The GTPase Activity and C-terminal Cysteine of the Escherichia coli MnmE Protein Are Essential for Its tRNA Modifying Function*

The Escherichia coli MnmE protein is a three-domain protein that exhibits a very high intrinsic GTPase activity and low affinity for GTP and GDP. The middle GTPase domain, when isolated, conserves the high intrinsic GTPase activity of the entire protein, and the C-terminal domain contains the only cysteine residue present in the molecule. MnmE is an evolutionarily conserved protein that, in E. coli , has been shown to control the modification of the uridine at the wobble position of certain tRNAs. Here we examine the biochemical and functional consequences of altering amino acid residues within conserved motifs of the GTPase and C-terminal domains of MnmE. Our results indicate that both domains are essential for the MnmE tRNA modifying function, which requires effective hydrolysis of GTP. Thus, it is shown for the first time that a confirmed defect in the GTP hydrolase activity of MnmE results in the lack of its tRNA modifying function. Moreover, the mutational analysis of the GTPase domain indicates that MnmE is closer to classical GTPases than to GTP-specific metabolic enzymes. Therefore, we propose that MnmE uses a conformational change associated with GTP hydrolysis -Galactosidase and glucose-6-phosphate dehydrogenase assays performed as described previously (21, 22). Protein Techniques and Production of Antisera— Overproduction and purification of GST fusion proteins was done in DEV16 transformed with plasmid pIC684 or derivatives. Protein purification, SDS-PAGE, and immunoblotting were carried out essentially as described (1). Be- fore use, the anti-MnmE antiserum (1) was affinity purified with nitro-cellulose-bound MnmE. The anti-CyoA antibody (obtained in our labo- ratory by J. C. Escudero) was produced by inoculating the purified C-terminal domain of CyoA (fused to GST) into New Zealand rabbits. NTPase and GTP Binding Assays— NTP hydrolysis was measured by a colorimetric assay for determination of P i release (23, 24). The puri- fied proteins (1.89 (cid:3) M ) were incubated with increasing concentrations (0–5 m M ) of GTP or XTP in 50 (cid:3) l of binding buffer (50 m M Tris-HCl, pH 7.5, 50 m M KCl, 2 m M MgCl 2 , 5% glycerol) at 37 °C during 10 min (a period in which the velocity of P i liberation was constant). Reactions were stopped by adding 200 (cid:3) l of daily prepared malachite green reagent. This reagent contained 2 volumes of 0.0812% malachite green, 2 volumes of bidistilled water, 1 volume of ammonium molybdate (5.72% in 6 N HCl), and 1 volume of 2.32% polyvinyl alcohol. After 2 min at room temperature, 25 (cid:3) l of 34% sodium citrate were added to stop the color development. The samples were read within the next 15 min in a Labsystems Multiskan ® Plus plate reader using a 690-nm filter. Blanks containing the corresponding nucleotide concentrations in binding buffer plus malachite green reagent and citrate were subtracted from each sample. To quantitate the amounts of enzymatically released P i , the samples were compared with a standard curve, which was prepared with dilutions of a 500 (cid:3) M KH 2 PO 4 solution in binding buffer, over a range from 0 to 10 nmol of inorganic phosphate. To determine values for V max and K m , the data were fitted to the Michaelis-Menten equation using non-linear regression (GraphPad Prism version 3.00 for Win-dows; GraphPad Software, Inc.). All nucleotides were from Sigma. The affinity carboxykinase (25, 26, 40). Our results indicate that the substitution of the invariant Asp of the G4 motif by an Asn residue changes specificity of MnmE from guanine to xanthine nucleotides, without affecting the mechanism of nucleoside triphosphate hydrolysis (Table II), as described for classical GTPases. Moreover, evidence that the putative G2 motif is important for the MnmE GTPase activity is provided by the strong effects of alanine substitutions. 2 These data support the idea that MnmE is closer to GTPases that use GTP hydrolysis to function as molecular switches than to GTP-specific metabolic enzymes. Classical GTPases, like Ras-related proteins, are in an active state when GTP-bound. The binding of GTP causes a conformational change in the proteins that allows interaction with a target (effector) molecule. Upon GTP hydrolysis, they become inac-tive. In contrast, other GTPases, like EF-G (41, 42) and dy-namins (43), have been proposed to use GTP hydrolysis to promote structural rearrangements that convert the protein to its active state. Thus, it is possible that GTP hydrolysis by MnmE leads to changes of the overall protein conformation, which, in turn, promote subsequent steps of the modification reaction, possibly led by the MnmE C-terminal domain and GidA (see below). This model does not exclude that MnmE, when GTP-bound, may recruit effector(s) of the modification pathway; these could be, for instance, the still unknown cmnm group donor, GidA, or tRNA.

The evolutionarily conserved MnmE (TrmE) protein of Escherichia coli is a GTPase that differs extensively from regulatory GTPases such as p21 (1). Thus, MnmE exhibits a very high intrinsic GTPase hydrolysis rate and low affinity for GTP and GDP, and it can form self-assemblies. MnmE has a molecular mass of 50 kDa and is organized as a multidomain protein (see Fig. 1) consisting of an ϳ220-amino acid N-terminal domain, probably required for self-assembly, a middle GTPase domain, of about 160 residues, and an ϳ75-amino acid C-terminal domain, which contains the only Cys residue present in the protein. Strikingly, the isolated GTPase domain roughly conserves the guanine nucleotide binding and GTPase activities of the intact MnmE molecule (1).
Null mnmE mutants are defective in the biosynthesis of the hypermodified nucleoside 5-methylaminomethyl-2-thiouridine (mnm 5 s 2 U 34 ) 1 (2), which is found in the wobble position (position 34) of tRNAs that read codons ending with A or G, in the mixed codon family boxes, specifically tRNAs for lysine and glutamic acid (3). The synthesis of mnm 5 s 2 U 34 is complex and requires several enzymatic reactions (4,5), shown below in Reaction 1. The mnmA gene product (6) carries out thiolation in the 2-position of the wobble uridine (U 34 ), whereas mnmE controls the first step of the modification in the 5Ј-position, but it is unclear how many steps precede the formation of cmnm 5 s 2 U 34 (2,4). Several data support that a second gene, named gidA, trmF, or mnmG, is also involved in the cmnm 5 group addition and that the MnmE activity precedes the activity of GidA (2,5). The mnmC gene product has two enzymatic activities that transform the cmnm 5 intermediate into the final mnm 5 modification (4). The 2-thio group, but not the cmnm 5 group, is important for aminoacylation of tRNA Glu and tRNA Lys (2,3), whereas modification in both 2-and 5-positions function in the codon recognition process (3,5,(7)(8)(9)(10)(11).
MnmE is also involved in modification of tRNA 1 Gln , but, in this case, the wobble base modification might be different from mnm 5 s 2 U 34 (3). In addition, the non-thiolated derivatives cmnm 5 U 34 and mnm 5 U 34 are present in tRNA 4 Leu and tRNA 4 Arg , respectively, and their formation may be also mediated by MnmE (7).
The precise role of MnmE in the cmnm group addition is unknown. The only biochemical activity of this protein reported to date is the GTPase activity (1), but whether it is required for the tRNA modifying function has not been analyzed yet. Given that MnmE is a multidomain protein, the possibility that different domains are involved in different cellular functions cannot be discarded. This would not be something new, because, for instance, the trmA gene product (RUMT) has been proposed to have two different activities associated with different regions of the protein, the synthesis of m 5 U 54 (a non-essential modification that is present in all E. coli tRNA species) and another, yet unknown, essential function (12,13). Subcellular fractionation followed by immunoblotting, as well as immunoelectron microscopy, revealed that MnmE is a cytoplasmic protein partially associated with the inner membrane (1). Interestingly, the C-terminal domain of MnmE and its homologues contain the tetrapeptide motif C(I/L/V)GK at the extreme C terminus, which shows good homology with the CAAX motif (where "A" represents an aliphatic residue, and "X" represents any residue) characteristic of the Ras proteins of eukaryotic cells (14). The C-terminal cysteine might be involved in the membrane association of MnmE, as it occurs with other GTPases containing C-terminal cysteines, or in the putative tRNA modifying activity of this protein, because cysteine residues have been shown to be responsible for methyl transferase activity of some tRNA modifying enzymes (15). To unravel its functional role, as well as to investigate whether the cycle of GTP/GDP exchange regulates the subcellular distribution of MnmE, the construction of point mutations affecting the C-terminal Cys or the GTPase activity of MnmE may be of great value.
MnmE is homologous to the mitochondrial protein encoded by MSS1, a nuclear gene of Saccharomyces cerevisiae, whose product has been proposed to play a role in some aspect of mitochondrial translation (16,17). In E. coli, we have found that null mnmE mutations are lethal depending on the genetic background (1). Interestingly, mutations in MSS1 render the yeast cells respiratory-deficient only in the presence of a paromomycin-resistance mutation (P R 454 ), which resides in the mi-tochondrial 15 S rRNA gene and may alter the conformational changes involved in the decoding process (16,18,19). Altogether these results led us to postulate that the lack of the mnmE gene product may be tolerated by E. coli cells depending on features of other elements involved in translation (1). However, because MnmE might be a multifunction protein, it is not sure that the lethality conferred by null mnmE mutations in certain genetic backgrounds is directly related with the impairment of the tRNA modifying function. Therefore, isolation of missense mnmE mutations affecting this function is important to determine whether it is essential for E. coli. The purpose of this study was to elucidate the role of the GTPase domain and the C-terminal Cys of MnmE in the tRNA modifying function of this protein by constructing specific missense mutations and to analyze the effect of such mutations on the subcellular localization of the protein and cell viability.
Media, Growth Conditions, and Enzyme Assays-LBT (Luria Bertani broth containing 40 g/ml thymine) and LAT (LBT containing 20 g of Difco agar per liter) were used for routine cultures and plating of E. coli. Antibiotics were added as recommended (21). IC4126 and derivatives were grown at 30°C and plated at 30 or 42°C for viability tests. ␤-Galactosidase and glucose-6-phosphate dehydrogenase assays were performed as described previously (21,22).
Protein Techniques and Production of Antisera-Overproduction and purification of GST fusion proteins was done in DEV16 transformed with plasmid pIC684 or derivatives. Protein purification, SDS-PAGE, and immunoblotting were carried out essentially as described (1). Before use, the anti-MnmE antiserum (1) was affinity purified with nitrocellulose-bound MnmE. The anti-CyoA antibody (obtained in our laboratory by J. C. Escudero) was produced by inoculating the purified C-terminal domain of CyoA (fused to GST) into New Zealand rabbits.
NTPase and GTP Binding Assays-NTP hydrolysis was measured by a colorimetric assay for determination of P i release (23,24). The purified proteins (1.89 M) were incubated with increasing concentrations (0 -5 mM) of GTP or XTP in 50 l of binding buffer (50 mM Tris-HCl, pH 7.5, 50 mM KCl, 2 mM MgCl 2 , 5% glycerol) at 37°C during 10 min (a period in which the velocity of P i liberation was constant). Reactions were stopped by adding 200 l of daily prepared malachite green reagent. This reagent contained 2 volumes of 0.0812% malachite green, 2 volumes of bidistilled water, 1 volume of ammonium molybdate (5.72% in 6 N HCl), and 1 volume of 2.32% polyvinyl alcohol. After 2 min at room temperature, 25 l of 34% sodium citrate were added to stop the color development. The samples were read within the next 15 min in a Labsystems Multiskan ® Plus plate reader using a 690-nm filter. Blanks containing the corresponding nucleotide concentrations in binding buffer plus malachite green reagent and citrate were subtracted from each sample. To quantitate the amounts of enzymatically released P i , the samples were compared with a standard curve, which was prepared with dilutions of a 500 M KH 2 PO 4 solution in binding buffer, over a range from 0 to 10 nmol of inorganic phosphate. To determine values for V max and K m , the data were fitted to the Michaelis-Menten equation using non-linear regression (GraphPad Prism version 3.00 for Windows; GraphPad Software, Inc.). All nucleotides were from Sigma.
The affinity of wild-type and mutant MnmE proteins for [␥-35 S]GTP (Ͼ37 MBq/mmol; Amersham Biosciences) was determined by using a nitrocellulose filter binding assay as described previously (1). The assays were done at 30°C with a protein concentration equal to 4.7 M in 25-l reactions. K d values were calculated by fitting the curves with non-linear regression (one binding site hyperbola), using GraphPad Prism software.
Limited Proteolysis-The purified proteins (1 g) were incubated with the indicated nucleotide concentrations in 18 l of digestion buffer (0.1 mM ammonium bicarbonate, pH 7.8, 2 mM MgCl 2 , 50 mM KCl) for 5 min at room temperature. Proteolysis reactions were started by adding 2 l of N␣-p-tosyl-L-lysine cloromethyl-ketone-treated chymotrypsin (12.5 ng/l; Sigma) at 30°C. After a 20-min incubation, reactions were stopped with the addition of phenylmethylsulfonyl fluoride (1 mM final concentration) and 2ϫ Laemmli sample buffer, heated 3 min at 95°C, and subjected to 20% SDS-PAGE. The gels were transferred to nitrocellulose membranes (Hybond ECL; Amersham Biosciences), and the resulting bands were identified by Western blot with anti-MnmE antibody. When using endopeptidase Glu-C (Sigma), the incubation was performed for 1 h at 37°C with 50 ng of the enzyme.
In Vivo Formaldehyde Cross-linking, Cellular Fractionations, and Immunoelectron Microscopy-Cross-linking with formaldehyde was carried out as described previously (1). For subcellular localization of MnmE proteins, cells were collected in 10 mM Tris buffer, pH 8.0, 10 mM MgCl 2 , 10 mM KCl plus a mixture of protease inhibitors (1 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, 1 M aprotinin, 10 M bestatin, 100 M leupeptin, 1 M pepstatin A) and disrupted by sonication. Cell debris were removed from the lysate by centrifugation (4,400 ϫ g for 15 min, 4°C). Cell envelopes were separated from the soluble fraction (containing cytoplasmic and periplasmic proteins) by ultracentrifugation (150,000 ϫ g for 60 min, 4°C). The pellet (membrane fraction) was washed once as above and resuspended in ice-cold 10 mM Tris, pH 8.0, 10 mM MgCl 2 , 10 mM KCl, and inhibitors before storage at Ϫ20°C. Fractions were tested by Western blot analysis with anti-MnmE antiserum (or anti-CyoA antiserum as a control for the membrane fraction). Immunoelectron microscopy and morphometric analysis were performed as described previously (1).
Readthrough Measurements-Misreading of the UAG stop codon carried by the lacZ105 gene was monitored by using the ␤-galactosidase assay (2). Cultures were grown in LBT containing 1 mM isopropyl-1thio-␤-D-galactopyranoside and 1 mM cAMP and maintained at least four generations in exponential growth by repeated dilution with the same medium before being assayed. ␤-Galactosidase assays were performed on SDS-chloroform-permeabilized cells as described (21). Special precautions for very low ␤-galactosidase activities were taken (2,5).
Analysis of tRNA Glu Modification by Northern Blots-Total tRNA from strains DEV16, IC4767, IC4768, IC4769, IC4770, and IC4640 grown to an A 600 of ϳ0.6 -0.8 was extracted and deacylated (3). The tRNA was quantified by absorbance measurement at A 260 nm with an Unicam UV-visible spectrophotometer (Helios-␤). Northern blots were performed essentially as described (3), with the following modifications: the electrophoresis was run for 2.5 h at 100 V and 40°C through a 10-cm-long, 0.75-mm-thick, 8% (w/v) polyacrylamide gel. The area between the two dyes was electroblotted onto a Nylon membrane positively charged. Membranes were pre-hybridized at 50°C for 1 h in DIG Easy Hyb (Roche Applied Science). To detect the tRNA Glu , the following digoxigenin-labeled probe was used: 5Ј-DIG-CCCCTGTTACCGCCGT-GAAAGGGCGGTGTC-3Ј. Hybridization was performed for 6 h at 50°C in DIG Easy Hyb containing the probe at a final concentration of 3.3 nM. After post-hybridization washes (2 ϫ 15 min at 25°C with 2ϫ SSC, 0.1% SDS and 2 ϫ 15 min at 48°C with 0.5ϫ SSC, 0.1% SDS), chemiluminescent detection with CDP-Star (Roche Applied Science) was carried out according to the manufacturer's protocol.

RESULTS
Biochemical Characterization of Mutants-To analyze the importance of the GTPase and C-terminal domains in the known functions and properties of MnmE, we separately changed three of the conserved amino acids in the GTPase domain, as well as the Cys located in the C-terminal domain (Cys-451; see Fig. 1). Three-dimensional structures of members from the G protein superfamily indicate that the conserved residues of the G domain are invariably involved in binding of guanine nucleotides, hydrolysis of GTP, or controlling the conformational switch (25,26). Residues in G1 (GXXXXGK(S/T)) are involved in binding the charged phosphate groups of the nucleotide. The sixth residue in this motif is always a Gly; a side chain in this position would interfere sterically with the guanine part of the nucleotide. The Asp residue of G3 (DXXG) participates in binding the magnesium ion via a water molecule. In GTP-binding proteins, Mg 2ϩ plays a crucial role in binding of substrate and catalytic activity; this ion is also needed for the GTP hydrolysis mediated by MnmE (1). Finally, the Asp residue of G4 ((N/T)KXD) is one of the elements responsible for the high specificity of G proteins for guanine nucleotides. The exchange of Asp for Asn leads to a lower affinity for GTP and to a higher affinity for XTP (25)(26)(27). Accordingly, we generated mutations G228A, D270A, and D338N in the G 1 , G 3 , and G 4 motifs, respectively, of the MnmE GTPase domain (see Fig. 1). In addition, we generated mutation C451S, because substitution of Cys by Ser at the C-terminal CAAX motif of Ras proteins is known to prevent their membrane association (14).
To determine the enzymatic parameters of GTP hydrolysis by purified MnmE proteins, the substrate concentration was varied from 0 to 5 mM, and the inorganic phosphate produced in the reaction was quantified. Fig. 2A indicates that proteins G228A and D270A are defective in the GTPase activity whereas D338N is able to work as a GTP hydrolase, especially at high GTP concentration. Protein C451S displays a GTPase activity similar to that of the wild-type MnmE protein, as expected.
To analyze whether the substitution of the invariant Asp of the G4 motif by an Asn residue indeed changes specificity of MnmE from guanine to xanthine nucleotides, as described for classical GTPases, we analyzed the substrate concentration dependence of the D338N NTP hydrolase activity. Fig. 2B shows that D338N efficiently hydrolyzes XTP and requires concentrations of XTP lower than GTP for half-maximal reaction velocity, which strongly suggests that affinity of D338N for XTP is higher than for GTP and that the NKAD tetrapeptide is the guanine specificity determinant of MnmE.
On the other hand, affinity of MnmE mutants for guanine nucleotides was determined by a nitrocellulose filter assay. Fig.  2C shows that proteins G228A and D338N are impaired for GTP binding, as expected, whereas D270A maintains a meaningful affinity for GTP, in contrast to predictions (see below). Binding of C451S to GTP was similar to that of the wild-type protein.
The top portion of Table II summarizes the kinetic parameters and the dissociation constants (K d ) for GTP of mutant and wild-type proteins. The V max values for G228A and D270A were extremely low, whereas the V max values for D338N and C451S were similar to those of the wild-type protein. Note, however, that D338N required very high substrate concentrations for half-maximal reaction velocity (K m ϳ2100 M), which is in agreement with the very low affinity of this mutant for GTP (K d ϳ2800 M). Kinetic analysis revealed that D338N hydrolyzes GTP and XTP with similar rate constants (10.5 and 9.5 min Ϫ1 , respectively; see Table II), which indicates that mutation D338N does not affect the mechanism of nucleoside triphosphate hydrolysis of MnmE. However, the K m for XTP hydrolysis was 10-fold lower than that observed for GTP hydrolysis (206 versus 2126 M, respectively), which supports the idea that affinity of D338N for XTP is higher than for GTP (specificity constants, K cat /K m , were 0.046 and 0.005 M Ϫ1 min Ϫ1 , respectively). The K cat for XTP hydrolysis by the wild-type MnmE protein was only moderately reduced in comparison with that for GTP hydrolysis (5.3 and 10.2 min Ϫ1 , respectively). Curiously, the specificity constants (K cat /K m ) for XTP and GTP exhibited by the wild-type protein were similar (ϳ0.02 M Ϫ1 min Ϫ1 ), which suggests that MnmE binds XTP better than expected. Usually, the affinity for GTP of the wild-type versions of other GTPases is significantly higher than that for XTP because of the characteristic bifurcated hydrogen bond between the Asp residue of the G4 motif and the guanine base (27)(28)(29)(30). Thus, it is possible that in the wild-type MnmE protein certain residues that remain to be identified may contribute to the binding of XTP and compensate for the unfitting between Asp-338 and XTP.
The top portion of Table II also shows that mutant D270A has an affinity for GTP in the same range as the wild-type protein. This is in contrast with data reported from the similar D57A mutant of p21 ras , in which the exchange of Asp for Ala in the G3 motif reduced the affinity for GTP about 100-fold (31). If, as it occurs with other GTPases, the Asp residue of the MnmE G3 motif (Asp-270) is involved in binding the magnesium ion via a water molecule, our data suggest that mutation D270A does not produce great disturbance of the nucleotide binding site but impairs the GTPase activity maybe by preventing proper coordination of Mg 2ϩ . It should be noted that the B max value for GTP␥S of D270A was lower than that of the wild-type protein (Fig. 2C). It is possible that improper coordination of Mg 2ϩ alters the kinetics of the interaction of D270A with GTP (although the equilibrium dissociation constant K d would remain unchanged). In this respect, it is worthy to point out that the similar D57A mutation of p21 ras was reported to increase both the dissociation and association rate constants of the protein with GTP, although the effect was greater on the first one (31). An increase in the GTP dissociation rate constant of the MnmE mutant D270A may cause a technical problem concerning the evaluation of its GTP binding capacity by means of the filter binding assay (32). Such an increase could favor partial loss of the nucleotide during filter washing and produce lower B max values. Alternatively, these values may be because of a higher instability of D270A in relation to the wild-type protein. If so, they would indicate that, compared with wildtype MnmE, only around 30% of D270A is active for GTP binding; accordingly, a correction factor should be applied to the K cat value indicated in Table II. However, note that even so, the K cat of D270A remains significantly lower than that of the wild-type protein. Therefore, independently of which were the cause(s) of the low B max value for D270A, our results suggest that this mutant can bind but not effectively hydrolyze GTP.
Finally, G228A was the mutation with the most drastic effect on the interaction between MnmE and GTP. In all binding experiments, protein G228A systematically showed an affinity for GTP even lower than that displayed by D338N (see Fig. 2C and the top portion of Table II), and this feature might explain its extremely low GTPase activity. We must realize that, given the rather low affinity of MnmE proteins for guanine nucleotides, the filter binding assay is not the best procedure to finely calculate the K d values, although, in our hands, it has been good enough to obtain a satisfactory picture on the GTP binding capabilities of the mutants used in this work (see next paragraph).
Nucleotide-induced Conformational Change as Assessed by Limited Proteolysis-We used limited proteolysis to detect conformational changes of the wild-type and mutant MnmE proteins associated with their binding to nucleotides. As shown in  4 and 5), which indicates that MnmE binds to guanine nucleotides, but not to ATP, as reported previously by means of a different approach (1). Note that patterns produced by binding of MnmE to GTP and GTP␥S (which were used at saturating concentrations) were similar and characterized by the appearance of a triplet of ϳ28 kDa and one band of ϳ21 kDa (lanes 4 and 5). Fig. 3B shows that the proteolysis pattern of mutants G228A, D270A, D338N, and C451S was, in the absence of GTP, similar to that produced by the wild-type MnmE protein, which suggests that the tertiary structure of these proteins is roughly equivalent. When GTP was present during proteolysis, only the pattern of G228A remained unchanged, supporting the idea that this mutant is impaired for binding to GTP. Fig. 3C shows that binding of the wild-type MnmE protein to 5 mM GTP or XTP produces identical proteolysis patterns (lanes 2 and 3), whereas binding of D338N to GTP produces a pattern weaker than that resulting from its binding to XTP (compare triplets and ϳ21-kDa bands of lanes 6 and 7), in agreement with the different affinity of this mutant for both nucleotides (see Table II). The low affinity of D338N for guanine nucleotides is also revealed by the proteolysis pattern produced by binding of this mutant to 1 mM GTP␥S, which is remarkably weaker than that obtained with the wild-type MnmE protein (compare lanes 4 and 8) or mutant D270A (compare lanes 8 and 10). These data are consistent with those from nucleotide binding assays showing that D270A has a higher affinity for GTP␥S than D338N (see Fig. 2C and Table II).
Altogether these results indicate that the capability to bind GTP is severely or partially impaired in mutants G228A and D338N, respectively, whereas it remains roughly unaltered in mutants D270A and C451S. Similar conclusions were achieved when susceptibility of MnmE proteins to endopeptidase Glu-C was examined (data not shown).
Assembly and Subcellular Localization of the MnmE Mutant Proteins-In a previous work (1), we showed that MnmE is able to multimerize in vivo. When wild-type cells treated with 1% formaldehyde were analyzed by SDS-PAGE and immunoblotting, two bands with apparent molecular mass of 100 and 150 kDa were detected. These bands could correspond to dimers and trimers of MnmE, because capability of this protein to form self-assemblies was shown by in vitro chemical cross-linking and gel filtration analysis (1). Alternatively, they could result from interaction of MnmE with other molecules present in the bacterial cell.
Here we have analyzed the effect that mutations G228A, D270A, D338N, and C451S transferred onto the E. coli chromosome have on the ability of MnmE to multimerize in vivo. As shown in Fig. 4A, the 100-and 150-kDa complexes were detected when strains carrying the mutated alleles as unique source of MnmE protein were cross-linked with 1% formaldehyde but not when the treatment was applied to strain DEV16,  11 and 12) and that its disappearance is not related with the detection of the complexes. Thus, the in vivo cross-linking experiments shown in Fig. 4A indicate that neither the GTPase activity of MnmE nor the single Cys of this protein are involved in the formation of the 100-and 150-kDa complexes. Curiously, the 150-kDa band was detected in both the wild-type and mutant cells immediately after the addition of formaldehyde, whereas detection of the 100-kDa complex requires longer incubation (1) (data not shown). We suggested previously (1) that the rapid rate of fixation of the 150-kDa complex by formaldehyde treatment of the wild-type strain may be indicative of a peripheral location of this complex in the cell. In fact, immunoelectron microscopy and subcellular fractionation followed by immunoblotting indicated that MnmE is localized in both the cytoplasm and, to a lesser but significant extent, the inner membrane (1). In this respect, the rapid rate of fixation of the 150-kDa complex in the mnmE mutants suggests that the subcellular distribution of MnmE does not depend on its GTPase activity or C-terminal Cys. Supporting this proposal, Fig. 4B shows that mutations G228A and C451S do not alter the subcellular fractionation pattern of MnmE in relation to that observed in the wild-type strain. It should be pointed out that activity of glucose-6-phosphate dehydrogenase, used as a cytoplasmic marker, was negligible in the membrane fraction (lanes 3), indicating little, if any, cross-contamination with the soluble fraction (lanes 2). Similar results were obtained from studies on the subcellular distribution of the wild-type, G228A, and C451S MnmE proteins by fractionation followed by immunoblotting and quantitative immunoelectron microscopy of strain DEV16 (null mutant) carrying the mnmE alleles on a low copy plasmid (data not shown). Therefore, we conclude that the cycle of GTP/GDP exchange does not regulate the subcellular distribution of MnmE and that the C-terminal Cys is not essential for association of MnmE to the cell inner membrane.
The GTPase Activity and C-terminal Cys of MnmE Are Needed for Its tRNA Modifying Function-MnmE-mediated modification is known to be required for efficient readthrough of stop codons by certain suppressor tRNAs (2,7). Thus, null mnmE mutants exhibit reduced readthrough of a leaky UAG stop codon present in a mutant lacZ gene (lacZ105), which is easily detected on XG plates as a change from pale-blue colonies to white colonies or by means of the ␤-galactosidase assay as a reduction in an already low ␤-galactosidase activity (2). We made use of this phenotype to investigate the role of the GTPase and C-terminal domains of MnmE in the tRNA modifying function of this protein; specifically, we compared the effect of mutations G228A, D270A, D338N, and C451S on the misreading of the lacZ UAG codon with that produced by mnmE null mutants. Substitution of the Q192X allele on the DEV16 chromosome by alleles G228A, D270A, and C451S did not produce any color change in the recovered colonies on XG plates (supplemented with kanamycin), whereas the same substitution by alleles D338N or wild-type resulted in pale-blue colonies. The effect of such substitutions was quantified by means of the ␤-galactosidase assay. As shown in Fig. 5A, mutations G228A, D270A, and C451S reduced the UAG misreading to the same extent than Q192X, whereas mutation D338N reduced it to a lesser extent (about 42% of the wild-type level). These results suggest that the tRNA modifying function of MnmE is completely impaired by mutations G228A, D270A, and C451S but only partially by D338N.
To test the above conclusion, we directly analyzed, by using electrophoresis on acidic polyacrylamide gel and Northern blot hybridization, the modification state of tRNA Glu isolated from wild-type and mutant mnmE strains. The mnm 5 group adds a positive charge to the tRNA at pH 5.0, and thus tRNA containing this group migrates slower than tRNA lacking it (3). As shown in Fig. 5B, tRNA Glu from the G228A, D270A, C451S, and Q192X mutants migrated faster than that isolated from the wild-type and D338N strains, indicating that only tRNA Glu from the last ones carries the mnm 5 group. In these experiments, tRNAs were deacylated by base treatment, but similar results were obtained with acylated tRNAs (data not shown). Western blot analysis indicated that the cellular levels of the mutant proteins were indistinguishable from that of the wildtype MnmE (Fig. 5C). Therefore, the effect produced by G228A, D270A, and C451S cannot be attributed to a lower accumulation of these proteins into the cell. Altogether these results lead us to conclude that both mutations G228A and D270A, which impair the GTPase activity of MnmE, as well as mutation C451S, which affects the C-terminal domain of MnmE, make this protein deficient for the tRNA modification function. On the contrary, mutation D338N, which partially impairs GTP binding but not the hydrolase activity, renders the cell capable of tRNA modification. We cannot discard the possibility that in strain D338N, modified and hypomodified forms of tRNA Glu co-exist but migrate as a sole band in our gels (that corresponding to the modified form; see Fig. 5B) or that the amount of the hypomodified tRNA may be under the limit of detection of our approach (note that we used a non-radioactive probe in blots shown in Fig. 5B). If so, these results would be compatible with those indicating that mutation D338N reduces the readthrough of the UAG codon to about 42% of the wild-type level (Fig. 5A).
Cell Viability of mnmE Mutants Is Associated with Their Capability to Modify tRNA-We have shown previously (1) that null mnmE mutations are lethal in the genetic backgrounds of strains JC7623 and V5701 but not in those of MC1000 and DEV16. To analyze whether synthetic lethality is associated with a defective tRNA modifying function of MnmE, and not to any other putative function of this protein, we first introduced the missense mnmE alleles on the IC3647 chromosome by means of a linear transformation method (1,33). For this purpose, we used plasmid pIC914 and derivatives, which carry the mnmE allele together with a Kan R determinant inserted into the adjacent tnaA gene. Plasmids were linearized and transformed into IC3647. This strain is JC7623 (recBC sbcBC) containing an additional copy of the wild-type mnmE gene on a prophage (IC718), which is inserted at minute 17 on the chromosome (34) and provides a functional MnmE protein.
Recombination between plasmids and chromosome allowed the formation of IC3647 Kan R derivatives carrying the different mnmE alleles placed into the mnmE locus, at minute 83 on the chromosome (34), together with the Kan R determinant. Then, bacteriophage P1 lysates grown on these strains were used to transduce each mnmE allele into JC7623 and V5701, as well as into their respective derivatives IC3647 and IC4126, which contain a second copy of the wild-type mnmE gene on a prophage (IC3647) or a plasmid (IC4126). Selection for resistance to kanamycin was performed; the co-transduction frequency of mnmE and the Kan R determinant was expected to be close to 100%, because both genes are separated by only ϳ1000 bp in our constructions (35). As shown in Table III, when JC7623 and V5701 were the recipients, Kan R transductants were solely obtained from infections with P1 lysates bearing the wild-type and D338N alleles. In contrast, more than 100 Kan R transductants were obtained in each case when transduction experiments were driven using IC3647 and IC4126 as recipients.
These results indicate that substitution of the wild-type mnmE allele by alleles G228A, D270A, and C451S may only occur in the presence of a second copy of the gene that provides a functional MnmE protein, which strongly suggests that these alleles are lethal in the genetic backgrounds of JC7623 and V5701. To further analyze whether survival of the IC4126 transductants carrying alleles G228A, D270A, and C451S requires the complementary mnmE ϩ gene on the helper plasmid, which is temperature-sensitive for replication, an appropriate clone from each transduction was selected, made recA Ϫ , and used to grow exponential cultures at 30°C. Dilutions of these cultures were then spread on LAT plates at 30 and 42°C. As shown in Table III, survival at 42°C of IC4126 derivatives carrying alleles G228A, D270A, and C451S was 3 orders of magnitude lower than that of derivatives carrying the wildtype and D338N alleles. Moreover, only ϳ5% of the survivors at 42°C from the last ones were Cm R (a plasmid-dependent phenotype), which indicates that pIC755 has been lost in most of them, as expected from strains carrying non-lethal alleles of mnmE. In contrast, most of the surviving colonies at 42°C recovered from plating IC4126 derivatives carrying alleles G228A, D270A, and C451S were Cm R , which means that they were able to grow at the non-permissive temperature only after integration of plasmid into the bacterial chromosome (note that plasmid integration occurred at very low frequency, because strains were recA Ϫ ). Furthermore, the few Cm s colonies that survived at 42°C were shown, by appropriate PCR analysis, to carry a wild-type mnmE gene, which indicated that this allele had been rescue from pIC755 by recombination, despite the recA Ϫ genotype of the strains. All these data strongly suggest that the plasmid-born mnmE ϩ copy is required for growth of V5701 derivatives carrying mutations G228A, D270A, and C451S.
In brief, our results support the notion that mutations G228A, D270A, and C451S are deleterious in the genetic backgrounds of JC7623 and V5701. Because these mutations impair the modifying function of MnmE (see Fig. 5, A and B), we conclude that the absence of this function is responsible for the synthetic lethality observed in the mentioned backgrounds.
It is of interest to mention that the V5701 derivative carrying mutation D338N on the chromosome grew more slowly than the wild-type isogenic strain (data not shown). This behavior could be related to the slight impairment of the tRNA modification function conferred by mutation D338N, as deduced from the readthrough experiments (Fig. 5A). Both features may result from the very low affinity of protein D338N for GTP. It is thought that the intracellular concentration of GTP in E. coli ranges from about 1 to 4 mM (28), but free concentrations of NTPs are not known with certainty (36). In this respect, viability of V5701 carrying mutation D338N (K m,GTP , 2126 M) suggests that either the intracellular concentration of GTP is high enough to allow protein D338N to work, or some other cellular factor positively regulates the binding of MnmE to GTP. In any case, because the behavior of D338N differs in some extent from the wild-type strain, it is possible that fully modified and hypomodified tRNAs co-exist in D338N mutants, and this could explain both the slow growth of V5701 D338N and the partial readthrough observed in the DEV16 derivative carrying allele D338N. DISCUSSION The E. coli MnmE protein is involved in the modification of the uridine in the wobble position (U34) of the tRNAs that read codons ending with A or G, in the mixed codon family boxes, i.e. tRNA Lys , tRNA Glu , tRNA 1 Gln , tRNA 4 Leu , and tRNA 4 Arg (7,37). In the modification pathway of these tRNAs, MnmE controls, together with GidA, the addition of the cmnm group in position 5 of the U34 (2,4,5). The phenotype conferred by gidA mutations to wild-type or mnmE strains has suggested that the GidA activity follows that of MnmE in the formation of the cmnm group (5).
mnmE and gidA are both evolutionarily conserved genes. In S. cerevisiae, homologues of MnmE and GidA (named Mss1p and Mto1p, respectively) are nuclear-encoded mitochondrial proteins that seem to form a complex (16,17), which supports the idea that both proteins act in the same pathway. Alignment of the E. coli MnmE and GidA proteins with their human homo- logues displays a high amino acid sequence conservation (39.7 and 49% of identity, respectively) and size similarity (38,39).
E. coli has been shown to be an excellent model organism for the study of evolutionarily conserved proteins and functions. Here, we have used the E. coli MnmE protein to gain information about the role played by the GTPase and C-terminal domains of this protein on its tRNA modifying function.
The tRNA Modifying Function of MnmE Is Coupled to GTP Hydrolysis-To determine the importance of the GTPase domain in the known functions and properties of the E. coli MnmE protein, and lacking any structural information, we chose to make several point mutations in the GTP binding motifs, concentrating on residues that are thought to be involved in GTP binding or, indirectly (through coordination of Mg 2ϩ ), in GTP catalysis (see Fig. 1).
As summarized in Table II, proteins G228A and D338N are completely or partially impaired for binding to GTP, respectively, whereas D270A maintains a meaningful affinity for GTP in comparison with mutant D338N (see also Fig. 2C). However, D338N is able to work as a GTP hydrolase whereas D270A is defective in the GTPase activity. The fact that the latter mutant does not modify tRNA (Fig. 5, A and B) supports the conclusion that effective GTP hydrolysis by MnmE, and not simply GTP binding, is necessary for tRNA modification.
MnmE could use energy from GTP hydrolysis to catalyze some step in the cmnm group addition reaction. However, this would make MnmE different from typical GTPases, which use hydrolysis of GTP to function as molecular switches that control cellular events. Alignment of GTP-binding proteins shows that MnmE has the typical arrangement of GTP binding motifs, with G1 through G4 (1). This does not occur in the few GTPases that use GTP as energy source, such as adenylosuccinate synthetase or phosphoenolpyruvate carboxykinase (25,26,40). Our results indicate that the substitution of the invariant Asp of the G4 motif by an Asn residue changes specificity of MnmE from guanine to xanthine nucleotides, without affecting the mechanism of nucleoside triphosphate hydrolysis (Table  II), as described for classical GTPases. Moreover, evidence that the putative G2 motif is important for the MnmE GTPase activity is provided by the strong effects of alanine substitutions. 2 These data support the idea that MnmE is closer to GTPases that use GTP hydrolysis to function as molecular switches than to GTP-specific metabolic enzymes. Classical GTPases, like Ras-related proteins, are in an active state when GTP-bound. The binding of GTP causes a conformational change in the proteins that allows interaction with a target (effector) molecule. Upon GTP hydrolysis, they become inactive. In contrast, other GTPases, like EF-G (41, 42) and dynamins (43), have been proposed to use GTP hydrolysis to promote structural rearrangements that convert the protein to its active state. Thus, it is possible that GTP hydrolysis by MnmE leads to changes of the overall protein conformation, which, in turn, promote subsequent steps of the modification reaction, possibly led by the MnmE C-terminal domain and GidA (see below). This model does not exclude that MnmE, when GTP-bound, may recruit effector(s) of the modification pathway; these could be, for instance, the still unknown cmnm group donor, GidA, or tRNA.
Cys-451 Is Critical for the tRNA Modifying Function of MnmE-The tetrapeptide motif 451 C(I/L/V)GK at the extreme C terminus of MnmE is present in all MnmE/Mss1 proteins of data banks, which suggests that it plays an important and conserved role.
Although the 451 C(I/L/V)GK motif is reminiscent of the CAAX motif of the Ras proteins (which anchors these proteins to cell membranes in eukaryotic cells), we found no evidence that Cys-451 is needed for association of MnmE to the inner membrane. Our experiments involved subcellular fractionation followed by immunoblotting (Fig. 4B), as well as immunoelectron microscopy (data not shown). Moreover, Cys-451 is not required for the in vivo formation of the 100-and 150-kDa complexes (Fig. 4A). Therefore, another role is likely to exist for this conserved residue.
Our results clearly show that the conversion of Cys-451 into Ser disrupts the tRNA modifying function of MnmE, even though the GTPase activity remained unaffected (see Fig. 5, A and B and Table II). Moreover, readthrough experiments like those shown in Fig. 5A have indicated that the conversion of Cys-451 into Ala also extensively impairs the MnmE function (data not shown). This rules out the possibility that the effect of C451S is because of a polarity change mediated by the serine hydroxyl group. Because the change Cys to Ser is structurally conservative, the inability of Ser to replace Cys-451 suggests a role for the thiol group of this cysteine. As far as we know, there is no evidence supporting that MnmE is directly involved in the modification reaction; however, it is tempting to propose that MnmE is a tRNA modifying enzyme and that Cys-451 functions as a catalytic residue in the modification reaction. The addition of the cmnm group to C5 of U34 could take place through a mechanism similar to that proposed for known pyrimidine C5 modifying enzymes such as thymidylate synthase and RUMT, which use an enzymatic cysteine to activate pyrimidine carbon 5 for nucleophilic attack (44). In the biosynthesis of DNA, the enzyme thymidylate synthase is solely responsible for the generation of dTMP from dUTP. This occurs through methylation of C5, utilizing 5,10 methylenetetrahydrofolate as the methyl donor. In the case of the MnmE-dependent reaction, the cmnm group originates from a source other than S-adenosylmethionine that has not been identified yet, although it might be formyltetrahydrofolate (2,4). Interestingly, the RUMTs from Streptococcus faecalis and Bacillus subtilis also use 5,10 methylenetetrahydrofolate as the methyl donor (44). In addition, both enzymes require FADH 2 to complete the methyl transfer, which suggests that they may be bifunctional enzymes. In this respect, it is important to point out that a small GidA protein identified in Myxococcus xanthus binds FAD and that the alignment of GidA proteins in the data base reveals a highly conserved dinucleotide binding motif near the N terminus (45). The presence of a FAD binding site in the N-terminal region of GidA suggests that this protein catalyzes an oxidation-reduction reaction, which might be required to complete the cmnm group transfer initiated by MnmE. Thus, the multistep mechanism underlying the cmnm group addition might require the sequential intervention of specific domains of MnmE and GidA. Obviously, other alternatives are also possible; however, the evolutionary conservation of the three-domain structure of MnmE supports the idea that mechanistic coordination of the putative roles played by each domain (complex assembly, GTP hydrolysis, and Cys-mediated catalysis) is important for the MnmE function.
Why Do Mutations That Impair the tRNA Modifying Function of MnmE Confer Synthetic Lethality?-Our results demonstrate that mutations abolishing the tRNA modifying function of MnmE (G228A, D270A, C451S) are lethal in the genetic backgrounds of E. coli strains JC7623 and V5701. Strikingly, the construction of a null gidA mutant from strain JC7623 has been reported to be unsuccessful (46), which suggests that null gidA mutations also confer synthetic lethality in this background. In S. cerevisiae, MSS1 or MTO1 null mutations impair mitochondrial protein synthesis only when the mitochondrial 2 M. Martínez-Vicente, unpublished results.