Mitochondrial translation requires folate-dependent tRNA methylation

Abstract

Folates enable the activation and transfer of one-carbon units for the biosynthesis of purines, thymidine and methionine. Antifolates are important immunosuppressive and anticancer agents. In proliferating lymphocytes and human cancers, mitochondrial folate enzymes are particularly strongly upregulated. This in part reflects the need for mitochondria to generate one-carbon units and export them to the cytosol for anabolic metabolism. The full range of uses of folate-bound one-carbon units in the mitochondrial compartment itself, however, has not been thoroughly explored. Here we show that loss of the catalytic activity of the mitochondrial folate enzyme serine hydroxymethyltransferase 2 (SHMT2), but not of other folate enzymes, leads to defective oxidative phosphorylation in human cells due to impaired mitochondrial translation. We find that SHMT2, presumably by generating mitochondrial 5,10-methylenetetrahydrofolate, provides methyl donors to produce the taurinomethyluridine base at the wobble position of select mitochondrial tRNAs. Mitochondrial ribosome profiling in SHMT2-knockout human cells reveals that the lack of this modified base causes defective translation, with preferential mitochondrial ribosome stalling at certain lysine (AAG) and leucine (UUG) codons. This results in the impaired expression of respiratory chain enzymes. Stalling at these specific codons also occurs in certain inborn errors of mitochondrial metabolism. Disruption of whole-cell folate metabolism, by either folate deficiency or antifolate treatment, also impairs the respiratory chain. In summary, mammalian mitochondria use folate-bound one-carbon units to methylate tRNA, and this modification is required for mitochondrial translation and thus oxidative phosphorylation.

Conflict of interest statement

Competing financial interests

The authors declare no competing financial interests.

Figures

Extended Data Figure 1. SHMT2 deletion-induced respiratory chain dysfunction in different cellular backgrounds and clones

a, Change in media color after 48 h cell growth. b, Lactate secretion and c, normalized NAD+/NADH ratio of HCT116 knockout cell lines (n = 6). c, Basal respiration of HEK293T folate 1C gene CRISPR/Cas9 knockout cell lines as measured by Seahorse XF analyzer (n = 3) and e, normalized NAD+/NADH ratio (n = 3). f, Normalized levels of TCA-cycle and associated metabolites (n = 3). g, Steady-state labeling fraction into citrate from [U-13C]-substrates glutamine and glucose, respectively (n = 3). h, Immunoblot of extracted mitochondria for subunits of respiratory chain complexes I-V and markers of mitochondrial mass. i, Mitochondrial complex I levels (NDUFS4) in independent HCT116 folate 1C gene knockout clones. Results are given as mean ± s.e.m; n indicates the number of biological replicates. Significance is calculated by two-tailed Student´s t-test. * P <0.01 (see Supplementary Table 7 for exact p values). Abbreviations: CI-V, respiratory chain complex I-V.

Extended Data Figure 2. Catalytically deficient SHMT2 constructs

a, Mapping of mutated amino acid residues on human SHMT1 (PDB code 1BJ4) using iCn3D and alignment of E. coli serine hydroxymethyltransferase (GLYA), H. sapiens mitochondrial serine hydroxymethyltransferase 2 (GLYM) and cytosolic serine hydroxymethyltransferase 1 (GLYC). Positions for GLYM are given with reference to GenBank NM_005412.5. b, Sanger sequencing traces of mutant constructs. c, Immunoblot for mitochondrial complex I levels (NDUFS4) in cell lines re-expressing catalytically deficient forms of SHMT2.

Extended Data Figure 3. Restoring SHMT2 catalytic activity is critical for normalizing one-carbon flux, respiratory chain expression, glycolytic activity, and cell growth

a, Immunoblot of re-expression of catalytically active SHMT2 and effects on mitochondrial complex I and II levels. b-f, Impact of re-expression of catalytically active and inactive forms of SHMT2 in the HEK293T background. b, Normalized NAD+/NADH ratio (n = 6) c, lactate secretion and glucose uptake (n = 6), d, cell proliferation (n = 6). e, Purine biosynthesis intermediate 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR) levels (n = 4) as an indicator of cytosolic folate 1C status, and f, [2,3,3-2H]-serine tracing to differentiate cytosolic from mitochondrial folate 1C unit production for incorporation into deoxythymidine triphosphate (n = 3). Results are given as mean ± s.e.m; n indicates the number of biological replicates. Significance is calculated by two-tailed Student´s t-test. * P <0.01 (see Supplementary Table 7 for exact p values).

Extended Data Figure 4. Oxidative phosphorylation defect is caused by a post-transcriptional mechanism independent of methionine formylation

a, Fraction of initiating amino acid (formyl-methionine vs. methionine) of mitochondrial-expressed MTCO1 peptide determined by high resolution liquid-chromatography mass spectrometry (WT n = 4, ΔSHMT2 n = 3, MTHFD2 n = 2). b, Lactate secretion (n = 3) upon sarcosine supplementation (1 mM). c, Relative mtDNA levels in HEK293T (n = 3). d, agarose gel of mtDNA long range-PCR products of HCT116 and HEK293T knockout cell lines. e, relative mRNA levels of mtDNA encoded respiratory chain subunits (n = 3). f, Gene expression levels in SHMT2 knockout cell lines compared to SHMT2 wild-type re-expressed lines by total RNA sequencing. Each dot represents mean gene expression as derived from two biological replicates of two independent knockout clones and matched re-expressed lines (n = 4). Genes linked to human OXPHOS function are highlighted in red. Significantly differentially expressed genes are listed in Supplementary Table 2. g, Position dependent next generation sequencing coverage of mtDNA in HEK293T wild-type, SHMT2 knockout and MTHFD2 knockout cell lines supports the absence of deletions due to SHMT2 loss. h, Corresponding variant position and frequency. Variant list is given in Supplementary Table 1. Bar graphs show mean ± s.e.m; n indicates the number of biological replicates. Significance is calculated by two-tailed Student´s t-test. * P <0.01 (see Supplementary Table 7 for exact p values). Abbreviations: mtDNA, mitochondrial DNA.

Extended Data Figure 5. Impairment of mitochondrial translation due to loss of SHMT2

a, SDS-PAGE of [35S]-methionine labeled mitochondrially translated proteins in wild-type (lane 1) and two SHMT2 knockout HEK293T cell lines (lane 2, 3). Decreased synthesis of COX1 and COX2/3 are evident upon short exposure and reduced synthesis of ND5 and ND6 is more easily visualized upon longer exposure. b, Absorbance at 254 nm upon sucrose gradient fractionation of Micrococcal nuclease digested cell lysates (Figure 3a). Fractions corresponding to 4 and 5 were collected for mitochondrial ribosome enrichment as shown on the matched immunoblot for mitochondrial ribosome subunit MRPL11. c, Read length distribution (top) and read length-dependent sub-codon read phasing (bottom) across the 13 mitochondrial protein coding transcripts. Panel c is based on data from the mitochondrial ribosome profiling experiment displayed in main text Figure 3 and displays the mean of two technical replicates of two independent samples.

Extended Data Figure 6. Mitochondrial ribosome stalling at guanosine-ending split codon box nucleotide triplets suggests deficient 5-taurinomethyluridine modification

a, Expanded version of main text Figure 3b, in this case showing the mean cumulative ribosome protected fragments of all mitochondrial protein-coding genes. b, Mean relative density of actively translating (i.e. not stalled) ribosomes for mitochondrial transcripts. a and b, two technical replicates of two independent samples. c, Enzymatic activities of citrate synthase and individual mitochondrial respiratory chain complexes from mitochondrial extracts (n = 5). Bar graphs show mean ± s.e.m. Significance is calculated by two-tailed Student´s t-test. * P <0.01 (see Supplementary Table 7 for exact p values). d, Mitochondrial genetic code table with split codon boxes depending on taurinomethylated tRNAs for translation highlighted in red. Codons decoded by anticodon formylcytidine-containing tRNAMet are highlighted in blue. e, Mean codon-specific mitochondrial ribosome occupancy of HCT116 SHMT2/MTHFD2 double knockout cell lines supplemented with sarcosine (1 mM). Codons highlighted in red are decoded by tRNAs carrying a 5-taurinomethyluridine modification. The supplementation with sarcosine prevents the stalling normally observed with SHMT2 deletion (n =2).

Extended Data Figure 7. tRNA modification status in ΔSHMT2 and effects of 5-taurinomethyluridine modification loss caused by human disease gene MTO1

a, Total ion chromatogram of 5-formylcytidine monophosphate in digested mitochondrial tRNAs upon loss of SHMT2. The same samples were analyzed for 5-taurinomethyluridine monophosphate (tm5U) in Figure 4b. The combined data demonstrate that SHMT2 deletion causes loss of tm5U but not 5-formylcytidine. b, levels of tm5U, 5-taurinomethyl-2-thiouridine monophosphate (tm5s2U) and 2-thiouridine monophosphate (s2U) in wild-type HCT116 and SHMT2 deletion lines normalized to 5-formylcytidine monophosphate (f5C) (n = 3). c, Taurine levels in HCT116 wild-type and SHMT2 knockout cell lines (n = 3). d, tm5U levels in digested mitochondrial tRNAs upon re-expression of SHMT2 (n = 1). e, tm5U, tm5s2U and s2U levels normalized to f5C in HCT116 SHMT2/MTHFD2 knockout lines after sarcosine supplementation and HCT116 upon loss of MTO1 (n = 2). For all panels data is presented as mean ± s.e.m. or individual data points only. f, Labeling pattern of 5-taurinomethyluridine and 5-formylcytidine monophosphate extracted from mitochondrial tRNAs after growth in media containing either [3-13C]-serine or [U-13C]-methionine. g, Mean cumulative count of ribosome protected fragments (RPF) mapping to mitochondrial protein coding transcripts upon ribosome profiling in HCT116 MTO1 knockout cell lines. Data was normalized to reads per million (RPM) (n = 2); n indicates the number of biological replicates. Significance is calculated by two-tailed Student´s t-test * P <0.01 (see Supplementary Table 7 for exact p values).

Extended Data Figure 8. Investigation of mRNA and protein secondary structure effects on mitochondrial ribosome stalling sites

a, Identification of mitochondrial RNA secondary structure based on analysis of the mitochondrial transcript data from the dimethyl sulphate sequencing dataset published by Rouskin S. et al. R-value and Gini difference were calculated to detect changes in nucleotide reactivity between the in vivo and denatured condition for the complete mitochondrial transcriptome. Colored points indicate structured regions given in Supplementary Table 4. b, Determination of ribosome stalling sites in HCT116 SHMT2 knockout cell lines. Data points represent individual codons of all 13 mitochondrial protein-coding transcripts. For each codon, the y-axis indicates the ribosome counts normalized to the gene-median in reads per million (RPM). The x-axis indicates the ratio of normalized counts in SHMT2 knockout to normalized counts in wild-type HCT116. Two and three standard deviations (SD) above the mean of all codons in the genome are indicated by the grey and black dotted line respectively. Highlighted in red are codons > 2 SD. c, Mapping of AAG/UUG codons from SHMT2 knockout-specific ribosome stalling sites (> 3 SD) on protein structures. For b and c, analysis based on ribosome profiling data presented in main text Figure 3 with two technical replicates of two independent samples. A list of identified codons and mapped AAG/UUG sites is given in Supplementary Table 5.

Extended Data Figure 9. Mitochondrial transcript codon occupancy from ribosome profiling of individual patient lines

a, Codon-specific mitochondrial ribosome occupancy ratio (patient/control fibroblasts) in individual patient derived cell lines (n = 1 for each individual patient, normalized to mean of n = 2 control fibroblast lines). Patients had either nuclear MTO1 missense mutations (pat_a c.[1261-5T>G];[1430G>A], pat_b c.[1222T>A];[1222T>A]) or were diagnosed with MELAS and carry the recurrent point mutation m.3243A>G in the mitochondrial gene for tRNA Leu1 (MT-TL1). b, Next-generation sequencing of mtDNA mutation load m.3243A>G (MT-TL1) in control fibroblasts and MELAS patient cell lines. Each bar shows one biological replicate for control and patient cell lines. Integrative genomics viewer sequencing raw data is shown on the right.

Extended Data Figure 10. Effects of targeting one-carbon metabolism on mitochondrial function

a, Mitochondrial complex I and II levels on immunoblot after growth in the absence of folate for 5 passages or in the presence of the indicated methotrexate concentration for 96 h. Ethidium bromide (250 nM) was used as a positive control. b, Cellular mtDNA levels in HCT116 cells upon folate depletion (with or without hypoxanthine and thymidine as rescue agents) or presence of methotrexate for 96 h (n = 3). c, In an effort to determine if the decrease in respiration due to methotrexate arises from methotrexate depleting mitochondrial DNA, impairing mitochondrial translation, or a combination, we compared the effects in HCT116 cells of methotrexate (50 nM) to ethidium bromide (250 nM = 100 ng/ml), which is classically used to deplete mitochondrial DNA, and to chloramphenicol (310 μM = 100 μg/ml), which blocks mitochondrial translation. After 48 h of treatment, methotrexate and ethidium bromide both decreased oxygen consumption and DNA content. Importantly, despite ethidium bromide much more strongly depleting mitochondrial DNA, methotrexate had an equivalent effect on oxygen consumption consistent with methotrexate’s effect on oxygen consumption being in part via mitochondrial translation inhibition. Data is normalized and compared to untreated control (all n = 3; except oxygen consumption methotrexate 96 h n = 6 and control n =4). Data are reported as mean ± s.e.m; n indicates the number of biological replicates. Significance is calculated by two-tailed Student´s t-test * P <0.01 (see Supplementary Table 7 for exact p values), Abbreviations: H/T, hypoxanthine (100 μM)/thymidine (16 μM); Mtx., methotrexate.

Figure 1. Mitochondrial respiratory chain function is dependent on SHMT2 catalytic activity

a, 1C pathway and known mitochondrial products. b, Lactate secretion of HCT116 knockout cell lines (n = 6). c, Oxygen consumption rate measured by Seahorse XF analyzer (n = 3). d, Immunoblot for mitochondrial respiratory complex I and II proteins (NDUFS4 and SDHA), 1C enzymes, and mitochondrial mass (VDAC1). e, Basal respiration (n = 3) upon re-expression of wild-type or catalytically deficient mutant forms of SHMT2 in HEK293T knockout cell lines. Results are given as mean ± s.e.m; n indicates the number of biological replicates, which for Seahorse experiments refers to independent plates on separate days. Significance is calculated by two-tailed Student´s t-test. * P <0.01 (see Supplementary Table 7 for exact p values). Abbreviations: methylene-THF, 5,10-methylene-THF; formyl-THF, 10-formyl-THF; THF, tetrahydrofolate; dTTP, deoxythymidine triphosphate; f-Met, n-formyl-methionine; Oligom., oligomycin; Rot./Antim., rotenone/antimycin; Cat. inact., catalytically inactive SHMT2; PLP bind., PLP binding-deficient SHMT2.

Figure 2. ΔSHMT2 induced respiratory chain deficiency is caused by mitochondrial 5,10-methylene-THF depletion but is unrelated to dTTP synthesis

a, Sarcosine serves as a SHMT2-independent source of mitochondrial methylene-THF. b, NAD+/NADH ratio (n = 6) and NDUFS4 (complex I) protein expression upon sarcosine supplementation (1 mM) in SHMT2 single and SHMT2/MTHFD2 double knockout cell lines compared to wild-type. c and d, Functional readouts for mitochondrial dTTP status based on c, mtDNA level (n = 3) determined by quantitative real-time PCR and d, gene expression in SHMT2 knockout and wild-type HEK293T cells. For d, each data point represents mean gene expression of two biological replicates of two independent knockout clones (n = 4) and two wild-type replicates (n = 2). Genes linked to OXPHOS function are highlighted in red (nuclear encoded) or blue (mitochondrial encoded). Significantly differentially-expressed genes are listed in Supplementary Table 2. Bar graphs show mean ± s.e.m; n indicates the number of independent biological replicates. Significance is calculated by two-tailed Student´s t-test. * P <0.01 (see Supplementary Table 7 for exact p values). Abbreviations: dTTP, deoxythymidine triphosphate; mtDNA, mitochondrial DNA; RPKM, reads per kilobase per million mapped reads.

Figure 3. Mitochondrial ribosome profiling reveals ΔSHMT2 cells are deficient in translating specific guanosine-ending codons

a, Workflow of mitochondrial ribosome profiling: translation was halted using chloramphenicol and immersion into liquid nitrogen, cells were lysed and RNA was digested using Micrococcal nuclease (MNase). Following sucrose-gradient enrichment for mitochondrial ribosomes (shaded in red), protected fragments were sequenced. b, Mean cumulative ribosome density along selected mitochondrial transcripts. Black is wild-type HCT116 and red is ΔSHMT2. Additional transcripts are given in Extended Data Fig. 6a. c, Mean codon-specific mitochondrial ribosome occupancy (HCT116ΔSHMT2/HCT116wt;). Data points highlighted in red correspond to codons that are decoded by tRNAs carrying the 5-taurinomethyluridine modification. Text labels highlighted in red correspond to the subset of those codons that end in guanosine and thus require wobble-base pairing. Methionine codons are highlighted in blue and show no increased codon occupancy. The insert shows mean normalized ribosome density relative to UUG and AAG codon position. b and c, two technical replicates of two independent samples. Abbreviations: RPF, ribosome protected fragment; RPM, reads per million mapped reads.

Figure 4. MTO1/GTPBP3 dependent tRNA methylation requires mitochondrial 5,10-methylene-THF

a, Interaction of tRNA position 34 anticodon loop modified base with messenger RNA codon three position A/G, forming a non-Watson-Crick base pair. b, Total ion chromatogram of 5-taurinomethyluridine monophosphate (m/z = 460.043) from digested mitochondrial tRNAs. 5-formylcytidine monophosphate was not altered (Extended Data Fig. 7a). c, Mean codon-specific mitochondrial ribosome occupancy for HCT116 MTO1 knockout cell lines and primary patient derived fibroblasts carrying MTO1 mutations or the MT-TL1 m.3243A>G MELAS variant (n = 2). Corresponding immunoblots are shown below. Individual patient data are in Extended Data Fig. 9a. d, Basal respiration rates measured using the Seahorse XF analyzer. Data were collected after growth in the absence of folate for 5 passages or in the presence of the indicated methotrexate concentration for 96 h (n = 3, except HCT116_WT n = 4 and Mtx 50 nM n = 6). Graphs show mean ± s.e.m; n indicates the number of biological replicates. Significance is calculated by two-tailed Student´s t-test. * P <0.01 (see Supplementary Table 7 for exact p values). Abbreviations: Mtx, methotrexate; H/T, hypoxanthine (100 μM)/thymidine (16 μM).