Mitochondrial protein synthesis quality control

Abstract

Human mitochondrial DNA is one of the most simplified cellular genomes and facilitates compartmentalized gene expression. Within the organelle, there is no physical barrier to separate transcription and translation, nor is there evidence that quality control surveillance pathways are active to prevent translation on faulty mRNA transcripts. Mitochondrial ribosomes synthesize 13 hydrophobic proteins that require co-translational insertion into the inner membrane of the organelle. To maintain the integrity of the inner membrane, which is essential for organelle function, requires responsive quality control mechanisms to recognize aberrations in protein synthesis. In this review, we explore how defects in mitochondrial protein synthesis can arise due to the culmination of inherent mistakes that occur throughout the steps of gene expression. In turn, we examine the stepwise series of quality control processes that are needed to eliminate any mistakes that would perturb organelle homeostasis. We aim to provide an integrated view on the quality control mechanisms of mitochondrial protein synthesis and to identify promising avenues for future research.

mitochondria, protein synthesis, RNA processing, MTRFR, OPA1, AFG3L2, OMA1, OXA1L, post-transcriptional, proteostasis, ribosomes, membrane morphology, ribosome quality control, fusion open reading frames, non-stop mRNA, cell stress, co-translational quality control

In humans, mitochondria are defined by the presence of a small multi-copy maternally inherited genome surrounded by two dynamic membranes. Faithful expression of mitochondrial DNA (mtDNA) is required for the organelle to function as a metabolic and signalling hub of the cell. Within the organelle, there is no physical barrier to separate transcription and translation, nor is there evidence that quality control surveillance pathways are active to prevent translation of faulty mRNA transcripts. Since all 13 proteins encoded in mtDNA are hydrophobic and require co-translational insertion into the inner membrane of the organelle (Fig. 1), aberrations in protein synthesis must be recognized and resolved to prevent impingements to the integrity of the inner membrane. Disruptions to the inner membrane can uncouple organelle function as cellular hub of metabolism and in the process impair cell fitness. Thus, quality control mechanisms of mitochondrial protein synthesis are paramount to maintain inner membrane integrity.

Figure 1

Overview of mitochondrial gene expression. (A) A map of human mitochondrial DNA. The two circles represent the heavy (H-strand) and light (L-strand) of the genome with corresponding genetic content colour coded for the oxidative phosphorylation complex, rRNA and tRNA. Mitochondrial tRNA genes are indicated by a single letter abbreviation. Arrows indicate the positions of the major promoter. HSP, H-strand promoter. LSP, L-strand promoter. (B) A table indicating the 3′ non-coding end of the coding sequence (CDS) for the indicated human mitochondrial mRNAs. Those transcripts that do not encode a stop codon are highlighted with a box. (C) A schematic of mitochondrial gene expression from transcription to translation and insertion of nascent chains into the inner membrane. Mitochondrial RNA processing and post-transcriptional events occur in RNA granules, which is shaded. No physical barrier prevents translation initiation on aberrant mRNAs. Image created with BioRender.com.

Here, we focus on the molecular processes by which aberrations in mitochondrial protein synthesis can arise and the mechanisms to resolve these errors. Other reviews in this series cover mtDNA replication, transcription, repair, post-transcriptional processing, mitochondrial ribosomes and translation. Protein synthesis defects can arise naturally due to the culmination of inherent mistakes of molecular events in gene expression that are amplified by orders of magnitude from the genome through transcription and to translation on ribosomes. In turn, pathogenic variants and small molecules that inhibit any steps in mitochondrial gene expression can exacerbate these errors, manifesting into human pathologies that differ in severity, age of onset, and tissue-specificity [1]. In this review, we will omit discussion on pathogenic mtDNA variants, such as deletions, rRNA and tRNA mutations, as well as tRNA charging by mitochondrial aminoacyl tRNA synthetases as these topics have been covered previously in considerable detail [2–4]. Instead, we seek to understand the process of mitochondrial protein synthesis quality control by tracing the origin of errors that can occur naturally in a wild type setting and get amplified with the vectorial path from the genome to the nascent chain. This process starts with synthesis of the mRNA templates, folding of the emerging nascent chain and the co-translational insertion into the inner membrane, followed by the stepwise series of molecular events that are needed to eliminate any aberrations that would perturb organelle homeostasis. We will also explore how mitochondrial ribosomes can act as a hub for coordinating proteostasis within the organelle matrix. Our goal is to provide a unifying and integrated view on the quality control of mitochondrial protein synthesis, the connection to human diseases when relevant and to identify promising avenues for future research.

Protein synthesis mistakes arising from transcription errors

Across metazoans, mtDNA is transcribed by a single subunit RNA polymerase from promoters in the non-coding region of the genome [5]. This transcription generates long polycistronic transcripts containing mRNAs, rRNAs and tRNAs. A stepwise series of post-transcriptional events (e.g. RNA processing and modifications) are needed to mature each molecular species in order to facilitate mitochondrial protein synthesis (Fig. 1). There is no mechanism to induce expression of specific transcripts of individual oxidative phosphorylation subunits or tRNAs and rRNAs. Generally, impairments to transcription initiation and elongation reduce the abundance of mRNAs, tRNAs and rRNA and will lead to an overall decrease in the synthesis of individual subunits of the oxidative phosphorylation complexes and their steady-state abundance in the membrane [6, 7]. However, another class of errors that occur during transcription as well in the post-transcriptional maturation steps can alter the reading frame of the genetic template and induce aberrations in protein synthesis on mitochondrial ribosomes, in particular misfolding of the nascent chain. Here, we explore the various means in which the fidelity of mitochondrial mRNAs can compromise protein synthesis and, in the process, trigger co-translational quality control mechanisms.

The first level of gene expression mistakes arises with the fidelity of transcription itself which has an intrinsically higher error rate compared to DNA replication by orders of magnitude. Until recently, only data from in vitro biochemical assays existed for the human mitochondrial RNA polymerase (POLRMT) (2 × 10⁻⁵) [8]. Whether this reflected the in vivo reality in the complex milieu of the mitochondrial matrix was an open question because of inherent methodological limitations in accurately sequencing RNA. Recently, these technical limitations have been overcome with a robust library construction and deep sequencing approach that revealed an error rate of 8 × 10⁻⁶/bp in cultured human fibroblasts [9]. Surprisingly, this error rate in mitochondrial transcription is higher than that for the nuclear RNA polymerases RNAPI and RNAPII, responsible for the synthesis of nuclear-encoded rRNAs and mRNAs, respectively [9]. These mutations can introduce non-synonymous alterations in the genetic code and to the resulting nascent polypeptide chain during synthesis with a propensity to induce protein misfolding and instability. Thus co-translational degradation by proteases is required.

Another variable that affects transcription fidelity is processivity. TEFM is a mitochondrial transcription elongation factor that interacts with POLRMT to ensure the long polycistronic transcripts are generated [10–12]. Whereas pathogenic variants in TEFM impair the ability to synthesize RNA molecules distal from the promoters [7], the processivity facilitated by TEFM can also introduce mutations during transcription [8]. While the conventional thinking is that in a healthy wild type setting the full polycistronic transcripts are generated regularly from the light and heavy strand promoters of the genome, this view does not consider low-level processivity mistakes whereby the POLRMT-TEFM complex disassociates from the genome template and generates a partial transcript. Evidence points to the synthesis of incomplete mRNA transcripts that are subsequently polyadenylated [13, 14]. Importantly, there appears to be no surveillance system to prevent translation initiation on mitochondrial mRNAs with aberrant 3′ ends [14, 15]. Thus, translation of these partial polyadenylated mRNAs would generate fusion open reading frames (ORFs) whose nascent chains would misfold during synthesis or in the inner membrane, impinging on the proteostasis of the organelle. The overall in vivo processivity of the POLRMT-TEFM complex across mtDNA in different cell types, in response to metabolic shifts and with ageing are open questions. Methodologies for deep sequencing nascent RNA synthesis [16] now provide tools to address these questions and will allow us to determine how frequently truncated aberrant mitochondrial mRNAs are generated during transcription that would impair protein synthesis.

Protein synthesis mistakes arising from post-transcriptional errors

Processing of the mitochondrial RNA polycistronic transcripts has a direct effect on the integrity of the nascent mRNA. Six of the protein coding genes in mtDNA do not encode a stop codon (Fig. 1B) and are directly flanked at the 3′ end by a tRNA sequence (Fig. 1A). Mitochondrial tRNAs act as beacons for the nuclease activity of the RNase P complex and ELAC2 at the 5′ and 3′ of tRNAs, respectively [17, 18]. Elegant research has reconstituted the processing events mediated by the RNase P complex composed of TRMT10C, HSD17B10, and PRORP on single tRNA substrates [17, 19, 20]. Recent experimental evidence, however, suggests that low level errors in the 5’ tRNA processing arise that would in turn affect the 3′ end of six mitochondrial mRNAs [14, 15]. Deep sequencing of translating mitochondrial ribosomes revealed that these transcripts consist of the mRNA with a fragment of the 3′ flanking tRNA sequence in frame (Fig. 1C). Importantly, these were not processing intermediates of the polycistronic transcript. None of these six mRNAs encode a stop codon, so translation of these mRNA-tRNA transcripts would generate fusion ORFs whose nascent chains will misfold during synthesis or membrane insertion. So far, the established model for 5′ processing only accounts for a single tRNA molecule. Within the mitochondrial genome there are tRNAs encoded in tandem (e.g. 2, 3, and 4) (Fig. 1A). The stepwise RNA processing events and methylation of mitochondrial tRNAs encoded in tandem remains unknown. Although mitochondrial tRNAs encoded on the light or heavy strand of the genome are transcribed in cis, the steady-state abundance of individual tRNA molecules varies and differs across cell types [21]. Understanding the temporal and spatial regulation of tRNA processing, particularly at the 5′ end, is critical to develop a deeper view on how often aberrant mitochondrial mRNAs are generated across cell types and with age.

According to the tRNA punctuation model proposed over 40 years ago [22], tRNA processing followed by polyadenylation would generate mRNAs with a stop codon for the six genes where it is not encoded (Fig. 1B and C). Recent deep sequencing data of mitochondrial mRNAs from translating ribosomes in cultured fibroblasts identified pools of mRNAs missing polyadenlyation [14]. This pattern has also been seen in mitochondrial total RNA isolated from human tissues such as the liver, heart and skeletal muscle [23], suggesting that the finding is not an aberration from culturing cells, but in fact reflects the biology of mitochondrial gene expression. Pathogenic variants in the RNA binding protein LRPPRC, also disrupt polyadenylation and mRNA stability in a transcript-specific manner [23–25]. It is important to emphasize the presence or absence of polyadenylation on its own does not alter the reading frame of the coding sequences (CDS) that do not encode a stop codon (Fig. 1). Translation of these mRNAs will generate a fully functional protein as long as termination occurs following decoding of the penultimate codon (see below).

Other types of regulatory post-transcriptional modifications of mitochondrial mRNAs include m1A, and pseudouridinylation. These can exert negative and positive effects on protein synthesis, respectively. The m1A has been detected on a number of mitochondrial mRNAs and the abundance appears to be developmentally regulated and differs across cell types [26, 27]. The m1A is situated on the Watson-Crick face of the mRNA transcript, which will disrupt base pairing and has been shown to repress translation by inducing ribosome stalling [26, 27]. Interestingly, the methyl transferase (TRMT10C) component of the RNase P complex can mediate the m1A addition on mitochondrial mRNAs [27]. Thus, this modification would then stall translation elongation in a site-specific manner during nascent chain synthesis. Prolonged ribosome stalling is a significant cell stress that would initiate a series of co-translational quality control mechanisms to terminate protein synthesis and recycle the ribosomes.

In contrast, the regulatory role of pseudouridine, which is widespread on mitochondrial mRNAs [28–30], is less clear. Loss of function studies with the enzymes that catalyse pseudouridylation have variable results. In some cases, loss of the pseudouridine on mitochondrial mRNA has no effect on the rate of translation elongation whilst, in other cases, there is a systemic reduction in protein synthesis and nascent chain instability [28, 29]. Thus, there is more complexity to the temporal and spatial regulation of mitochondrial RNA processing and post-transcriptional modifications that can exert direct effects on the fidelity of mitochondrial protein synthesis.

Mitochondrial ribosomes as a hub for protein quality control

Since there appears to be no surveillance mechanism to prevent translation of mRNAs with aberrant 3′ ends [14, 15], understanding how defects in translation elongation arise, are recognized, and resolved is critical for developing an integrated view of mitochondrial protein synthesis quality control. The mitochondrial ribosome provides an ideal hub to coordinate these events. First, we focus on how protein synthesis termination occurs on aberrant mRNAs.

Termination of protein synthesis is commonly understood as a process in which the decoding centre of the translating ribosome arrests on the stop codon of the mRNA because no tRNA recognizes the mitochondrial UAA or UAG (Fig. 2). The arrested ribosome is then recognized by a class I release factor that catalyses hydrolysis of the nascent chain from the tRNA in the ribosomal P-site [31]. Together, these steps facilitate release of the nascent chain, recycling of the ribosome into the small and large subunits, and dissociation of the mRNA.

Figure 2

Translation termination of mitochondrial protein synthesis for wild type and aberrant mRNAs. (A) A schematic of translation termination with an mRNA encoding a stop codon. The mitochondrial class I release factor MTRF1L structurally recognizes the stop codon in the mRNA [31]. (B) A model for translation termination on truncated nonstop mRNAs based upon a working model for the class I release factor MTRFR that is based upon that developed for the bacterial rescue factor ArfB. (C) A model for inducing translation termination when cellular stress induces mitochondrial ribosome stalling in the middle of the mRNA. Image created with BioRender.com.

In the absence of a stop codon, a translating mitochondrial ribosome will arrest at the last codon position, leaving the mRNA channel empty. We refer to these transcripts as non-stop mRNA. A working model is being established on how translation termination could arise on mitochondrial non-stop mRNAs. True to its origins as an alpha-proteobacteria, this mitochondrial mechanism appears to be similar to that of a single component bacterial pathway orchestrated by the release factor ArfB [32–35]. At the structural level, ArfB consists of a globular domain with a conserved GGQ catalytic motif and a C-terminal alpha helix [32, 35]. The release factor is not a structural component of the ribosome, rather it constantly probes the mRNA channel with its C-terminal alpha helix. When the translating ribosome has less than 9 nucleotides of the mRNA from the A-site in the channel, ArfB can insert its C-terminus into the mRNA channel [33–35] thereby inducing a conformational change that repositions the conserved GGQ motif to the peptidyl transferase centre (PTC) to catalyse hydrolysis of the nascent chain from the tRNA. Considering that mature bacterial mRNAs used as templates in translation lack polyadenylation, ArfB provides an effective backup mechanism to terminate protein synthesis on transcripts that leave the mRNA channel empty.

In vertebrates there are two mitochondrial homologues of ArfB: MTRFR (formerly known as C12orf65) and mL62 (formerly known as ICT1) [36, 37]. Structurally, both proteins are similar to ArfB [14, 31, 38]. However, mL62 has been incorporated as a core structural subunit of the large ribosome and is located too distal from the peptidyl transferase centre (PTC) to be functional in nascent chain release [37, 39, 40]. In contrast, MTRFR is not a structural subunit of the ribosome and its catalytic GGQ motif is required for mitochondrial protein synthesis [14, 41]. There is a genotype-phenotype correlation for MTRFR between the length of the C-terminal alpha helix and the effect on mitochondrial protein synthesis [14]. This would be consistent with the ArfB model, whereby the C-terminal alpha helix is required for insertion into the mRNA channel. Although some types of non-stop mRNAs accumulate with MTRFR deficiency [14], which mRNA specifically require MTRFR for termination still needs to be determined. Recent ribosome profiling data from across a diversity of human cell types indicates that the rates of mitochondrial protein synthesis for subunits of the oxidative phosphorylation complexes differ considerably [42]. Identifying which types of mRNAs most require MTRFR function will, therefore, require considerable effort. It is also worth speculating whether mL62 can be upregulated in a cell-specific context, but not assemble into the large subunit to function in an analogous mechanism for the termination of mitochondrial protein synthesis. Elegant in vitro research suggests mL62 can function in such a manner [31], thus finding the biological context for this mechanism in vivo will be important to discover. Collectively, MTRFR provides an effective ribosome quality control step to catalyse termination of protein synthesis on non-stop mRNAs and in the process prevent the accumulation stalled ribosome complexes.

How does protein synthesis terminate if a mitochondrial ribosome stalls in the middle of the mRNA during translation elongation? This stalling could arise because of the mRNA template itself, nascent chain misfolding, and/or other co-translational events. Under the working model for MTRFR, a truncated transcript would need to be generated for access to the mRNA channel (Fig. 2B and C). Thus, a nuclease would first need to be recruited to the mitochondrial ribosome to cleave the mRNA. Whether such a mechanism is operational within mitochondria as part of a ribosome quality control step still needs to be discovered.

Co-translational degradation of mitochondrial nascent chains to protect against membrane stress

Degradation of misfolded nascent chains synthesized by mitochondrial ribosomes is mediated by a membrane bound protease complex of bacterial origin. AFG3L2 subunits form a AAA (ATPases Associated with diverse cellular Activities) protease complex as a homo-hexamer or as a hetero-hexamer with paraplegin (SPG7) subunits [43, 44]. The complex is anchored by the prohibitin membrane scaffold [45] and uses the energy of ATP hydrolysis to unfold substrates, threading them into the internal proteolytic cavity for hydrolysis into small peptide fragments [44, 46]. Remarkably, the structure of this proteolytic mechanism has been conserved from the bacterial counterpart FtsH [44, 47]. Over the last 20 years, evidence has been accumulating on the intersection between AFG3L2/paraplegin function and oxidative phosphorylation function, mitochondrial protein synthesis and organelle membrane dynamics [48–59]. These cellular phenotypes are core features associated with pathogenic variants in AFG3L2 and SPG7, and are key to the molecular pathogenesis. The connection between these disparate molecular observations has only recently been identified [41, 60–62]. Impaired quality control of mitochondrially synthesized nascent chains triggers a linear series of molecular events that progressively remodels the organelle membranes, followed by activation of an RNA and ribosome decay pathway that ultimately impairs the ability to synthesize subunits of the oxidative phosphorylation complexes (Fig. 3).

Figure 3

Membrane-associated quality control of mitochondrial protein synthesis and ensuing stress responses. (A) Wild type setting, OXA1L interacts co-translationally with mitochondrial ribosomes to mediate nascent chain insertion into the inner membrane. Nascent chains that misfold are rapidly degraded by the AFG3L2 protease complex. (B) Defects in nascent chain insertion via OXA1L leads to protein misfolding followed by rapid degradation by the AFG3L2 protease complex. (C) Short-term AFG3L2 dysfunction results in the accumulation of nascent chains, which triggers a proteotoxic stress response on the inner membrane. This activates the OMA1 protease, which cleaves the membrane anchored OPA1 to remodel first the inner membrane followed by the outer membrane morphology. (D) Progressive AFG3L2 dysfunction maintains the membrane stress, which in turn triggers a mitochondrial ribosome and mRNA decay pathway. This negative feedback response leads to the overall reduction of oxidative phosphorylation complexes with time. Image created with BioRender.com.

To illustrate the mechanism, it is easiest to follow the vectorial path of nascent chains exiting the mitochondrial ribosome into the lipid bilayer of the inner membrane (Fig. 3). Mammalian mitochondrial ribosomes have an exit tunnel that is narrower in diameter compared to the bacterial version, preventing the premature folding of nascent chains [63]. Thus, a newly synthesized polypeptide chain only begins to fold as it emerges from the ribosome. The mitochondrial insertase OXA1L interacts with the translating ribosome to mediate co-translational protein insertion into the inner membrane [63]. Mitochondrial protein synthesis can still occur in the absence of OXA1L, however, selected nascent chains such as MT-ATP6, MT-ATP8, and MT-CO2 are the first to be rapidly degraded by the AFG3L2 protease complex [62]. Moreover, this protease complex has also been shown to degrade nascent truncated MT-CO1 due to nonsense mutations [64] and MT-ND1 when there are early defects in Complex I assembly [65]. Together this points to the role of the AFG3L2 protease complex in safeguarding the proteostasis of the inner membrane to prevent aberrant nascent chains from over accumulating.

Mitochondrial inner membrane remodelling is a well-coordinated inducible response mediated by the OMA1 stress activated protease, which cleaves the dynamin related GTPase OPA1 [53, 66–68]. This proteolytic processing releases OPA1 from its inner membrane anchor into a soluble pool (Fig. 3), leading to cristae remodelling and fragmentation of the overall morphology of the organelle. Most of the mechanistic insights for this process were obtained using severely toxic compounds to induce the response. Linking AFG3L2 dysfunction to this pathway [53] provided the first clue to the endogenous activator of this mitochondrial stress response. Using specific antibiotic inhibitors of mitochondrial protein synthesis combined with genetic approaches, revealed a molecular cause and effect whereby the OMA1 activation and OPA1 processing was due specifically to the accumulation of misfolded mitochondrial nascent chains [41, 61, 69–71]. Inhibiting mitochondrial protein synthesis with chloramphenicol, a translation elongation inhibitor [72], completely blocked the stress induced OPA1 processing with AFG3L2 dysfunction. Furthermore, the progressive loss of mitochondrial ribosomes that has been observed with AFG3L2 dysfunction [52, 54, 56] could be totally blocked when mitochondrial protein synthesis was inhibited with chloramphenicol [61]. Lastly, defects in the co-translational quality control of a mtDNA encoded fusion ORF could also induce the same stress response pathway to remodel the membrane and progressively degrade mitochondrial ribosomes and mRNA [61, 62]. Together, the data point to the toxic effect that mitochondrial nascent chains can exert on the organelle form and function when their quality control is disrupted.

Future perspectives

Our understanding of the mechanisms regulating the quality control of mitochondrial protein synthesis has improved dramatically over the last decade due to technological advances in functional genomics and cryo-EM. We now have tools to investigate homeostatic processes quantitatively within the mitochondrial compartment to understand the underlying fidelity of gene expression and the associated quality control processes needed to maintain the form and function of the organelle.

Many fundamental questions, however, remain. Are there polysomes for mitochondrial protein synthesis? Their presence has been speculated over the years, but direct evidence is lacking. In the cytosol, nascent chain proteotoxicity can induce ribosome collision and activation of the integrated stress response in the matter of minutes [73], a process linked to RNA quality control [74]. The cytosolic ribosome quality control pathways to resolve errors differ fundamentally from the emerging MTRFR model. Thus, it is important to determine how much functional redundancy exists for mitochondrial ribosome quality control. Also, how does nascent chain misfolding on the membrane activate OMA1 and how are misfolded nascent chains delivered to the AFG3L2 protease complex? And what mechanisms regulate the ribosome and RNA decay within the organelle? Understanding these processes will be key to reveal how pathogenic variants that impair mitochondrial protein synthesis can exert such enormous clinical heterogeneity [1] and help us as a field develop an integrated understanding of mitochondrial gene expression with organelle form and function.

Acknowledgements

We thank Sarah O’Keefe, Kah Ying Ng, and Güleycan Lutfullahoglu Bal for discussion.

Funding

Research in BJB lab is supported by funding from the Sigrid Juselius Foundation Senior Investigator Award, Research Council of Finland, donations from the Hereditary Neuropathy Foundation and Lindsey Flynt, National Ataxia Foundation, and the Magnus Ehrnrooth Foundation.

Conflict of interest statement: The authors declare no conflict of interest.