The role of mitochondrial mRNA translation in cellular communication Open Access

Over the past two decades, our view of mitochondria has changed profoundly. It is now well established that these organelles, which house the respiratory chain and other metabolic pathways, are not isolated or homogeneous entities. Rather, they are highly dynamic, heterogeneous, and capable of adapting their function to cellular needs and environmental changes. Mitochondria sense a variety of external inputs, integrate information through their network and produce output signals that regulate physiological processes. As key signalling hubs, mitochondria dictate cellular responses to nutrients, stress, infection and pathological conditions. They also influence cell fate by modulating the epigenome, cell proliferation, differentiation and survival (Picard and Shirihai, 2022; Suomalainen and Nunnari, 2024).

To meet the demands of cells and tissues, mitochondria reconfigure their network and metabolic output. Under conditions such as starvation or hypoxia, these adaptations occur rapidly, suggesting the involvement of post-transcriptional and post-translational mechanisms. In addition, mitochondria exposure to external inputs can vary depending on their subcellular location. For example, mitochondria at the leading edge of migrating cells or in synaptic terminals of neurons are challenged by nutrient fluctuations more rapidly than those in the perinuclear region. Finally, tightly regulated quality control mechanisms are employed to manage dysfunctional organelles. Changes in mitochondrial network dynamics, regulation of import mechanisms and activation of intra-organellar proteolysis are strategies to adapt mitochondrial function to changing conditions, which have been extensively reviewed elsewhere (Deshwal et al., 2020; Fenton et al., 2021; Song et al., 2021).

Mitochondria are composed of proteins encoded by both mitochondrial DNA (mtDNA) and nuclear DNA (nDNA). Whereas mtDNA encodes 13 proteins that form part of the electron transport chain, eukaryotic cells contain more than 1100 nuclear-encoded mitochondrial (NEM) proteins (Morgenstern et al., 2021; Rath et al., 2021). Although many of these proteins are considered housekeeping, the mitochondrial proteome is extremely heterogeneous in different tissues, explaining the tissue-specific morphology and metabolic functions of mitochondria (Baker et al., 2024). Furthermore, the mitochondrial proteome does not always reflect the corresponding transcriptome, especially in neurons (Buccitelli and Selbach, 2020; Jung et al., 2023). This discrepancy suggests that the regulation of which NEM proteins are synthesised, where they are produced and at what levels, plays a key role in adapting mitochondrial function to cellular needs, providing metabolic flexibility and enabling mitochondrial specialisation. The translation of NEM mRNAs serves as a crucial communication node, integrating incoming signals and output cellular responses (Fig. 1). In this Review, we first discuss how signalling pathways and metabolic states crosstalk with the translational machinery and specific RNA-binding proteins (RBPs) to regulate NEM protein synthesis. We then focus on the regulatory role of spatial localisation of the translation of NEM mRNAs. Finally, we explore how aberrant translation of NEM mRNAs can act as a signal to promote mitochondrial turnover and other stress responses.

Translation is an energy-demanding process, estimated to consume ∼30% of cellular resources in mammalian cells (Buttgereit and Brand, 1995). Consequently, it is tightly coupled to signalling pathways that sense the amount of nutrients available, such as mammalian target of rapamycin complex 1 (mTORC1) (Box 1). Through the 4E-BP axis, mTORC1 promotes the translation initiation of mRNAs encoding core constituents of the translation machinery, such as ribosomal proteins and translation factors, as well as several NEM mRNAs (Morita et al., 2013, 2017). mTORC1-dependent translational regulation increases the synthesis of mitochondrial ribosomal components, respiratory complexes, mitochondrial transcription factor A (TFAM) – which is essential for mitochondrial DNA transcription and replication – and the mitochondrial fission factor MTFP1 (Morita et al., 2013, 2017). Thus, in highly proliferating cells with abundant nutrients, mTORC1 promotes a feed-forward mechanism that maintains sufficient ATP levels for ongoing translation. In contrast to mRNAs for ribosomal subunits, NEM mRNAs lack a TOP motif (see Box 1), but are characterised by a short 5′ untranslated region (UTR) and a translation initiator of short 5′ UTR (TISU) element (Gandin et al., 2016; Sinvani et al., 2015). Remarkably, under conditions of nutrient scarcity, such as falling glucose levels, the translation of NEM mRNAs containing TISU elements can be preserved despite mTORC1 inhibition, as a mechanism to counteract stress. In this context, a cooperation between eIF1 and eIF4G1 is required to initiate cap-dependent translation (Sinvani et al., 2015), illustrating how different translation initiator factors participate in metabolic reprogramming of mitochondria.

Translation elongation – the process by which the nascent protein chain is extended by one amino acid during each elongation cycle – is also subject to metabolic regulation. One example is the activity of the eukaryotic translation initiation factor eIF5A. Although initially described as an initiation factor, eIF5A also promotes elongation of proteins with amino acid sequences prone to translation slowdown, such as polyproline stretches (Schuller et al., 2017; Faundes et al., 2021). Notably, several NEM proteins fall into this category, as demonstrated by studies in yeast, flies, senescent cells, macrophages and liver (Barba-Aliaga et al., 2024; Liang et al., 2021; Jiang et al., 2024; Puleston et al., 2019; Zhou et al., 2022). eIF5A activity is uniquely regulated by hypusination, a post-translational modification in which spermidine – a key metabolite in the polyamine synthesis pathway – is added to a lysine residue by two consecutive enzymatic reactions. Therefore, eIF5A activity is positively correlated with the rate of polyamine biosynthesis, which increases in cancer and decreases with ageing. For example, in yeast, the translation of a specific subunit of the inner mitochondrial translocase, TIM50, which contains a polyproline stretch, is regulated by eIF5A. In the absence of eIF5A, translation stalls at the polyproline stretch, hampering TIM50 synthesis and leading to mitochondria import defects. This ultimately triggers a stress response, resulting in global suppression of mitochondrial protein translation (Barba-Aliaga et al., 2024).

In conclusion, the synthesis of NEM proteins is under tight translational control closely linked to metabolism. Interactions between specific components of the translational machinery and individual sequences within NEM mRNAs or protein products are crucial to orchestrate mitochondrial function and tailor it to metabolic needs. It is likely that many more cell- and tissue-specific mechanisms will be uncovered in the future, particularly during stress responses when cap-dependent translation is inhibited, and alternative initiation complexes form.

The availability of an mRNA for translation is determined by the set of RBPs that bind it throughout its lifetime. This so-called RBP code impacts splicing, nuclear export, stability, transport and the final location of mRNAs, indirectly contributing to translation efficiency (Harvey et al., 2018). Furthermore, RBPs can either repress or activate translation by acting at various steps of the process. By responding to activating or inhibiting signalling pathways, RBPs shape the translational landscape of NEM mRNAs and rewire the organelle metabolic output during crucial metabolic shifts, such as abrupt changes in oxygen levels or limited glucose availability. Several RBPs have been described to bind and regulate NEM mRNAs at various stages of their lifecycle. Here, we focus on RBPs that have a role as broad regulators of NEM–mRNA translation in response to signalling cascades (Table 1).

The yeast Puf3 protein, a member of the Pumilio family, was the first RBP identified as specific for NEM mRNAs. Puf3 is a crucial regulator of mitochondrial biogenesis given that it regulates the fate of mRNAs encoding proteins involved in mitochondrial import, translation and respiratory complex assembly (Olivas and Parker, 2000; Garcia-Rodriguez et al., 2007; Saint-Georges et al., 2008; Lapointe et al., 2018). Remarkably, Puf3 promotes the decay of its bound mRNAs under fermentative conditions favouring glycolysis over oxidative phosphorylation (OXPHOS) (Olivas and Parker, 2000). However, during a metabolic shift to oxidative conditions, Puf3 mediates the binding of its target mRNAs to mitochondria and facilitates their localised translation on the organelle surface (Miller et al., 2014; Lee and Tu, 2015) (Table 1). This functional switch in Puf3 activity is driven by its phosphorylation via nutrient-responsive kinases following glucose depletion (Lee and Tu, 2015). Indeed, a Puf3 phosphomutant that cannot undergo phosphorylation, fails to associate with translating polysomes in response to glucose levels. Instead, it shows a dominant-negative phenotype, compromising yeast growth in glucose-depleted media. Furthermore, following glucose depletion, the phosphomutant localises in punctate foci, called PUF bodies, which might trap Puf3 and its target mRNAs, thus inhibiting its function (Lee and Tu, 2015).

The mammalian Pumilio homologues, PUM1 and PUM2, mainly act as translational repressors by binding the 3′ UTR of target mRNAs (Wickens et al., 2002). Among these targets, PUM2 inhibits the translation of the mRNA encoding the mitochondrial fission factor (MFF) (Table 1). This regulation is particularly relevant during ageing, as increased levels of PUM2 repress mitochondrial fission and mitophagy, the selective removal of mitochondria by autophagy (D'Amico et al., 2019). However, more studies are needed to understand whether other NEM mRNAs are regulated by PUM2, and whether this RBP always inhibits translation of its targets or can also be regulated through phosphorylation.

Another RBP that dynamically regulates a subset of NEM mRNAs is the mammalian clustered mitochondrial homolog (CLUH), and its orthologues CluA in Dictyostelium discoideum, Clu1 in yeast, Friendly mitochondria (FMT) in Arabidopsis thaliana and Clueless in Drosophila melanogaster. These proteins harbour a highly conserved Clu domain of yet unclear function, followed by tetratricopeptide repeat (TPR) motifs, and share a characteristic mitochondrial clustering phenotype upon knockout (Gao et al., 2014; Fields et al., 1998; Zhu et al., 1997; Cox and Spradling, 2009; El Zawily et al., 2014). Most importantly, they bind several NEM mRNAs and regulate their translation (Gao et al., 2014; Sen and Cox, 2016; Hemono et al., 2022b). In mammals, CLUH-target mRNAs encode proteins involved in several metabolic pathways, including the tricarboxylic acid (TCA) cycle, amino acid degradation, ketogenesis and OXPHOS (Gao et al., 2014). CLUH is pivotal during metabolic transitions to oxidative conditions. This is exemplified by the phenotype of constitutive Cluh-knockout mice, which die shortly after birth due to an inability to maintain glucose levels during the glycolysis-to-OXPHOS transition that occurs in several tissues (Schatton et al., 2017). CLUH also plays fundamental roles in instructing the liver to mount a ketogenic response to starvation (Schatton et al., 2017), to promote adipogenesis (Cho et al., 2019) and maintain peripheral axons (Zaninello et al., 2024) (Fig. 2A). In Drosophila, Clueless mutants display phenotypes such as uncoordinated movement, reduced lifespan and sterility. These mutants show mitochondrial impairments in fly muscle, along with decreased ATP production (Cox and Spradling, 2009).

The precise molecular function of the CLUH orthologues when bound to NEM mRNAs remains unclear. However, several findings suggest they have a crucial role in translational regulation. Besides binding NEM mRNAs, CLUH and its orthologues interact with ribosomal and translational components and are often found near mitochondrial polypeptides or overexpressed mitochondrial-targeting sequences (Sen and Cox, 2016; Ayabe et al., 2021; Zaninello et al., 2024; Hemono et al., 2022a,b; Antonicka et al., 2020). Moreover, the association of CLUH with polysomes has been shown in human, fly and yeast cells (Schatton et al., 2022; Sen and Cox, 2016; Miller-Fleming et al., 2024 preprint). In mammals, CLUH stabilises its target mRNAs (Schatton et al., 2017) (Table 1), although the increased decay of these mRNAs in the absence of CLUH might result from secondary effects of translational dysregulation. In fact, in CLUH-deficient axons, decreased levels of the Atp5a1 mRNA can be rescued by overexpression of ATP-binding cassette sub-family E member 1 (ABCE1), a factor involved in translation initiation and ribosomal quality control (see also last section of this review) (Zaninello et al., 2024).

The physiological conditions requiring CLUH and its orthologs in multicellular organisms are consistently characterised by metabolic rewiring. A feature of CLUH proteins is their ability to respond to changing nutrient conditions by relocating to granular structures (Table 1, Fig. 2B). For example, in Drosophila, Clu particles form in the ovaries upon insulin stimulation but are dissolved by stress (Sheard et al., 2020), whereas in yeast they are triggered by shifting the growth conditions from fermentation to respiration and by mitochondrial stress (Miller-Fleming et al., 2024 preprint). In primary hepatocytes, CLUH adopts a punctate cytosolic pattern, but upon starvation stress, it coalesces with target mRNAs in larger granules. These granules contain mTORC1 and the Ras-GTPase-activating protein (GAP)-binding protein 1 (G3BP1), a marker of stress granules. They restrict mTORC1 activation under limiting nutrient conditions. In fact, loss of CLUH leads to hyperactivation of mTORC1, and inhibition of mTORC1 using torin or rapamycin ameliorates the clustering phenotype observed in CLUH-deficient mouse liver and human cell cultures (Pla-Martin et al., 2020). The functional significance of these granules, what regulates their dynamic behaviour and whether they serve similar roles across different organisms remains unclear. Unlike stress granules, the CLUH/Clu1/Clueless granules are resistant to cycloheximide treatment and show conserved evidence of being translationally active (Pla-Martin et al., 2020; Miller-Fleming et al., 2024 preprint; Hwang et al., 2024 preprint). However, in hepatocytes, their translational activity depends on the target mRNA and the metabolic status (Pla-Martin et al., 2020) (Table 1). Unravelling the signalling cascades that regulate the dynamic subcellular localisation of CLUH and understanding the interplay between the translation of its targets and granule formation will be crucial. Given the evolutionary conservation, CLUH likely plays a fundamental role in the translation of a subgroup of NEM mRNAs.

Whereas mTORC1 signalling promotes anabolic pathways, the AMP-activated protein kinase (AMPK) acts as a key sensor of falling energy levels, becoming activated by increased AMP/ATP and ADP/ATP ratios. AMPK signalling positively regulates catabolic pathways and induces autophagy to restore ATP levels (Herzig and Shaw, 2018). The balance between mTORC1 and AMPK plays a role in controlling the relocation of RBPs and their bound NEM mRNAs. For example, in U2OS cells, the balance between mTORC1 and AMPK regulates Smaug1 (also known as SAMD4A), an RBP implicated in the translational regulation of at least two NEM mRNAs, SDHB and UQCRC1 (Fernández-Alvarez et al., 2022). Smaug1 forms membrane-less bodies that dissolve upon treatment with the complex I inhibitor rotenone and the mTORC1 inhibitor rapamycin. However, this dissolution is inhibited by compound C, an AMPK inhibitor. Dissolution of Smaug1 bodies elicits the release of mRNAs and their translational activation (Fernández-Alvarez et al., 2022) (Table 1). These findings suggest that nutrient-responsive signalling cascades might regulate the redistribution of RBPs and their bound NEM mRNAs in membrane-less organelles and could represent a common principle controlling translation in a temporal-spatial manner.

Recent findings have implicated AMPK in the translational regulation of the mRNA encoding the mitochondrial Phosphatase and tensin homolog (PTEN)-induced kinase 1 (PINK1), which is a crucial regulator of mitophagy (Narendra and Youle, 2024), and is involved in autosomal recessive familiar forms of Parkinson's disease (PD) (Valente et al., 2004). In neurons, Pink1 mRNA, together with other NEM mRNAs, is bound by synaptojanin-2a (SYNJ2a, an isoform of SYNJ2), an inositol 5-phosphatase containing an RNA recognition element. SYNJ2a is tethered to the mitochondrial surface through synaptojanin-2 binding protein (SYNJ2BP) (Harbauer et al., 2022) (Table 1 and Fig. 3). Tethering of Pink1 mRNA to mitochondria is crucial for its transport to dendrites and axons, and subsequent translation. AMPK phosphorylates SYNJ2BP within its PDZ domain, a modification required for SYNJ2a binding and Pink1 mRNA recruitment to mitochondria. Surprisingly, insulin signalling, which inhibits AMPK, leads to SYNJ2BP dephosphorylation and the subsequent release of SYNJ2a and Pink1 mRNA from mitochondria. This release promotes Pink1 translation and mitophagy (Hees et al., 2024b). SYNJ2BP recruits other NEM mRNAs to the outer mitochondrial membrane (OMM) in non-neuronal cells, facilitating their local translation after recovery from stress (Qin et al., 2021). It remains an open question as to whether AMPK-dependent regulation of SYNJ2BP and its interaction with SYNJ2a is conserved and extends to additional targets beyond Pink1.

Mitochondrial adaptations driving metabolic reprogramming also influence cancer development and response to therapy. Certain cancers, such as acute myeloid leukaemia, rely on OXPHOS for transformation, whereas others depend on anabolic pathways residing in mitochondria. A recent study identified the expression of a short cytosolic form of the transcription factor RREB1, called RREB1S, in leukaemia stem cells (Table 1). RREB1S specifically binds to a conserved motif in the 3′ UTR of several NEM mRNAs encoding respiratory chain subunits and enhances their translation through an unclear mechanism that appears to involve interaction with the elongation factor eEF1A1 (Han et al., 2024). Interestingly, CLUH might also play a role in cancer metabolism. CLUH binds both the mRNA and the protein product of the SPAG5 mRNA. SPAG5 encodes astrin, a protein involved in mitotic progression and highly upregulated in several cancers. CLUH promotes the translation and stabilises the full length astrin1 protein, ensuring that cell cycle progression is coupled with mitochondrial function (Schatton et al., 2022). However, the specific role of the CLUH–astrin axis in different cancer types remains to be elucidated.

In conclusion, RBPs dynamically interact with their target NEM mRNAs, controlling their fate and translational proficiency in response to an intricate signalling network that depends on cellular conditions. Through this mechanism, cells adapt mitochondrial biogenesis to changing environmental and metabolic conditions, surviving rapid metabolic transitions or stress situations. In the next section, we will discuss regulatory mechanisms influencing the spatial localisation of NEM mRNA translation and their functional significance.

The correct localisation of mRNAs provides an additional regulatory layer for their translation. Localised mRNA translation is a widespread phenomenon, and is particularly important in polarised organisms and cells (Martin and Ephrussi, 2009; Das et al., 2021; Bourke et al., 2023). Localised translation offers several benefits, including preventing protein misfolding, aggregation and degradation before they reach their target compartment, reducing the energy costs associated with protein trafficking and enabling translation at protected sites where it can be rapidly switched on or off as required. Where are NEM mRNAs translated? A wealth of studies across different organisms and cell types paints a complex picture, identifying the mitochondrial surface and contact sites between mitochondria and other organelles as major locations where NEM proteins are synthesised.

Groundbreaking work by Ron Butow's group in yeast identified the association of 80S cytosolic ribosomes with the OMM, which occurs in various degrees depending on the metabolic conditions of the cell (Kellems et al., 1974; Kellems and Butow, 1972, 1974). Some of these ribosomes could only be removed from mitochondria by inhibiting the association of the nascent polypeptide chain (NPC) with the ribosome through treatment with puromycin, which triggers the release of the mRNA with the nascent chain from ribosomes (Kellems et al., 1974). More recently, electron cryo-tomography has been used to reveal clusters of ribosomes on the yeast mitochondrial surface that depend on interactions between the NPC and the translocase of the outer membrane (TOM) complex (Gold et al., 2017). Consistent with this, the association of many mRNAs and RBPs with mitochondria relies on Tom20 (also known as Tomm20) (Eliyahu et al., 2010; Saira et al., 2024 preprint). That translation can occur at the OMM has been shown by proximity-specific ribosome profiling for many inner mitochondrial membrane proteins in S. cerevisiae (Williams et al., 2014). Furthermore, one study suggests a possible ribosome specialisation for the translation of NEM mRNAs (Segev and Gerst, 2018), a concept that represents an emerging area of investigation.

The extent to which these findings in yeast can be transferred to other organisms is still being debated. Mapping RNA localisation to different cellular compartments in Hek293 cells using proximity labelling of RNA by the peroxidase APEX2 unravelled many NEM mRNAs enriched at the OMM (Fazal et al., 2019). Remarkably, for several of these mRNAs, OMM enrichment increased when translation was stalled with cycloheximide – a drug that preserves the mRNA–ribosome–NPC association – whereas it decreased when the mitochondrial membrane potential (MMP) was dissipated or upon puromycin treatment. These findings suggest that their OMM localisation depends on the NPC–TOM interaction, similar to what occurs in yeast (Kellems et al., 1974). Conversely, other transcripts, including those encoding OXPHOS components and mitochondrial ribosomal proteins, maintained their OMM localisation even when the interaction with ribosomes was disrupted by puromycin, hinting to a ribosome- and NPC-independent targeting mechanism (Fazal et al., 2019). However, a study in zebrafish has found that many mitochondrial proteins are synthesised on cytosolic polysomes, whereas a few transcripts characterised by long 3′ UTRs and encoding large proteins with transmembrane domains are present in the mitochondrial fraction (Uszczynska-Ratajczak et al., 2023). Together, these findings support the hypothesis that at least a fraction of NEM proteins, probably carrying specific features, are synthesised locally on the mitochondrial surface and co-translationally imported into mitochondria.

The research summarised above suggests that NEM mRNAs can be targeted to the OMM by two mechanisms: one group of mRNAs depends on ribosome and NPCs, whereas another group possibly relies on binding directly to a OMM receptor or RBP. Research in S. cerevisiae has been successful in identifying targeting factors that localise translating ribosomes and mRNAs to mitochondria (Fig. 3). One such factor is the OMM protein OM14, which acts as a mitochondrial receptor for the nascent chain-associated complex (NAC) (Lesnik et al., 2014). NAC is a chaperon that binds ribosomes and newly synthesised proteins at the ribosome exit tunnel. Alternatively, RBPs can recruit NEM mRNAs directly to the OMM. As already mentioned, one of these is Puf3, which targets a subset of NEM mRNAs to the mitochondrial surface, promoting their translation under oxidative conditions (Miller et al., 2014). In yeast, another mechanism that enhances NEM mRNA localisation to mitochondria is the geometric constraint induced by the reduction in cell volume during the shift from glycolytic to oxidative conditions. This mechanism would be particularly relevant for mRNAs that are not found at the OMM under basal conditions (Tsuboi et al., 2020).

In multicellular organisms, a homologue of OM14 has not been identified so far. However, important players on the mitochondrial surface are A-kinase-anchoring proteins (AKAPs). In Drosophila ovaries, the RBP La ribonucleoprotein 1 (Larp1) interacts with MDI (also known as Spoon, the orthologue of human AKAP1) on the OMM and promotes the local translation of a subset of NEM mRNAs required for mtDNA replication (Zhang et al., 2016). This mechanism might be conserved in mammals, since another La-related RBP, LARP4, has recently been linked to the translational regulation of NEM mRNAs (Lewis et al., 2024), and is recruited on OMM by AKAP1 (Gabrovsek et al., 2020). AKAP1 binds regulatory subunits of protein kinase A, potentially acting as a scaffold to coordinate signalling pathways and the translational machinery on mitochondria.

Local translation of mitochondrial proteins is fundamental in highly polarised cells, such as neurons, where it plays a central role in growth cones, synaptic spines and axons (Hafner et al., 2019; Yoon et al., 2012; Rangaraju et al., 2019). NEM mRNAs have been found in axons and dendrites (Aschrafi et al., 2016; Saal et al., 2014; Yoon et al., 2012; Zappulo et al., 2017). However, there is conflicting evidence regarding whether NEM transcripts are enriched or underrepresented in axons compared to the soma (Aschrafi et al., 2016; Maciel et al., 2018; Nijssen et al., 2018; Jung et al., 2023; Zappulo et al., 2017; Briese et al., 2016; Shigeoka et al., 2016). Notably, many NEM mRNAs belong to a class of low abundance, but highly translated, axonal mRNAs (Jung et al., 2023). Thus, for a subset of NEM proteins, the regulation of local translation is crucial for their synthesis. Consistent with this, the loss of RBPs regulating NEM mRNAs is often characterised by neuronal phenotypes. PUM1 and PUM2 are highly expressed in the soma of developing neurons, and PUM2 mediates the retention of target mRNAs in the soma, preventing their transport into neurites (Martínez et al., 2019). Remarkably, concomitant downregulation of PUM2 and PUM1 in hippocampal neurons leads to increased levels of not only known target mRNAs encoding synaptic proteins but also NEM mRNAs (Randolph et al., 2024). These changes, reflected by increased synapse density and maturation, suggest a role for PUM1 and PUM2 in regulating synapse maturation during development (Randolph et al., 2024). Whereas low levels of PUM1 and PUM2 are associated with increased NEM mRNA levels in neurite, knockout of Cluh in murine neural cell progenitors causes severe depletion of its target NEM mRNAs and respective protein levels from axons. This depletion is accompanied by an axonal degeneration phenotype in long peripheral motor axons and defects in axonal translation (Zaninello et al., 2024).

Owing to their polarised architecture, neurons are a model of choice for visualising the trafficking of mRNAs and localised translation. Studies in neurons have shed light into novel principles and new players in the translation of NEM mRNAs. Using single-molecule fluorescence in situ hybridisation (smFISH) or a fluorescent tagging system to express and detect specific mRNAs, several studies have shown that not only mitochondria (Cohen et al., 2022; Harbauer et al., 2022) but also early endosomes (EEs) (Schuhmacher et al., 2023), late endosomes (LEs) (Cioni et al., 2019) and lysosomes (Liao et al., 2019; De Pace et al., 2024) function as carriers for mRNAs, including NEM mRNAs, in dendrites and axons (Fig. 3). The interactions between mRNAs and these organelles depend on specific adaptors, such as the five-subunit endosomal Rab5 and RNA–ribosome intermediary (FERRY) complex on EEs (Schuhmacher et al., 2023), and the previously mentioned SYNJ2BP–SYNJ2a complex on mitochondria (Harbauer et al., 2022). Notably, thousands of mRNAs, including many NEM mRNAs encoding subunits and accessory components of the mitochondrial F₁F_O–ATPase, are trafficked along the endosomal system in Ustilago maydis, a filamentous fungus. This emphasises that the trafficking of mRNAs along endosomes is a general principle (Olgeiser et al., 2019). In other cases, as shown for the Cox7c mRNA, the coding sequence of the mRNA itself contains information required for its association with mitochondria via a yet unclear mechanism, allowing its co-trafficking in motor neuron axons (Cohen et al., 2022).

The localisation of NEM mRNAs to the endo-lysosomal system in neurons is important, as it acts as a translational hub. Translation activity has been identified on LEs, frequently in proximity to mitochondria (Cioni et al., 2019). Furthermore, the FERRY complex on EEs associates with ribosomes, and FERRY-positive endosomes loaded with mRNAs are found near mitochondria (Schuhmacher et al., 2023). Knockout of the lysosomal kinesin adaptor biogenesis of lysosome-related organelles complex (BLOC)-one-related complex (BORC), which inhibits anterograde lysosomal trafficking, reduces levels of mRNA and translation of ribosomal and some NEM proteins involved in OXPHOS in axons of human induced pluripotent stem cell (iPSC)-derived neurons (De Pace et al., 2024). LEs and lysosomes also mediate the delivery of granules containing the fragile X messenger ribonucleoprotein (FMRP; also known as FMR1) and the Mff mRNA to the mitochondrial midzone (Fenton et al., 2024). This process mediates the local translation of MFF, promoting fission at the centre of the mitochondria and thereby preserving mtDNA content (Fenton et al., 2024).

Contact sites among endosomes, the endoplasmic reticulum (ER) and mitochondria might also play a role in regulating the translation of NEM mRNAs. The delivery of NEM proteins from the ER to mitochondria, a process named ER-SURF, was originally identified in yeast (Hansen et al., 2018) (Fig. 3). During this process, chaperons on the ER interact with NEM precursors and transfer them onto receptor proteins of the mitochondrial surface. ER-SURF relies on close ER–mitochondria contact sites and is particularly important for hydrophobic mitochondrial precursor proteins (Koch et al., 2024). This process is not simply a re-routing of proteins targeted to the wrong compartment, but rather an active and beneficial process that involves binding to the ER chaperon Djp1 (Hansen et al., 2018). For how many NEM mRNAs this pathway is of relevance remains unclear. However, it has been proposed to occur in neurons for the Pink1 mRNA. As mentioned above, this mRNA is tethered to the mitochondrial surface by the SYNJ2a–SYNJ2BP complex (Harbauer et al., 2022) but is released upon insulin signalling (Hees et al., 2024b). Once released, the Pink1 mRNA associates with ribosomes on the endo-lysosomal system and the ER, where translation occurs. The nascent chain is bound by the ER membrane-bound chaperon DNAJB6, the mammalian orthologue of Djp1, which transports PINK1 back to the mitochondria for import (Hees et al., 2024a preprint).

Interorganelle contact sites are also important for orchestrating translation rewiring during the activation of cellular stress responses, such as the integrated stress response (ISR), which represses cap-dependent translation by inducing phosphorylation of eIF2α (EIF2S1), the factor that delivers the initiation methionine tRNA (Brito Querido et al., 2024). Surprisingly, when ER stress triggers the ISR, the translation of NEM mRNAs is protected. This safeguard mechanism occurs through the interaction between the PERK kinase and the mitochondrial ATPase ATAD3 at the ER–mitochondrial contact sites. This interaction generates local domains with a low concentration of phosphorylated eIF2α, where the translation of certain beneficial mitochondrial proteins can take place (Brar et al., 2024).

Taken together, the mechanisms controlling the spatial translation of NEM mRNAs are complex, highly regulated and likely specific to the transcript, the cell type and even the subcellular compartment. Several questions also remain unanswered. Can processes identified in neurons be extended to other cell types or subcellular locations? What determines the transition between a transported, translationally repressed NEM mRNA and its active translation at a distinct time and location? Furthermore, it is still unclear whether the association of mRNAs with different organelles serves distinct roles, and how the interaction between a specific mRNA and a given organelle is regulated.

The translation of NEM mRNAs not only responds to cellular metabolic signals but also conveys information about the functional status of mitochondria. In fact, the translation of NEM mRNAs is continuously surveyed by a sophisticated set of ribosome-associated quality control (RQC) factors. These factors contribute not only to resolving and counteracting translational defects but also to inducing downstream stress responses (Fig. 4).

The evolutionary conserved RQC pathway is active in all cells to cope with ribosome stalling, with collision between adjacent ribosomes on the same mRNA, and with aberrant template mRNAs or translation products (Brandman and Hegde, 2016; Joazeiro, 2019). The interface between collided ribosomes is typically sensed by the E3 ubiquitin ligase ZNF598 (Hel2 in yeast), which mono-ubiquitylates specific ribosomal 40S constituents (Garzia et al., 2017; Juszkiewicz et al., 2018). This ubiquitylation profile acts as a signal through which further RQC components are recruited to the stalled ribosomal complex to initiate the appropriate response. The first players in this process are protein pelota homolog (Pelo), HBS1L and ABCE1 (Dom34, Hbs1 and Rli1 in yeast), which dissociate the 40S and 60S ribosomal subunits and mediate their recycling (Pisareva et al., 2011; Gouridis et al., 2019). After ribosomal splitting, endonucleases and the exosome complex degrade the template mRNA. Meanwhile, a series of factors, including NEMF, listerin, VCP and ANKZF1 (Rqc2 or Tae2, Ltn1, Cdc48 and Vms1, respectively, in yeast) are recruited to the 60S ribosomal subunit containing the NPC (Doma and Parker, 2006; Joazeiro, 2019). Upon NEMF binding, the E3 ubiquitin ligase listerin polyubiquiylates the NPC and targets it for proteasomal degradation (Brandman et al., 2012; Bengtson and Joazeiro, 2010). If the NPC does not expose lysine residues for ubiquitylation, NEMF can extend it outside the ribosome exit tunnel by incorporating C-terminal alanine and threonine extensions (CAT tails) (Shen et al., 2015). This phenomenon is called C-terminal extension (CTE) and increases the chances of poly ubiquitylation by listerin or other E3 ligases (Shen et al., 2015). Ultimately, proteasomal degradation of the NPC is facilitated by ANKZF1-mediated endonucleolytic cleavage of the tRNA carrying the NPC (Su et al., 2019) and NPC extraction through VCP (Defenouillère et al., 2013).

RQC mechanisms at the mitochondrial surface face specific challenges, especially under conditions that demand increased mitochondrial function, which could overburden the RQC machinery itself. For instance, the ubiquitylation of extended mitochondrial precursors might be obstructed by the concomitant import of NPCs into the TOM channel. Consequently, faulty variants of components of the respiratory chain might be imported into mitochondria and incorporated into dysfunctional respiratory complexes (Fig. 4). Alternatively, extended precursors might form intra-mitochondrial aggregates that sequester chaperones and RQC factors (Izawa et al., 2017; Wu et al., 2019) (Fig. 4). In yeast, these harmful effects can be prevented by blocking the CTE process, thus counteracting the formation of aggregation-prone peptides. Under stress conditions, Vms1 localises to the stalled ribosomes at the mitochondrial surface and induces the displacement of Rqc2 (Izawa et al., 2017; Su et al., 2019; Zurita Rendón et al., 2018). In Drosophila tissues and HeLa cells, persistent stress can impair the surveillance of the translation termination at the mitochondrial surface, resulting in the synthesis of extended NEM proteins, a variant of the co-translational CTE phenomenon, known as ‘mitochondrial-stress-induced translational termination impairment and protein C-terminal extension’ (MISTERMINATE) (Wu et al., 2019). These C-terminal extended proteins can be harmful not only if imported into mitochondria, but also if they aggregate in the cytosol. Counteracting this phenomenon is possible by potentiating co-translational RQC, for example, by overexpressing the ribosome recycling factor ABCE1, or the eukaryotic translation termination factor eRF1.

Genes involved in the RQC pathway are transcriptionally downregulated in the brains of individuals with PD (Wu et al., 2018), and increased levels of CTE variants of specific NEM proteins have been identified in fibroblasts from individuals with PD (Wu et al., 2019). In line with this, loss-of-function mutations in PINK1, a well-known PD causative condition, lead to a reduction of specific RQC components, including ABCE1 and eukaryotic release factor 1 (eRF1) in both Drosophila tissues and HeLa cells (Wu et al., 2019). Similarly, loss of CLUH in motor neurons induces depletion of ABCE1 specifically from axons. Overexpression of ABCE1 rescues the translation defects of Cluh-deficient axons and restores the axonal levels of the Atp5a1 mRNA (Zaninello et al., 2024). These findings suggest that RBPs for NEM mRNAs strongly cross-interact with the RQC pathway and highlight the importance of co-translational surveillance at the mitochondrial surface in the pathogenesis of neurodegenerative diseases. This opens up the interesting possibility that RQC components could serve as therapeutic targets.

Increasing evidence links components of the RQC pathway with mitochondrial turnover. An example of this interplay involves PINK1 and has been shown in Drosophila. Upon loss of the MMP, PINK1 accumulates on the OMM, where, together with Parkin, it recruits NEM mRNAs, such as the complex I subunit C-I30 (also known as ND-30), in an attempt to repair mitochondrial damage by promoting their local translation (Gehrke et al., 2015). This occurs through the displacement of translational repressors, such as Pumilio, and the recruitment of translational activators, such as eIF4 (Gehrke et al., 2015). However, a persistent and substantial drop in MMP leads to a marked slowdown in the import of NPCs translated at the OMM, resulting in ribosome stalling and activation of RQC pathway. A key event in this process, aimed at amplifying the RQC response, is the polyubiquitylation of ABCE1 by the ribosome-associated E3 ligase CCR4–NOT transcription complex subunit 4 (CNOT4). Polyubiquitylated ABCE1 serves as a signal that shifts the function of PINK1 from a translation enhancer to a mitophagy promoter (Fig. 4). Indeed, upon accumulation of polyubiquitylated ABCE1 on the OMM, several autophagy receptors are recruited to the stalled messenger ribonucleoprotein complex, leading to the autophagic removal of the damaged organelle (Wu et al., 2018). In this scenario, mitochondrial stress-induced modifications to the local RQC machinery play an important role in signalling mitochondrial damage to the cell and in triggering the appropriate clearance response. The same authors identified another mechanism through which the mitochondrial RQC enhances its performance via CNOT4. In this case, CNOT4-mediated polyubiquitylation of the ribosome collision sensor ZNF598 induces its accumulation on the OMM (Fig. 4). The increased presence of ZNF598 leads to a more efficient suppression of stalled mitochondrial translation and faulty peptide production (Geng et al., 2024). These studies reveal how the mitochondrial RQC system sensitively responds to stress conditions and enhances its own activity through the production of key signalling molecules.

Interestingly, a recent study has suggested that cells could exploit the high sensitivity of the RQC at the OMM to limit excessive translation of NEM mRNA. This has been shown in yeast, where overexpression of the mRNA encoding the mitochondria matrix factor 1 (Mmf1) triggers a mechanism named mito-ENcay. In this process, the stalled co-translational nascent chain is targeted at the OMM, where the mitochondrial RQC is promptly triggered, leading to the degradation of the MMF1 mRNA in order to prevent new Mmf1 synthesis and accumulation (Chen et al., 2023). A central player in this pathway is the Not4 ubiquitin ligase (CNOT4 in mammals), which stands out as a key element in QC pathways at the OMM that culminate in autophagy. The proposed model is that, together with Hel2, Not4 ubiquitylates the ribosome nascent chain complex to trigger autophagic removal. Notably, yeast does not contain a PINK1 homologue, suggesting that different species evolved alternative strategies to cope with aberrant translation at the OMM.

In conclusion, increasing evidence highlights translational stress at the mitochondrial surface as an important signalling node that dictates mitochondrial turnover. This coupling not only prevents the accumulation of dysfunctional organelles but also allows the biogenesis of new mitochondria, with a complement of proteins tailored to survive the stress conditions.

The translational regulation of NEM mRNAs reflects an intricate intersection of cellular signalling pathways, enabling cells to dynamically adapt mitochondrial function in response to metabolic and environmental cues. NEM mRNAs, specific RBPs and components of the translational machinery act as both responders and executioners of cellular signals, serving as versatile control nodes that coordinate cellular energetics, metabolism and mitochondrial quality control. Despite tremendous progress in elucidating these mechanisms in recent years, our current knowledge represents the tip of the iceberg.

Exciting new research will not only uncover additional RBPs involved in this process but will also help us understand their complex and tissue-specific regulation, as well as how they interact with the individual NEM mRNA sequences. Several examples outlined above suggest that certain mRNAs are regulated by specific mechanisms, but we still do not fully understand the commonalities among these mRNAs or how they are recognised. Another promising area of research will focus on the significance of the relocalisation of mRNAs and translational components to granular compartments during stress, either to select mRNAs for translation or protect them from degradation. A better understanding of the interplay between the spatial localisation of translation for specific NEM mRNAs and the import mechanism of the corresponding proteins is needed. Finally, it will be crucial to determine whether coupling translational dysregulation to mitochondrial turnover is a mechanism for rapidly renewing and reprogramming a specific mitochondrion when required. One of the most exciting developments in the field is the promise to reveal how intra-mitochondrial variability is achieved within a single cell.

In conclusion, research into the translational regulation of NEM mRNAs is rapidly evolving, providing a better understanding of mitochondria as central signalling organelles. Importantly, discoveries related to the spatial-temporal mechanisms controlling the translation of NEM mRNAs have implications for human diseases. Alterations in these processes have been linked to neurodegeneration, cancer, and metabolic disorders, revealing new potential therapeutic targets.

The authors thank members of the Rugarli lab for helpful discussions.