|Home | About | Journals | Submit | Contact Us | Français|
The metazoan mitochondrial translation machinery is unusual in having a single tRNAMet that fulfills the dual role of the initiator and elongator tRNAMet. A portion of the Met-tRNAMet pool is formylated by mitochondrial methionyl-tRNA formyltransferase (MTFMT) to generate N-formylmethioninetRNAMet (fMet-tRNAmet), which is used for translation initiation; however, the requirement of formylation for initiation in human mitochondria is still under debate. Using targeted sequencing of the mtDNA and nuclear exons encoding the mitochondrial proteome (MitoExome), we identified compound heterozygous mutations in MTFMT in two unrelated children presenting with Leigh syndrome and combined OXPHOS deficiency. Patient fibroblasts exhibit severe defects in mitochondrial translation that can be rescued by exogenous expression of MTFMT. Furthermore, patient fibroblasts have dramatically reduced fMet-tRNAMet levels and an abnormal formylation profile of mitochondrially translated COX1. Our findings demonstrate that MTFMT is critical for efficient human mitochondrial translation and reveal a human disorder of Met-tRNAMet formylation.
Of the ~90 protein components of the oxidative phosphorylation (OXPHOS) machinery, 13 are encoded by the mitochondrial DNA (mtDNA) and translated within the organelle. Defects in mitochondrial protein synthesis lead to combined OXPHOS deficiency. Although the mtDNA encodes the ribosomal and transfer RNAs, all remaining components of the mitochondrial translational machinery are encoded by nuclear genes and imported into the organelle. To date, mutations in more than 10 different nuclear genes have been shown to cause defective mitochondrial translation in humans. However, molecular diagnosis by sequencing these candidates in patients with defects in mitochondrial translation is far from perfect (Kemp et al., 2010), underscoring the need to identify additional pathogenic mutations underlying these disorders.
Translation within metazoan mitochondria is reminiscent of the bacterial pathway, initiating with N-formylmethionine (fMet) (Kozak, 1983). Unlike bacteria, which encode distinct tRNAMet molecules for translation initiation and elongation, metazoan mitochondria express a single tRNAMet that fulfills both roles (Anderson et al., 1981). After aminoacylation of tRNAMet, a portion of Met-tRNAMet is formylated by mitochondrial methionyl-tRNA formyltransferase (MTFMT) to generate fMet-tRNAMet. The mitochondrial translation initiation factor (IF2mt) has high affinity for fMet-tRNAMet, which is recruited to the ribosomal P-site to initiate translation (Spencer and Spremulli, 2004). In contrast, the mitochondrial elongation factor (EF-Tumt) specifically recruits Met-tRNAMet to the ribosomal A-site to participate in polypeptide elongation. Synthesized proteins can then be deformylated by a mitochondrial peptide deformylase (PDF) and demethionylated by a mitochondrial methionyl aminopeptidase (MAP1D)(Serero et al., 2003; Walker et al., 2009).
Here, we applied targeted exome sequencing to two unrelated patients with Leigh syndrome and combined OXPHOS deficiency to discover pathogenic mutations in MTFMT. Fibroblasts from these patients have impaired Met-tRNAMet formylation, peptide formylation, and mitochondrial translation. Despite studies in yeast suggesting that MTFMT is not essential for mitochondrial translation (Hughes et al., 2000; Li et al., 2000; Vial et al., 2003), we show that in humans this gene is required for efficient mitochondrial translation and function.
We studied two unrelated patients with Leigh syndrome and combined OXPHOS deficiency (Fig. 1A). Clinical summaries for Patient 1 (P1) and Patient 2 (P2) are provided in Supplementary Data. Patient fibroblasts had reduced synthesis of most mtDNA-encoded proteins as assayed by [35S]-methionine labeling in the presence of inhibitors of cytosolic translation (Fig. 1B). This correlated with reduced steady state protein levels as detected by immunoblotting (Fig. 1C), and, at least for ND1, was not due to reduced mRNA (Supplementary Fig. S1). Collectively, these data suggest a defect in translation of mtDNA-encoded proteins.
To elucidate the molecular basis of disease in P1 and P2, we performed next-generation sequencing of coding exons from 1034 nuclear-encoded mitochondrial-associated genes and the mtDNA (collectively termed the “MitoExome”). DNA was captured using an in-solution hybridization method (Gnirke et al., 2009) and sequenced on an Illumina GA-II platform (Bentley et al., 2008). Details are provided in Supplementary Data and Supplementary Table S1.
We identified ~700 single nucleotide variants (SNVs) and short insertion or deletion variants (indels) in each patient relative to the reference genome, and prioritized those that may underlie a severe, recessive disease (Fig. 2A). We first filtered out likely benign variants present at a frequency of >0.005 in public databases which left ~20 variants in each patient We then prioritized variants that were predicted to have a deleterious impact on protein function (Calvo et al., 2010), leaving ~12 variants. Focusing on genes that fit autosomal recessive inheritance, having either homozygous variants or two different variants in the same gene, only one candidate gene, MTFMT, remained in both patients (Fig. 2A).
We identified three distinct heterozygous variants in our patients (Fig. 2B). Both patients harbor a c.626C>T mutation. The c.626C site is 20bp upstream of the 3’ end of exon 4 and is predicted to eliminate two overlapping exonic splicing enhancers (GTCAAG, TCAAGA) (Fairbrother et al., 2002) and to generate an exonic splicing suppressor (GTTGTT) (Wang et al., 2004). Skipping of exon 4 results in a frameshift and premature stop codon (p.R181SfsX5). The second mutation in P1 is a nonsense mutation (c.382C>T, p.R128X), while the second mutation in P2 changes a highly conserved serine to leucine in the catalytic core of MTFMT (c.374C>T, p.S125L) (Supplementary Fig. S2). An affected cousin of P1 also carries the c.382C>T and c.626C>T mutations.
As predicted by in silico analysis, the shared c.626C>T mutation caused skipping of exon 4 (Fig. 2C). qRT-PCR analysis revealed that P1 had only 9% full length MTFMT transcript compared to controls (Supplementary Fig. S2), the majority of which carries the c.382C>T nonsense mutation and lacks the c.626C>T splicing mutation (Fig. 2C). P2 had 56% full length MTFMT transcript (Supplementary Fig. S2), all of which appears to carry the c.374C>T mutation and to lack the c.626C>T splicing mutation (Fig. 2C). Collectively, these results confirm compound heterozygosity of the MTFMT mutations and almost complete exon skipping due to the c.626C>T mutation.
We used cDNA complementation to prove that the translation defect in these patients is due to mutations in MTFMT. Fibroblasts from both patients showed reduced levels of the mtDNA-encoded complex IV subunit, COX2, consistent with a defect in mitochondrial translation, and of the nuclear-encoded complex I subunit, NDUFB8, reflecting instability of complex I in the absence of mtDNA-encoded proteins (Fig. 2D). Lentiviral transduction of MTFMT cDNA caused a significant increase of COX2 and NDUFB8 in both patients (Fig. 2E). In contrast, lentiviral transduction of a control cDNA, C8orf38, caused no change of these subunits (Fig. 2E). These data confirm that an MTFMT defect is responsible for the combined OXPHOS deficiency in these patients.
To directly analyze the mitochondrial tRNAMet pools (Fig. 3A), we used a modified protocol of acid-urea PAGE followed by Northern blotting (Enriquez and Attardi, 1996; Köhrer and RajBhandary, 2008; Varshney et al., 1991) (Fig. 3B). We were able to separate the mitochondrial uncharged tRNAMet, Met-tRNAMet and fMet-tRNAMet from total RNA isolated from fibroblasts and to show that two independent wild-type cell lines contained uncharged tRNAMet and fMet-tRNAMet, but very little Met-tRNAMet (Fig. 3B, lanes 1 – 7). In striking contrast, fibroblasts from P1 and P2 lacked detectable fMet-tRNAMet and contained mostly Met-tRNAMet along with traces of the uncharged tRNAMet (Fig. 3B, compare lanes 8 – 13 to control lanes 5 – 7). We also observed a 2.7 fold increase of the overall mitochondrial tRNAMet signal in patient fibroblasts compared to control (Fig. 3B, top panel; compare lanes 8 and 11 to control lane 5), while the cytoplasmic initiator tRNAiMet showed constant signal throughout (Fig. 3B, bottom panel). The analysis of the mitochondrial tRNAMet pools clearly shows a defect in tRNAMet formylation.
Although fibroblasts from P1 and P2 have severely impaired mitochondrial translation, they do retain residual activity (Fig. 1B). This residual activity could be due to (i) low activity of mutant MTFMT generating a small amount of fMet-tRNAMet that is rapidly consumed in translation initiation and, therefore, undetectable by Northern blot analyses; and/or (ii) the human IF2mt recognizing, albeit weakly, the non-formylated Met-tRNAMet species to support translation initiation. Translation through the first mechanism would produce formylated protein, while translation through the second mechanism would produce unformylated protein.
To investigate these two possibilities, we used semi-quantitative mass spectrometric analysis to simultaneously measure three possible N-terminal states of mitochondrially translated COX1: formylated (Fig. 4A), unformylated (Fig. 4B) and demethionylated (des-Met) (Fig. 4C). We applied this method to complex IV immunoprecipitated from fibroblasts from P1 and P2 and two independent wild-type cell lines (Fig. 4D). Although no fMet-tRNAMet was detected in patient fibroblasts by Northern blotting (Fig. 3B), the dominant COX1 peptide in all four samples is the formylated species as estimated from total ion current of each form (Fig. 4E). Thus, patient fibroblasts retain residual MTFMT activity. The expression of mitochondrial PDF and MAP1D was normal in patient fibroblasts (Supplementary Fig. S3).
Here we describe the first human patients with mutations in MTFMT. The causal mutations were verified by rescuing the mitochondrial translation defects in patient fibroblasts via lentiviral transduction of MTFMT. Analysis of the tRNAMet pools in patient fibroblasts revealed severe MTFMT dysfunction. To our knowledge, this is the first time the human mitochondrial tRNAMet profile has been analyzed. It is interesting to note that control fibroblasts lack detectable Met-tRNAMet, suggesting that it is utilized as quickly as it is produced; either converted to fMet-tRNAMet or used to donate Met to the growing polypeptide chain. Strikingly, patient fibroblasts lack detectable levels of fMet-tRNAMet and contain mostly Met-tRNAMet.
Drastically decreased fMet-tRNAMet levels prevent efficient mitochondrial translation as demonstrated by the reduced translation observed in patient fibroblasts. Although fibroblasts from P1 and P2 have severely impaired mitochondrial translation, they do retain some residual activity. To understand the origin of this activity, we measured the relative distribution of three possible N-terminal states of mitochondrially translated COX1 by mass spectrometry. While this is not the first targeted study of the N-termini of COX1 (Escobar-Alvarez et al., 2010), to our knowledge, this is the first time all three states are measured, and the first observation of mitochondrial methionine excision activity, which is detectable albeit weak.
Formylated COX1 is the dominant species in patient fibroblasts, indicating residual MTFMT activity. Assuming P1's nonsense mutation has a full loss of function, then the allele harboring the shared c.626C>T mutation must confer MTFMT activity. Transcript that has not undergone skipping of exon 4 encodes an MTFMT variant harboring a p.S209L missense mutation. Residue p.S209 is moderately conserved and lies on the periphery of MTFMT based on homology with the bacterial enzyme.
Similarly, P2's residual MTFMT activity must originate from enzyme variants carrying the p.S209L mutation and/or the p.S125L mutation located in the active site.
Studies in bacteria and yeast have raised questions about the absolute requirement for Met-tRNAMet formylation. Formylation is not essential in all bacteria (Newton et al., 1999) and in yeast disruption of FMT1 causes no discernible defect in mitochondrial protein synthesis or function (Hughes et al., 2000; Li et al., 2000; Vial et al., 2003). Additionally, bovine IF2mt is able to restore respiration in a yeast mutant lacking both IF2mt and FMT1 (Tibbetts, 2003), suggesting that bovine IF2mt, like yeast IF2mt, can initiate protein synthesis without fMet-tRNAMet. However, a number of studies in mammals indicate that formylation of mitochondrial Met-tRNAMet is required for translation initiation. Bovine IF2mt has a 25-50 fold greater affinity for fMet-tRNAMet than for Met-tRNAMet in vitro (Spencer and Spremulli, 2004) and 12 of the 13 bovine mtDNA-encoded proteins retain fMet at the N-terminus (Walker et al., 2009). The presence of some Met instead of fMet at the N-terminus of COX1 from patient fibroblasts (Fig. 4E) could, however, suggest that unformylated Met-tRNAMet can initiate mitochondrial translation although less efficiently than fMet-tRNAMet.
One of the factors that could allow Met-tRNAMet to initiate mitochondrial translation is its increased concentration in patient fibroblasts. In Salmonella typhimurium, amplification of initiator tRNA genes compensates for a lack of methionyl-tRNA formyltransferase activity and allows translation initiation without formylation of the initiator tRNA (Nilsson et al., 2006). The “up-regulation” of the mitochondrial tRNAMet in patient fibroblasts (Fig. 3B) could, in principle, be a compensatory response due to limited fMet-tRNAMet.
In summary, we have used MitoExome sequencing to identify MTFMT as a gene underpinning combined OXPHOS deficiency associated with Leigh syndrome. We have shown that patient fibroblasts have a striking deficiency of fMet-tRNAMet leading to impaired mitochondrial translation. Despite studies in yeast suggesting that MTFMT is not essential for mitochondrial translation (Hughes et al., 2000; Li et al., 2000; Vial et al., 2003), we show here that in humans this gene is required for efficient mitochondrial translation and function. More generally, this study demonstrates how MitoExome sequencing can reveal insights into basic biochemistry and the molecular basis of mitochondrial disease.
Cells were grown at 37°C and 5% CO2 in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) supplemented with 10% (v/v) fetal bovine serum (FBS, Invitrogen).
Spectrophotometric analysis of mitochondrial OXPHOS activity was performed as described previously (Kirby et al., 1999). Investigations were performed with informed consent and in compliance with ethics approval by the Human Research Ethics Committee of the Royal Children's Hospital, Melbourne.
MtDNA-encoded proteins in patient fibroblasts were labeled using 35S-methionine/35S-cysteine (EXPRE35S35S Protein Labeling Mix; Perkin Elmer Life Sciences) prior to mitochondrial isolation and analysis of translation products by SDS-PAGE as previously described (McKenzie et al., 2009).
Immunoblotting was performed as previously described (Calvo et al., 2010). Proteins were detected with the following antibodies: complex II α-70kDa subunit monoclonal antibody (MitoSciences, MS204), ND1 polyclonal antibody (kind gift from Dr Anne Lombes, Paris), α-complex IV subunit I monoclonal antibody (Invitrogen, 459600), α-complex IV subunit II monoclonal antibody (Invitrogen, A6404), Total OXPHOS Human WB Antibody Cocktail containing αNDUFB8 and αCOX2 (MitoSciences, MS601) and either α-mouse or α-rabbit IgG horseradish peroxidase (HRP; DakoCytomation).
We used an in-solution hybridization capture method (Gnirke et al., 2009) to isolate target DNA, that was sequenced on the Illumina GA-II platform (Bentley et al., 2008). The 4.1Mb of targeted DNA included the 16kb mtDNA and all coding and untranslated exons of 1381 nuclear genes, including 1013 mitochondrial genes from the MitoCarta database (Pagliarini et al., 2008), 21 genes with recent strong evidence of mitochondrial association, and 347 additional genes. All analyses were restricted to the mtDNA and coding exons of the 1034 genes with confident evidence of mitochondrial association (1.4Mb). Detailed methods for target selection, sequencing, alignment and variant detection are submitted elsewhere (Calvo et al. unpublished data).
Differences in DNA sequence between each individual and the GRCh37 human reference assembly were identified. Nuclear variants that passed quality control metrics were prioritized according to three criteria: i) SNV allele frequency <0.005 in public databases [dbSNP (Sherry et al., 2001) version 132 and the 1000 genomes project (Durbin et al., 2010) released November 2010] or indels absent in the 1000 genomes data; ii) variants predicted to modify protein function as previously described (Calvo et al., 2010); iii) variants consistent with recessive inheritance (homozygous variants or two heterozygous variants in the same gene). We also prioritized mtDNA variants annotated as pathogenic in MITOMAP (Ruiz-Pesini et al., 2007). Detailed methods are submitted elsewhere (Calvo et al. unpublished data).
DNA isolation, RNA isolation, cDNA synthesis, inhibition of nonsense mediated decay, and sequencing of PCR products were performed as described previously (Calvo et al., 2010).
The MTFMT open reading frame (ORF) was purchased in a pCMV-SPORT6 vector (Clone ID: BC033687.1, Open Biosystems) and was cloned into the lentiviral vector pF_5x_UAS_MCS_SV40_puroGEV16-W vector (Yeap et al., 2010). MTFMT viral particles were generated and patient fibroblasts were transduced as described previously (Calvo et al., 2010). Three independent transductions were performed and cells were harvested after 10-12 days selection with 1μg/mL puromycin.
Total RNA was isolated under acidic conditions using TRIzol (Invitrogen) following manufacturer's instructions. Acid washed glass beads (0.5 mm diameter; Sigma) were added during extraction. Total RNAs were separated by acid-urea PAGE as described previously (Enriquez and Attardi, 1996; Köhrer and RajBhandary, 2008; Varshney et al., 1991) with modifications. Briefly, 0.1 A260 units of each RNA sample were loaded onto a 6.5% polyacrylamide gel containing 7 M urea and 0.2 M sodium acetate pH 5.0. Individual tRNAs were detected by Northern blotting (Köhrer and RajBhandary, 2008) using the following hybridization probes: 5’TAGTACGGGAAGGGTATAA3’ (mitochondrial tRNAMet) and 5’TTCCACTGCACCACTCTGCT3’ (cytoplasmic initiator tRNAiMet). Northern blots were quantified by PhosphorImager analysis using ImageQuant software (Molecular Dynamics). Experiments were performed in duplicate.
Aminoacyl-tRNAs and formylaminoacyl-tRNAs were deacylated by base treatment in 0.1 M Tris-HCl (pH 9.5) at 65°C for 5 min followed by incubation at 37°C for 1 h. Aminoacyl-tRNAs were selectively deacylated by treatment with copper sulfate as described (Schofield and Zamecnik, 1968). Total RNA was aminoacylated in vitro with methionine using E. coli MetRS as previously described with minor modifications (Köhrer and RajBhandary, 2008).
Briefly, complex IV was immunoprecipitated from control and patient fibroblasts using MitoSciences’ complex IV Immunocapture kit (MS-401) and separated by gel electrophoresis on a NuPAGE 4-12% Bis-Tris gel (Invitrogen). A band corresponding to the MW of COX1 was excised and subjected to ingel Lys-C digestion (Kinter & Sherman). Extracted peptides were separated on C18 column using a 1200-Series nano-LC pump (Agilent) and run on a LTQ-Velos-Orbitrap mass spectrometer (ThermoFisher) set to scan and to targeted MS/MS for m/zs corresponding to the z=3 states of the unformylated, formylated and des-Met species of the N-terminal peptide of COX1 (MFADRWLFSTNHK).
Extracted ion chromatographs (XICs) were generated from the Orbitrap survey scans based on the z=3 states of the unformylated, formylated and des-Met species of the N-terminal peptide of COX1 using XCalibur software (ThermoFisher Scientific). The identities of the peaks corresponding to these species were verified using the accompanying static MS/MS spectra (Fig. 4A-C). The areas under these peaks were integrated using the Genesis peak detection algorithm included in XCalibur with all standard defaults. Peak areas were further normalized to the peak area of a distal peptide of COX1 (VFSWLATLHGSNMK, m/z 795.9085, z=2) to ensure that comparisons allowed for the different amounts of COX1 in control and patient samples. Full methods are in the Supplemental Data.
Two-way repeated measures analysis of variance (ANOVA) was used for comparisons of groups followed by post hoc analysis using the Bonferroni method.
We thank Drs J. Silke and P. Ekert for providing the pF 5x_UAS_MCS_SV40_puroGEV16_W vector, C. Guiducci, C. Sougnez, L. Ambrogia, J. Wilkinson, for assistance with sample preparation and sequencing, T. Fennel, M. DePristo, E. Banks, and K. Garimella for assistance with bioinformatic analysis, S. Flynn for assistance with IRBs, and the subjects and referring physicians who participated in the study. This work was supported by an Australian Postgraduate Award to E.J.T., a National Defense Science and Engineering Graduate Fellowship to S.G.H., an Australian National Health and Medical Research Council (NHMRC) Career Development Award to M.M., an NHMRC Principal Research fellowship to D.R.T., by the Victorian Government's Operational Infrastructure Support Program, and by grants from the Ramaciotti Foundation and the James and Vera Lawson Trust to M.M., the NHMRC to M.M., M.T.R. and D.R.T., and the National Institutes of Health to U.L.R. (GM17151) and to V.K.M. (GM077465 and GM097136).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.