The Yeast Translation Release Factors Mrf1p and Sup45p (eRF1) Are Methylated, Respectively, by the Methyltransferases Mtq1p and Mtq2p*

The translation release factors (RFs) RF1 and RF2 of Escherichia coli are methylated at the N5 -glutamine of the GGQ motif by PrmC methyltransferase. This motif is conserved in organisms from bacteria to higher eukaryotes. The Saccharomyces cerevisiae RFs, mitochondrial Mrf1p and cytoplasmic Sup45p (eRF1), have sequence similarities to the bacterial RFs, including the potential site of glutamine methylation in the GGQ motif. A computational analysis revealed two yeast proteins, Mtq1p and Mtq2p, that have strong sequence similarity to PrmC. Mass spectrometric analysis demonstrated that Mtq1p and Mtq2p methylate Mrf1p and Sup45p, respectively, in vivo. A tryptic peptide of Mrf1p, GGQHVNTTDSAVR, containing the GGQ motif was found to be ∼50% methylated at the glutamine residue in the normal strain but completely unmodified in the peptide from mtq1-Δ. Moreover, Mtq1p methyltransferase activity was observed in an in vitro assay. In similar experiments, it was determined that Mtq2p methylates Sup45p. The Sup45p methylation by Mtq2p was recently confirmed independently (Heurgue-Hamard, V., Champ, S., Mora, L., Merkulova-Rainon, T., Kisselev, L. L., and Buckingham, R. H. (2005) J. Biol. Chem. 280, 2439–2445). Analysis of the deletion mutants showed that although mtq1-Δ had only moderate growth defects on nonfermentable carbon sources, the mtq2-Δ had multiple phenotypes, including cold sensitivity and sensitivity to translation fidelity antibiotics paromomycin and geneticin, to high salt and calcium concentrations, to polymyxin B, and to caffeine. Also, the mitochondrial mit– mutation, cox2-V25, containing a premature stop mutation, was suppressed by mtq1-Δ. Most interestingly, the mtq2-Δ was significantly more resistant to the anti-microtubule drugs thiabendazole and benomyl, suggesting that Mtq2p may also methylate certain microtubule-related proteins.

and RF2, are required for recognition of the translation termination stop codons; RF1 recognizes UAG and UAA and RF2 recognizes UAA and UGA (12,13). In eukaryotes a single protein, eRF1, recognizes all three translation stop codons (14); thus eukaryotic RFs are structurally and functionally distinct as compared with the prokaryotic counterparts. The corresponding genes of the class I release factors, RF1, RF2, and eRF1, are essential for viability in bacteria and yeast (11).
RF3 (class II RF) enhances the activity of RF1 and RF2 in a GTP-dependent manner and catalyzes the dissociation of RFI from the ribosome following peptide release (15)(16)(17), although this has not been established in eukaryotes. RF3 has been shown not to be essential in E. coli (18); however the eukaryotic release factor, eRF3 (Sup35p in yeast), is an essential gene and is required for the activity of the release factor complex at all three stop codons (19 -21). Also, eRF3 forms a ribosome-bound complex with eRF1 that in bacteria could be detected only in presence of ribosomes (16,22). Organellar release factors have been identified in the mitochondria of rat, yeast, and humans (23)(24)(25). The only release factor that acts at the translation termination stop codons in mitochondria is mRF1 (24,26). In the yeast S. cerevisiae, the MRF1 gene encodes a protein more similar to the prokaryotic RF1 than to RF2, and the MRF1 gene is required for proper translation in mitochondria (27). Mrf1p recognizes only the termination codons UAA and UAG on mitochondrial and bacterial ribosomes (25), the same codons recognized by RF1. The UGA codon is not a stop signal in yeast mitochondria (28).
Bacterial RF1, mitochondrial RF1, and eRF1 are homologous proteins (supplemental Fig. 1), and although they are functionally similar, they belong to the different families and display differences in the protein domain structure (10). However, one tripeptide motif, the GGQ motif, is the most conserved feature of the release factors in prokaryotes, eukaryotes, and Archaea (14,29). This motif was shown to be important for the hydrolysis of peptidyl-tRNA (22), and it has been proposed that the glutamine residue in the release factor GGQ motif is involved in the coordination of the water molecule necessary for the hydrolysis of the peptidyl-tRNA ester bond (29) and for normal functioning of the class I release factors. S. cerevisiae Mrf1p and Sup45p (eRF1) have extensive sequence similarities to the bacterial RF1, including the predicted site of methylation, GGQ.
The GGQ motif is essential for the function of both prokaryotic and eukaryotic RFs, and mutations of the glutamine residue in E. coli and S. cerevisiae result in lethality (29,30). RF1 and RF2 proteins of E. coli are post-translationally methylated at the glutamine residue by PrmC (HemK) methyltransferase (31,32). It has been confirmed in another study that the glutamine residues of the GGQ motif at position 235 in RF1 and at position 252 in RF2 are indeed N 5 -methylated (8). Moreover, methylation of glutamine at position 252 in the GGQ motif of E. coli RF2 correlates with increased efficiency of translation termination (8). The E. coli ribosomal protein L3 is the only other protein known to have an N 5 -methylglutamine modification (33).
In addition to high similarities of the protein sequences for translation release factors in bacteria and yeast, including the conserved GGQ motif, the proteins similar to E. coli PrmC methyltransferase could be identified in yeast as well. The protein methyltransferases responsible for these modifications generally have conserved functional domains, responsible for binding of the cofactor AdoMet. The orthologs of these protein methyltransferases share common functional domains, and consequently potential protein methyltransferases can be identified by sequence comparisons. To identify those methyltransferases that may modify the yeast translation release factors, we have searched the yeast S. cerevisiae proteome for potential candidates; this search revealed the two yeast proteins, Ynl063wp and Ydr140wp, that had strong sequence similarity to the E. coli PrmC (Fig. 1). The corresponding genes will be referred to as MTQ1 and MTQ2, respectively.
In this study we demonstrated that the yeast S. cerevisiae protein methyltransferases, Mtq1p and Mtq2p, methylate Mrf1p and Sup45p (eRF1), respectively, in vivo. We also tested the methyltransferase activity of Mtq1p and Mtq2p in vitro and the phenotypes of the corresponding deletion mutants. Although mtq1-⌬ strains had only slightly diminished growth on YPG medium, the mtq2-⌬ strains showed multiple phenotypes, including cold sensitivity and sensitivity to translation fidelity antibiotics paromomycin and geneticin, to high salt and calcium concentrations, to polymyxin B, and to caffeine. Most interestingly, the mtq2-⌬ mutant was significantly more resistant to the anti-mitotic drugs thiabendazole and benomyl as compared with the normal strain. While this work was in progress, Heurgue-Hamard et al. (35) independently found that YDR140w was capable of methylating eRF1.

EXPERIMENTAL PROCEDURES
Yeast Strains-The strains of S. cerevisiae used in this study are listed in Table 1. The analysis of N 5 -glutamine methylation was carried out with two isogenic series, which were derived from the following parental strains: B-8114 (MATa CYC1 cyc7-67 lys5-10 ura3-52) and B-14276 (MAT␣ his3⌬0 leu2⌬0 lys⌬0 ura3⌬0). Several deletion mutant strains, which were used as a source to generate the PCR disruption products, and the MRF1::TAP strain were purchased from Invitrogen or Open Biosystems (Huntsville, AL).
Construction of the Deletion Mutants-Standard molecular biological procedures were used to generate deletions. The MTQ1, MTQ2, and MRF1 genes were disrupted by replacing portions of the genes with the URA3 gene of Kluveromyces lactis or the kanMX4 gene produced by PCR and then using the appropriate fragment for yeast transformation. For example, primers Oligo1 and Oligo2 (Table 2) and plasmid pAB2630 (pBS1539) were used to prepare the PCR fragment, which were used in the construction of the mtq1-⌬::URA3 disruption. After yeast transformation, the correct disruptions among the transformants were identified by PCR, using the set of primers Oligo3 and Oligo4. Similarly, the DNA fragment required for producing the mtq2-⌬::URA3 disruption was prepared with Oligo5 and Oligo6 and plasmid pAB2630. The transformants were screened by PCR with Oligo7 and Oligo8. To obtain mtq1, mtq2, and mrf1 deletion strains with the marker kanMX4 in the B-8114 background, yeast genomic DNAs, which were prepared from the corresponding ORF::kanMX4 deletion strain (Invitrogen), were used as template for PCR with primers Oligo3 and 4, Oligo7 and 8, and Oligo14 and 15, respectively.
Construction of the SUP45::TAP Yeast Strain-The SUP45::TAP strain was made by a PCR-based technique with Oligo9 and Oligo10 and with plasmid pAB2630 (pBS1539) as a template (Table 2). B-8114 was transformed with the resulting PCR product containing the SUP45::TAP::URA3 DNA fragment, and the transformants were screened by amplifying the SUP45 locus with Oligo11 and Oligo12. The SUP45::TAP from the positive clones was amplified by PCR, and the insert was sequenced to confirm the correct in-frame fusion. In addition, the cell extract of the positive strain, B-15543, was probed with peroxidase-anti-peroxidase complex antibody that recognizes the TAP tag within proteins. Because B-15543 grew similar to the normal strain, without a detectable phenotype, the SUP45::TAP fusion was inferred to be completely functional. It is important to note that SUP45, unlike MRF1, is an essential gene, and the sup45 deletion mutant is not viable. Also, MRF1::TAP was completely functional, although the mrf1-⌬ strain does not grow on nonfermentable carbon sources.
Test for Suppression of the Readthrough Accuracy Using the cox2-V25 Mutant-Testing gene deletions for readthrough accuracy in mitochondrial translation was performed by using strains B-15594, normal, and B-15593 that contain the cox2-V25 allele, which has an early premature stop codon that interrupts the COX2 gene (37). These tester strains were obtained from Dr. M. Boguta (Institute of Biochemistry and Biophysics, Polish Academy of Sciences, Warsaw, Poland). mtq1-⌬ deletions were made in both strains B-15594 and B-15593 by replacing the MTQ1 gene with the kanMX4 gene, as described above, to produce B-15598 and B-15599, respectively (Table 1). Similarly, mtq2-⌬ strains B-15891 and B-15890 were made

Open reading frame Oligonucleotide
from B-15594 and B-15593, respectively. Subsequently, all strains were tested on YP0.1%DG medium. The level of mitochondrial Cox2p expression was determined by Western blot analysis, using cell extracts of the yeast strains grown on YPDG and using mouse monoclonal anti-Cox2p antibody (a gift from Dr. T. Fox, Cornell University). We also used actin antibody as loading control (rabbit polyclonal antibody was a gift from Dr. A. Bretscher, Cornell University). The following isogenic series of strains were used: normal B-15594 (mit ϩ ); B-15593 (cox2-V25); and each of the strains containing mtq1-⌬ or mtq2-⌬ (Table 1). Yeast cell extract preparation, protein separation by SDS-PAGE, Western blotting, and probing with antibody were performed by standard techniques (38,39).
Total yeast RNA was extracted by standard procedures with phenol: chloroform treatment of the glass-bead disrupted cells. The RNA samples were normalized after measuring their absorbance at 280/260 nm and were tested by electrophoresis on 1% agarose gel. COX2 and ACT1 cDNAs were obtained with ThermoScript RT-PCR System kit (Invitrogen) using gene-specific oligonucleotides 20 and 23, respectively (Table  2). Subsequently, specific DNA fragments were amplified by PCR with nested oligonucleotides 21 and 22 for COX2 and oligonucleotide 24 and 25 for ACT1. The expected PCR products of 550 bp for COX2 and 574 bp for ACT1 were observed in 1% agarose gel.
MTQ1 and MTQ2 Overexpression in Bacterial System-Cloning of the MTQ1 and MTQ2 genes for overexpression in the E. coli pET system was performed by using PCR with oligonucleotides Oligo16 and Oligo17 for MTQ1 and Oligo18 and Oligo19 for MTQ2 (Table 2). Genomic DNA prepared from the normal yeast strain served as template for PCR, and the primers were designed to introduce XhoI and BamHI restriction sites to both ends of the resulting PCR products ( Table 2). This allowed for the direct cloning of DNA fragments containing the modified MTQ1 and MTQ2 genes into the pET15b vector (Novagen, Madison, WI) that is used for making polyhistidine-tagged proteins. The positive transformants containing inserts were sequenced to confirm the correct in-frame fusion for MTQ1 and MTQ2 genes; subsequently the corresponding cell extracts were tested with anti-polyhistidine antibody (Sigma). The plasmids containing N-His 6 -MTQ1 and -MTQ2 were transformed into the E. coli strain BL21 that is designed for production of proteins.
Purification of N-His 6 -Mtq1p and N-His 6 -Mtq2p Proteins-The bacterially overexpressed Mtq1p and Mtq2p His 6 fusion proteins were purified by immobilized metal affinity chromatography using TALON IMAC resin (Clontech). Bacterial cell extract preparation and affinity purification were carried out according to the manufacturer's protocol supplied with the TALON resin. Additional purification conditions were used to optimize protein purification of the Mtq1p and Mtq2p His 6 fusion proteins. Four sets of Wash Buffers and corresponding Elution Buffers were made according to the manufacturer's protocol, using the following different reagents in the amounts indicated: 1% Nonidet P-40; 0.5% SDS and 0.1% deoxycholic acid; 4 M urea; and 0.8 M NaCl. The manufacturer's protocol was followed in all other aspects of affinity purification. The resultant eluted proteins were dialyzed against methyltransferase buffer (see below).
Purification of the MRF1::TAP and SUP45::TAP Fusion Proteins-The TAP fusion proteins were purified by the TAP method (38). SUP45 was fused to the TAP tag by PCR-based tagging using plasmid pAB2630; MRF1::TAP was made by a similar procedure and was obtained from Open Biosystems. Neither of the tags, ProtA and CBP, impaired protein functions. To increase the protein yield of the TAP fusions, the MRF1 and SUP45 genes from the corresponding strains, B-15403 and B-15543, were amplified by PCR, cloned into the multicopy plasmid pAB2324 (pBEVY-U), and transformed into the normal yeast strain B-8114. The resultant strains, B-15610 and B-15612, were used for protein purification. B-15610 was grown in 2% raffinose medium, and B-15612 was grown in a synthetic medium lacking uracil.
Cells from an ϳ2-liter culture were homogenized by using a glass beads technique in IPP150 buffer (10 mM Tris, pH 8.0, 150 mM NaCl, and 0.1% Nonidet P-40). After 10 cycles of 30 s vortexing/30 s on ice, the cell homogenate was centrifuged at 1000 ϫ g for 5 min, and the glass beads were washed once with IPP150 buffer, vortexed for an additional 4 cycles of 30 s vortexing/30 s on ice, and subsequently centrifuged at 1000 ϫ g for 5 min. The combined supernatants were spun at 12,000 ϫ g for 20 min, and the cleared extracts were mixed by rotation with IgG-Sepharose beads (Amersham Biosciences) at 4°C for 2 h. The IgG beads were pelleted at 700 ϫ g for 5 min, washed three times with IPP150 buffer, and transferred to Bio-Rad biomini columns (1.2 ml). The columns were washed with tobacco etch virus protease cleavage buffer (same as IPP150 but with 0.5 mM EDTA and 1 mM DTT), and the bound proteins were cleaved with 100 units of tobacco etch virus protease (Invitrogen) at 16°C for 12 h. The solutions containing Sup45p-CBP and Mrf1p-CBP were eluted by gravity flow, filtered through a 0.2-m filter, and concentrated by vacuum speed centrifugation. The purified proteins were checked by electrophoresis in a 4 -20% gradient Tris-glycine SDS-polyacrylamide gel, and the protein bands were visualized by PageBlue protein staining solution (Fermentas, Vilnius, Lithuania). Portions of the proteins were used for tryptic digests, mass spectrometry (MS) analysis, and methyltransferase assays.
In Situ Gel Protein Tryptic Digests and Peptide MS Analysis-After staining, the protein bands were excised from the gel, washed, reduced, alkylated, and then incubated with trypsin essentially as described by Polevoda et al. (39). Briefly, the gel slices were washed in 500 l of 100 mM NH 4 HCO 3 for 1 h, then incubated in 150 l of 100 mM NH 4 HCO 3 with 10 l of 45 mM DTT at 60°C for 30 min, followed by the addition of 10 l of 100 mM iodoacetamide, and treatment for 30 min at room temperature in the dark. The solvent was discarded, and the gel slices were washed in 50% acetonitrile, 100 mM NH 4 HCO 3 for 1 h with shaking. The gel slices were cut into smaller pieces and dehydrated with 50 l of acetonitrile followed by vacuum speed evaporation. Subsequently, the gel pieces were re-swelled with 10 l of 25 mM NH 4 HCO 3 containing sequencing grade trypsin (Promega, Madison, WI) in a substrate to enzyme ratio of ϳ10:1. After 15 min, an additional portion of 25 mM NH 4 HCO 3 was added to cover the gel pieces in order to keep them hydrated. The samples were incubated overnight at 37°C, and the supernatants containing the peptide mixture were transferred to new tubes and, if necessary, concentrated by vacuum.  FEBRUARY 3, 2006 • VOLUME 281 • NUMBER 5 products were precipitated by the addition of 125 l of bovine serum albumin at 10 mg/ml and 950 l of 20% trichloroacetic acid on dry ice. Subsequently the samples were thawed and spun for 20 min at 4°C; the pellets were washed with 1 ml of 5% trichloroacetic acid and dissolved in 150 l of 88% formic acid. A total of 3 ml of EcoscintH (National Diagnostics, Atlanta, GA) was added to the vials, and the samples were counted in a Beckman liquid scintillation counter. To visualize the products of the assays, reactions were precipitated with acetone and NaCl (50 mM) at Ϫ70°C, washed with 80% acetone, and loaded onto an SDS-polyacrylamide gel, and proteins were separated. The gel was stained by Coomassie Blue, treated with Amplify solution (Amersham Biosciences), dried, and exposed to Hyperfilm (Amersham Biosciences). The 3 H-labeled protein bands were revealed on autoradiograms.

Mtq1p and Mtq2p Methylate Translation Release Factors
Protein Sequence Alignment and Phylogenic Analysis-Multalin version 5.4.1 (multiple sequence alignment with heirarchical clustering) was used to align protein homologs. MegAlign sequence analysis software (DNAStar) was used to construct a phylogenic tree.

Identification of the Mrf1p and Sup45p as the Potential Substrates for the Methyltransferases Mtq1p and Mtq2p
, Respectively-The protein sequence alignments of Mrf1p and Sup45p (supplemental Fig. 1) with multiple RF sequences from different species showed significant similarity, including the E. coli region of the conserved GGQ motif containing the methylated glutamine residue. Moreover, our search for PrmC orthologs in S. cerevisiae proteome revealed the two yeast proteins, Mtq1p (YNL063w) and Mtq2p (YDR140w), that have strong sequence similarity to the E. coli PrmC methyltransferase (Fig. 1). Also, both Mtq1p and Mtq2p have conserved functional domains responsible for binding of the cofactor, AdoMet, that is generally characteristic for protein methyltransferases (40). This high protein sequence similarity suggested that Mrf1p and Sup45p may have glutamine residues methylated at the positions corresponding to the E. coli RFs, and those reactions could be carried out by Mtq1p and Mtq2p, respectively. Mrf1p was considered as a potential Mtq1p substrate because Mrf1p is a mitochondrial protein, and Mtq1p was also predicted to reside in mitochondria by PSORT II (psort.nibb.ac.jp), although Sup45p is known to be in the cytoplasm. Moreover, in high throughput experiments GFP-Mtq1p fusion was also detected primarily in mitochondria (41).
Mrf1p Is Methylated in Vivo by Mtq1p-To determine whether Mtq1p methylates Mrf1p in vivo, we determined the methylation status of Mrf1p prepared from the normal strain B-15403 and the corresponding mtq1-⌬ deletion mutant strain B-15452 (Table 1). The mtq1-⌬ deletion strain was made by replacing a portion of the MTQ1 open reading frame in the MRF1::TAP fusion strain with the marker kanMX4. The MRF1::TAP fusion was completely functional as the corresponding strain B-15403 grew well on glycerol, whereas the mrf1-⌬ deletion mutant was unable to grow on YPG. The Mrf1p-TAP proteins from normal and mtq1-⌬ strains were purified by tandem affinity purification protocol, separated on SDS-polyacrylamide gel, cut from the gel, and cleaved by trypsin as described under "Experimental Procedures." The resultant peptides were analyzed by MALDI-TOF mass spectrometry (MS) (Fig. 2). A complete tryptic digest of Mrf1p is predicted to produce the 1341-Da peak, corresponding to unmodified peptide with the amino acid sequence GGQHVNTTDSAVR, in which Q is the proposed methylation site. Two peaks with the molecular masses of 1341 and 1355 Da were observed in the sample prepared from the normal strain, whereas only one peak of 1341 Da was detected in Mrf1p sample from the mtq1-⌬ strain. The 14-Da increase in the molecular mass of the second peptide, 1355 Da, corresponds to the molecular mass of a methyl group. The results clearly indicate that this peptide is methylated only in the normal strain but not in the mtq1-⌬ mutant. Apparently, Mrf1p is only partially methylated in the normal strain, with about 50% of the protein modified.
We also determined the amino acid residue in the GGQHVNTTD-SAVR peptide that was methylated. The tryptic digest of the Mrf1p-TAP protein was subjected to LC-MS/MS mass spectrometry ( Table 3). The mass/charge value (ϩ1) of the Y ion for the peptide derivative having glutamine at the end was higher (141.75 m/z) than if that glutamine were unmodified (on average 128 m/z). This indicates that the position of the glutamine residue is methylated in that peptide. How-ever, it should be noted that the relative abundance of the unmodified peptide was higher as compared with the methylated one.
Mtq1p Is Sufficient to Methylate Mrf1p in Vitro-Mtq1p was shown to be sufficient for methylating Mrf1p in vitro. The substrate Mrf1p was purified by TAP protocol as described under "Experimental Procedures," and Mtq1p was overexpressed in the pET bacterial system as a His 6 -tagged protein and was purified by a common affinity chromatography procedure. An in vitro assay was performed with partially purified Mtq1p in the presence of [ 3 H]AdoMet and with Mrf1p prepared from the normal strain or from the mtq1-⌬ deletion strain (Fig. 3). The products of the methyltransferase assay were separated on an SDS-polyacrylamide gel, and the protein bands that incorporated 3 H were visualized by autoradiography. 3 H-Labeled Mrf1p was observed in the samples containing the protein purified from both the normal strain and the mtq1-⌬ deletion strain (Fig. 3); however no such product was seen in the sample lacking the substrate. Two bands of significantly lower and higher molecular masses probably correspond to the nonspecific proteins because of incomplete purity of the Mtq1p sample. Also, Mtq1pdependent methylation of the normal Mrf1p could be explained by the fact that the protein is only ϳ50% methylated in vivo (Fig. 2).

TABLE 3 Amino acid sequence of the GGQ-containing peptide of Mrf1p
The amino acid sequence was determined by LC-MS/MS mass spectrometry. Ion series for charge ϩ1 is presented. The position of the methylated glutamine residue is marked with an arrow.   FEBRUARY 3, 2006 • VOLUME 281 • NUMBER 5

JOURNAL OF BIOLOGICAL CHEMISTRY 2567
Mtq2p Methylates Sup45p-Similar to Mrf1p/Mtq1p experiments, the Sup45p (eRF1) protein was found to be methylated by Mtq2p. TAPpurified proteins were prepared from the normal and mtq2-⌬ strains and analyzed by MS, generating partially modified peptide GGQSALR with the molecular mass of 702 Da only in the sample from the normal strain as compared with unmethylated peptide of 688 Da in the mtq2-⌬ (data not shown). Moreover, it should be noted that methylation of Sup45p by Mtq2p both in vivo and in vitro was recently confirmed by Heurgue-Hamard et al. (35).
Phenotypic Analysis of mtq1-⌬ and mtq2-⌬ Yeast Strains-To determine whether the deletions of MTQ1 or MTQ2 would cause any defect in cell proliferation or abnormal sensitivities to various agents, we tested the growth of the corresponding mutants on a number of different media and at different temperatures (for a list of media, see "Experimental Procedures"). Phenotypic analysis using 1/10 serial dilutions of the deletion mutants mtq1-⌬ (B-15545) and mtq2-⌬ (B-15567) revealed multiple phenotypes for the mtq2-⌬ deletion mutant (Fig. 4). Although the mtq1-⌬ deletion strain grew normally or near normally on most media, the mtq2-⌬ deletion strain grew slower on common yeast media and was cold-sensitive as it grew substantially slower than the normal strain at 16°C. The mtq2-⌬ mutant was sensitive to the translation fidelity drug paromomycin at all temperatures and to another similar antibiotic of the aminoglycoside group, geneticin (G418). Increased sensitivity to paromomycin was observed at the higher temperature. The mutant strain mtq2-⌬, was also sensitive to high salt and calcium concentrations, such as 1 M KCl, 0.3 M CaCl 2 , as well as caffeine (0.15%). In addition, the mtq2-⌬ strain was sensitive to polymyxin-B sulfate, which leads to an enhanced permeability of yeast (42), which is an inhibitor of protein kinase C of prokaryotic type membranes, and in mammals was shown to inhibit the activity of glycerol-3-phosphate acyltransferase that catalyzes the initial and rate-limiting step of triaglycerolipid synthesis. Most interesting, the mtq2-⌬ mutant was resistant to the anti-microtubule drugs thiabendazole (25 g/ml) and benomyl (30 g/ml), although the normal strain and the mtq1-⌬ were highly sensitive to those agents. Some of the phenotypes mentioned above, like paromomycin and anti-microtubule drugs sensitivity, were reported previously for SUP45 mutants (43).
There were no significant phenotypes observed for the mtq1-⌬ mutant on solid media. However, a quantitative growth analysis of the normal and mtq1-⌬ mutant strains in liquid cultures with the nonfermentable carbon source, ethanol or glycerol, revealed that the mtq1-⌬ deletion strain grew slower (Fig. 5A). By using a cell counting chamber, we determined that the doubling time during the exponential phase of growth in YPG medium of the mtq1-⌬ strain was ϳ395 min as compared with 340 min for the normal strain. This suggests that mtq1-⌬ affects mitochondrial function.
Testing the mtq1-⌬ Deletion for Suppression of the Readthrough Accuracy in the cox2-V25 Mutant-The mitochondrial translation readthrough accuracy in the mtq1-⌬ strain was investigated with the mitochondrial mit Ϫ mutation, cox2-V25, containing a premature stop mutation in the COX2 gene (37). The following four isogenic strains were tested on media containing nonfermentable carbon sources: The products of the reactions were separated on an SDS-polyacrylamide gel. The gel was strained with Coomassie Blue, treated with Amplify reagent, dried, and exposed to Hyperfilm. The position of Mrf1p-CBP is marked on the left, and the molecular markers are on the right.  16,30, and 37°C on the following media: YPD; YPG; YPD with 800 g/ml paromomycin; YPD with 75 g/ml geneticin (G418); YPD with 0.3 M CaCl 2 ; YPD with 1 M KCl; YPD with 30 g/ml benomyl (Beno); YPD with 25 g/ml thiabendazole (TBZ); and YPD with 0.15% caffeine (Caff). Unless indicated otherwise, the strains were tested at 30°C. The mtq2-⌬ strain grew slower on YPD, was cold-sensitive, and was also sensitive to paromomycin and geneticin, to high salt and calcium concentrations, and to caffeine and polymyxin B but was significantly more resistant to the microtubule-destabilizing drugs benomyl and thiabendazole. B-15594 (normal); B-15593 (cox2-V25); B-15598 (cox2-V25 mtq1-⌬); and B-15599 (mtq1-⌬). Only the B-15593 (cox2-V25) was unable to grow on media with nonfermentable carbon sources because COX2, which is translated in mitochondria, is essential for mitochondrial function (37). Further tests were carried out with YPDG medium that contains only a limited amount of fermentable carbon source, 0.1% dextrose. After initial quick growth phase utilizing dextrose, the cells then are forced to shift to glycerol, a nonfermentable carbon source. The growth of the B-15593 (cox2-V25) strain was only marginal, whereas the B-15598 (cox2-V25 mtq1-⌬) strain grew to a high density in approximately 2 days and was comparable with the normal strain B-15594 (Fig.  5B), indicating that mtq1-⌬ suppresses the cox2-V25 nonsense mutation.
We also directly tested the Cox2p level in the cell extracts by Western blot analysis of the pertinent strains grown in YPDG medium. The Cox2p level was approximately the same in the normal COX2 strain (Fig. 6A, lane 2), the mtq1-⌬ (lane 4) and mtq2-⌬ (data not shown) derivatives, and no Cox2p was detected in the cox2-V25 mutant or in the cox2-25 mutant, which also contains a nonsense mutation in the COX2 gene (Fig. 6A, lanes 1 and 5, respectively). However, a minor but detectable level of Cox2p, corresponding to approximately several percent of the wild type level, was observed in the mtq1-⌬ cox2-V25 strain (Fig. 6A, lane 3). On the other hand, the levels of actin in the samples (loading control) showed very little variation. Similar results were observed when the same set of strains was grown on YPEG medium. In contrast, the mtq2-⌬ cox2-V25 strain lacked Cox2p, similar to the cox2-V25 strain (results not shown). These results established that mtq1-⌬ suppressed the cox2-V25 defect at the translational level, probably by causing readthrough of the nonsense mutation. Furthermore, we determined that the level of COX2 mRNA was not diminished by premature nonsense mutations. RT-PCR experiments revealed the same or very similar levels of COX2 mRNA in comparison to ACT1 mRNA from the following strains: (Fig. 6B).

DISCUSSION
In this study we uncovered the two yeast protein methyltransferases, Mtq1p (YNL063w) and Mtq2p (YDR140w), that had extensive sequence similarity to PrmC of E. coli (Fig. 1). Although overall sequence homology is relatively low, these proteins are highly homologous in the regions responsible for binding AdoMet (40), particularly motif I (44) with consensus sequence hh(D/E)hGXGXG (where h is a hydrophobic residue) and the NPPY motif, which were both previously thought to be characteristic of DNA methyltransferases (45)(46)(47). Multiple sequence alignments and phylogenic analysis demonstrated that E. coli PrmC and S. cerevisiae Mtq1p and Mtq2p, as well as their various orthologs, could be classified into two related subfamilies of protein methyltransferases (supplemental Fig. 2).
Mass spectrometry analysis of the peptides produced from the tryptic digests of the mitochondrial translation release factor Mrf1p of S. cerevisiae from normal and mtq1-⌬ mutant strains revealed that Mrf1p was methylated only in the normal strain and not in the mtq1-⌬ deletion strain (Fig. 2). Biochemical assays demonstrated that Mtq1p was sufficient to methylate Mrf1p (Fig. 3). Also, MS analysis of cytoplasmic translation release factor Sup45p proteins from the normal and mtq2-⌬ strains revealed that Mtq2p methylates Sup45p. Apparently both Mrf1p and Sup45p are only partially (ϳ50%) methylated in vivo. Furthermore, both Mrf1p and Sup45p methylated peptides included the highly conserved GGQ motif with glutamine residue as a potential methylation site. In addition, using LC-MS/MS mass spectrometry, we demonstrated that the GGQ-containing peptide of Mrf1p is methylated at the conserved glutamine position ( Table 3). The specific methyltransferase activities of Mtq1p and Mtq2p are similar to the previously observed glutamine methyltransferase activity of E. coli PrmC toward RF1 and RF2 (31,32). Recently, it was independently shown that Mtq2p methylates the glutamine residue in the conserved GGQ motif of the eRF1 (Sup45p) of S. cerevisiae (35).
Although Mtq1p modifies the Mrf1p, the protein that is essential for mitochondrial function (27), the mtq1-⌬ mutant had no significant phenotypes on solid media (Fig. 4). However, the quantitative growth experiment in nonfermentable liquid medium (YPE, ethanol) showed that the mtq1-⌬ deletion strain grew slower than the normal strain (Fig.  6A). This is indicative of a biological role for Mrf1p related to the proper functioning of mitochondria and to mitochondrial translation that is required for normal growth on YPE. Moreover, mtq1-⌬ deletion also affected the accuracy of readthrough in the cox2-V25 strain. The double mutant cox2-V25 mtq1-⌬ was able to grow to a high culture density in a medium limited on fermentable carbon source as compared with the cox2-V25 strain (Fig. 5B). Direct evidence for Mtq1p involvement in readthrough activity of the mitochondrial release factor was obtained when we examined the level of Cox2p in cox2-V25 mtq1-⌬ and related strains (Fig. 6). The mtq1-⌬ deletion in the cox2-V25 strain caused a minor but clearly detectable level of Cox2p as compared with the strain with the wild type MTQ1. In contrast, the mRNA levels were very similar in both the normal and cox2-V25 strain and their corresponding mtq1-⌬ mutants (Fig. 6B). Although very low, such an increase in Cox2p protein translation in the cox2-V25 mtq1-⌬ strain is comparable with the level of suppression of nonsense mutations of chromosomal genes by various SUP45 mutations that encode altered forms of the cytoplasmic translation release factor eRF1 (21).
In contrast, the mtq2-⌬ deletion mutant showed multiple phenotypes (Fig. 4). First, the mtq2-⌬ strain was sensitive to the translation fidelity drugs paromomycin and geneticin, indicating that the methylation of Sup45p (eRF1) is important for its function and affects the translation accuracy in cytoplasm. This is also consistent with the role of the universally conserved FIGURE 6. The mtq1-⌬ deletion enhances the readthrough activity of the mitochondrial release factor Mrf1p. A minor but detectable level of Cox2p protein is translated in the cox2-V25 mtq1-⌬ double mutant strain as compared with the cox2-V25 mutant (A). The relative levels of Cox2p were determined by Western blot analysis of cell extracts prepared from the normal and cox2-V25 strains. Protein cell extracts were prepared from the cells grown in YPDG. Equal amounts, corresponding to 25 g of total protein, were loaded onto 4 -20% linear gradient SDS-PAGE. The proteins were transferred to an ECL nitrocellulose membrane. The membrane was incubated with mouse monoclonal anti-Cox2p antibody (␣-Cox2p) or rabbit polyclonal anti-actin (␣-Act1p) antibody. The positive protein bands were visualized with ECL Plus reagents (Amersham Biosciences). The following yeast strains were used (Table 1) FEBRUARY 3, 2006 • VOLUME 281 • NUMBER 5 GGQ motif in the translation peptide release. In addition, mtq2-⌬ was cold-sensitive, grew slower at normal temperature, and was sensitive to high salt and calcium concentrations and to polymyxin and caffeine. Cold sensitivity in bacteria is often associated with defects in ribosomal assembly (2). Most interestingly, the lack of ribosomal protein L3 methylation caused a cold-sensitive phenotype (9,31), and it was suggested that L3 methylation may serve as one of the "assembly factors" for E. coli ribosome (2). Although methylation of the yeast L3 protein was described, the position of the methylated residue is not known (48). Moreover, the yeast cytoplasmic L3 protein does not contain the QNQ motif that is methylated in E. coli L3. Both bacterial methyltransferases PrmC, which methylates RFs, and PrmB, which methylates L3, have extensive sequence homology to the yeast Mtq1p and Mtq2p; nevertheless, considering the other findings presented above, it is not clear whether yeast L3 protein is methylated by any of the Mtq1p or Mtq2p and whether a cold-sensitive phenotype of mtq2-⌬ deletion is related to L3.

Mtq1p and Mtq2p Methylate Translation Release Factors
It is possible that the cold-sensitive phenotype is related to another mtq2-⌬ defect, particularly the significantly increased resistance to anti-microtubule drugs benomyl and thiabendazole, which effect chromosome segregation at anaphase. Most interestingly, some mutants of SUP45 and SUP35 have been observed to have increased sensitivity to benomyl (43). This suggests that Mtq2p may be involved in the methylation of certain microtubule-related proteins and that Mtq2p may have an additional substrate or that the lack of Sup45p methylation affects its interaction with another protein that acts beyond the translation release. Recently, it has been suggested that Sup45p and Sup35p have functions outside of translation related to cytoskeleton organization and cell cycle regulation (49). Moreover, eRF1 was found to interact directly with myosin light chain protein Mlc1p to effect cytokinesis (50). The C-terminal domain of Sup35p is homologous to the translation elongation factor EF-1␣, which binds aminoacyl-tRNAs and is the most abundant actin-binding protein in eukaryotic cells (51). Repression of SUP35 caused damaging effects, not related to translation, including the disappearance of actin cytoskeletal structures, defects in mitotic spindle formation, nuclei division, and segregation in mitosis (49). Accumulation of unbudded cells with high DNA content was noticed when the SUP45 gene was repressed (49). The observed benomyl and thiabendazole resistance in the mtq2-⌬ mutant supports such function outside of translation for Sup45p or Sup35p. The benomyl and thiabendazole resistance of the mtq2-⌬ mutant strain can be explained by the effect of glutamine methylation, or lack of thereof, on any of these functions and interactions of Sup45p and Sup35p. It is tempting to speculate that methylation of Sup45p or Sup35p is required for interaction with Mlc1p or another microtubule-related protein.
It is also possible that the lack of Sup45p methylation causes a secondary effect. For example, methylation could affect binding to Sup35p, which has a ribosome-dependent GTPase activity (16,17). However, Sup35p could also function as a GTP exchange factor involved with microtubule proteins known to utilize GTP as an energy source. The GTPase activity of Sup35p may have an additional function related to cytoskeletal proteins.
The protein structure of Sup45p shows that the GGQ motif is located on the surface of the molecule and extended from a backbone, where it is easily accessible for modification. One approach would be to look for a similar fold within the pool of the known or predicted yeast protein structures containing the conserved GGQ motif that potentially could be methylated. It remains to be seen whether Mtq2p has any additional substrate besides Sup45p.