Uncovering human METTL12 as a mitochondrial methyltransferase that modulates citrate synthase activity through metabolite-sensitive lysine methylation

Lysine methylation is an important and much-studied posttranslational modification of nuclear and cytosolic proteins but is present also in mitochondria. However, the responsible mitochondrial lysine-specific methyltransferases (KMTs) remain largely elusive. Here, we investigated METTL12, a mitochondrial human S-adenosylmethionine (AdoMet)-dependent methyltransferase and found it to methylate a single protein in mitochondrial extracts, identified as citrate synthase (CS). Using several in vitro and in vivo approaches, we demonstrated that METTL12 methylates CS on Lys-395, which is localized in the CS active site. Interestingly, the METTL12-mediated methylation inhibited CS activity and was blocked by the CS substrate oxaloacetate. Moreover, METTL12 was strongly inhibited by the reaction product S-adenosylhomocysteine (AdoHcy). In summary, we have uncovered a novel human mitochondrial KMT that introduces a methyl modification into a metabolic enzyme and whose activity can be modulated by metabolic cues. Based on the established naming nomenclature for similar enzymes, we suggest that METTL12 be renamed CS-KMT (gene name CSKMT).

Thus far, only a single human mitochondrial KMT has been characterized, namely ETF␤-KMT (gene name ETFBKMT; alias METTL20), which is a member of MTF16 and targets Lys-200 and Lys-203 in the ␤-subunit of electron transfer flavoprotein (ETF␤) (17,19). However, many mitochondrial proteins have been shown to contain methylated lysine residues, and several uncharacterized MTases have a predicted mitochondrial localization, suggesting the existence of additional mitochondrial KMTs (22,23). One interesting candidate is METTL12, which shows sequence similarity to established KMTs and has a predicted mitochondrial localization (20,23).
Citrate synthase (CS) resides in the mitochondrial matrix where it catalyzes the condensation of oxaloacetate (OAA) with acetyl-CoA to form citrate and CoA, i.e. the first committed and irreversible step of the Krebs cycle. Numerous studies have investigated the regulation of CS activity (24 -26). It is generally accepted that the overall rate of the CS-catalyzed reaction is mainly determined by the availability of the substrates acetyl-CoA and OAA, which are usually present in mitochondria at concentrations much lower than those required to saturate CS. In addition, CS was reported to be inhibited by citrate and succinyl-CoA (27,28). So far, regulation of CS activity by post-translational modification has not been reported. However, mammalian CS has been reported to be subject to lysine methylation; CS from both human and pig were reported to be trimethylated at Lys-395 (22,29) (amino acid numbering according to UniProt), which is localized in the CS active site (30).
In the present study we have taken an activity-based approach to establish the function of METTL12. We detected a single substrate for recombinant METTL12 in METTL12 knock-out (KO) cell extracts, and we found this substrate to be CS. Through extensive in vitro and in vivo studies we demonstrate that METTL12 is responsible for methylation of Lys-395 in CS. We also found that CS activity was diminished by methylation and that METTL12-dependent methylation was inhibited by OAA and AdoHcy.

Cloning and mutagenesis
Cloning of human open reading frames and mutagenesis were performed as described previously (17). All constructs were sequence-verified.

Bioinformatics analysis
NCBI Basic Local Alignment Search Tool (BLAST) was used to identify protein sequences homologous to human METTL12 (31). Multiple sequence alignments were generated using algorithms embedded in the Jalview (v2.8) interface (32).

Cell cultivation and generation of stable cell lines
Human cell lines were grown at standard conditions, typically in DMEM GlutaMAX medium supplemented with 10% (v/v) fetal bovine serum (FBS), 100 units/ml of penicillin, and 0.1 mg/ml streptomycin (P/S). Human HAP1 cells were grown in Iscove's modified Dulbecco's GlutaMAX medium (IMDM) supplemented with 10% FBS and P/S. The generation of METTL12 KO HAP1 cells was initiated as a custom (nonexclusive) project with Horizon Genomics (Vienna, Austria), and these cells are now commercially available (Horizon Discovery HZGHC000536c011). Genomic ablation of the METTL12 gene was performed using the CRISPR-Cas9 technology, with guide RNA designed to target the METTL12 gene upstream of motif "Post I" (Fig. 2A). Individual clones were selected by limiting dilution, and frame-shifting events within the METTL12 gene were determined by sequencing of genomic DNA. Complementation of METTL12 KO cells was performed by transfecting cells with a p3ϫFLAG-CMV-14-derived plasmid that encoded either wild-type or D107Amutated C-terminally 3ϫFLAG-tagged METTL12 using the FuGENE6 Transfection Reagent (Roche Applied Science). Transfected cells were selected with 1 mg/ml Geneticin (Gibco) and expanded in medium containing Geneticin. Individual clones were screened by Western blot for the presence of 3ϫFLAG tag, citrate synthase (CS), and GAPDH (as loading control) using anti-FLAG (Sigma, F1804), anti-citrate synthase (Proteintech, 16131-1-AP), and anti-GAPDH (Abcam, ab9485) antibodies, respectively.

Transient transfection and fluorescence microscopy
HeLa cells were transiently transfected with pEGFP-N1derived plasmids (Clontech), encoding human full-length METTL12 or only its N terminus (consisting of amino acids 1-30) fused to the N terminus of enhanced green fluorescent protein (EGFP). Cells were analyzed by confocal fluorescence microscopy 24 h after transfection. Living cells were stained with 50 nM MitoTracker Deep Red FM (Life Technologies) and 0.5 g/ml Hoechst 33258 (Sigma) to visualize the mitochondria and the nuclei, respectively. Cells were imaged using an Olympus FluoView 1000 (Ix81) confocal fluorescence microscopy system with a PlanApo 60ϫ NA 1.1 oil objective (Olympus). The different fluorophores were excited at 405 nm (Hoechst), 488 nm (EGFP), and 635 nm (MitoTracker), and Kalman averaging (n ϭ 3) was used to record multichannel images. The fluorescent signals emitted from EGFP, MitoTracker, and Hoechst were acquired through green, red, and blue channels, respectively, and merged.
Proteins were Ͼ90% pure as assessed by SDS-PAGE and Coomassie Blue staining. Protein concentrations were measured using Pierce BCA protein assay kit (Thermo Fisher Scientific).

Preparation and fractionation of cell extracts
Crude mitochondria-enriched (mitochondrial) fraction was prepared using the Mitochondria Isolation Kit for Cultured Cells (Thermo Fisher Scientific) according to the manufacturer's instructions by following the reagent-based protocol.
Human cells or mitochondrial fractions were lysed for 5 min at 4°C in lysis buffer supplemented with 1 mM dithiothreitol and protease inhibitor mixture (catalog no. P8340; Sigma), and the resulting lysates were cleared by centrifugation. Samples of frozen pig organs were fragmented mechanically, lysed in lysis buffer (supplemented as above), sonicated, and cleared by centrifugation.
Mitochondrial or whole-cell extracts were fractionated at 4°C by ion-exchange chromatography using the Pierce Strong Cation Exchange (S) Spin Column (Thermo Fisher Scientific). First, NaCl concentration was reduced to 50 mM by diluting cell lysates with dilution buffer (50 mM Tris-HCl, pH 7.4, 5% glycerol), and then extracts were applied onto S-column equilibrated in dilution buffer. Material bound to the S-column was eluted in 100-l aliquots by a step gradient of increasing NaCl concentrations prepared in dilution buffer, similarly as previously described (33).

In vitro methyltransferase assay using [ 3 H]AdoMet
To test the MTase activity of METTL12 on cellular material, 10-l reactions were assembled on ice containing 1ϫ storage buffer, 50 pmol of recombinant METTL12, 40 -60 g of protein from cell extracts, and 0.5 Ci of [ 3 H]AdoMet (Perkin-Elmer Life Sciences) ([AdoMet] total ϭ 0.64 M, specific activity ϭ 78.2 Ci/mmol). Reaction mixtures were incubated at 30°C for 1 h and analyzed by SDS-PAGE and fluorography, similarly as previously described (18). Typically, fluorography experiments were performed three times, with similar results, and data from a representative experiment 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 nonradioactive AdoMet (1 mM). When indicated, the methylation reaction contained additionally 1 unit/ml AdoHcy hydrolase (AdoHcyase) from rabbit erythrocytes (Sigma). CS present in extracts was up-concentrated by loading cleared lysates on the S-column and eluting bound material with elution buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5% glycerol) with added protease inhibitor mixture. Proteins were resolved by SDS-PAGE and stained with Coomassie, the portion of gel containing the protein of interest was excised and subjected to in-gel trypsin (Sigma) or chymotrypsin (Roche Applied Science) digestion, and the resulting proteolytic fragments were analyzed by liquid chromatography MS, similarly as previously described (18).
MS data were analyzed using an in-house-maintained human and pig protein sequence databases 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, cysteine carbamidomethylation, and lysine and arginine methylation were selected as variable modifications. MS/MS spectra of peptides corresponding to methylated CS were manually searched by Qual Browser (v2.0.7).

Citrate synthase activity assay
The activity of CS was assayed by reaction of 5,5-dithiobis-(2-nitrobenzoic acid) (DTNB) with the free thiol group of CoA (generated as result of citrate formation from OAA and acetyl-CoA), which leads to formation of yellow 5-thio-2nitrobenzoic acid (TNB) that is detected spectrophotometrically at 412 nm (34). To determine the effect of METTL12mediated methylation on CS enzymatic activity, 0.5 M CS was incubated for 1 h at 30°C with 2 M METTL12, either wild-type or D107A-mutated, in the absence or presence of 1 mM AdoMet or AdoHcy, all in 1ϫ storage buffer. Next, the incubation mixture was diluted 250ϫ in storage buffer supplemented with 100 M DTNB and 300 M acetyl-CoA. Samples were left to equilibrate at room temperature for 1 min, and then the reaction was started by adding 300 M OAA through manual mixing. For titration experiments the reaction was started by adding varying amounts of OAA. The kinetics of TNB formation was monitored continuously at 412 nm for 1 min using a Shimadzu UV-1601 spectrophotometer, and the initial rate of reaction was calculated from the slope of the linear part of the kinetic curve and expressed as change in absorption at 412 nm/min (⌬A 412 /min). When indicated, the results from individual experiments were normalized for the activity of CS in the appropriate control and averaged.

Human METTL12 methylates mitochondrial citrate synthase Statistical analysis
The independent two-sample Student's t test was used to evaluate the probability (p value) that the means of two populations are not different.

METTL12 is a mitochondrial 7BS MTase mainly found in vertebrates
We recently reported that the uncharacterized human 7BS MTase METTL12 is likely a KMT due to its sequence similarity with other established human KMTs (20). Also, METTL12 contains a putative MTS (23) and because only a single mitochondrial KMT has been characterized thus far, we found it of particular interest to investigate the function of METTL12.
Protein sequence searches revealed that putative METTL12 orthologues are mainly restricted to vertebrates, where they show a somewhat scattered evolutionary distribution, but they are also present in some invertebrate animals (Fig. 1A). Among mammals, METTL12 can be found in human, cow, and pig, but it is present only in a subset of rodents, e.g. it is found in guinea pig, but is absent in rat and mouse. An alignment of METTL12 orthologues revealed the presence of conserved hallmark motifs characteristic for 7BS MTases, such as Motif I, Post I, and Motif II (Fig. 1A) (35), as well as the specificity-associated motif Post II, which is very similar to that found in eEF1AKMT2, eEF1AKMT4, and METTL13 (20).
By using the MitoProt algorithm (36) we found that METTL12 is likely localized to mitochondria, as the initial ϳ28

Human METTL12 methylates mitochondrial citrate synthase
N-terminal amino acids of METTL12 were predicted to represent a MTS. Indeed, both the full-length METTL12 and the isolated putative MTS were, when expressed as N-terminal fusions with GFP, able to target GFP to mitochondria, thus confirming that METTL12 is a mitochondrial protein (Fig. 1B).

Generation of METTL12 KO cells and demonstration of protein MTase activity
To functionally characterize METTL12, the METTL12 gene was disrupted in the haploid HAP1 cells using CRISPR/Cas9 technology. Abrogation of the METTL12 gene function was assured by designing a guide RNA to target a sequence located upstream of motif Post I, which contains a catalytically important acidic residue (i.e. Asp-107 in METTL12) that is crucial for AdoMet binding in 7BS MTases (12,15,17,20). Through DNA sequencing, a clone was identified that carried a 1-bp insertion at the target site, resulting in a shifted reading frame. This mutant gene encodes a predicted protein encompassing the 51 N-terminal amino acids of METTL12 followed by 66 residues resulting from out-of-frame translation. The mutant protein is thus severely truncated relative to the wild-type protein (240 amino acids) and highly likely inactive ( Fig. 2A).
Previously, we showed that many recombinant human KMT enzymes can methylate their respective substrates in a cellular extract (14,15,17,18). In several cases, such methylation was much more prominent in extracts from corresponding KMT KO cells, due to the substrate being exclusively in an unmethylated state. Consequently, we tested the ability of recombinant METTL12 to methylate proteins in a mitochondria-enriched extract from METTL12 KO cells. For this purpose we used a truncated version of METTL12 that lacked the 28 N-terminal amino acids predicted to represent the mature, MTS-less version of the protein. The methylation reactions were performed in the presence of [ 3 H]AdoMet and then analyzed by SDS-PAGE and fluorography, enabling the visualization of methylated proteins as distinct bands. Interestingly, we detected radiolabeling of a protein with an apparent molecular mass of ϳ48 kDa upon incubation of METTL12 with the extract from KO cells. No such labeling was observed when the wild-type (WT) HAP1 cells were used (Fig. 2B), indicating that the 48 kDa band represents a true METTL12 substrate. A putatively enzymatically inactive mutant of METTL12 (D107A) was used as a negative control, and, reassuringly, we did not detect any label-

Human METTL12 methylates mitochondrial citrate synthase
ing of the ϳ48 kDa substrate when this mutant enzyme was used, excluding the possibility that the observed labeling was mediated by an E. coli-derived contaminant (Fig. 2C). In addition we detected after longer exposures weak automethylation of METTL12 similar to what has been observed for many other KMTs (12,17,20,33). In summary, these results show that METTL12 is a mitochondrial enzyme with protein MTase activity.

Identification of citrate synthase as a likely substrate of human METTL12
To reveal the identity of the ϳ48-kDa substrate, we incubated mitochondrial extracts from METTL12 KO cells with [ 3 H]AdoMet and recombinant METTL12 and then subjected the reaction mixtures to fractionation on a cation-exchange (S) column (Fig. 3A). A parallel sample where wild-type recombinant METTL12 had been replaced by the enzymatically inactive mutant D107A was subjected to the same fractionation scheme, thus representing a negative control. We found that the ϳ48-kDa 3 H-labeled protein bound to the S-column was optimally eluted between 0.05 and 0.15 M NaCl (denoted as the 0.15 fraction) (Fig. 3B). To identify the METTL12 substrate, the METTL12 KO extract was fractionated (in the absence of radioactivity), the 0.15 fraction was resolved by SDS-PAGE, and the ϳ48-kDa region was excised from the gel (Fig. 3C) followed by trypsin digestion and protein identification by MS. Among several mitochondrial proteins that were identified (Table 1), only one, namely citrate synthase, had been reported in the literature as being methylated on lysine. CS contains a trimethylated lysine residue in position 368 in the mature protein corresponding to Lys-395 in the CS precursor (29). The latter numbering is used in protein databases such as UniProt and Phos-phoSitePlus (post-translational modifications) and will also be used here by us. The molecular weight of mature CS (49 kDa) matched that of the methylated substrate detected by fluorography, and when an anti-CS antibody was used to detect CS throughout fractionation, it was found that the elution profile of CS was indistinguishable from that of the ϳ48-kDa substrate (compare Fig. 3, B and C).
To further investigate the potential ability of METTL12 to methylate CS, we incubated the METTL12 KO cell extract with recombinant METTL12, then subjected the extract to fractionation, and the methylation status of Lys-395 in CS from the 0.15 fraction was assessed by MS. Strikingly, we found that Lys-395 was exclusively unmethylated in the KO cells (Fig. 3, D and E), whereas treatment with recombinant METTL12 shifted Lys-395 to the trimethylated (Me3) and dimethylated (Me2) states (Fig. 3, D and E). This demonstrates that METTL12 can methylate Lys-395 in CS in vitro and suggests that the enzyme is responsible for CS methylation also in vivo.

METTL12 catalyzes methylation of Lys-395 in CS in vitro
To further investigate CS as a substrate for METTL12-mediated methylation in vitro, we expressed and purified mature (MTS-less) CS in E. coli, and the resulting recombinant protein was then tested as a substrate for METTL12-mediated methylation. Clearly, METTL12 methylated recombinant CS in vitro, but a rather high amount of enzyme relative to substrate (Ͼ2-fold excess) was required to obtain maximal methylation (Fig.  4A). Importantly, protein mass spectrometry demonstrated the presence of trimethylated Lys-395 in recombinant CS that had been treated with METTL12, whereas Lys-395 was found to be unmethylated in untreated CS (supplemental Fig. S1). We also replaced Lys-395 by arginine in recombinant CS, which completely abolished methylation, indicating that Lys-395 is the only methylation site (Fig. 4B). Moreover, we considered the possibility that also a neighboring lysine residue, Lys-393, may be subject to methylation. However, this appears not to be the case, as methylation was not affected by mutation at this position (Fig. 4B) and because we were unable to find evidence for such methylation in the MS data. Taken together, the above data demonstrate that METTL12 catalyzes methylation of CS at Lys-395 in vitro.

Processivity, turnover, and metabolite modulation of METTL12-mediated methylation
Many KMTs that introduce trimethylation on their target lysines have a distributive mode of action, i.e. they introduce a single methyl group per binding event, thus generating a mixture of methylation states at lower enzyme concentrations. In contrast, other trimethylating KMTs are processive, i.e. they introduce all three methyl groups during a single binding event and thus exclusively generate the trimethylated product. To investigate the mode of action of METTL12, we incubated recombinant CS with different amounts of METTL12 and assessed the methylation status of Lys-395 by MS using extracted ion chromatograms corresponding to the various methylation states of a Lys-395-encompassing peptide (Leu-389 -Leu-408). Treatment of CS with a large excess of METTL12 predominantly yielded the trimethylated state, but when equimolar amounts of CS and METTL12 were used a mixture of the methylation states was observed, clearly demonstrating that METTL12 has a distributive mode of action (Fig. 5A, upper panel).
We noted that METTL12 enzyme was relatively inefficient and appeared incapable of performing catalytic turnover; at limiting enzyme concentrations, one MTase molecule mediated, on average, the incorporation of (approximately) a single methyl group (Fig. 5B). Many AdoMet-dependent MTases are inhibited by the byproduct AdoHcy (the demethylated counterpart of AdoMet) (3), and we, therefore, considered the possibility that inhibition of METTL12 by AdoHcy may prevent catalytic turnover. To test this we added AdoHcyase, which catalyzes the hydrolysis of AdoHcy, to the reaction mixture for METTL12-mediated methylation of CS. Clearly, the addition of AdoHcyase dramatically increased the efficiency of methylation and improved the catalytic turnover (Fig. 5, A and B). These results indicate that AdoHcy generated in the methylation reaction has a strong inhibitory effect on METTL12.
To further investigate the influence of AdoHcy on METTL12 activity, we performed in vitro methylation reactions where the concentrations of CS, METTL12, and AdoMet were held constant (at 2 M, 2 M, and 32.6 M respectively) whereas the concentration of AdoHcy was varied (0 -300 M). The results showed that METTL12-mediated methylation of CS was strongly inhibited by the addition of AdoHcy (Fig. 5C); visible inhibition was observed already at 1 M AdoHcy, and methyl-

Human METTL12 methylates mitochondrial citrate synthase
ation was almost completely abolished by 100 M AdoHcy. Importantly, the addition of AdoHcyase greatly alleviated the inhibitory effect of AdoHcy, and in agreement with the results in Fig. 5B, AdoHcyase increased methylation also in the absence of AdoHcy. The above results indicate that METTL12 has a much higher affinity for AdoHcy than for AdoMet and that AdoHcy, therefore, acts as a strong inhibitor of the enzyme, preventing enzymatic turnover.
We previously found for other 7BS KMTs that the efficiency of lysine methylation could be modulated by the specific binding of various ligands to the substrate (15,18,20). We, therefore, tested the effect of relevant metabolites, i.e. the CS substrates OAA and acetyl-CoA, as well as the product citrate on METTL12-mediated CS methylation. We found that OAA had a strong inhibitory effect on METTL12-dependent methylation of CS, i.e. methylation was inhibited by ϳ80% at 200 M OAA (Fig. 6A), whereas neither acetyl-CoA nor citrate had any effect (Fig. 6B). Taken together, the data presented above indicate that METTL12 methylates CS in a non-processive reaction, which is inhibited by AdoHcy and OAA.

Modulation of CS activity by METTL12 in vitro
The crystal structure of pig CS shows that Lys-395 is localized close to the active site of the enzyme (30), suggesting that CS activity may be affected by METTL12-mediated methylation. To test this, we incubated CS with METTL12 in the presence of AdoMet and then measured the enzymatic activity of CS. Corresponding samples, where AdoMet had been omitted or replaced by AdoHcy, were included as negative controls. Interestingly, when CS was incubated with METTL12 in the presence of AdoMet, a substantial (ϳ30%) reduction in CS activity was observed relative to the negative controls (Fig. 7A). Notably, no such effect was observed when the inactive METTL12 mutant (D107A) was used (we confirmed that this mutant enzyme was also inactive on recombinant CS; see supplemental Fig. S2). Taken together, these experiments indicate that METTL12-mediated methylation causes a small, but significant, reduction in CS activity.
To determine how METTL12-mediated methylation of CS affected the kinetics of the CS reaction, we measured the velocity of the CS-catalyzed reaction as a function of increasing OAA concentration while keeping the concentration of acetyl-CoA constant. These titration experiments were performed in parallel for methylated CS (incubated with AdoMet and METTL12 wild type) and unmethylated CS (incubated with AdoMet and METTL12 D107A mutant). Although the overall shape of the saturation curves was similar in the two cases, the apparent maximum velocity of the reaction was reduced for methylated CS compared with unmethylated CS (Fig. 7B). These results indicate that METTL12-mediated methylation primarily affects the rate constant of the CS-catalyzed reaction while having less effect on the affinity of CS for OAA.

METTL12 is responsible for CS methylation in vivo
We showed that recombinant METTL12 catalyzed methylation of CS at Lys-395 in vitro, both on recombinant CS and on

Figure 4. METTL12-mediated methylation of Lys-395 in recombinant CS in vitro. A, methylation of recombinant CS is dependent on METTL12. CS was incubated with [ 3 H]AdoMet and increasing amounts of METTL12, and [ 3 H]methyl incorporation was visualized by fluorography (top) of Ponceau S-stained membrane (bottom). B, evaluation of CS point mutants as substrates for METTL12-mediated methylation. CS, either wild-type (WT) or mutant (K395R or K393R), was incubated with [ 3 H]AdoMet and METTL12, and [ 3 H]-methyl incorporation was visualized by fluorography as in A.
Human METTL12 methylates mitochondrial citrate synthase

Human METTL12 methylates mitochondrial citrate synthase
CS from METTL12 KO extracts. Taken together with the published observation that CS from pig heart is trimethylated on Lys-395,thisindicatesthatMETTL12isresponsibleforCSmethylation also in vivo. To firmly establish this, we assessed the methylation status of Lys-395 in CS in "wild-type" and METTL12 KO HAP1 cells as well as in KO cells that had been complemented (supplemental Fig. S3) with ectopically expressed genes encoding either wild-type METTL12 or the enzymatically inactive mutant (D107A). We found that Lys-395 was predominantly trimethylated in the WT cells, with only small amounts of the un-, mono-, and dimethylated forms being detected (Fig. 8A). In contrast and as already shown, Lys-395 was exclusively observed in the unmethylated state in the METTL12 KO cells (Fig. 8A). Reassuringly, complementation with an expression plasmid carrying the wild-type METTL12 cDNA restored methylation in the KO cells, whereas no such effect was observed with the D107A-mutated cDNA. Taken together with the results from in vitro experiments, these data firmly establish that enzymatic activity of METTL12 is necessary and sufficient for CS methylation in vivo.

Methylation level of Lys-395 in CS differs between cell lines and various organs
Our in vitro experiments indicated that CS activity may be regulated by METTL12-mediated methylation in response to alterations in metabolite levels. To further address whether CS methylation is variable and, thus, likely to play a regulatory role, we analyzed CS Lys-395 methylation status in a panel of human cell lines as well as in several organs from pig. To this end we partially purified CS from the cells/organs using the purification scheme described previously (depicted in Fig. 3, A and B) and then assessed the methylation status of CS by MS. In the majority of cell lines tested, we observed high levels of Lys-395 methylation, with the trimethylated form (Me3) being predominant (Ͼ80%) (Fig. 8B). However, in HEK293 cells, CS showed a considerably lower methylation level, and a mixture of all four possible methylation states of Lys-395 was detected, i.e. unmethylated (Me0: ϳ36%), mono-(Me1: ϳ28%), di-(Me2: ϳ15%), and trimethylated (Me3: ϳ21%). In all analyzed pig organs (heart, muscle, liver, kidney, brain, ovary, uterus, small intestine, bone marrow, spleen) CS Lys-395 was exclusively found in the trimethylated form (data not shown).

Discussion
In the present study we have unraveled the biochemical function of the human MTase METTL12 using both in vitro and in vivo approaches. We found METTL12 to be localized to mitochondria and to methylate CS at Lys-395 in a non-processive fashion. Moreover, methylation reduced the activity of CS and was inhibited by the CS substrate OAA as well as by AdoHcy, suggesting that methylation may regulate CS function in response to altered metabolite levels.
We found that recombinant METTL12 methylated a single substrate of ϳ48 kDa in mitochondrial extracts. Partial purification of the extract followed by MS analysis identified several candidate substrates. These candidates included CS, which we subsequently established as a bona fide METTL12 substrate. Mitochondrial elongation factor Tu (TUFM) is another interesting candidate (Table 1) because it represents the mitochondrial ortholog of the cytosolic eEF1A, which is subject to methylation by the closest characterized mammalian homologues of METTL12, i.e. eEF1A-KMT2 and eEF1A-KMT4 (20,21). Consequently, we have tested the activity of METTL12 on recombinant TUFM but were unable to detect any methylation, indicating that TUFM is not a substrate of METTL12.
While this manuscript was under preparation, a study from another group implicated METTL12 in methylation of CS (37). Using an antibody that recognized trimethylated lysine residues, various trimethyllysine-containing mitochondrial proteins were detected, and only one of these, subsequently identified as CS, lacked this modification in HAP1-derived METTL12 KO cells. MS analysis demonstrated that Lys-395 in CS was unmethylated in the KO cells. Also, ectopic overexpression of METTL12 in HEK293 cells, where CS is relatively hypomethylated on Lys-395, resulted in a substantial increase in methylation. Although this study linked METTL12 to CS Cellular MTases differ greatly with respect to their affinities for the substrate AdoMet (expressed as the K m value) and the potentially inhibitory product AdoHcy (expressed as the K i value). Some MTases have an approximately equal affinity for AdoMet and AdoHcy (K m (AdoMet) Ϸ K i (AdoHcy)) and others have a higher affinity for AdoMet (K m (AdoMet) Ͻ K i (AdoHcy)), whereas others bind AdoHcy more avidly than AdoMet (K m (AdoMet) Ͼ K i (AdoHcy)) (3). The latter category of MTases will be particularly prone to product inhibition by AdoHcy and sensitive to alterations in the [AdoMet]/[AdoHcy] ratio. Clarke and Banfield have compared K m (AdoMet) and K i (AdoHcy) values for 23 mammalian MTases (3), and the ones with the relatively highest affinity for AdoHcy displayed K m (AdoMet)/K i (AdoHcy) ratios in the range 25-30, which is comparable to METTL12 (the addition of only 1 M AdoHcy inhibited METTL12 activity substantially in the presence of 32.6 M of AdoMet; Fig. 5C). Thus, METTL12 appears, compared with most other mammalian MTases, to be highly sensitive toward alterations in the cellular [AdoMet]/[AdoHcy] ratio. This was also supported by the observation that METTL12 was unable to perform catalytic turnover unless AdoHcyase was added to the reaction mixture. Importantly, the AdoMet and AdoHcy concentrations used in our in vitro experiments are comparable with those found inside cells; typical cellular concentrations of AdoMet and AdoHcy lie within the ranges 10 -100 M and 0.1-20 M, respectively (38). Interestingly, a number of recent studies have demonstrated that alterations in cellular metabolism leads to changes in the cellular AdoMet/AdoHcy ratio, which again causes changes in histone lysine methylation, thus providing an interesting link between metabolism and epigenetics (2,40,41). Similarly, our in vitro data suggest that CS activity may be modulated by AdoHcysensitive, METTL12-mediated lysine methylation in response to metabolic changes.
We found that the CS substrate OAA strongly inhibited METTL12-mediated methylation of CS, whereas no such effect was observed with either the co-substrate acetyl-CoA or with the product citrate. X-ray crystallographic studies have determined the structure of pig CS in complex with various metabolites, and CS exists in an "open" conformation in the absence of ligands or when complexed with citrate, whereas it adopts a "closed" conformation when complexed with OAA (30,42). Only a minor portion of the CS structure is affected by the conformational switch between the closed and open states, but the methylation site Lys-395 is part of a segment that is shifted substantially (Fig. 9). This suggests that OAA exerts its inhibitory effect through inducing the closed conformation and, thereby, a repositioning of Lys-395 and surrounding residues.
We found that METTL12-mediated methylation caused a small, but significant, reduction in CS activity. In contrast, Rhein et al. (37) did not detect any difference in activity between CS isolated from wild-type and METTL12 KO HAP1 cells, and another study did not detect significant differences in  Fig. S3. B, methylation status of Lys-395 in CS from various human cell lines. CS was enriched by cation-exchange (S-column) chromatography (as in Fig. 3, A and B), chymotrypsin-digested, and methylation of CS-derived peptides, encompassing Lys-395, was analyzed by MS. Shown are the relative intensities of MS signals gated for different methylation states of the peptides with error bars indicating the range of values from at least three independent experiments.

Human METTL12 methylates mitochondrial citrate synthase
catalytic properties between recombinant pig CS, which is unmethylated, and CS from pig heart, which is trimethylated (43). However, in these two studies, the activity of CS from two different sources was compared (CS from WT versus KO cells; recombinant CS versus CS from pig heart), and this approach may be prone to uncertainties (e.g. variable loss of activity during enzyme isolation or inaccuracies in protein concentration determination) that could hinder the detection of small differences. In contrast, we used the same preparation of recombinant CS and then studied the effect of different additions (METTL12, AdoHcy, AdoMet), thus likely avoiding the above mentioned sources of error and allowing us to demonstrate a small, but robust, effect of methylation on CS activity; an effect that required the presence of both AdoMet and catalytically active METTL12. The small observed effect agrees well with the sketchy distribution of METTL12 in mammals, e.g. its absence in rats and mice, which also indicates that methylation is not strictly required for CS function but rather serves an optimizing or regulatory function that is beneficial under certain conditions. We observed that CS was predominantly trimethylated in the tested human cell lines, except in HEK293 cells, where CS showed a much lower methylation level (Fig. 8B). Strikingly, we detected exclusively fully trimethylated CS in all analyzed pig organs, indicating that trimethylation of CS may represent the default state in METTL12-containing organisms. These observations further supports that methylation is important for CS function, but also raises the question of whether altered methylation actively regulates CS activity in vivo. It should be noted, however, that the analyzed organs came from a healthy, well fed pig, grown in a controlled environment. Thus, one may speculate that METTL12-mediated methylation may regulate CS activity under certain stress conditions (e.g. starvation or extensive exercise), thus playing a role in the adaptation to such conditions. Bioinformatics analysis identified METTL12 as a member of a family of 7BS MTases, which also encompasses the human MTases eEF1A-KMT2 (gene name EEF1AKMT2; alias METTL10), eEF1A-KMT4 (gene name EEF1AKMT4; alias ECE2), and METTL13 (20). Including the present work, three of the four human members of this family have now been established as KMTs. This indicates that the remaining uncharacterized human member of this family, METTL13, is likely also a KMT. Numerous human 7BS MTases have been established as KMTs during the last years, and all of these are highly specific, i.e. only a single substrate has been identified, or in case of METTL21A, a group of highly related substrates (various Hsp70 proteins) (8,13,15,39). Consequently, a naming nomenclature based on the substrate specificity has emerged for these enzymes. For example, the MTases enzymes METTL20 and FAM86A were found to target the ␤-subunit of the mitochondrial electron transfer flavoprotein (ETF␤) and eukaryotic elongation factor 2 (eEF2), respectively, and were consequently redubbed ETF␤-KMT and eEF2-KMT (gene names EFTBKMT and EEF2KMT) (14,17). Thus, we suggest that METTL12 is renamed CS-KMT (gene name CSKMT) in keeping with this established nomenclature.