The METTL20 Homologue from Agrobacterium tumefaciens Is a Dual Specificity Protein-lysine Methyltransferase That Targets Ribosomal Protein L7/L12 and the β Subunit of Electron Transfer Flavoprotein (ETFβ)*

Human METTL20 is a mitochondrial, lysine-specific methyltransferase that methylates the β-subunit of electron transfer flavoprotein (ETFβ). Interestingly, putative METTL20 orthologues are found in a subset of α-proteobacteria, including Agrobacterium tumefaciens. Using an activity-based approach, we identified in bacterial extracts two substrates of recombinant METTL20 from A. tumefaciens (AtMETTL20), namely ETFβ and the ribosomal protein RpL7/L12. We show that AtMETTL20, analogous to the human enzyme, methylates ETFβ on Lys-193 and Lys-196 both in vitro and in vivo. ETF plays a key role in mediating electron transfer from various dehydrogenases, and we found that its electron transferring ability was diminished by AtMETTL20-mediated methylation of ETFβ. Somewhat surprisingly, AtMETTL20 also catalyzed monomethylation of RpL7/L12 on Lys-86, a common modification also found in many bacteria that lack METTL20. Thus, we here identify AtMETTL20 as the first enzyme catalyzing RpL7/L12 methylation. In summary, here we have identified and characterized a novel bacterial lysine-specific methyltransferase with unprecedented dual substrate specificity within the seven β-strand class of lysine-specific methyltransferases, as it targets two apparently unrelated substrates, ETFβ and RpL7/L12. Moreover, the present work establishes METTL20-mediated methylation of ETFβ as the first lysine methylation event occurring in both bacteria and humans.

Methylation is a common biochemical reaction catalyzed by various methyltransferases (MTases) 2 and involving transfer of a methyl group from a methyl donor, usually S-adenosyl-L-methionine (AdoMet), to a wide range of acceptor molecules, including small metabolites and macromolecules such as RNA, DNA, and proteins (1). Proteins can be methylated on several amino acid residues but most commonly on lysines and arginines (2)(3)(4). Lysine residues can be mono-, di-, and trimethylated on the ⑀-amino group in reactions catalyzed by lysine (K)specific protein MTases (KMTs). Lysine methylation leaves the overall charge unaffected but decreases the potential for hydrogen bond formation and increases the bulkiness. Lysine methylation can modulate many different aspects of protein function, such as stability, intracellular localization, enzymatic activity, and the ability to interact with other molecules (5).
The human genome has been predicted to encode ϳ200 MTases, and the two main groups are the seven-␤-strand (7BS) MTases, which have a characteristic core fold of seven ␤-strands, and the SET proteins, which contain a defining SET domain (6). Whereas SET proteins mainly encompass KMTs, many of which target histones (7), the 7BS MTases have been shown to methylate a wide range of small molecules and macromolecular substrates, including lysine and arginine residues in proteins. In recent years, several of the human members of a group of related 7BS MTases, denoted "methyltransferase family 16" (MTF16), were established as KMTs, and these include CaM-KMT that methylates calmodulin (8), VCP-KMT that targets the ATP-dependent chaperone valosin-containing protein (9), HSPA-KMT (METTL21A) that methylates several HSPA (Hsp70) proteins (10,11), METTL22 that targets Kin17 (10), and eEF2-KMT (FAM86A) that methylates eukaryotic elongation factor 2 (12). Another human MTF16 member is METTL20 (ETF␤-KMT), a mitochondrial enzyme that methylates the ␤-subunit of electron transfer flavoprotein (ETF␤) (13,14). Interestingly, putative orthologues of human METTL20 are found in some bacteria, contrasting with other human MTF16 members, which are restricted to eukaryotes.
Eukaryotic ETF is a mitochondrial protein composed of two subunits, ␣ and ␤ (ETF␣ and ETF␤), and it binds two cofactors, FAD and AMP (15). ETF mediates transfer of electrons from several mitochondrial FAD-containing dehydrogenases (DHs) to the ETF:quinone oxidoreductase (ETF-QO), leading to reduction of the quinone pool of the mitochondrial respiratory chain (15). In humans, there are ϳ13 ETF-dependent DHs, including sarcosine dehydrogenase and dimethylglycine dehydrogenase (DMGDH), which are involved in choline metabolism, as well as several acyl-CoA dehydrogenases (ACADs) involved in degradation of amino acids or in ␤-oxidation of fatty acids (e.g. medium-chain acyl-CoA dehydrogenase (MCAD)) (16 -18). Structural studies have shown that the so-called "recognition loop" in ETF␤, encompassing residues 191-200, is important for its interaction with MCAD, and it has been suggested that this loop is also involved in the interaction with the remaining DHs (19). We and others recently showed that human METTL20 methylates ETF␤ on Lys-200 and Lys-203, which partly overlap with the recognition loop, and that methylation affects electron transfer from DHs to ETF (13,14). Interestingly, bacterial METTL20 homologues are mostly limited to ␣-proteobacteria (13), and ETFs from ␣-proteobacteria are, in contrast to most other bacterial ETFs, highly similar to their human counterparts (20). As ␣-proteobacteria are considered the evolutionary precursors of mitochondria (21), this suggested that ETF␤ methylation may be conserved from bacteria to eukaryotes.
Protein lysine methylation appears much less common in bacteria than in eukaryotes, and many of the reported methylation events involve components of the protein synthesis machinery, e.g. ribosomal proteins and translation factors (22). For example, the 50S ribosomal protein RpL11 is methylated at multiple sites by the MTase PrmA (23). Another example is RpL7/L12, which is part of the multimeric ribosomal stalk shown to be important for recruiting translational factors to the ribosome (24,25). RpL7/L12 is methylated at a single lysine residue (Lys-82 in Escherichia coli RpL7/L12) in a wide range of prokaryotic species (26 -29). However, the identity of the responsible enzyme(s) as well as the functional significance of the methylation remains elusive.
In the present study, we have taken an unbiased approach to identify the substrates of the putative METTL20 orthologue from Agrobacterium tumefaciens (AtMETTL20). Through extensive in vitro and in vivo studies, we demonstrate that AtMETTL20 is a dual specificity KMT that targets both ETF␤ and RpL7/L12. By analyzing several ETF-dependent dehydrogenases from A. tumefaciens, we show that methylation of ETF␤ impairs the ability of ETF to extract electrons from these dehydrogenases.

Experimental Procedures
Cloning and Mutagenesis-Genes encoding bacterial proteins were amplified from genomic DNA isolated from A. tumefaciens strain C58 (ATCC 33970) or Rhizobium etli strain CFN 42 (ATCC 51251) using Phusion DNA Polymerase HF (Thermo Fisher Scientific). cDNA generated from HeLa cells was used to amplify ORFs for human proteins. Mutagenesis was performed using mutagenic primers designed with the PrimerX program. All constructs were sequence-verified.
Cell Cultures-A. tumefaciens strains (C58 and GV3101 pM90) were grown at 28°C in LB medium supplemented with 10 -100 g/ml rifampicin (Rif) and 10 -50 g/ml gentamycin (Gent). The growth rate of different strains of A. tumefaciens at 28°C in LB or minimal medium with 0.2% glucose was followed by monitoring absorbance at 600 nm, starting from cultures with identical A 600 (ϳ0.2).
Generation of A. tumefaciens Strain with AtMETTL20 Gene Knock-out-A. tumefaciens strain GV3101 pM90 (tetracyclinesensitive) was used to generate bacteria with AtMETTL20 gene knock-out (KO) using the TargeTron Gene Knock-out System (Sigma-Aldrich) according to the manufacturer's protocol with some modifications. In short, the pBL1 plasmid (32) was mutated to contain an intron that would self-insert into A. tumefaciens METTL20 gene between nucleotides 261 and 262, thus introducing a premature stop codon, located downstream of the SGSG sequence within Motif I (Fig. 1A). The resulting plasmid (pBL1-MT20(261)KO) was transformed into bacteria that were selected for Rif (100 g/ml), Gent (50 g/ml), and tetracycline (12.5 g/ml) resistance. Positive colonies were grown at 28°C in MGL medium (5 g/liter tryptone, 2.5 g/liter yeast extract, 5 g/liter mannitol, 1 g/liter monopotassium L-glutamate, 1 g/liter biotin, 250 mg/liter KH 2 PO 4 , 100 mg/liter NaCl, 100 mg/liter MgSO 4 ϫ7H 2 O) supplemented with Rif, Gent, and tetracycline until A 600 reached 0.3, and then intron insertion was induced with 5 mM m-toluic acid for 3 h. Cells were harvested, plated, and screened for the presence of the AtMETTL20 gene with a 900-bp intron insertion. The pBL1-MT20(261)KO plasmid was removed (cured) from the bacteria by overnight growth at 28°C in LB medium (Rif ϩ Gent), plating on LB-agarose (Rif ϩ Gent), and screening for tetracycline-sensitive colonies. Finally, A. tumefaciens colonies with AtMETTL20 gene knock-out were tested for absence of pBL1 plasmid (by colony PCR) and sequenceverified to contain mutated AtMETTL20 gene with properly inserted intron.
Expression and Purification of Recombinant Proteins-Human ⌬38-METTL20 (with the N-terminal 38 amino acids deleted) and human ETF␣/␤ heterodimer were cloned, expressed, and purified as described previously (13). The gene encoding A. tumefaciens ETF␣ was cloned into pETDuet-1, whereas other relevant bacterial genes were cloned into pET28a. All proteins were expressed as N-terminally His 6tagged proteins in the E. coli strain BL21-CodonPlus(DE3)-RIPL (Agilent Technologies). Protein expression was routinely carried out at 16°C (for 18 h) by induction with 0.1 mM isopropyl 1-thio-␤-D-galactopyranoside except in the case of RpL7/ L12 from A. tumefaciens, which as indicated was expressed at either 16°C or 37°C (for 4 h). Cells were harvested by centrifugation and frozen at Ϫ20°C.
Unless indicated otherwise, all proteins were purified according to the same purification protocol as described below. Bacteria were lysed in Lysis Buffer 1 (50 mM Tris-HCl, pH 7.4, 500 mM NaCl, 5% glycerol, 30 mM imidazole) supplemented with 2 mM ␤-mercaptoethanol, 1ϫ Complete (EDTA-free) protease inhibitor mixture (Roche Applied Science), and 10 units/ml Benzonase nuclease (Sigma-Aldrich). His-tagged proteins were loaded onto Ni-NTA-agarose column (Qiagen) equilibrated in Lysis Buffer 1, washed extensively with Lysis Buffer 1, and finally eluted with Lysis Buffer 1 supplemented with 200 mM imidazole. During purification of FAD-containing proteins, Lysis Buffer 1 was additionally supplemented with 2 M FAD, and during purification of ETF-QO, Lysis Buffer 1 was Dual Specificity METTL20 Homologue from A. tumefaciens supplemented with 2 M FAD and 1% Triton X-100. Proteins eluted from Ni-NTA-agarose were buffer-exchanged to Storage Buffer (50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 10% glycerol) using centrifugal concentrators with a molecular mass cutoff of 10 -50 kDa (Sartorius, Goettingen, Germany) and stored at Ϫ20°C, except recombinant A. tumefaciens METTL20 and RpL7/L12, which were stored at Ϫ20°C in Storage Buffer containing 25% glycerol. All proteins were 90 -95% pure as assessed by SDS-PAGE and Coomassie Blue staining except for human METTL20, which was ϳ50% pure. Protein concentration was determined using a Pierce BCA Protein Assay kit (Thermo Fisher Scientific). Theoretical molecular mass of recombinant proteins was used to calculate molar concentrations.
Purification of Recombinant A. tumefaciens ETF␣/␤ Heterodimer-E. coli cells expressing His 6 -tagged ETF␣ and His 6tagged ETF␤ were lysed separately (in Lysis Buffer 1 supplemented with 2 M FAD) and then mixed together, and the mixed lysate was used to purify ETF␣/␤ heterodimer on Ni-NTA-agarose similarly as described in the previous section. The eluate from Ni-NTA-agarose was buffer-exchanged to Storage Buffer using centrifugal concentrators with a molecular mass cutoff of 50 kDa and frozen at Ϫ20°C. After thawing on ice, the precipitated protein, containing mostly free ETF␤, was removed by centrifugation, whereas the supernatant, containing mostly ETF␣/␤ heterodimer, was again buffer-exchanged to Storage Buffer. Finally, the amount of glycerol in the sample was increased to 25%, and the protein was stored at Ϫ20°C. For ETF␣/␤ and other FAD-containing enzymes, the A 275 /A 436 ratio was ϳ7, indicating that FAD remained strongly associated with these proteins throughout the purification procedure.
Preparation and Fractionation of Bacterial Cell Extracts-A. tumefaciens cells were lysed at 4°C for 10 min in Lysis Buffer 2 (50 mM Tris-HCl, pH 6.5, 100 mM NaCl, 1% Triton X-100, 5% glycerol, 0.5 mM DTT, 1ϫ Complete (EDTA-free) protease inhibitor mixture), and the lysate was sonicated and cleared by centrifugation. Cell extracts were fractionated at 4°C by ion exchange chromatography using either Pierce Strong Cation Exchange (S) or Pierce Strong Anion Exchange (Q) Spin Columns (Thermo Fisher Scientific). First, the NaCl concentration was reduced to 50 mM by diluting cell lysates with Dilution Buffer (50 mM Tris-HCl, pH 6.5, 1% Triton X-100, 5% glycerol), and then extracts were applied onto the S-column equilibrated with Dilution Buffer. Material bound to the S-column was eluted by a step gradient of increasing NaCl concentrations prepared in Dilution Buffer. Typically, four fractions were collected: 0.15S (eluted between 0.05 and 0.15 M NaCl), 0.3S (eluted between 0.15 and 0.3 M NaCl), 0.5S (eluted between 0.3 and 0.5 M NaCl), and 0.75S (eluted between 0.5 and 0.75 M NaCl). Unbound material, which was released in the S-column flow-through, was reapplied onto the Q-column equilibrated with Dilution Buffer. The unbound material, present in the Q-column flow-through, was designated FTQ, whereas material bound to the Q-column was eluted using Dilution Buffer with a step gradient of increasing NaCl concentrations, similarly as for the S-column, in four fractions, 0.15Q, 0.3Q, 0.5Q, and 0.75Q.
Isolation of Ribosomes from A. tumefaciens Cells-Ribosomal particles were isolated from A. tumefaciens cells at 4°C following a previously published procedure (27) with some modifications. Cells were lysed in Ribosome Isolation Buffer (20 mM Tris-HCl, pH 7.5, 50 mM magnesium acetate, 100 mM NH 4 Cl, 1 mM EDTA) supplemented with fresh 2 mM DTT, 1 unit/ml DNase I (Qiagen), and 1ϫ Complete (EDTA-free) protease inhibitor mixture and sonicated. Cellular debris was removed by centrifugation, and the supernatant was filtered (0.2 m). Three milliliters of clear lysate was carefully layered on top of 1.5 ml of 1.1 M sucrose in Ribosome Isolation Buffer and centrifuged at 100,000 ϫ g for 24 h using an Optima Max ultracentrifuge with MLS-50 rotor (Beckman-Coulter). The supernatant and the sucrose cushion were carefully removed, and the ribosomal pellet was washed, finally resuspended in Ribosome Isolation Buffer, aliquoted, and stored at Ϫ80°C. Typically, ribosome solutions had an A 260 /A 280 ratio of ϳ1.9. Ribosomes were quantified by RNA content, assuming a molecular extinction coefficient at 260 nm of ⑀ 260 ϭ 39.1 ϫ 10 6 M Ϫ1 cm Ϫ1 (33).
In Vitro Methyltransferase Assays Using [ 3 H]AdoMet-MTase activity of METTL20 was tested by setting up on ice 10-l reactions containing 1ϫ MTase Assay Buffer (50 mM Tris-HCl, pH 7.4, 50 mM NaCl, 50 mM KCl, 1 mM MgCl 2 , 10% glycerol), 3-5 M recombinant substrate or 20 -60 g of proteins from cell extracts (and/or equivalent fractions obtained from ion exchange chromatography), 50 -80 pmol of METTL20, and 0.5 Ci of [ 3 H]AdoMet (PerkinElmer Life Sciences) ([AdoMet] total ϭ 0.64 M; specific activity ϭ 78.2 Ci/mmol). Reaction mixtures were then incubated at 28°C for 2 h. Proteins were resolved by SDS-PAGE, transferred to a PVDF membrane, and stained with Ponceau S. The membrane was dried, sprayed with EN3HANCE spray (PerkinElmer Life Sciences), and exposed to Kodak BioMax MS film (Sigma-Aldrich) at Ϫ80°C for 1-30 days. Precision Plus Protein Dual Color Standards (Bio-Rad) were used to evaluate the size of polypeptides after SDS-PAGE, and a Glow Writer autoradiography pen (Sigma-Aldrich) was used to mark the position of the standards on PVDF membrane, enabling their visualization in fluorography. When recombinant protein substrates were used, [ 3 H]AdoMet was diluted with non-radioactive AdoMet (New England Biolabs) ([AdoMet] total ϭ 32.6 M; specific activity ϭ 1.53 Ci/mmol). All fluorography experiments were performed three times with similar results, and results of representative experiments are shown.
Preparation of Samples for MS Analysis-In vitro methylation of recombinant or cellular (in extract) proteins for the purpose of mass spectrometry (MS) analysis was performed as in the above section except that [ 3 H]AdoMet was replaced with non-radioactive AdoMet (1 mM). Proteins were resolved by SDS-PAGE and stained with Coomassie Blue; the portion of the Dual Specificity METTL20 Homologue from A. tumefaciens APRIL 29, 2016 • VOLUME 291 • NUMBER 18 gel containing the protein of interest was cut out and subjected to in-gel trypsin (Sigma-Aldrich), Asp-N (Roche Applied Science), or Glu-C (Promega) digestion; and the resulting proteolytic fragments were analyzed as described previously (9,11). MS data were analyzed with an in house-maintained A. tumefaciens protein sequence database using SEQUEST TM and Proteome Discoverer TM (Thermo Fisher Scientific). The mass tolerances of a fragment ion and a parent ion were set as 0.5 Da and 10 ppm, respectively. Methionine oxidation and cysteine carbamidomethylation were selected as variable modifications. MS/MS spectra of peptides corresponding to methylated ETF␤ and RpL7/L12 were manually searched by Qual Browser (v2.0.7).
2,6-Dichloroindophenol (DCIP) and CoQ 1 Reduction Assays-The ability of A. tumefaciens ETF␣/␤ to mediate electron transfer from FAD-containing dehydrogenases was tested either by following reduction of DCIP (18)  . Reaction mixtures were allowed to equilibrate at room temperature for 10 min, and then the reaction was started by adding, through manual mixing, acyl-CoA (50 M) or dimethylglycine/sarcosine (10 mM). The kinetics of DCIP reduction (bleaching) was monitored continuously at 600 nm for 1-2 min using a Shimadzu UV-1601 spectrophotometer. The initial rate of DCIP reduction was calculated from the slope of the linear part of the kinetic curve using a molecular extinction coefficient of DCIP at 260 nm of ⑀ 260 ϭ 21,700 M Ϫ1 cm Ϫ1 and expressed in M of DCIP reduced/ min. The kinetics of CoQ 1 reduction was monitored continuously at 275 nm for 2 min, and the rate of CoQ 1 reduction was calculated from the slope of the linear part of the kinetic curve using a molecular extinction coefficient of CoQ 1 (oxidized minus reduced) of ⑀ 275 ϭ 12,500 M Ϫ1 cm Ϫ1 (34) and expressed in M of CoQ 1 reduced/min. Student's t test was used to evaluate the probability (p values) that two populations are the same.
GTPase Activity Assay-GTPase activity of ribosomes and translation factors was assayed with an EnzChek Phosphate Assay kit (Thermo Fisher Scientific) where purine-nucleoside phosphorylase uses inorganic phosphate to convert 2-amino-6mercapto-7-methylpurine riboside into the ribose 1-phosphate and 2-amino-6-mercapto-7-methylpurine with resulting characteristic changes in the absorption spectra (35). Typically, 350-l mixtures were prepared in Ribosome Isolation Buffer containing 0.2 mM 2-amino-6-mercapto-7-methylpurine riboside, 1 unit/ml purine-nucleoside phosphorylase, 0.5 mM GTP, and varying amounts of ribosomes. Reaction mixtures were allowed to equilibrate at room temperature for 5 min, and then the reaction was started by adding, through manual mixing, either IF2 (1 M), EF-G (0.1 M), RF3 (1 M), or the corresponding volumes of Storage Buffer. The kinetics of phosphate release was monitored continuously at 360 nm for 2 min. The initial rate of phosphate (P i ) release was determined from the slope of the linear part of the kinetic curve using an appropriate conversion coefficient for P i that was determined from a standard curve prepared in parallel. Initial rates are expressed in M of P i released/min.

Results
Bacterial METTL20 Homologues Have Protein Methyltransferase Activity-During our efforts to characterize the human MTase METTL20 (13), we noticed the presence of proteins displaying high sequence homology to METTL20 in a few bacteria. The bacterial METTL20 proteins are primarily found in ␣-proteobacteria, especially in the Rhizobiales order, where they seem to be ubiquitous. In addition, some scattered members of ␤-proteobacteria, ⑀-proteobacteria, and actinobacteria also have a METTL20-like protein. Interestingly, the majority of these proteins have already been annotated as "50S ribosomal protein L11 methyltransferase" because they were found to display some sequence homology to the lysine-specific MTase PrmA from E. coli that targets ribosomal protein L11 (36). However, as illustrated by the sequence alignment shown in Fig. 1A, they are much less similar to PrmA than to human METTL20, suggesting that their function is different from that of PrmA. Besides containing the 7BS MTase hallmark motifs (I, Post I, and II), the bacterial METTL20 proteins have the To investigate the function of bacterial METTL20 proteins, we expressed and purified such enzymes from A. tumefaciens and R. etli as recombinant proteins in E. coli. Reasoning that an A. tumefaciens cell extract may contain substrates for bacterial METTL20 proteins, we incubated such extracts with recombinant enzymes in the presence of 3 H-labeled AdoMet and then analyzed the samples by SDS-PAGE and fluorography. Interestingly, extracts incubated with AtMETTL20 showed radiolabeling of two proteins with apparent molecular masses of ϳ28 and ϳ15 kDa (Fig. 1B, lane 5). Two putatively inactive mutant AtMETTL20 enzymes, D81A and D105A, where key catalytic residues important for AdoMet binding had been mutated (indicated in Fig. 1A) were included as negative controls. Extracts incubated with these AtMETTL20 mutants showed no labeling of the ϳ28and ϳ15-kDa substrates, thus excluding the possibility that the labeling was due to an E. coli-derived contaminant rather than the activity of the recombinant enzyme (Fig. 1B). Also, both the ϳ28and ϳ15-kDa substrates were methylated by the METTL20 homologue from R. etli, whereas only the ϳ28-kDa protein was labeled by human METTL20 (Fig. 1C). These results indicate that bacterial homologues of METTL20 are protein MTases with a substrate specificity partially overlapping with that of human METTL20.
Identification of ETF␤ and RpL7/L12 as Likely Substrates of Bacterial METTL20 -To reveal the identity of the ϳ28and the ϳ15-kDa substrates of bacterial METTL20 proteins, A tumefaciens extracts were fractionated using ion exchange chromatography, and the fractions were subjected to methyla-tion by AtMETTL20 and analyzed by fluorography ( Fig. 2A). The ϳ28-kDa substrate bound poorly to an S-column and was found mostly in the S-column flow-through, whereas the ϳ15-kDa substrate bound partially to the S-column and could be eluted predominantly between 0.15 and 0.3 M NaCl (0.3S) ( Fig.  2A, lane 5). When the S-column flow-through fraction was reapplied onto a Q-column, both the ϳ28and the ϳ15-kDa proteins were found to partially bind and were predominantly eluted between 0.05 and 0.15 M NaCl (0.15Q) (Fig. 2A, lane 9). For the 0.15Q fraction, regions of interest were excised from a Coomassie-stained protein gel (Fig. 2B) and subjected to trypsin digestion followed by protein identification through MS. Because human METTL20 was previously shown to methylate ETF␤ (13,14) and because the ϳ28-kDa substrate from A. tumefaciens was methylated also by human METTL20, we anticipated this protein to be A. tumefaciens ETF␤ (AtETF␤), which has a molecular mass of ϳ28 kDa. Indeed, AtETF␤ was identified as the predominant protein in the ϳ28-kDa region of the 0.15Q fraction (Table 1). Also, Lys-193 in AtETF␤, which corresponds to one of the target sites of human METTL20 in human ETF␤, was found to be methylated in A. tumefaciens extracts and to exist in several different methylation states (un-, mono-, and dimethylated) (Fig. 2C).
Ribosomal protein RpL7/L12 from A. tumefaciens (AtRpL7/ L12) was identified as the most abundant protein in the ϳ15-kDa region of the 0.15Q fraction (Table 2), and also for this Dual Specificity METTL20 Homologue from A. tumefaciens APRIL 29, 2016 • VOLUME 291 • NUMBER 18

JOURNAL OF BIOLOGICAL CHEMISTRY 9585
protein, a methylated lysine residue was identified, namely Lys-86, which was found primarily in the monomethylated state (Ͼ97% monomethylation) (Fig. 2D). E. coli RpL7/L12 has been reported to be associated with ribosomes (37), and to further investigate whether AtRpL7/L12 is likely to be the ϳ15-kDa substrate, ribosomes isolated from A. tumefaciens were subjected to in vitro methylation. Reassuringly, labeling of the ϳ15-kDa substrate, but not of the ϳ28-kDa substrate, was observed (Fig. 2E, lane 5). We used MS to establish the identity of the ϳ15-kDa substrate present in ribosomes, and again AtRpL7/L12 was found to be the most abundant protein in the ϳ15-kDa region (Table 3). In summary, the above results suggested two proteins, namely AtETF␤ and AtRpL7/L12, as likely substrates of AtMETTL20.
Lys-193 and Lys-196 in A. tumefaciens ETF␤ Are Methylated by METTL20 in Vitro-Based on sequence similarity, ETFs have been categorized into three main groups: I, II, and III (20). Interestingly, group I includes ETF from humans and other eukaryotes but also ETF from a few bacteria, including those possessing METTL20 proteins. In contrast, other bacteria have rather different ETFs belonging to group II or III. Thus, a sequence alignment of the region encompassing the putative methylation sites in ETF␤ (Lys-193 and Lys-196 in AtETF␤) shows a high degree of sequence homology between ETF␤ from human and A. tumefaciens, whereas the sequence of the E. coli ETF␤ homologue FixA (also known as YaaQ), which belongs to group III, diverges from these two (Fig. 3A).
To verify that ETF␤ is a substrate of AtMETTL20, we generated recombinant ETF␣/␤ dimers to be used in enzymatic assays. To also investigate the cross-species compatibility of these enzymes and substrates, we investigated the activity of METTL20 from both human and A. tumefaciens on ETF␣/␤ from both these organisms. The results showed that human and bacterial METTL20 were both active on ETF␤ from these two organisms, although the enzymes were slightly more active on their cognate substrates (Fig. 3B). In accordance with the absence of a METTL20 homologue in E. coli, AtETF␤ was unmethylated at Lys-193 when expressed in this host (Fig. 3C). Most importantly, incubation of E. coli-expressed AtETF␤ with AtMETTL20 shifted the state of Lys-193 into predominantly mono-and dimethylation, showing that AtMETTL20 can indeed methylate this residue (Fig. 3C).
Our previous studies of human METTL20 had demonstrated that human ETF␤ is methylated on two neighboring residues, namely Lys-200 and Lys-203, which correspond to Lys-193 and Lys-196 in the A. tumefaciens protein, respectively (Fig. 3A). However, the MS analysis of tryptic peptides described above did not yield information on the methylation state of Lys-196. To address whether AtETF␤, similarly to its human counterpart, is methylated at both Lys-193 and Lys-196, we mutated  (Table 3). FT, flow-through.   Dual Specificity METTL20 Homologue from A. tumefaciens APRIL 29, 2016 • VOLUME 291 • NUMBER 18 these residues together or individually into arginine and investigated the effect on METTL20-mediated methylation. Indeed, whereas methylation was substantially reduced (to ϳ60%) when these residues were mutated individually, methylation was completely abolished in the K193R/K196R double mutant (Fig. 3, D and E). To substantiate these findings, we analyzed by MS an Asp-N-generated peptide encompassing Lys-193 and Lys-196 in ETF␤ before and after treatment with AtMETTL20.
In the absence of AtMETTL20 treatment, this peptide displayed in the case of all the ETF␤ variants a mass corresponding to a complete lack of methylation (Fig. 3F). After AtMETTL20 treatment, the peptide from wild-type ETF␤ displayed primarily tri-and tetramethylation, whereas the K193R and K196R mutants were found predominantly in the dimethylated state (Fig. 3F). (Note that in principle independent confirmation of these findings could have been provided by MS/MS sequencing of the Asp-N peptide, but we were unable to obtain such data). Taken together, the above results demonstrate that AtMETTL20 specifically and independently methylates Lys-193 and Lys-196 in AtETF␤ and indicate that AtMETTL20 is a non-processive enzyme that typically introduces two methyl groups at each of these sites.

Dual Specificity METTL20 Homologue from A. tumefaciens
METTL20 Modulates the Ability of ETF to Receive Electrons from FAD-containing Dehydrogenases-In humans, ETF acts as a mobile electron carrier that shuttles electrons from several mitochondrial DHs to ETF-QO. Interestingly, the residues targeted by METTL20 (Lys-200 and Lys-203) were shown to be a part of the recognition loop that interacts with one of these DHs, namely MCAD (19). Sequence homology with characterized ETF-dependent DHs from other organisms indicates the presence of more than 10 such enzymes in A. tumefaciens, but these remain uncharacterized experimentally. To study the possible effects of methylation on the ability of AtETF to receive electrons from DHs, we cloned the genes encoding various putative ETF-dependent DHs from A. tumefaciens, and we were able to express and purify several of the corresponding recombinant proteins from E. coli. To assess the ETF dependence and substrate specificity of these DHs, we used a colorimetry-based assay where DCIP served as the final electron acceptor. ETF-dependent activity of the DHs was monitored as reduction-induced bleaching of DCIP (Fig. 4A). For the enzymes belonging to the ACAD family, the activity of the recombinant protein was tested against a panel of relevant CoA-containing substrates. We found one of the DHs (NP_356237.1), which already is annotated as an A. tumefaciens glutaryl-CoA dehydrogenase, to display the expected specificity toward glutaryl-CoA (Fig. 4B). Similarly, we identified another enzyme (NP_357132.2) that preferred isovaleryl-CoA, thereby representing an A. tumefaciens isovaleryl-CoA dehydrogenase (Fig. 4B). A third enzyme (NP_357135.2) denoted A. tumefaciens short-chain acyl-CoA dehydrogenase oxidized both butyryl-and isobutyryl-CoA, thereby displaying activities characteristic of both short-chain acyl-CoA dehydrogenase and isobutyryl-CoA dehydrogenase. Finally, a fourth enzyme (NP_353529.2)denotedA. tumefacienslong-chainacyl-CoAdehydrogenase had octanoyl-and palmitoyl-CoA as preferred substrates, thereby displaying both long-chain acyl-CoA dehydrogenase-and MCAD-like activities. In addition to these putative ACAD family members, we characterized another enzyme (NP_354625.1) denoted AtDMGDH as it showed high sequence homology to human DMGDH. The enzyme indeed represented a bona fide DMGDH as it was highly active in oxidizing dimethylglycine and preferred it over sarcosine (Fig. 4C). Importantly, all the above DHs showed the expected dependence on AtETF (Fig. 4D).
Our previously published results (13) indicated that methylation of human ETF by METTL20 can modulate its ability to extract electrons from ACADs. To test that this was also true in the A. tumefaciens system, we measured the ability of AtETF to extract electrons from AtACADs in the absence or presence of AtMETTL20. In these experiments, we included as a negative control an AtMETTL20 mutant, D81A, where a key catalytic residue was mutated and for which no enzymatic activity could be detected (Fig. 4E). We observed that in the presence of wildtype AtMETTL20, but not the D81A mutant, AtETF displayed reduced ability to extract electrons from all tested AtACADs (Fig. 4F). In contrast, the presence of wild-type AtMETTL20 seemed to have a negligible effect on the ability of AtETF to extract electrons from AtDMGDH as monitored by the DCIP reduction assay (Fig. 4G, left panel).
We also considered the possibility that methylation of ETF may not only influence its ability to receive electrons from the various DHs but also the ability to further transfer electrons to ETF-QO. Thus, we also purified as a recombinant protein a putative ETF-QO orthologue from A. tumefaciens (NP_354018.1) and tested its ability to mediate transfer of electrons from AtETF to the electron acceptor CoQ 1 as monitored through a change in absorption (275 nm) of CoQ 1 (Fig. 4A). Somewhat surprisingly, although AtETF-mediated electron transfer to DCIP was observed with a number of different DHs (see above), the reduction of CoQ 1 through the DH/ETF/ ETF-QO system was only observed with AtDMGDH but not with any of the tested AtACADs. Notably, we found that wild-type AtMETTL20, but not the inactive D81A mutant, reduced the ability of AtETF to extract electrons from AtD-MGDH and shuttle them to AtETF-QO (Fig. 4G, right  panel). In summary, the above results indicate that methylation of AtETF␤ by AtMETTL20 impairs the ability of AtETF to mediate electron transfer between ETF-dependent dehydrogenases and ETF-QO.
METTL20 Methylates A. tumefaciens RpL7/L12 on Lys-86 in Vitro-The results shown in Fig. 2 suggested AtRpL7/L12 to be a substrate for AtMETTL20. To further investigate this, we expressed and purified recombinant AtRpL7/L12 from E. coli, assessed AtMETTL20-mediated methylation of AtRpL7/ L12 in an MTase assay using [ 3 H]AdoMet, and detected protein methylation by fluorography. Reassuringly, we observed AtMETTL20-mediated methylation of wild-type AtRpL7/L12 but not of the K86A mutant where the putative methylation site had been mutated (Fig. 5A). RpL7/L12 from E. coli shows high sequence similarity to AtRpL7/L12 (Fig. 5B) and has been shown to be monomethylated on the corresponding site, Lys-82, by a yet unidentified MTase (29). Accordingly, we observed by MS analysis that AtRpL7/L12 expressed in E. coli was already partially methylated on Lys-86 (Fig. 5, C and D). Interestingly and in agreement with previous studies of endogenous E. coli RpL7/L12 (38), we found methylation at Lys-86 of recombinant E. coli-expressed AtRpL7/L12 to decrease dramatically with increasing growth temperature; the protein was mostly monomethylated when expressed at 16°C (ϳ85% monomethylated), whereas it existed predominantly in the unmethylated state when expressed at 37°C (ϳ95% unmethylated) (Fig. 5C). Importantly, in either case, the level of Lys-86 methylation could be further increased by treatment of recombinant AtRpL7/L12 with AtMETTL20 (Fig. 5C). In summary, the above results clearly demonstrate that AtMETTL20 can methylate Lys-86 in AtRpL7/L12 in vitro and show that E. coli expresses an MTase capable of targeting the same site and that likely is also responsible for methylation of Lys-82 in endogenous RpL7/L12 from E. coli.
Because E. coli possesses an enzyme capable of methylating AtRpL7/L12, the possibility existed that the apparent activity of recombinant AtMETTL20 on AtRpL7/L12 originated from an E. coli-derived contamination rather than from the recombinant enzyme itself. However, no methylation of AtRpL7/L12 was observed either with Homo sapiens (Hs) METTL20 or with the inactive AtMETTL20 D81A mutant, thus excluding this possibility (Fig. 5E).
To further investigate the observed dual specificity of AtMETTL20 toward AtRpL7/L12 and AtETF␤, we assessed the relative activity of the enzyme on these two substrates. Thus, in vitro methylation reactions were performed using a constant concentration of recombinant AtETF␣/␤ or AtRpL7/L12 (expressed at 37°C) and varying concentrations of AtMETTL20. Although equal amounts of substrates were used, a substantially higher methylation level (ϳ2.5-fold higher) was observed for AtETF␤ compared with AtRpL7/12 (Fig. 5F), which may be explained by the higher number of methyl groups received by AtETF␤ (compare Figs. 3F and 5C). Otherwise, the titration curves for the two substrates were similar, supporting the notion that methylation of both substrates is biologically relevant.
AtMETTL20 Is Responsible for in Vivo Methylation of AtETF␤ and AtRpL7/L12 in A. tumefaciens-Next, we set out to investigate whether AtMETTL20 is also responsible for in vivo methylation of AtETF␤ and AtRpL7/L12 in A. tumefaciens as well as to further investigate functional consequences of AtMETTL20-mediated methylation. For these purposes, we generated an A. tumefaciens strain with AtMETTL20 gene KO. AtMETTL20 KO cells were viable, and their growth in LB or glucose-supplemented minimal medium was indistinguishable from that of wild-type cells (Fig. 6A). To indirectly assess AtRpL7/L12 and AtETF␤ methylation status in KO versus wildtype bacteria, extracts were subjected to AtMETTL20-mediated methylation. Substantially higher levels of METTL20-induced methylation of both AtRpL7/L12 and AtETF␤ were observed with the KO bacteria, indicating that these proteins are hypo-or unmethylated in the KO bacteria (Fig. 6B). A similar pattern was observed when human METTL20 was used except that this enzyme methylated AtETF␤ but not AtRpL7/ L12 as observed previously.
The methylation status of AtETF␤ from wild-type and AtMETTL20 KO bacteria was also directly assessed by MS analysis of the Asp-N-generated peptide encompassing both . Whereas these two residues together carried up to four methyl groups in the wild-type bacteria, they were exclusively found in the unmethylated state in the KO bacteria (Fig. 6C). The methylation status of AtRpL7/L12 present in the ribosomal fraction was directly assessed by MS analysis of the Glu-C-generated peptide containing Lys-86. AtRpL7/L12 was found predominantly in the monomethylated state in the wild-type bacteria but was found to be completely unmethylated in the KO bacteria (Fig. 6D). Taken together, the above experiments firmly establish AtMETTL20 as the enzyme responsible for lysine methylation of AtETF␤ and AtRpL7/L12 in vivo.
Several translation factors with GTPase activity bind to the ribosome through the so-called RpL7/L12 stalk, and the GTPase activity has been shown to be stimulated by RpL7/L12 and ribosome binding (39,40) Thus, we tested ribosomes from wild-type and METTL20 KO bacteria for their ability to stimulate such GTPase activity and found that they were similarly active in stimulating the GTPase activity of EF-G (Fig. 6E) as well as of IF2 and RF3 (data not shown). In summary, these results demonstrate that AtMETTL20 is a dual specificity KMT and the sole enzyme responsible for lysine methylation of ETF␤ and RpL7/L12 in A. tumefaciens.

Discussion
We have demonstrated here, using a wide range of in vitro and in vivo approaches, that METTL20 from A. tumefaciens is a novel bacterial KMT with unprecedented dual specificity, targeting both ETF␤ and RpL7/L12. Taken together with the corresponding results obtained with METTL20 from R. etli, this indicates that other bacterial METTL20-like proteins also are likely to display similar activities. Moreover, we have here established lysine methylation of ETF␤ as the first example of a lysine-specific modification that occurs both in bacteria and humans.
Protein lysine methylation is much more prevalent in eukaryotes than in prokaryotes. For example, the SET domain KMTs, which constitute close to a third of the MTases in humans (6), are rare in bacteria where they do not target endogenous bacterial proteins but rather act as effectors that modify histone proteins in the infected host (41). Accordingly, almost all the bacterial KMTs acting on endogenous proteins belong to the 7BS MTase family. Of these, PrmA, which targets ribosomal protein RpL11, has been studied in the most detail and appears to be present in all bacteria (36). Pseudomonas and certain other species of ␥-proteobacteria and firmicutes possess a 7BS MTase, EftM (EF-Tu-modifying enzyme), which methylates Lys-5 in translation elongation factor EF-Tu, thus promoting bacterial infectivity (42) Similarly, two related 7BS MTases, PKMT1 and PKMT2, that methylate the outer membrane protein OmpB on several Lys residues in Rickettsia subspecies, are also important for bacterial virulence (43,44). Thus, bacterial KMTs seem to primarily target components of the translational machinery as well as determinants of bacterial virulence. The AtMETTL20 enzyme reported here is apparently unrelated (beyond the shared 7BS fold) to the above enzymes, thus representing a new type of bacterial KMT.
We found that AtETF␤ is methylated by AtMETTL20 on both Lys-193 and Lys-196, corresponding to the previously reported target sites of the human enzyme on HsETF␤, namely Lys-200 and Lys-203 (13,14). These residues are part of the recognition loop, which has been shown to interact with human FIGURE 4. AtMETTL20 impairs the ability of AtETF to receive electrons from FAD-containing dehydrogenases. A, scheme of DCIP and CoQ 1 reduction assays. B, identification of acyl-CoA substrate specificity for the A. tumefaciens ACADs NP_356237.1, NP_357132.2, NP_357135.2, and NP_353529.2. For each AtACAD, the ability to oxidize the indicated acyl-CoA was monitored by the DCIP reduction assay, and the activity is indicated relative to that obtained with the optimal substrate. Data from three independent experiments (run in duplicate or triplicate) were averaged. Sample variation is expressed as S.D. (error bars) (n ϭ 7). C, identification of A. tumefaciens protein NP_354625.1 as DMGDH. Shown is the ability of NP_354625.1 to oxidize dimethylglycine (DMG) relative to sarcosine (SAR) monitored as in B. Replicates and statistical analysis are as in B (n ϭ 7). D, substrate oxidation by A. tumefaciens dehydrogenases is AtETF-dependent. Oxidation of the indicated substrates by particular AtDHs was monitored by DCIP reduction assay in the presence of increasing concentrations of AtETF. The experiment was performed three times with duplicate samples. Data from a representative experiment are shown. E, inability of the AtMETTL20 mutant D81A to mediate methylation of AtETF␤. AtETF was incubated with [ 3 H]AdoMet and either WT or D81A mutated AtMETTL20, and ETF␤ methylation was assessed by fluorography (1-day exposure). F, AtMETTL20 impairs the ability of AtETF to extract electrons from AtACADs. AtETF was incubated with AdoMet and AtMETTL20 (WT or D81A mutant), and the ability of AtETF to receive electrons from AtACADs utilizing the indicated acyl-CoA substrates was monitored by the DCIP reduction assay. Activity is expressed relative to untreated ETF. Sample variation is expressed as S.D. (error bars) (n ϭ 8). * denotes a p value Ͻ0.01; n.s. denotes "not significant." G, AtMETTL20 impairs the ability of AtETF to shuttle electrons from AtDMGDH to AtETF-QO (NP_354018.1). AtETF was incubated with mutant or WT AtMETTL20 as in F, and its ability to mediate electron transfer from AtDMGDH (utilizing dimethylglycine as substrate) to either DCIP (left) or CoQ 1 (right) was monitored. Activity is expressed relative to untreated ETF, and statistical analysis was performed as in F (n ϭ 8). SCAD, short-chain acyl-CoA dehydrogenase; LCAD, long-chain acyl-CoA dehydrogenase; IVD, isovaleryl-CoA dehydrogenase; GCDH, glutaryl-CoA dehydrogenase.
MCAD (19), and we have previously shown that methylation of HsETF␤ by human METTL20 impairs the ability of HsETF to extract electrons from human MCAD and glutaryl-CoA dehydrogenase (13). In the present work, we report similar findings, i.e. that enzymatically active AtMETTL20 inhibits electron flow from various A. tumefaciens ETF-dependent dehydrogenases to ETF.
Three different classes of ETFs are found in bacteria (20), and interestingly, ␣-proteobacteria, which represent the evolutionary precursors of mitochondria, have an ETF that shows high similarity to eukaryotic ETFs (AtETF␤ and HsETF␤ show ϳ60% sequence identity). In contrast, the E. coli equivalent of ETF␤, FixA, is more distantly related to HsETF␤, and we did not observe AtMETTL20-mediated methylation of FixA in vitro (data not shown). Although a eukaryotic-like ETF␤ appears to be present in all ␣-proteobacteria, a METTL20-encoding gene is found only in a subset of these, thus resembling the scattered distribution of METTL20 orthologues in Eukaryota. Taken together, this indicates that Lys methylation is not required for basic ETF␤ function but rather has a biologically relevant modulatory role that gives a selective advantage in certain organisms.
We here report AtMETTL20-catalyzed lysine methylation of AtRpL7/L12 at Lys-86, and the corresponding residue appears to be commonly methylated in prokaryotes. Proteomic studies of bacterial ribosomal proteins have reported di-or monomethylation at this residue in E. coli (29), Caulobacter crescentus (28), and Rhodopseudomonas palustris (27), and the reported molecular mass of Thermus thermophilus RpL7/L12 also suggests methylation (26). However, this modification appears not to be present in all bacteria as Bacillus subtilis RpL7/L12 was found to be unmodified (45). Interestingly, except for R. palustris, none of the bacteria with reported RpL7/L12 methylation possess a putative METTL20 orthologue. This indicates that bacteria independently have evolved at least two types of KMTs for methylation of RpL7/L12, i.e. METTL20 and the yet unidentified KMT(s) predicted to exist in the METTL20-less bacteria with methylated RpL7/L12, such as E. coli. This likely example of convergent evolution strongly suggests that RpL7/L12 methylation is functionally important, although we were not able to detect any functional consequences of AtMETTL20 gene knock-out. Notably, the extent of RpL7/L12 methylation in E. coli has been reported to be enhanced at low temperature (38), and in agreement with this, we found that E. coliexpressed recombinant AtRpL7/L12 showed a dramatically higher methylation level when the bacteria were grown at 16°C relative to 37°C. This suggests that RpL7/L12 methylation may play an adaptive role in regulating protein translation in response to external cues.
The majority of the 7BS KMTs characterized so far methylate a single residue in a unique substrate (or in a group of highly related substrates), but some enzymes, e.g. the bacterial PrmA and PKMT1/PKMT2 as well as human METTL20, have been shown to methylate several Lys residues within the same substrate (13,14,23,43). However, AtMETTL20 is the first 7BS KMT shown to methylate two different substrates, and thus it possesses unprecedented dual substrate specificity. The two substrates methylated by AtMETTL20, namely AtETF␤ and AtRpL7/L12, appear functionally and structurally unrelated. Thus, it is pertinent to ask whether both these methylation events are biologically relevant, and as elaborated above, we believe this is likely the case. This is further supported by our observations that both substrates were methylated in vivo and were similarly amenable to AtMETTL20-mediated methylation in vitro.
As METTL20-mediated RpL7/L12 methylation occurs in ␣-proteobacteria, the evolutionary precursors of mitochondria, the possibility existed that HsMETTL20 methylates mRpL12, the mitochondrial counterpart of bacterial RpL7/L12. However, human mRpL12 does not have a lysine residue in the position corresponding to Lys-86 in AtRpL7/L12 (Fig. 5A), and we were unable to detect any in vitro MTase activity of HsMETTL20 on either AtRpL7/L12 (Fig. 5E) or mRpL12 (data not shown). Taken together, this suggests that during evolution FIGURE 6. AtETF␤ and AtRpL7/L12 are methylated by AtMETTL20 in vivo. A, growth rates of WT and AtMETTL20 KO A. tumefaciens strains. The bacteria were grown at 28°C in LB or minimal medium with 0.2% glucose, and the bacterial growth was monitored by measuring A 600 . The experiment was performed three times. Averaged data from a representative experiment (run in triplicate) are shown. B, increased AtMETTL20-mediated methylation in extracts from AtMETTL20 KO cells indicates hypomethylation of AtETF␤ and AtRpL7/L12. Extracts from WT or KO cells were incubated with [ 3 H]AdoMet and either AtMETTL20 or HsMETTL20, and methylation was assessed by fluorography (3-day exposure). C, AtETF␤ is unmethylated in AtMETTL20 KO cells. AtETF␤ from WT or KO cells was Asp-N-digested, and methylation of the AtETF␤-derived peptide, encompassing residues 177-199, was analyzed by MS. Shown are the mean relative intensities of signals gated for the different methylation states of the peptide with error bars indicating the range of values from three independent experiments. Red color indicates the location of the methylated residues Lys-193 and Lys-196 within the peptide sequence. D, AtRpL7/L12 is unmethylated in AtMETTL20 KO cells. AtRpL7/L12 from ribosomes of WT or KO cells was Glu-C-digested, and methylation of AtRpL7/L12-derived peptide, encompassing residues 77-87, was analyzed by MS. Shown are the mean relative intensities of signals corresponding to the different methylation states of Lys-86 in AtRpL7/L12 with error bars indicating the range of values from three independent experiments. Red color indicates the location of the methylated residue Lys-86 within the peptide sequence. E, stimulation of GTPase activity of recombinant AtEF-G by ribosomes isolated from WT or KO cells. Experiment was run twice in duplicate; all data points are shown. eukaryotic METTL20 has lost the ability to methylate RpL7/ L12 or alternatively that bacterial METTL20 proteins acquired the ability to methylate RpL7/L12 after the endosymbiotic event that led to the generation of mitochondria.