexpressed tRNA Leu cmnm5UmAA purified from a double DtrmL/DmnmG mutant as a substrate in the in vitro modification assay, the MnmEG-mediated synthesis of nm 5 s 2 U was once again observed ( Figure 6F) . Moreover, the MnmEGmodified tRNA Leu cmnm5UmAA was a good substrate for the in vitro synthesis of mnm 5 through MnmC(m) ( Figure 6F ). Altogether, these results suggest that the ammonium pathway of MnmEG is ineffective on tRNA Leu cmnm5UmAA in vivo, even though it efficiently functions on this tRNA in the in vitro assay.

We also assessed the activity of MnmC(o) on the tRNA Leu cmnm5UmAA purified from a trmL strain and on the in vitro transcribed tRNA Leu cmnm5UmAA previously modified by MnmEG in vitro via the glycine pathway (i.e. on tRNA molecules carrying cmnm 5 U). The synthesis of nm 5 from cmnm 5 was not observed under any condition (data not shown). Therefore, we concluded that tRNA Leu cmnm5UmAA is a substrate in vitro for MnmC(m) ( Figure 6E and F) but not MnmC(o).

Notably, MnmEG also catalyzed the ammoniumdependent synthesis of nm 5 U in tRNA Glu mnm5s2UUC , tRNA Arg mnm5UCU and tRNA Gly mnm5UCC obtained by in vitro transcription (Supplementary Figure S4) . We suspect that modification of these tRNAs follows a pattern similar to that observed in tRNA Lys mnm5s2UUU , because all these tRNA species appear to contain mainly the mnm 5 group of U34. However, this hypothesis should be examined in future studies.

The overexpressed tRNA Lys mnm5s2UUU and tRNA Gln cmnm5s2UUG purified from the mnmC-W131stop mutant during the stationary phase exhibited different cmnm 5 s 2 U/nm 5 s 2 U ratios ($35/65 and $90/10, respectively; Figure 5A and B). Moreover, the ratio in the overexpressed tRNA Lys mnm5s2UUU ($35/65) differed from that found in the total tRNA obtained from the mnmC-W131stop strain during exponential growth ($69/31; Table 2 ). These results suggest that the glycine pathway (which produces cmnm 5 s 2 U) could be less effective on tRNA Lys mnm5s2UUU in the stationary phase. These observations prompted us to carefully investigate the efficiency of the glycine and ammonium pathways in total tRNA and specific native tRNAs during the growth curve.

First, we determined the modification status of total tRNA in the wild-type strain and mnmC mutants. In the wild-type or mnmC(m)-G68D strain (Figure 7 , panels A and B), we could not identify which pathway contributes to the synthesis of the final modification (mnm 5 s 2 U or nm 5 s 2 U, respectively) because the activity of MnmC(o) converges both pathways by transforming cmnm 5 s 2 U into nm 5 s 2 U (Figure 1) . However, the strain carrying an mnmC-W131stop or mnmC(o) null mutation (Figure 7 , panels C and D) revealed that the level of cmnm 5 s 2 U and nm 5 s 2 U or mnm 5 s 2 U in bulk tRNA depends on the growth phase; nm 5 s 2 U (or mnm 5 s 2 U) accumulates as optical density increases, whereas cmnm 5 s 2 U exhibits the opposite trend and becomes the minor component at an OD 600 of $2.0. As the data in Figure 7 represent the relative distribution of each nucleoside with respect to the sum of the peak areas of the two nucleosides, it is important to point out that the net decrease in the cmnm 5 s 2 U peak area along the growth curve was concomitant with the net increase in the nm 5 s 2 U or mnm 5 s 2 U area.

Interestingly, when the DmnmC(o) strain was grown in minimal medium instead of LBT, cmnm 5 s 2 U was consistently observed as the major intermediate, irrespective of the growth phase (Table 4) . Therefore, the growth conditions (growth medium and growth phase) affect the pathway responsible for modification at position 5 of U34.

We subsequently analyzed the behavior of native tRNAs in both the exponential and stationary phases (Table 5 and Supplementary Figure S5 ). During the exponential growth of the DmnmC strain, cmnm 5 s 2 U was the major intermediate accumulated in native tRNA Lys mnm5s2UUU (94%), but this nucleoside became the minor intermediate (4%) in the stationary phase. In contrast, HPLC analysis revealed that cmnm 5 s 2 U was consistently the major intermediate present in native tRNA Gln cmnm5s2UUG (>90%), irrespective of the growth phase. Altogether, these results indicate that tRNA Gln cmnm5s2UUG functions different from tRNA Lys mnm5s2UUU and total tRNA when the DmnmC strain is grown to stationary phase in LBT (i.e. in growth conditions that favor the ammonium pathway). Therefore, the accumulation of cmnm 5 or nm 5 depends on the specific characteristics of the tRNA.

The limited data available regarding the effect of mnmC mutations on growth rate are contradictory. No differences between mnmC and wild-type strains were initially reported (22, 45) . However, more recently, Pearson and Carell (26) knockout strain from the Keio collection and observed a larger growth constant for the wild-type strain, in agreement with the hypothesis that accumulation of intermediates in the synthesis of mnm 5 (s 2 )U affects the translation process negatively. We investigated the effect of mnmC mutations on growth rate in different genetic backgrounds and did not observe significant differences between the wild-type strain and any derivative carrying an mnmC allele [DmnmC, mnmC-W131stop, DmnmC(o) and mnmC(m)-G68D; Table 6 ]. We cannot explain why our results differ from those reported by Pearson and Carell (26) , even when the same background (BW25113) was used. Perhaps, the dilution method used to maintain the pre-cultures in exponential growth could explain the different results. We took great care to not dilute the precultures below an OD 600 of 0.25 and to not exceed an OD 600 of 0.5 in order to avoid major changes under the exponential growth conditions. The lack of effect in our hands of mnmC mutations during the exponential growth of cells contrasts with the behavior of mnmE-and mnmG-knockout strains, which, as we have previously demonstrated, grow slower than the wild-type strain (12, 32) . These data suggest that the lack of any modification at the wobble uridine is more deleterious to the cell than the presence of any intermediate of mnm 5 (s 2 ) synthesis. Supporting this conclusion, we also observed that the mnmC mutations, unlike the mnmE mutations, did not affect resistance to acid stress of E. coli (Figure 8 ). This resistance depends on several systems, one of which involves proteins whose translation likely depends on tRNAs modified by MnmEG (39, 46) .

The MnmA protein catalyzes the last step of thiolation at position 2 of U34 in tRNA Lys mnm5s2UUU , tRNA Glu mnm5s2UUC , and tRNA Gln cmnm5s2UUG (Figure 1 ). Inactivation of the mnmA gene has been reported to increase the doubling time (31, 47) . Interestingly, combination of the mnmA and mnmC mutations had a synergic effect (Table 6 ), revealing that loss of any MnmC activity is detrimental under specific conditions. Mutation mnmC(m)-G68D appeared to slow down the growth rate of the mnmAknockout strain slightly more than the DmnmC mutation.

We also performed competition experiments between BW25113 (which is a wild-type, tetracycline-sensitive strain) and TH48 (wild-type), TH49 (mnmC(m)-G68D) and TH69 (mnmC-W131stop), which are tetracyclineresistant strains (Table 7) . This type of experiment compares not only the growth in exponential phase, but also the ability to sustain and recover from stationary phase. Although TH48 and its derivatives were outcompeted by BW25113, it is clear that the impairment of the MnmC functions reduces the ability of the strain to compete. Curiously, in these experiments, mutant mnmC-W131stop was less competitive than mnmC(m)-G68D.

This study aimed to further explore the activities and specificities of the MnmEG complex and the MnmC domains, the ability of specific tRNAs to follow the ammonium-or glycine-MnmEG pathway under different growth conditions and the effect of mnmC mutations on cell growth.

The fulfillment of the two first aims was facilitated by the cloning and separate expression of the MnmC domains. We demonstrate for the first time that the MnmC(o) domain can fold independently of MnmC(m) and that both domains are catalytically active when expressed separately. Importantly, the domains exhibit similar kinetic properties to those observed in the entire MnmC protein, which suggests that the two domains operate independently of each other. Moreover, MnmC(o) and MnmC(m) differ in terms of their specificity for tRNA substrates. Thus, MnmC(o) modifies tRNA Lys mnm5s2UUU , but not tRNA Gln cmnm5s2UUG and tRNA Leu cmnm5UmAA . In contrast, all three tRNAs (tRNA Lys mnm5s2UUU , tRNA Gln cmm5s2UUG and tRNA Leu cmnm5UmAA ) serve as substrates for MnmC(m) not only in vitro, but also in vivo at least in the cases of tRNA Lys mnm5s2UUU and tRNA Gln cmm5s2UUG . These data strongly support the notion that the MnmC(o) and MnmC(m) domains function independently. The presence of mnm 5 s 2 U in in vivo synthesized tRNA Gln cmnm5s2UUG suggests that the nomenclature of this tRNA should be changed to tRNA Gln ðcÞmnm5s2UUG . Sequence searches of complete genomes have revealed that the distribution of the putative orthologs of the E. coli bi-functional MnmC protein are conserved only in g-proteobacteria, with some additional members in other bacterial classes (24) . Interestingly, in several genomes, potential orthologs of a single domain have been identified (24) , but the biological function has been experimentally demonstrated only in the case of the Aquifex aeolicus DUF752 protein, an MnmC(m) homolog (41) . It is unclear from the phylogenetic analysis whether the independent putative orthologs represent the ancestral or derived versions of the bi-functional MnmC enzyme (24) . The ability of the E. coli MnmC(o) and MnmC(m) domains to function independent of each other, as shown in this work, suggests that the origin of the full MnmC protein present in g-Proteobacteria (24) likely occurred by domain fusion. The MnmC(o)-MnmC(m) fusion in E. coli suggests a functional advantage of the spatial proximity of both domains. The fusion could increase the stability of the resulting protein. In fact, our results indicate that the entire MnmC protein is substantially more stable than the separate domains ( Figure 2C ) or other RNA modification proteins, such as RsmG or MnmG, which have half-lives of $17 and 31 min, respectively (48) . The possibility that the fusion also facilitates the functional cooperation between the domains during modification of tRNAs following the 'canonical' MnmEG-MnmC(o)-MnmC(m) pathway (cmnm 5 U!nm 5 U!mnm 5 U) cannot be ruled out. A better understanding of how the MnmC protein functions will entail determining whether the binding of a tRNA molecule to one domain exerts a negative allosteric effect on the other domain. In this respect, the structural characterization of MnmC-tRNA complexes could prove helpful.

Biosynthesis of complex modifications like mnm 5 s 2 U, cmnm 5 s 2 U, cmnm 5 Um and mnm 5 U requires the participation of at least two specific enzymes ( Figure 1 ). Our in vivo data suggest that MnmEG and MnmC activities are kinetically tuned to produce the final modification mnm 5 s 2 U on native tRNA Lys mnm5s2UUU because no intermediates (cmnm 5 s 2 U and nm 5 s 2 U) were observed in a wild-type strain ( Figure 4) . Apparently, the intermediates are not subject to active degradation as they were shown to be stable in strains DmnmC and mnmC(m)-G68D. Notably, neither s 2 U nor the non-thiolated nucleosides cmnm 5 U and nm 5 U were found in native tRNA Lys mnm5s2UUU purified from the wild-type strain (Figure 4 ; data not shown). Therefore, we believe that the MnmEG-MnmC and MnmA pathways are also coordinated by tuning the activities and relative abundances of the tRNA-modifying enzymes. Modifications of U34 by MnmEG and MnmA occur independently of each other [(30,49); Figure 4D ], but whether the presence of the s 2 group facilitates the modification of position 5 in tRNA Lys mnm5s2UUU , tRNA Glu mnm5s2UUC and tRNA Gln ðcÞmnm5s2UUG by modulating the electron density distribution in the uridine ring remains unclear.

There are very few data available on the identity determinants required by TrmL. Methylation of the ribose The ratio was calculated as CFU/ml recovered on LAT supplemented with tetracycline versus CFU/mL recovered on LAT.

2 0 -OH group by this enzyme occurs as a late step in tRNA Leu CmAA maturation given that the enzyme fails to methylate tRNA Leu CmAA without the prior addition of N 6 -(isopentenyl)-2-methyladenosine (ms 2 i 6 A) at position 37 (21) . Nucleoside ms 2 i 6 A is also present in tRNA Leu cmnm5UmAA . Hence, it is feasible that TrmL also requires ms 2 i 6 A as a positive identity determinant to modify tRNA Leu cmnm5UmAA . Here our data indicate that MnmEG modifies tRNA Leu cmnm5UmAA in the absence of 2 0 -O-methylation (Figures 5D and E; 6E and F), but we do not know whether TrmL requires the presence of the cmnm 5 group to act on tRNA Leu cmnm5UmAA . More experiments are needed to unravel the order of action of the modifying enzymes working on U34 and the identity elements required by each enzyme.

Another interesting question concerning tRNA maturation is whether physical interactions among tRNA modification enzymes contribute to the efficiency of complex pathways. In Aquifex aeolicus, an interaction between the E. coli MnmC(m) homolog, protein DUF752 and the A. aeolicus MnmE and MnmG proteins has been observed (41) . In E. coli, we did not detect interaction of MnmC with MnmE or MnmG by either fast-performance liquid chromatography (FPLC) analysis ( Figure 2F ) or SPR (data not shown). However, we cannot rule out the possibility of the MnmEG complex being able to interact with MnmC in vivo. It appears reasonable to hypothesize that the spatial clustering of tRNA-modifying enzymes within the cell can optimize the coordination of their work. Therefore, one challenging aim for future studies is to determine whether tRNA modifying enzymes are spatially organized, as observed for other enzymes involved in RNA biology, such as the components of the RNA degradosome (50) .

The use of strains carrying the DmnmC(o) or DmnmC mutation has allowed us to study the ability of MnmEG to use the ammonium or the glycine pathway under different growth conditions and thus, to detect the reprogramming of U34 base modification in both bulk and specific tRNAs (Tables 4 and 5; Figure 9 ). The analysis of bulk tRNA indicates that when DmnmC(o) or DmnmC strains are grown in LBT, the glycine pathway prevails over the ammonium pathway in the exponential phase as cmnm 5 s 2 U is the most abundant nucleoside ( Figure 9A ). However, a gradual change in the output of both pathways is observed along the growth curve so that the nucleosides produced by the ammonium pathway become prevalent during entry into the stationary phase ( Figure 9B ). In contrast, when cells are grown in minimal medium, the glycine pathway is prevalent in both the exponential and stationary phases ( Figure 9A) . A preference by MnmEG for the ammonium or the glycine pathway to catalyze tRNA modification may depend on the availability of these substrates, which, in turn depends on the growth medium and cell metabolism.

The nature of tRNA species also appears crucial for the net output of the MnmEG pathways (Table 5 and Figure 9 ). Thus, whereas tRNA Lys mnm5s2UUU follows the pattern observed in bulk tRNA and is preferably modified via the ammonium pathway during the transition to the stationary phase, tRNA Gln ðcÞmnm5s2UUG fails to be effectively modified through this pathway, as indicated by the low accumulation of nm 5 s 2 U in the DmnmC mutant. Moreover, the ammonium pathway proves ineffective in modifying tRNA Leu cmnm5UmAA at any phase of growth as no nm 5 U or mnm 5 U is detected in a trmL strain ( Figure 5D and E). Therefore, neither tRNA Leu cmnm5UmAA nor tRNA Gln ðcÞmnm5s2UUG is a good substrate for the ammonium pathway in vivo, even though both tRNAs appear rather well modified by this pathway in vitro ( Figure 6B and E; Figure 9C and D).

Why the behavior of tRNA Gln ðcÞmnm5s2UUG and tRNA Leu cmnm5UmAA is different from that of tRNA Lys mnm5s2UUU remains unresolved. The use of ammonium or glycine by MnmEG could be regulated by interactions of this complex with its tRNA substrates and unknown factors in vivo. tRNA Gln ðcÞmnm5s2UUG and tRNA Leu cmnm5UmAA may be poor substrates for the ammonium pathway and thus out-competed by other tRNAs in vivo. Alternatively, tRNA Gln ðcÞmnm5s2UUG and tRNA Leu cmnm5UmAA carrying nm 5 or mnm 5 , but not cmnm 5 may be unstable and specifically degraded by some nuclease(s) (5, 51) . Additional experiments are required to test these hypotheses and to elucidate the mechanism underlying the selection by MnmEG of the glycine or the ammonium pathway.

Our results highlight the crucial role of MnmC(o) and Mnm(m) in the biosynthesis of mnm 5 (s 2 )U in accordance with growth conditions (Figure 9 ). During the exponential phase in LBT and at any phase in minimal medium, the oxidoreductase activity of MnmC(o) is required to drive the intermediate generated by the more active glycine pathway [cmnm 5 (s 2 )U] toward the MnmC(m)-dependent synthesis of the final modification in tRNA Lys mnm5s2UUU and, presumably, in tRNA Glu mnm5s2UUC , tRNA Arg mnm5UCU and tRNA Gly mnm5UCC ( Figure 9A ). In contrast, MnmC(o) plays a minor role when the ammonium pathway prevails; i.e. during the transition to stationary phase of cultures growing in LBT. In this case, the methylase activity of MnmC(m) functions fairly well alone [i.e. without the participation of MnmC(o)] to drive the production of the final modification mnm 5 (s 2 )U by transforming the intermediate nm 5 (s 2 ) generated by the more active ammonium pathway (Table 4 and Figure 9B ). Thus, the incorporation of MnmC(o) and MnmC(m) into the enzyme content of E. coli extends the adaptability of this bacterium to changing growth conditions. How our findings in E. coli apply to other bacteria, especially those lacking certain MnmC activity, is an issue that merits future research.

The lack of modifications at the wobble uridine due to the presence of mnmE or mnmG mutations leads to impaired growth, high sensitivity to acidic pH and defects in translational fidelity (12, 35, 46, 48, (52) (53) (54) (55) . Here our functional studies suggest that MnmC impairment is less detrimental than the inactivation of the MnmEG complex given that mnmC mutations neither reduce the growth rate in rich medium (Table 6 ) nor affect resistance to acidic pH ( Figure 8) . Nevertheless, impairment of any MnmC activity has a biological cost. Thus, mnmC mutations diminish the ability of cells to compete ( Table 7) , suggesting that fully modified tRNAs and, therefore the activities of MnmC, are required for efficient translation of mRNAs involved in the stationary phase stress response. In addition, our data indicate that mnmC mutations slow the growth rate of a mnmA strain, revealing a synergistic effect of mutations mnmC and mnmA. As mnmC mutations alone do not affect the growth rate, we hypothesize that the presence of the MnmA-dependent thiolation at position 2 of U34 in a subset of tRNAs can compensate for the lack of MnmC modifications at position 5 during exponential growth.

In an mnmA background, accumulation of nm 5 U produced by the mnmC(m)-G68D mutation appears slightly more detrimental than accumulation of cmnm 5 caused by a null mnmC mutation (Table 6 ). In line with this, it has been demonstrated that the reading of an amber codon by an ochre-suppressing derivative of tRNA Lys mnm5s2UUU (supG) is also more affected by the Figure 9 . Summary scheme of the reprogramming of U34 base modification in both bulk and specific tRNAs. The main and minor modification pathways are indicated by thick and thin arrows, respectively; pathways whose activity was only observed in vitro are indicated by dotted arrows. Activity of TrmL on tRNA Leu cmnm5UmAA has been omitted in panel D. Labels in each panel indicate the type of tRNA (specific or bulk tRNA) following the pathways, and the growth conditions under which the tRNA was obtained. LogP, logarithmic phase; StatP, stationary phase; MM, minimal medium; LBT, LBT medium. mnmC(m)-G68D mutation than by the null mnmC-W131stop mutation, although this effect is dependent on the codon context (56) . Therefore, it appears that cmnm 5 (s 2 ) confers the cell some advantage over nm 5 (s 2 )U to translate specific mRNAs during exponential growth.

Strikingly, in the competition experiments (Table 7) , the mnmC-W131stop mutation appears to be more disadvantageous than the mnmC(m)-G68D allele, which reveals an opposite trend to that observed in the growth rate determination experiments (Table 6 ). During the transition to the stationary phase, nm 5 (s 2 )U is more abundant in mnmC(m)-G68D than in mnmC-W131stop (Figure 7 , panels B and C). Therefore, the slightly better ability of the mnmC(m)-G68D mutant to compete might indicate that nm 5 (s 2 )U is more effective than cmnm 5 (s 2 )U in translating mRNAs required to sustain stationary phase stress conditions. Thus, it appears that the advantages offered by each intermediate, nm 5 (s 2 )U and cmnm 5 (s 2 )U, depend on the nature of the mRNAs to be translated which, in turn, depend on the growth conditions. Obviously, the synthesis of the final modification mnm 5 s 2 U is more advantageous to the cell as the strains expressing a fully active MnmC protein grow better under stressing conditions (i.e. those related to lack of MnmA and competition with other bacteria) than those strains lacking any MnmC activity. In conclusion, adaptation of MnmEG to use ammonium or glycine, together with the acquisition of MnmC(o) and MnmC(m) activities, provide E. coli with significant advantages to synthesize mnm 5 s 2 U and to survive under different conditions.

Transfer tRNA modification (posting data). Escherichia Coli and Salmonella

Nucleic acids are not boring long polymers of only four types of nucleotides: a guided tour. DNA and RNA Modification Enzymes: Structure, Mechanism, Function and Evolution

Biosynthesis and Function of Posttranscriptional Modifications of Transfer RNAs

Bringing order to translation: the contributions of transfer RNA anticodon-domain modifications

tRNA biology charges to the front

Acute infantile liver failure due to mutations in the TRMU gene

The 2-thiouridylase function of the human MTU1 (TRMU) enzyme is dispensable for mitochondrial translation

Mutations of the mitochondrial-tRNA modifier MTO1 cause hypertrophic cardiomyopathy and lactic acidosis

Reprogramming of tRNA modifications controls the oxidative stress response by codonbiased translation of proteins

Elongator complex influences telomeric gene silencing and DNA damage response by its role in wobble uridine tRNA modification

Enzymology of tRNA modification in the bacterial MnmEG pathway

Further insights into the tRNA modification process controlled by proteins MnmE and GidA of Escherichia coli

Evolutionarily conserved proteins MnmE and GidA catalyze the formation of two methyluridine derivatives at tRNA wobble positions

The Escherichia coli trmE (mnmE) gene, involved in tRNA modification, codes for an evolutionarily conserved GTPase with unusual biochemical properties

The structure of the TrmE GTP-binding protein and its implications for tRNA modification

Crystallization and preliminary X-ray diffraction analysis of the tRNA-modification enzyme GidA from Aquifex aeolicus

Structure-function analysis of Escherichia coli MnmG (GidA), a highly conserved tRNA-modifying enzyme

Crystal structures of the conserved tRNA-modifying enzyme GidA: implications for its interaction with MnmE and substrate

The tRNA-modifying function of MnmE is controlled by post-hydrolysis steps of its GTPase cycle

Mechanistic insights into sulfur relay by multiple sulfur mediators involved in thiouridine biosynthesis at tRNA wobble positions

YibK is the 2'-O-methyltransferase TrmL that modifies the wobble nucleotide in Escherichia coli tRNA

Genetic mapping and cloning of the gene (trmC) responsible for the synthesis of tRNA (mnm5s2U)methyltransferase in Escherichia coli K12

Transfer RNA(5-methylaminomethyl-2-thiouridine)-methyltransferase from Escherichia coli K-12 has two enzymatic activities

Identification of a bifunctional enzyme MnmC involved in the biosynthesis of a hypermodified uridine in the wobble position of tRNA

Sequence-structurefunction analysis of the bifunctional enzyme MnmC that catalyses the last two steps in the biosynthesis of hypermodified nucleoside mnm5s2U in tRNA

Assay of both activities of the bifunctional tRNA-modifying enzyme MnmC reveals a kinetic basis for selective full modification of cmnm5s2U to mnm5s2U

Crystal structure of the bifunctional tRNA modification enzyme MnmC from Escherichia coli

A Short Course in Bacterial Genetics: A laboratory manual and handbook for Escherichia coli and related bacteria

One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products

Novel E. coli mutants deficient in biosynthesis of 5-methylaminomethyl-2-thiouridine

The cysteine desulfurase IscS is required for synthesis of all five thiolated nucleosides present in tRNA from Salmonella enterica serovar typhimurium

Effects of mutagenesis in the switch I region and conserved arginines of Escherichia coli MnmE protein, a GTPase involved in tRNA modification

The Escherichia coli RlmN methyltransferase is a dual-specificity enzyme that modifies both rRNA and tRNA and controls translational accuracy

In vitro study of E.coli tRNA(Arg) and tRNA(Lys) identity elements

Ribonucleoside analysis by reversed-phase high-performance liquid chromatography

Growth rate dependence of transfer RNA abundance in Escherichia coli

A generic protocol for the expression and purification of recombinant RNA in Escherichia coli using a tRNA scaffold

Chaplet column chromatography: isolation of a large set of individual RNAs in a single step

The Era-like GTPase TrmE conditionally activates gadE and glutamate-dependent acid resistance in Escherichia coli

Acetylornithinase of Escherichia coli: partial purification and some properties

Characterization and structure of the Aquifex aeolicus protein DUF752: a bacterial tRNA-methyltransferase (MnmC2) functioning without the usually fused oxidase domain

Recombinant RNA technology: the tRNA scaffold

Characterization of GTPase activity of TrmE, a member of a novel GTPase superfamily, from Thermotoga maritima

A ''gain of function'' mutation in a protein mediates production of novel modified nucleosides

Isolation and partial characterization of three Escherichia coli mutants with altered transfer ribonucleic acid methylases

Mechanisms of acid resistance in Escherichia coli

The modification of the wobble base of tRNAGlu modulates the translation rate of glutamic acid codons in vivo

Regulation of expression and catalytic activity of Escherichia coli RsmG methyltransferase

Antisuppressor mutation in Escherichia coli defective in biosynthesis of 5-methylaminomethyl-2-thiouridine

The social fabric of the RNA degradosome

Do all modifications benefit all tRNAs?

Translational misreading: a tRNA modification counteracts a +2 ribosomal frameshift

The GTPase activity and C-terminal cysteine of the Escherichia coli MnmE protein are essential for its tRNA modifying function

Improvement of reading frame maintenance is a common function for several tRNA modifications

Expression of a coronavirus ribosomal frameshift signal in Escherichia coli: influence of tRNA anticodon modification on frameshifting

Undermodification in the first position of the anticodon of supG-tRNA reduces translational efficiency

We are very grateful to M. Villarroya, S. Prado, R. Ruiz-Partida and E. Knecht for much valuable advice. We also thank the National BioResource Project (NIG, Japan) for providing some of the E. coli strains used in this study. 

Supplementary Data are available at NAR Online.Conflict of interest statement. None declared.