Role of METTL20 in regulating β-oxidation and heat production in mice under fasting or ketogenic conditions

METTL20 is a seven-β-strand methyltransferase that is localised to the mitochondria and tri-methylates the electron transfer flavoprotein (ETF) β subunit (ETFB) at lysines 200 and 203. It has been shown that METTL20 decreases the ability of ETF to extract electrons from medium-chain acyl-coenzyme A (CoA) dehydrogenase (MCAD) and glutaryl-CoA dehydrogenase in vitro. METTL20-mediated methylation of ETFB influences the oxygen consumption rate in permeabilised mitochondria, suggesting that METTL20-mediated ETFB methylation may also play a regulatory role in mitochondrial metabolism. In this study, we generated Mettl20 knockout (KO) mice to uncover the in vivo functions of METTL20. The KO mice were viable, and a loss of ETFB methylation was confirmed. In vitro enzymatic assays revealed that mitochondrial ETF activity was higher in the KO mice than in wild-type mice, suggesting that the KO mice had higher β-oxidation capacity. Calorimetric analysis showed that the KO mice fed a ketogenic diet had higher oxygen consumption and heat production. A subsequent cold tolerance test conducted after 24 h of fasting indicated that the KO mice had a better ability to maintain their body temperature in cold environments. Thus, METTL20 regulates ETF activity and heat production through lysine methylation when β-oxidation is highly activated.


Results
ETFB is a major substrate for METTL20 in the mitochondria, and its catalytic activity is regulated by lysine methylation. Recently, a mitochondrial MTase, METTL20, was identified as an ETFB MTase 5,6 . We also attempted proteomic identification of METTL20 substrates in mitochondria. A proteomic analysis using a synthetic S-adenosylmethionine (AdoMet) analogue, ProSeAM 12,15 , has been applied to identify the substrate (Fig. 1a). HEK293T cells were cultured in DMEM with 10% of dialyzed foetal calf serum (FCS) containing either light isotopic 12 C 6 -Lys or heavy 13 C 6 -Lys. Each mitochondrial fraction was incubated with ProSeAM 12,13,15 in the absence or presence of His-METTL20. After biotinylation of ProSeAM-labelled proteins, a small part (5%) of the protein was subjected to western blotting with streptavidin conjugated with horseradish peroxidase (HRP; Fig. 1b). The samples were combined and mixed, and subsequent proteomic analysis revealed that ETFB was a major mitochondrial substrate for this enzyme (Fig. 1c).
We confirmed that ETFB was modified by ProSeAM ( Supplementary Fig. S2a) and labelled with 14 C-AdoMet ( Supplementary Fig. S2b) in the presence of METTL20 in vitro. Because histones contain many methylated lysine residues, we also tested recombinant histone H3 (hereafter: H3) in an experiment and demonstrated that METTL20 can methylate both ETFB and H3 in vitro. It should be noted that because histones are nuclear proteins, they are possibly not a physiological target of METTL20, which is localised to mitochondria. Nevertheless, it is still possible that METTL20 has one or more additional mitochondrial substrates aside from ETFB. We also confirmed that the methylation site in ETFB is exactly the same as in other reports, i.e., K200 and K203 ( Supplementary Fig. S2c-f) 5,6 . We then checked if this methylation is conserved between humans and mice. As shown in Fig. 1d, methylation of murine ETFB was abolished in K200/203R mutant, indicating that methylation of ETF is well conserved between these two species.
ETF functions as an electron transfer protein for acyl-CoA dehydrogenases. Given that the methylation of ETFB inhibits the catalytic activity 5,6 , we also checked the mouse ETF catalytic activity in vitro. The recombinant ETF complex was incubated with METTL20, with or without AdoMet for 3 h, after which the ETF complex was mixed with recombinant MCAD, octanoyl-CoA, and 2,6-Dichloroindophenol (DCPIP) as described previously 16,17 . The methylation status of ETFB after the methylation reaction was confirmed (Fig. 1e). As depicted in Fig. 1f, methylated ETF showed more reduced activity (51% ± 2%) than unmethylated ETF did (100% ± 5%). It can be inferred that either the KR mutant did not affect catalytic activity (98% ± 5%) or the methylation reaction did not influence its activity (105% ± 6%). These results confirmed that methylation of lysine in ETFB by METTL20 inhibits mouse ETF catalytic activity, in line with the results from other reports 5,6,18 . Creation of Mettl20 KO mice by CRISPR/Cas9 genome editing. To understand the physiological function of METTL20 in vivo, Mettl20 KO mice were generated via the CRISPR/Cas9 genome editing technology 19,20 . Guide RNAs were designed against exon 4, which encodes the MTase domain of METTL20 (Fig. 2a). The guide RNA sequence contains the MscI site, which is convenient for genotyping. A Mettl20 KO mouse carries a 17 bp deletion, which results in a frameshift mutation in the MTase domain (Fig. 2b). Expression of mMETTL20 in several organs of WT and KO mice was examined by western blotting with an anti-METTL20 antibody. As shown in Fig. 2c, mMETTL20 protein expression levels were undetectable in all the tissues tested in KO mice but were the highest in the WT liver. Because proteins METTL20 and ETFB were highly expressed in the mouse liver, which is an important organ for fatty acid metabolism, we checked the ETFB methylation level in the liver of WT and KO mice. ETFB in the liver was immunoprecipitated with an anti-ETFB antibody, and methylation was detected with an anti-di-methyl lysine antibody (diMeLys) or anti-tri-methyl lysine antibody (triMeLys, Fig. 2d). Tri-methylation of ETFB was not detected in KO mice, suggesting that mMETTL20 was crucial for the tri-methylation. Of note, di-methylation of ETF was detected in neither WT nor KO mice. To verify this result, the ETFB immunoprecipitated from WT and Mettl20 KO liver samples was subjected to liquid chromatography-mass spectrometry (LC-MS). As shown in Fig. 2e, the K200 and K203-containing ETF peptide had an additional molecular weight (MW) of 28 Da in WT mice, in agreement with the MW observed after 2× trimethylation. In contrast, this additional MW was completely lost in KO mice. These observations implied that K200 and K203 were fully tri-methylated in ETFB of WT mice but remained completely unmethylated in Mettl20 KO mice. Mettl20 KO mice show increased ETF activity, oxygen consumption, and heat production when fed a ketogenic diet (KD) or when fasted. We measured ETF activity in Mettl20 KO mice as reported elsewhere 17 , with a slight modification. Briefly, the mitochondrial fractions from the WT or KO murine liver were sonicated, and the supernatant was mixed with recombinant MCAD, octanoyl-CoA, and DCPIP. As presented in Fig. 3a, the ETF activity in Mettl20 KO mouse mitochondria was ~1.5-fold greater than that in the WT. This result is consistent with that of in vitro experiments, suggesting that methylation of ETFB inhibited ETF activity (Fig. 1f) because ETFB in the WT mitochondria was fully methylated (Fig. 2d and e).
The growth of Mettl20 KO mice was normal and indistinguishable from that of the WT when these mice were fed a normal diet (ND) or KD. Additionally, the differences in the body weight and several parameters such as Figure 1. ETFB methylation is conserved in human and mouse, which regulates the catalytic activity. (a) A schematic drawing of screening for METTL20 substrates. Cells were cultured in either light isotope labeled Lys containing medium (L) or heavy isotope labeled Lys containing medium (H). The mitochondrial lysates were reacted with ProSeAM with (H) or without (L) His-METTL20. After the in vitro modification, biotin tags were introduced to the modified residues via click reaction. Then the two samples were mixed together, and biotinylated proteins were pull-down with Streptavidin beads. After Lys-C digestion, peptides were analyzed by LC-MS/MS. (b) Confirmation of ProSeAM labeling. After the ProSeAM mediated alkylation and biotinylation via click reaction, small aliquot of the protein samples were analyzed by SDS-PAGE followed by western blotting with Streptavidin-HRP and anti-COXIV antibody as a loading control. (c) METTL20 substrates identified in the screening. Summary of two independent experiments were shown. Note that only ETFB was the protein identified in both cases. (d) Recombinant ETF complex (ETFA WT/ ETFB WT or K200/203R) and His-METTL20 were incubated with 14C-labeled AdoMet for 2 h at 30 °C. The autoradiography was detected with the image analyzer BAS-5000. (e) 4 µg of the ETF complex were mixed with 4 µg of METTL20 in the presence or absence of AdoMet and incubated at 30 °C for 3 h. After the reaction, they were mixed on ice with 8 µg of His-MCAD, 50 µM octanoyl-CoA and 70 µM DCPIP. Initial rate of the reduction of DCPIP was calculated from OD600 measured at 30 °C for every 30 sec for 10 min with a microplate reader (SpectraMax 190, Molecular Devices). Average of three independent experiments, n = 3 for each experiments, mean ± SEM. Student's T-test **p < 0.01. (f) Small aliquot of the ETF used in A were separated with SDS-PAGE, and their methylation were detected with anti-triMeLys antibodies (top) and anti-ETFB antibodies (bottom, as loading control).   results of liver function testing (LDH, AST, ALT, and GGT), renal function testing (TP, ALB, UN, CRE), lipid metabolism analyses (T-Bil, HDL-C, LDL-C, T-CHO), glucose, IPGTT, and serum adipocytokine tests were statistically insignificantly different between WT and KO mice (data not shown).
To examine the possibility that Mettl20 KO mice were better at generating energy from fat, a calorimetric analysis was performed on the mice fed the ND or KD, because the KO mice showed increased ETF activity (Fig. 3a). Both strains of mice had similar food intake and motor activity throughout the experiments (data not shown). As shown in Fig. 3b and c, the Mettl20 KO mice fed the KD exhibited increased oxygen consumption and heat production (p < 0.05), whereas these parameters were unchanged when these mice were fed the ND ( Supplementary Fig. S3).
Next, we checked heat production in a cold environment ( Fig. 3d and e). The mice were either fed or fasted for 24 h before the experiment, given that fatty acid metabolism is known to be elevated during fasting. The cold tolerance test indicated that Mettl20 KO mice had a better ability to maintain their body temperature than WT mice did during fasting (Fig. 3d, KO: 32.3 °C ± 0.3 °C vs. WT: 30.7 °C ± 0.2 °C, p < 0.01 after a 1 h cold test). Again, no difference was observed during feeding (Fig. 3e). These results indicated that mMETTL20 is involved in the regulation of β-oxidation under ketogenic or fasting conditions, where fatty acid metabolism is known to be highly accelerated.
Metabolomic analysis of serum fatty acids in Mettl20 KO mice. Because Mettl20 KO mice showed accelerated ETF activity and heat production during ketosis or fasting, it is possible that serum fatty acid contents are different between the WT and KO mice. To test this possibility, a metabolomic assay was performed to quantify fatty acids and acyl carnitines in WT or KO mouse serum samples. Overall, the levels of 177 metabolites detected in the assay were not significantly different between the two groups (<2-fold difference, data not shown). Moreover, levels of free fatty acids (FFAs) and acyl carnitine were similar between the two strains of mice ( Fig. 4a and b). To confirm these results, we measured the total FFA concentration in serum (Fig. 4c). Again, there was no difference in the serum FFA concentration between the WT and KO mice.
METTL20 is tolerant to loss-of-function (LoF) mutations. The METTL20/ETFBKMT homozygous loss-of-function mutation has been identified in an Icelandic population 21 , suggesting that this enzyme is not essential for viability of humans 18 . To further address the biological significance of genetic variants of METTL20 in the human population, we searched for genomic mutations within the exons of METTL20 in the ExAC Browser, which stores exome DNA sequence data for 60,706 individuals with diverse ancestries 22 . As shown in Fig. 5a, METTL20 was classified as a gene tolerant to LoF (if pLI = 0.00 or pLI < 0.01, then the gene is thought to be tolerant to LoF 22 ). In addition to the frameshift mutation that was reported in the Icelandic population 21 , individuals exist that are homozygous for certain missense mutations in METTL20 (ref. 22 and Supplementary  Table S1). Among them, the c.562G > A mutation (rs143179970) induces the D188N substitution in Motif Post II of the METTL20 MTase domain (Fig. 5b), which is crucial for the catalytic activity of seven-β-strand MTases such as VCPKMT 23 . To evaluate the enzymatic activity of the corresponding mutant protein, recombinant METTL20 D188N and another mutant (D188G), which is also present as a heterozygous mutation in the ExAC database, were expressed in bacteria; an MTase activity assay was conducted in vitro (Fig. 5c). These mutant enzymes could not methylate ETFB, implying that the loss of catalytic activity by METTL20 does not cause an obvious disadvantage for survival of humans. These findings confirmed that METTL20 is a gene tolerant to LoF.

Discussion
Approximately 40 MTase genes may participate in mitochondrial functions 4 ; however, only a few of these 'mitochondrial' MTase genes have been characterised so far. METTL20 has been identified as an ETFB MTase 5,6 . We also identified ETFB as a METTL20 substrate by a unique method that involves the synthetic co-factor ProSeAM 12,15 . During the screening, ETFB was found to be the sole candidate for METTL20 (Fig. 1c), suggesting that ETFB is a major target of METTL20 in mitochondrial lysates. In the in vitro experiments, METTL20 methylated ETFB at K200 and K203, and we could detect peptides carrying up to five methyl groups, but not fully methylated ones (Supplementary Fig. S2f). This observation is interesting because ETFB is fully methylated in the mouse liver (Fig. 2d,e). It is possible that other proteins or co-factors are required for full methylation of ETFB in vivo.
Given that methylation sites are located around the recognition loop, which is required for the interaction with MCAD, it is probable that methylation of mouse ETFB affects its catalytic activity (Fig. 1f), and this notion is consistent with the results from other reports 5,6 . In this study, we generated Mettl20 KO mice (Fig. 2) and demonstrated that methylation of ETF was lost in KO mice. The catalytic activity of ETF was higher in KO mice (Fig. 3a), suggesting that mMETTL20 has a regulatory role in fatty acid oxidation. Accordingly, Mettl20 KO mice also had a higher oxygen consumption rate and heat production when they consumed the KD (Fig. 3b and c). Furthermore, the KO mice showed greater cold tolerance than the WT did during fasting although there was no difference during ad libitum feeding, where glucose metabolism is known to be a primary energy source (Fig. 3d and e). From these results, we can conclude that β-oxidation activity is regulated by the methylation of ETFB by METTL20, when fatty acids serve as a primary energy source, i.e. β-oxidation is highly activated. It should be noted that serum fatty acid Fed. WT and Mettl20 KO mice were either fed on ND (e) or fasted for 24 h (d) before the experiment. Mice in the same cage per conditions/genotype were transferred to cold room (4 °C), and measured their core body temperatures after 1 h, 3 h and 5 h. n = 5, mean ± SEM. Student's T-test **p < 0.01. concentrations were not affected by the Mettl20 KO (Fig. 4), pointing to the existence of a compensatory mechanism ensuring lipid homeostasis. Further research is needed to reveal such a compensatory mechanism.
The fact that Mettl20 KO mice grew to be normal and undistinguishable from WT mice (data not shown) implies that this gene is not essential. To test whether this is true for humans, the ExAC database was searched for genomic variations in Mettl20 exons. The analysis revealed that METTL20 is classified as a gene tolerant to LoF (Fig. 5a). The identified homozygous missense mutation (D118N) completely eliminated the METTL20 catalytic activity (Fig. 5c). These results further indicate that METTL20-mediated protein methylation is not essential in humans but more likely modulates energy production during fasting or under other metabolic conditions where β-oxidation is up-regulated. One possible benefit of this regulation is that too much energy production during fasting is avoided. This mechanism may help animals survive long-term starvation. Besides, it is likely that methylation of ETFB dynamically changes according to the metabolic environment, which, in turn, regulates metabolism adaptability. In both cases, further experiments are needed to test these possibilities.

Materials and Methods
Animals. All the experiments involving mice were carried out according to protocols approved by the Animal Experiment Committee of the RIKEN Brain Science Institute. The animals were maintained in a 12 h light/dark cycle with ad libitum access to food and water. The temperature and humidity were maintained at 22-23 °C and 50-60%, respectively. Animal health was checked by the animal facility staff five times per week. The mice were subdivided into two groups and fed a standard chow (ND) containing 5% fat and 25% protein (CLEA Japan, Inc., CE-2) or a KD containing 91% fat and 9% protein (Harlan Teklad, TD96355) from the age of 4 weeks to 26 weeks.
Purification of recombinant proteins. BL21 (pLysS) strains were transformed with pET19b plasmids, and the bacteria were cultured in the 2xYT medium with ampicillin (100 µg/ml) and 0.

Stable isotopic labelling of cells and mitochondrial purification. For METTL20 target screening,
HEK293T cells were cultured for at least six cell doublings in DMEM supplemented with 10% of dialysed FCS containing either a light ( 12 C 6 ) or stable heavy ( 13 C 6 ) isotope in lysine after processing with the SILAC labelling kit (Pierce, cat. # 89983). The mitochondrial fractions of each cell type were purified as described previously 28 . Briefly, cell pellets were resuspended in SEM buffer (10 mM 4-[2-hydroxyethyl]-1-piperazineethanesulfonic acid [HEPES]-KOH pH 7.5, 0.2 M mannitol, 50 mM sucrose, 10 mM KCl, 1 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride), and homogenised by means of the Dounce tissue grinder (2 ml, Pestle A, Fisher Scientific, USA). After that, the supernatants were centrifuged at 700× g for 10 min to separate nuclear pellets from the supernatants containing cytosolic materials and mitochondria. The supernatants were again centrifuged at 700× g to remove residual pellets, and then at 7,000× g for 10 min to separate the mitochondrial pellet from the cytosol-enriched supernatant. The mitochondrial pellets were next washed twice with SEM buffer, and the resulting mitochondrial fractions were stored at −80 °C.
ProSeAM labelling. Labelling of mitochondrial proteins with ProSeAM was carried out by a previously described procedure 15 , with a slight modification. Briefly, 10 ul of lysis buffer (50 mM Tris-HCl pH 8.0, 50 mM KCl, 10% glycerol, 0.1% of Tween 20) containing 150 μg of either light-isotope-labelled mitochondria ( 12 C 6 -Mito) or heavy-isotope-labelled mitochondria ( 13 C 6 -Mito) was incubated with 0.4 mM ProSeAM, with ( 13 C-Mito) or without ( 12 C-Mito) His-METTL20 (10 μg) (protein:enzyme at 15:1) in 50 ul of MTase reaction buffer (50 mM Tris-HCl pH 8.0) at 20 °C for 2 h. The reaction was stopped by adding four volumes of ice-cold acetone. The reaction tube was centrifuged at 15,000× g for 5 min, and the precipitate was washed once with ice-cold acetone. The pellet was dissolved in 58.5 µl of 1× PBS containing 0.2% of SDS, after which 15 µl of 5× click reaction buffer and 1.5 µl of 10 mM Azide-PEG4-Biotin (Click Chemistry Tools) were added; the reaction mixture was incubated for 60 min at room temperature. The click reaction was stopped with four volumes of ice-cold acetone. The pellet was dissolved in 75 µl of binding buffer (1× PBS, 0.1% of Tween-20, 2% of SDS, 20 mM dithiothreitol) and sonicated for 10 s. 12 C-Mito and 13 C-Mito samples were mixed in a tube; 450 µl of IP buffer (Tris Buffered Saline (TBS) with 0.1% Tween 20) containing 3 mg of Dynabeads M-280 Streptavidin (Life Technologies Japan Ltd., Minato-ku, Tokyo, Japan) was added into the tube, and it was incubated for 30 min at room temperature (the final SDS concentration in the reaction was 0.5%). The protein-bound beads were washed thrice with wash buffer (1× PBS, 0.1% of Tween 20, 0.5% of SDS) and twice with 100 mM ammonium bicarbonate buffer, and were analysed by western blotting or MS.

Quantitative tandem MS (MS/MS) analysis for target protein identification.
Acetonitrile (1/10 volume) and dithiothreitol (20 mM) were added to protein-bound Dynabeads in 100 mM ammonium bicarbonate buffer, and the mixture was incubated for 30 min at 56 °C. Then, iodoacetamide was added, and the mixture was incubated for 30 min at 37 °C in the dark. Next, the protein samples were digested with 0.5 µg of Lys-C (Promega). The protein fragments were applied to a liquid chromatograph (EASY-nLC 1000; Thermo Fisher Scientific, Odense, Denmark) coupled to a Q Exactive Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Fisher Scientific, Inc., San Jose, CA, USA), with a nanospray ion source in positive mode. The peptides derived from the protein fragments were separated on a NANO-HPLC C18 capillary column (0.075 mm inner diameter × 150 mm length, 3 mm particle size; Nikkyo Technos, Tokyo, Japan). Mobile phase A was composed of water with 0.1% of formic acid, and mobile phase B consisted of acetonitrile with 0.1% of formic acid. Two different slopes were used for gradient elution for 120 min at a flow rate of 300 nl/min: 0-30% of phase B during 100 min and 30-65% of phase B for 20 min. The mass spectrometer was operated in top-10 data-dependent scan mode. The operating parameters of the mass spectrometer were as follows: spray voltage, 2.3 kV; capillary temperature, 275 °C; mass-to-charge (m/z) ratio, 350-1800; normalised collision energy, 28%. Raw data were acquired via the Xcalibur software (Thermo Fisher Scientific). The MS and MS/MS data were subjected to searches in the Swiss-Prot database by means of Proteome Discoverer 1.4 (Thermo Fisher Scientific) in the MASCOT search engine software version No. 2.4.1 (Matrix Science, London, United Kingdom). The search parameters were as follows: enzyme, Lys-C; quantitation, SILAC K (+6); static modifications, carbamidomethyl (Cys); dynamic modifications, oxidation (Met); precursor mass tolerance, ± 6 ppm; fragment mass tolerance, ± 20 mDa; and maximum missed cleavages, 1. The proteins were considered identified when their false discovery rate was less than 1%. For substrate identification, proteins manifesting at least a 2-fold increase in levels in two independent experiments were defined as positive hit proteins.
An in vitro MTase assay. Recombinant histone H3.1 was purchased from NEB. One microgram of the substrate was incubated in 10 ul of 1× Reaction buffer (50 mM Tris-HCl pH 8.0) with His-METTL20 (1 µg) and 14 C-labelled AdoMet (0.01 µCi, Perkin Elmer) from 2 h to overnight at 30 °C. The reaction was stopped by adding Laemmli SDS sample buffer. Proteins were resolved by SDS-PAGE in a 12.5% acrylamide gel. An imaging plate (Fuji-Film) was exposed to the dry gel for 48 h, and the autoradiograph was obtained on a BAS-5000 Image analyser (Fuji-Film).
An in vitro ETF activity assay. ETF enzymatic activity was determined by a previously described DCPIP dye assay 16,17 . The recombinant WT and mutant ETF complex were prepared with the pETDuet-1-His-ETFB/ ETFA. Four micrograms of the ETF complex was mixed with 4 µg of METTL20 with or without AdoMet and incubated at 30 °C for 3 h in 20 ul of MTase reaction buffer. After that, 4 µg of the ETF complex was mixed on ice with 8 µg of His-MCAD, 50 µM octanoyl-CoA, and 70 µM DCPIP in potassium phosphate buffer (pH 7.6) containing 0.2 mM N-Ethylmaleimide (NEM). For the mitochondrial ETF activity assay, 300 µg of the mitochondrial fraction from the murine liver was sonicated, and the supernatant was mixed with 3 µg of recombinant MCAD, 80 µM octanoyl-CoA, and 70 µM DCPIP in potassium phosphate buffer (pH 7.6) containing 0.2 mM NEM. The initial rate of reduction of DCPIP was calculated from optical density at 600 nm measured at 30 °C every 30 s for 10 min on a microplate reader (SpectraMax 190, Molecular Devices). Data were obtained in triplicate for each biological sample.
Creation of Mettl20 KO mice. Guide RNAs that target exon 4 of mouse Mettl20 were screened with CRISPR design tools (http://crispr.mit.edu/). The T7 promoter sequence 5′-TTAATACGACTCACTATAGG-3′ and the 20-mer Mettl20 guide RNA sequence 5′-GAGTGGAGCATCAAAGATCT-3′ were cloned into the AflII site of the gRNA_Cloning Vector, a gift from George Church (Addgene #41824) 19 . Oligos Insert-F (5′-TTTCTTGGCTTTATATATCTTAATACGACTCACTATAGGAGTGGAGCATCAAAGATCT-3′) and Insert-R (5′-GACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAACAGATCTTTGATGCTCCACTC-3′) were annealed and extended to construct a 100-bp double-stranded DNA fragment using Phusion polymerase (New England Biolabs, Japan). The gRNA_Cloning Vector was linearised with AflII and the inserts were incorporated via Gibson assembly (NEB, Japan, E2611S). To prepare hCas9 mRNA and Mettl20 guide RNAs, an in vitro transcription reaction kit with T7 RNA polymerase (mMESSAGE mMACHINE T7 Ultra Kit, Life Technologies) was used; the transcribed RNAs were purified with the MEGAclear Kit (Life Technologies). The quality of the guide RNA was checked by an in vitro Cas9 cleavage assay with recombinant hCas9 nuclease (NEB) and a 1 kb PCR product around the guide RNA target site. The Cas9 mRNA and guide RNA were injected into fertilised eggs (mouse strain C57BL/6J), and the mutated Mettl20 gene versions were screened by PCR after MscI digestion, because the MscI site TGGCCA is located immediately after the guide RNA target site. One in 11 heterozygous (+/−) males carried the expected 17 bp deletion. The heterozygous mouse was crossed with a WT C57BL/6J mouse, and the male (+/−) and female (+/−) mice were crossed to obtain Mettl20 KO (−/−) mice and WT (+/+) mice.

Calorimetric analysis.
We conducted this analysis according to the standard operating procedure for the indirect calorimetry test (IMPC_CAL_001, http://www.mousephenotype.org/impress/protocol/86) of the International Mouse Phenotyping Consortium 29,30 . Briefly, mice were acclimated to the test room (in a 12 h lightdark cycle; lights on at 8:00 h) and the calorimetry cage 24 h before testing. Into the calorimetry cage, a single mouse was placed; food and water were made available ad libitum during acclimation and calorimetric analysis. The measurements were started 5 h before lights were switched off for 21 h. Input and output levels of gases oxygen (O 2 ) and carbon dioxide (CO 2 ) were measured by an Oxymax indirect calorimetry system (Columbus Instruments, Columbus, OH). The changes in the CO 2 and O 2 levels between the input and output were used to compute O 2 consumption (VO 2 ; ml/h per animal) and CO 2 production (VCO 2 ; ml/h per animal).
The respiratory exchange ratio (RER), calorific value (CV), and heat production (KJ/h per animal) were computed from VCO 2 and VO 2 as follows: RER = VCO 2 /VO 2 , CV = 3.815 + (1.232 × RER); heat production (KJ/h per animal) = 4.184 × CV × VO 2 . Water intake was measured by the Food and Water Intake Monitor (FWI-16M, O'Hara Co., Ltd., Tokyo, Japan). Locomotor activity of the mice was quantified by the Infrared XY beam activity system for mice (IA-16M, O'Hara Co., Ltd.). Cold tolerance test. This test was performed as described previously 31 , with a slight modification. WT and Mettl20 KO mice were either fed the ND or fasted for 24 h before the experiment. Mice in the same cage in terms of conditions and genotype were transferred to the cold room (4 °C), and their core body temperature was measured with a BAT-7001H Thermometer (Physitemp Instruments, Inc.) after 1, 3, and 5 h. Metabolomic analysis. WT and Mettl20 KO male mice (n = 10 for each condition) were fed the ND containing 5% of fat and 25% of protein (CLEA Japan, Inc., CE-2) or the KD containing 91% fat and 9% protein (Harlan Teklad, TD96355) from the age of 4 weeks to 26 weeks. Twenty-six-week-old mice were killed, and the serum samples of mice from each group were pooled and kept at −80 °C until measurements. All LC-TOF MS analyses were conducted at Human Metabolome Technologies Inc., Tsuruoka, Japan. Briefly, 500 μl of pooled serum was mixed with 6 μM internal standards in 1.5 ml of the mixture of 1% formic acid and 99% acetonitrile. Samples were centrifuged at 2,300× g and 4 °C for 5 min, and the supernatants were collected and filtered through a solid-phase extraction column. Flow-through fractions were dried and dissolved in 100 μl of 50% 2-propanol. The samples were applied to a liquid chromatograph (Agilent 1200 series RRLC system SL, Agilent Technologies, Inc.) coupled to an Agilent LC/MSD TOF Mass Spectrometer (Agilent Technologies, Inc), with electrospray ionisation in both positive and negative modes. The samples were separated on an ODS column (2 mm inner diameter × 50 mm length, 2 μm particle size). Mobile phase A comprised water with 0.1% of formic acid, and mobile phase B comprised 65% of isopropanol, 30% of acetonitrile, 5% of H 2 O with 0.1% of formic acid, and 2 mM HCOONH 4 . Three slopes were used for gradient elution for 20 min at a flow rate of 300 μl/min: 0-1% of phase B for 0.5 min, 1-100% of phase B during 13.5 min, and then 100% of phase B for 20 min. The settings of the mass spectrometer were as follows: nebuliser pressure, 40 psi; dry gas flow, 10 l/min; dry gas temperature, 350 °C; capillary voltage, 3500 V; scan range, m/z 100-1700. The obtained data were subjected to searches in the HMT database by means of MasterHands ver.2.17.1.11 (ref. 32 ) (Keio University). The overall data processing flow consisted of the following steps: noise filtering, baseline removal, migration time correction, peak detection, and integration of the peak area from a 0.02 m/z-wide slice of the electropherograms.
Fatty acid measurement. WT and Mettl20 KO male mice (n = 10 for each condition) were fed the ND or KD from the age of 4 weeks to 23 weeks. The blood samples were collected from the orbital sinus during ad libitum feeding. Concentrations of serum FFAs were analysed with a kit (Abcam, cat. # ab65341). The obtained data were presented as mean ± SEM, n = 10 per condition.