Aim-less translation: loss of Saccharomyces cerevisiae mitochondrial translation initiation factor mIF3/Aim23 leads to unbalanced protein synthesis

The mitochondrial genome almost exclusively encodes a handful of transmembrane constituents of the oxidative phosphorylation (OXPHOS) system. Coordinated expression of these genes ensures the correct stoichiometry of the system’s components. Translation initiation in mitochondria is assisted by two general initiation factors mIF2 and mIF3, orthologues of which in bacteria are indispensible for protein synthesis and viability. mIF3 was thought to be absent in Saccharomyces cerevisiae until we recently identified mitochondrial protein Aim23 as the missing orthologue. Here we show that, surprisingly, loss of mIF3/Aim23 in S. cerevisiae does not indiscriminately abrogate mitochondrial translation but rather causes an imbalance in protein production: the rate of synthesis of the Atp9 subunit of F1F0 ATP synthase (complex V) is increased, while expression of Cox1, Cox2 and Cox3 subunits of cytochrome c oxidase (complex IV) is repressed. Our results provide one more example of deviation of mitochondrial translation from its bacterial origins.

absent 11 while a suite of specific accessory factors, so-called translational activators, promote yeast mitochondrial translation in an mRNA-specific manner 12 . Duplication and subsequent divergence of elongation factor EF-Tu in arthropods has led to paralogs that are specialized for delivery of specific tRNA species of highly unusual architecture 13,14 , and most eukaryotes carry two copies of EF-G, which have become specialized for one of EF-G's two roles in ribosome recycling and translocation 14,15 . Polypeptide release in human mitochondria is mediated by four release factors-RF1Lmt/mtRF1a, RF1 mt, C12orf65 and ICT1-with the latter being an integral component of the mitochondrial ribosome 16 .
Translation initiation in mitochondria is orchestrated by bacteria-like general initiation factors mIF2 and mIF3, which in yeast are aided by the mitochondria-specific translational activators [17][18][19][20] . For many years, mIF3 was thought to be absent in the yeast S. cerevisiae, due to a homologue not being found using standard sequence searching methods. Recently however, identification of S. cerevisiae mitochondrial protein Aim23 as an mIF3 orthologue using more sensitive searching and phylogenetic analysis has paved the way for genetic investigations of mIF3 function in this model organism 11 . High-throughput screening assays following yeast growth on non-fermentable media together with determination of petite frequencies have demonstrated that the aim23 gene is required for mitochondrial functionality, and Aim23 has been hypothesized to be involved in assembly of respiration complexes 21 . Supporting this hypothesis is a lack of observable membrane potential and dramatically decreased oxygen consumption in an aim23∆ strain 11 . The mitochondrial defect in the aim23∆ strain is complemented by the expression of human 17 and Schizosaccharomyces pombe 11 mIF3 and-partially-of Escherichia coli IF3 17 , demonstrating that Aim23 is a bona fide mIF3.
To date the role of mIF3/Aim23 in S. cerevisiae mitochondrial translation has not been directly tested. The bacterial homolog of mIF3/Aim23, IF3, is important for tRNA and mRNA selection during translation initiation 22 . In addition to its role in translation initiation, IF3 participates in ribosomal recycling after completion of the polypeptide chain -it prevents re-association of ribosomal subunits dissociated by translational factors EF-G and RRF 23 and promotes subsequent dissociation of tRNA and mRNA from the small subunit 24 . Given the central role of IF3 in the ribosomal functional cycle it is not surprising that the gene infC encoding IF3 in Escherichia coli is essential 25 . Moreover, a decrease in IF3 cellular level results in dramatic reduction of the polysomal fraction, indicating abrogation of cellular protein biosynthesis 25 .
In this report we have investigated the role of mIF3/Aim23 in mitochondrial functionality and protein synthesis in yeast S. cerevisiae. Surprisingly, mIF3/Aim23 is partially dispensable for mitochondrial functionality and mitochondrial protein synthesis; S. cerevisiae lacking the AIM23 gene can still grow on non-fermentable carbon sources requiring mitochondrial respiration, and the mitochondrial translational system can synthesize the full protein repertoire encoded in the mtDNA. However, the absence of mIF3/Aim23 causes a pronounced misbalance in the relative levels of mitochondrially encoded proteins and significant retardation of growth on non-fermentable media requiring respiration. These results underscore the differences in translation initiation in mitochondria, where only a handful of mRNA molecules are translated with a help of numerous specialized factors, and bacteria, where translation of a vast variety of mRNAs is orchestrated by three essential canonical initiation factors IF1, IF2 and IF3.

Results
Effects of AIM23 disruption on S. cerevisiae growth on non-fermentable carbon sources and mitochondrial functionality. We have previously characterized the growth of a S. cerevisiae aim23∆ strain on solid media with glycerol as a non-fermentable carbon source-a common test for yeast mitochondrial functionality-and concluded that the strain is incapable of respiration 11,17 . However, inspection of plates incubated for 72 hours at 30 °C reveals that the AIM23-deficient strain does, eventually, form detectable colonies on both glycerol (Fig. 1A, upper panel) and ethanol (Fig. 1A, lower panel), although growth is significantly retarded in comparison to the parental strain. This growth delay is the likely reason why the phenomenon went unnoticed by us as well as Hess and colleagues who identified AIM23 as a gene necessary for mitochondrial functionality in earlier high-throughput assays 21 .
The S288C-based BY4741 background we originally used for creating the aim23∆ strain 11 carries several polymorphisms in mitochondrial DNA polymerase MIP1, calcium-dependent mitochondrial ADP/ATP carrier SAL1 and mitochondrial inner membrane protein involved in ubiquinone biosynthesis CAT5, which negatively affects the stability of mitochondrial DNA 26 (mtDNA). To make sure that the observed phenotype of the aim23∆ strain is not linked to these polymorphisms, we have recreated aim23∆ in a D273-10B background devoid of them 17 . Deletion of AIM23 leads to growth retardation in liquid YPGly regardless of the strain background (Fig. 1B). The effect is more pronounced in BY4741, possibly due to the cumulative effect of AIM23 loss on mitochondrial functionality and polymorphisms in MIP1, SAL1 and CAT5 destabilizing the mtDNA. The pronounced lag phase of the aim23∆ strains could be due to outgrowth on the glycerol media mediated by a fraction of the inoculum carrying secondary compensatory mutations that restore mitochondrial functionality. However, this is unlikely to be the case as the serial dilutions of the aim23∆ strain give similar numbers of colonies as the wild type, suggesting that re-growth is not mediated by only a fraction of mutant cells (Fig. 1A). Therefore, we conclude that the lag is caused by slow adaptation of the aim23∆ strain to non-fermentable media that requires mitochondrial respiration. Subsequent experiments were performed in a D273-10B background, as its increased mtDNA stability makes it more suitable for investigation of mitochondrial functions.
We have characterized the aim23∆ D273-10B strain using several functional tests assessing mitochondrial functionality during cell growth on liquid fermentable media supplemented with galactose (YPGal). Galactose, like glucose, is metabolized by fermentation, bypassing the need for mitochondrial respiration; but unlike glucose, it does not suppress mitochondrial function 27 , therefore allowing detection of defects in mitochondrial functionality. Analysis of the pair of congenic wild type and aim23∆ strains with phase contrast and fluorescent microcopy using DNA staining with DAPI (4′ , 6-diamidino-2-phenylindole) 28 revealed the presence of mtDNA in both strains ( Fig. 2A). This assay cannot, however, distinguish between normal mtDNA and that containing deletions. Such deletions are indicative of underlying defects in mitochondrial function causing mtDNA instability and lead to the formation of small colony variants defective in mitochondrial respiration, so-called 'petites' 29,30 . We assessed the petite frequency in the aim23∆ strain using a standard colony count assay 30 (Fig. 2B). In good agreement with the results of Hess and colleagues 21 , petite incidence in the aim23∆ strain is increased approximately three times in comparison to the congenic wild type strain. The effect is present when cells are grown using either glucose or galactose as a carbon source; however, growth on galactose leads to a further increase in the proportion of 'petites' , indicating unsuppressed defective mitochondrial activity as an underlying cause. To assess mitochondrial functionality directly, we followed oxygen consumption using a Clark-type oxygen electrode 31 . Consumption was measured in yeast cultures pre-grown in liquid culture with a non-fermentable carbon source, glycerol, for 6 and 24 hours (Fig. 2C). These time points correspond to the lag phase and exponential growth of the aim23∆ strain, and to exponential phase and consequent cessation of growth of the wild type, respectively (Fig. 1B). In wild type cells the oxygen consumption dropped significantly after 24 hours of growth, coinciding with the end of rapid exponential growth at that point. In the aim23∆ strain, O 2 consumption increased from near-absent after 6 hours of growth on glycerol to levels close to wild type after 24 hours, again in good agreement with the growth measurements.
Taken together, our results demonstrate at least partial mitochondrial functionality in the aim23∆ strain. Since mitochondrial protein synthesis is indispensible for mitochondrial functionality, the ability of the aim23∆ strain to grow on glycerol and to respire suggests that, surprisingly, yeast mitochondrial translation can operate even in the absence of mIF3/Aim23.

Effects of AIM23 disruption on S. cerevisiae mitochondrial translation and mitochondrial mRNA levels.
To investigate the effects of mIF3/Aim23 on mitochondrial translation, we followed incorporation of 35 S-methionine in the presence of 0.2 mg/ml of the antibiotic cycloheximide that specifically inhibits cytoplasmic translation 32 (Fig. 3A). S. cerevisiae mtDNA encodes only eight protein genes 4 , seven of which encode components of mitochondrial OXPHOS complexes, with the remaining gene encoding a ribosomal protein, Var1 4 . Therefore, we can follow synthesis of all the eight individual polypeptides by resolving them on an SDS PAGE gel.
The overall efficiency of mitochondrial protein synthesis in the aim23∆ strain is similar to that of wild type (Fig. 3A); a highly surprising observation given that the bacterial ortholog of mIF3/Aim23, IF3, is crucial for two steps of the ribosomal cycle: initiation and ribosome recycling. This result is in stark contrast to the near-complete inhibition of translation observed upon thermal inactivation of a temperature-sensitive version of S. cerevisiae mitochondrial mtRRF -a specialized factor mediating ribosomal recycling in bacteria 33 and mitochondria 34 . The relative abundance of the mitochondrially synthesized proteins is, however, altered in the aim23∆ strain, affecting both the kinetics of protein production (Fig. 3A, Supplementary Figure 1) and the protein levels after a 5 minute-long (Fig. 3B) and 20 minute-long (Supplementary Figure 2) pulse labeling. To quantify the effects of knocking out AIM23 we have normalized the relative expression levels, i.e. presented individual protein expression as a fraction of total expression. The kinetics of 35 S-methionine incorporation has a pronounced biphasic nature (Fig. 3A, Supplementary Figure 1): during the initial linear phase up to around the 5 minute time point 35 S-methionine incorporation into newly synthesized cytochrome c oxidase subunits Cox1, Cox2 is reduced about two times in the aim23∆ strain, while synthesis of 35 S-methionine-labelled ATP synthase subunit Atp9 is promoted. In later time points (10, 15 and 20 minutes), labeling deviates from linearity and is saturating for some protein species (Supplementary Figure 1). As a result, in addition to the abovementioned effects on Cox1, Cox2 and Atp9, 35 S-methionine labeling of nascent Atp6 and Atp8 is increased in the aim23∆ strain relative to the wild type, while labeling levels of Cox3 is decreased (compare  The observed imbalance in protein production can potentially be brought about by changes in transcription, translation, or stability of mRNA or protein. Several quality control systems in mitochondria recognize and degrade unfolded individual proteins or properly assembled proteins complexes 35 , convoluting the effects on synthesis and stability. We assessed protein stability by means of a chase experiment: after the initial labeling with 35 S-methionine for 15 minutes, synthesis of radiolabeled proteins was stopped by the addition of an excess of 'cold' methionine and the levels of labeled proteins were followed over a 2 hour-long time course (Fig. 3C, see also Supplementary Figure  3 for two additional biological replicates). With the exception of Atp8 in the wild type strain, all other mitochondrial proteins are stable over the 120 minute-long time course, while 35 S-methionine incorporation experiments show clear effects already after 5 minutes of labeling. This suggests that alterations in protein stability cannot be the cause for the unbalanced protein synthesis in the aim23∆ strain.
To assess the steady state levels of mitochondrially synthesized proteins, which are defined by both the rate of the protein's synthesis and degradation, we performed Western blot detection of Cox1, Cox2, Cox3, Atp6, Atp9 and Cob along with mitochondrial outer membrane protein Porin 1 as a control (Fig. 3D, Supplementary Figure 4). The results are in agreement with the 35 S-methionine incorporation data for cytochrome c oxidase subunits. Cox1 decrease being considerably more dramatic as judged by the Western blot. Lower Cox3 levels as detected by Western blot are in better agreement with the 20-minute 35 S-methionine incorporation data ( Supplementary Figures 1 and  2) than with 5 minute labeling time point (Fig. 3B). In the case of Atp6 and Atp9, the expression levels detected by Western blot are similar in wild type and the aim23∆ strains. A possible reason for the discrepancy between this result and the 35 S-methionine labeling data could be the atypical biphasic 35 S-methionine incorporation kinetics into newly synthesized Atp6 (Supplementary Figure 1). At up to 5 minutes the rate of 35 S-Atp6 synthesis is similar in wild type and aim23∆ strains, and the differences observed in the later time points could reflect the differential depletion of cellular factors necessary for Atp6 synthesis in the two backgrounds rather then a true difference in protein production in the context of a live cell.
Since the observed imbalance in mitochondrial protein production could, in principle, be caused by altered mRNA levels rather than by a direct effect on translation, we compared the mRNA levels in the wild type and aim23∆ strains using Northern blot hybridization (Fig. 4). With the exception of a lower level of the bicistronic mRNA encoding Atp6 and Atp8, we detect no significant differences in the mRNA levels between the wild type and aim23∆ strain. The decreased mRNA Atp6/Atp8 level is surprising, given that the expression of both proteins is increased in the aim23∆ strain (Fig. 3). Aim23 associates with the small subunit of the mitochondrial ribosome as demonstrated by Western blot analysis of lysed isolated S. cerevisiae mitochondria fractionated in a sucrose gradient under buffer conditions that induce separation of mitochondrial ribosomal subunits (Fig. 5). This further supports the direct involvement of Aim23 in mitochondrial protein synthesis.

Discussion
Mitochondrial translation is of significant importance from a human health perspective; defects in mitochondrial translation are associated with a number of human genetic diseases 36,37 . It is also a potential drug target; In aim23∆ cells, O 2 consumption increased from 9.7 nmol/ml/min per OD 600 after 6 hours of growth on glycerol to 78.5 nmol/ml/min per OD 600 after 24 hours. the disabling of mitochondrial machinery with antibiotics may be a promising therapeutic tool in anti-cancer efforts 38 , but can also be a detrimental and undesirable side effect of antibacterial treatments 39 . S. cerevisiae is a valuable model organism for studying mitochondrial disease mechanisms and discovering therapies due to its amenability to mitochondrial and nuclear genome manipulation, as well as the ability to survive in the absence of functional mitochondrial oxidative phosphorylation (see recent review by Lasserre and colleagues 40 ). Therefore, understanding the molecular mechanisms of mitochondrial protein synthesis and its regulation in this organism is of great importance for both basic and applied research.
Maintenance of the appropriate relative levels of mitochondrial expression is crucial for the assembly of functional OXPHOS complexes, and numerous auto-regulatory mechanisms are in place to ensure this balance in S. cerevisiae (reviewed by Fontanesi 41 ). Assembly of the genome-encoded F 1 subunit of F 1 F 0 ATP synthase is the key regulator of the expression of Atp6 and Atp8 subunits 42 . Similarly, the expression of cytochrome b, Cob, a core component of complex III, is controlled by an analogous autoregulatory feedback mechanism 43,44 . Loss of mIF3/ AIM23 does not, surprisingly, result in a general defect of translation; instead, it perturbs the stoichiometry of mitochondrially synthesized proteins (Fig. 3, Supplementary Figures 1 and 2). The mechanism behind the misbalance is unclear. It is principally possible that mIF3 acts as an mRNA-specific translational factor; however this is unlikely given that the core functions of the bacterial general translational factor IF3, i.e. subunit anti-association and stimulation of the initiator fMet-tRNA i fMet binding to programmed ribosome, are preserved in human mIF3 19 . Therefore, we hypothesize that the misbalance is a consequence of the differential efficiency of ribosomal recruitment to different mRNAs in the absence of mIF3, resulting in a new 'pecking order' among the individual mRNAs competing for ribosomes.   57 . Aconitase (Aco1) was used as soluble protein control and it stays in the top of the gradient. Separated ribosomal subunits migrate into the gradient and are detected using antibodies against small subunit protein Mrps5 and large subunit protein Mrpl36. T is a loading control corresponding to 10% of the starting material applied on the gradient, Mw is molecular weight in kDa.
Scientific RepoRts | 6:18749 | DOI: 10.1038/srep18749 The modest requirement for mIF3 in mitochondrial protein synthesis in S. cerevisiae is somewhat surprising given that IF3 is absolutely essential in E. coli 25 . At the same time, a disruption of the mti3 gene encoding mtIF3 in S. pombe does not even lead to a severe phenotype 45 . S. pombe is a petite-negative yeast, i.e. it can not survive without functional mitochondria 46 , which, in turn absolutely requires functional mitochondrial translation 47 . Therefore, one can conclude that the absence of mIF3 does not abrogate mitochondrial translation in S. pombe either. Moreover, mIF3 is seemingly naturally missing in a handful of organisms, including yeast Yarrowia lipolytica 11 . All these lines of evidence suggest that mIF3 is, indeed, dispensable for mitochondrial translation, but the importance of the functionality of the protein varies from organism to organism.
The effects of mIF3 loss on the functionality of mammalian -and specifically human -mitochondria have not been studied. Mutations destabilizing human mIF3 mRNA are associated with Parkinson's disease [48][49][50] . The connection between Parkinson's and mtDNA instability is well established (for a review see 51 ), which fits well with the elevated petite frequency in the aim23∆ S. cerevisiae strain, an indirect readout of mtDNA stability. Follow-up experiments in a mammalian system are necessary to directly address the role of mIF3 in humans.
Analysis of yeast growth rates in liquid cultures. Single colonies of D273-10B (MATα mal) and D273-10B aim23∆ (MATα mal AIM23::kanMX4) yeast strains were transferred from a YPD (2% bacto-peptone, 1% yeast extract, 2% glucose, 1.5% agar) agar plate to 5 mL of liquid medium of the same composition (minus agar) and were grown at 30 °C for 15-18 hours, reaching final OD 600 of 4.0-5.0. Cells were gently pelleted, washed with water and inoculated into liquid YPGly media at OD 600 of 0.1. Growth rates were monitored at 30 °C every 18 min for 4.5 days in a TECAN microplate reader equipped with a temperature control unit.

Monitoring of mitochondrial translation in vivo.
Pulse labelling of mitochondrial proteins with 35 S-methionine was carried out in whole cells in the presence of 0.2 mg/ml cycloheximide according to Gouget and colleagues 32 . Equal amounts of total cell proteins were separated on a 17.5% PAGE gel, subjected to autoradiographic analysis and quantification using ImageJ 52 . A 20 minute-long incubation as the end-point in the 35

S-methionine incorporation experiments was chosen as per Herrmann and colleagues 53 .
Western blot analysis of isolated mitochondria. Equal amounts of mitochondrial proteins were loaded on SDS-PAGE, transferred to nitrocellulose membrane and immunodecorated with corresponding antisera (Cox1 1:300, Cox3 1:300, Tom70 1:3000, all gifted by Roland Lill; Cox2 1:3000, ABCAM; Atp6 1:10000, Atp9 1:5000, gifted by Marie-France Giraud; Porin 1 1:3000, ABCAM) and secondary anti-rabbit antibodies (ABCAM). Atp9 forms ring-like oligomers 54 that are stable under standard conditions of preparation of the protein extracts for Western-blotting analysis. Therefore in order to promote the formation of Atp9 monomers total protein extracts were precipitated with 10% TCA, and then the pellets were resuspended in 10 mM Tris-HCl pH 7.0, 5% SDS (Marie-France Giraud, personal communication). Porin 1 was used as a control for equal loading.
RNA detection by Northern blot hybridization. Mitochondria were isolated from 1 l of yeast cultures after cultivation in liquid YPGly media for 16 hours (starting OD 600 of 0.5, final OD 600 of 3 for wild type and 1.5 for aim23∆). Total RNA was extracted from mitochondria with TRIzol reagent (Life Technologies) according to the standard procedure. About 5 μ g of RNA was separated on the 1% denaturating MOPS-formaldehyde agarose gel 55 and transferred to the Hybond-N+ (GE Healthcare) membrane according to the manufacturer's instructions. Membranes were hybridized with the oligonucleotide probes according to Mager-Heckel and colleagues 56 . Visualization was performed using Storm Imager 685 (GE Healthcare), and mRNA levels were normalized to 21S rRNA. Oligonucleotides used for Northern blotting, 5′ -3′ : VAR1 GACCAATCCGGTGAACAACCGGATTGGC, COX1 GCACCCATTGATAATACATAGTGAAAATGTCCCACCACGTAG, COX2 AACTCAGAACATGCTCCATAGAAGACAC, COX3 TACCAGCATAGAATACTGAACCATAAACAC, COB AGTATTACCTCTTACTACACTTCTATCAGTA, ATP6/8GAATCATTAATAAGAAACCATATGTTAATTGATTCATAAAATAAAATGGAACTAATTGTGGC, ATP9 TACTAGGTCTTTAATTGATGGGTTTCTT, 21S rRNA CTATATTACCCTGTTATCCCTAGCGTAACT.
Fluorescence microscopy of live yeast cells. Nuclear and mitochondrial DNA staining with DAPI (4′,6-diamidino-2-phenylindole) was performed according to the protocol of Amberg and colleagues 28 .
Analysis of "petite" frequency. Yeast cultures were grown until OD 600 of 3-4 in YPD liquid medium, diluted with fresh medium and grown to OD 600 of 0.8-1, washed with water and plated onto solid YPGly media supplemented with 0.1% glucose. The number of small colonies ("petites") was scored after 5 days of growth at 30 °C and expressed as a percentage of the total colony count.