The mitochondrial intermembrane space – a permanently proteostasis-challenged compartment

1 Introduction: the dynamic life of proteins takes place under the surveillance of sorting and quality control machineries

The life of a protein unfolds as a dynamic journey, beginning as a nascent polypeptide emerging from the ribosome and undergoing a series of interconnected modifications, folding steps, and quality control measures to achieve its functional state (Figure 1). From the outset, co-translational processes sculpt the protein’s fate. Enzymatic modifications like N-terminal cleavage by methionine aminopeptidase and peptidases such as DPP8 and DPP9 (Cui et al. 2022; Zolg et al. 2024), as well as acetylation by N-terminal acetyltransferases (Ree et al. 2018), refine and stabilize the polypeptide chain (Varland et al. 2015). Simultaneously, ribosome-associated chaperones, including the nascent polypeptide-associated complex (NAC) (Deuerling et al. 2019; Gamerdinger et al. 2019; Kirstein-Miles et al. 2013) and the ribosome-associated complex (RAC, formed by Ssz1 and Zuo1)(Otto et al. 2005; Zhang et al. 2020), act near the ribosome exit tunnel to stabilize emerging polypeptide chains, preventing premature folding or aggregation (Kramer et al. 2019) (Figure 1, step 1).

Figure 1:

The life-cycle of proteins involves organelle-specific proteostasis networks. (1) Nascent polypeptides are recognized by ribosome-bound chaperones such as Hsp70, NAC and RAC, (2) stabilized and folded with the help of Hsp70 and chaperonins such as CCT/TRiC and (3) imported, either co- or post-translationally but mostly in an unfolded form, into organelles such as the ER, mitochondria, peroxisomes or chloroplasts. Organelle-localized chaperones (e.g. mtHsp70 and Hsp60/Hsp10 in mitochondria) support intra-organellar protein folding. (4) Many proteotoxic stresses such as elevated temperature or oxidative stress affect proteins in all cellular compartments, but intensity, duration and the specific combination of stresses differ. (5) Consistently, chaperone systems such as Hsp90, small heat shock proteins or disaggregases exist in cytosolic but also organellar variations. (6) If (re)folding fails, cells resort to containing and removing denatured proteins. While proteolysis already plays a role during maturation, proteases (such as HTRA1 in the cytsosol and HTRA2 in mitochondria) also remove damaged proteins. The proteasome, as part of the ubiquitin-proteasome system (UPS) is most important in the cytosol, although nuclear localization and activity has been reported. Autophagy processes remove protein aggregates or even organelles. (7) Proteotoxic stress almost invariantly induces a downregulation of translation to reduce the burden to the proteostatic machinery. This response exists across eukaryotic kingdoms and is termed ISR in mammals. However, stress- and organelle-specific responses modulate this response in many ways, for example to induce expression of organellar chaperone machinery or anti-oxidative factors. As part of homeostatic systems, such responses also become active at a stage when proteotoxic stress did not accumulate but can be prevented or at least anticipated.

Proper folding into functional conformations is further supported by diverse chaperone systems. Hsp70 chaperones prevent aggregation and assist in refolding misfolded proteins (Kramer et al. 2019; Rosenzweig et al. 2019; Willmund et al. 2013) (Figure 1, step 2). The Hsp90 system plays a critical role in the final stages of folding, particularly for large complexes and signaling proteins (Prodromou et al. 2023). Chaperonins like GroEL/GroES (and the mitochondrial Hsp60/10 complex) provide enclosed environments thereby shielding folding intermediates from interference by unwanted interactions. Small heat shock proteins bind partially folded intermediates to stabilize them for further folding (Reinle et al. 2022), while peptidyl-prolyl isomerases resolve rate-limiting cis-trans isomerizations of proline residues to ensure efficient folding (Schiene-Fischer 2015). Post-translational modifications further shape protein function and stability (Zhang and Schroeder 2025). In mitochondria, as well as in the endoplasmic reticulum (ER), disulfide bond formation and other oxidative modifications play key roles (Narayan 2021; Wang and Wang 2023; Zarges and Riemer 2024). Similarly, proteolytic cleavage occurs in both compartments as part of the regulated processing of newly synthesized proteins.

The integration of targeting signals early in translation ensures that proteins are directed to their cellular destinations (Gamerdinger and Deuerling 2024) (Figure 1, step 3). For mitochondria, targeting sequences are recognized by import machineries such as the translocases of the outer and inner membrane and their receptors (TOM and TIM, respectively). Similar import complexes exist in chloroplasts (TOC/TIC) (Liu et al. 2023) and the ER (e.g. Sec61 translocon) (Itskanov and Park 2023). The translocation process often requires proteins to remain unfolded, a state maintained by chaperones that shepherd nascent proteins en route to their destination, to channel them through narrow import pores. Chaperones involved in the targeting of proteins to the ER have been reviewed elsewhere (Kramer et al. 2019; Wang and Hegde 2024). Although mitochondrial proteins can be imported both co- and post-translationally (Becker et al. 2019), they are still required to remain or become unfolded for import. To prevent misfolding of the mitochondrial precursors in the cytosol, the NAC as well as the Hsp70 and Hsp90 systems play important roles and cooperate with a network of import complexes, as recently reviewed in (Brave et al. 2024) and further described below.

Despite the complex network of co-translational safeguards, environmental stressors like reactive oxygen species (ROS), thermal fluctuations or interactions with unfolded proteins can destabilize proteins, causing misfolding or aggregation (Figure 1, step 4). It has been proposed that nascent polypeptides in particular pose a threat to proteostasis due to a high rate of misfolding (Filbeck et al. 2022; Schubert et al. 2000). Although the extent of this problem remains unsolved (Eisenack and Trentini 2023), misfolding of newly synthesized proteins certainly poses a risk to cellular integrity, as has been demonstrated recently in the context of ageing (Stein et al. 2022). In particular, unfolded proteins en route to their cellular destination represent a liability for proteostasis (Vazquez-Calvo et al. 2023).

To further mitigate such risks, cellular quality control systems monitor protein integrity after translation (Figure 1, step 5). Chaperones either refold damaged proteins or sequester them to prevent aggregation (Saibil 2013). Irreparable proteins in the cytosol are degraded by the ubiquitin-proteasome system (UPS), which tags damaged proteins with ubiquitin chains for proteasomal degradation, or by autophagy-lysosome pathways, which degrade aggregates and organelles (Figure 1, step 6). In mitochondria, distinct proteases such as LONP1 and CLPXP in the matrix or YME1L and HTRA2 in the intermembrane space (IMS) contribute to degradation in a ubiquitin-independent manner. Stress responses, including the mitochondrial unfolded protein response (UPR^mt), cytosolic heat shock response (Lang et al. 2021), and ER unfolded protein response (Hetz et al. 2020) as well as the integrated stress response (ISR) (Pakos-Zebrucka et al. 2016) balance proteostasis by reducing general translation, increasing chaperone production, and enhancing degradation pathways.

While the protein life cycle adheres to many generalizable concepts, different compartments have found specific solutions caused by their specific tasks (Figure 1, steps 4–6). In this review, we describe the challenges and maintenance mechanisms for proteostasis specifically in the IMS of mitochondria. This subcompartment is a trafficking hub for many mitochondrial proteins, not just those that remain soluble within the IMS itself, as we will describe below. We will highlight unusual environmental factors that proteins passing through or populating the IMS are exposed to, including a low volume-to-membrane ratio, elevated temperature as well as variable pH and ROS levels. Subsequently, we will describe the systems that are in place to prevent or react to proteotoxic stress within the IMS. We will finish with a discussion of (sub)compartment-specific stress responses (Figure 1, step 7), including models that have been employed to elicit proteotoxic stress in the IMS. Of note, we will not extensively discuss stress responses linked to proteostasis defects in the mitochondrial matrix, except for those that have also been linked to IMS processes. These responses are the subject of a plethora of insightful reviews including (Bilen et al. 2022; Eckl et al. 2021; Kim and Lee 2024; Münch 2018; Mukhtar et al. 2023).

2 Mitochondrial protein precursors traverse the IMS during import and constitute a liability for proteostasis

The vast majority of mitochondrial proteins - all but 13 of at least 1136 proteins according to the latest mitochondrial protein inventory (MitoCarta 3.0; Rath et al. 2021) - are synthesised in the cytosol and imported into mitochondria. Mitochondrial import pathways have been described in several excellent reviews including (Backes and Herrmann 2017; Brave et al. 2024; Mukhtar et al. 2023; Neupert 2015; Pfanner et al. 2025; Schmidt et al. 2010; Wiedemann and Pfanner 2017). To illustrate the role of protein import for IMS proteostasis, the following paragraphs provide details on the various routes that import substrates take through the IMS, the factors that keep these proteins in a folding- and transport-competent state and challenges that these newly imported proteins may present to proteostasis (overview in Figure 2C). Insight into these processes was established to a large degree by studies using the model eukaryote Saccharomyces cerevisiae. To differentiate findings in baker’s yeast from the processes described in mammals, we will apply the corresponding nomenclatures, by which yeast proteins are written with a single capital letter (“Mia40”), whereas mammalian proteins are fully capitalized (“MIA40”). However, protein complexes are fully capitalized in both species (“TIM23”). Since these processes are in general well conserved, we will highlight aspects that differ, instead of providing species-specific descriptions.

Figure 2:

Mitochondrial import pathways traverse the IMS and pose threats to IMS proteostasis. (A) Mitochondrial targeting signals determine sub-organellar localization. (B) Alpha-helical OMM proteins are (1) likely stabilized by cytosolic chaperones such as Hsp70, recognized by TOM-associated MTCH2 in mammals or the MIM complex in yeast, inserted into the OMM and then laterally released. (3) MTCH2 and MIM also insert single-spanning alpha-helical proteins without associating with the TOM complex. Insertion is possible in both orientations. (C) Overview of import pathways, which traverse the TOM complex in the OMM. (D) Beta barrel proteins destined to the OMM are first (1) stabilized by HSP70, Hsp90 and Hsp40 proteins in the cytosol, then (2) imported via TOM and recognized by the Sam50 subunit of the SAM complex, which is linked to the TOM complex via cytosolic contacts (3). (4) At the trans side of the OMM in yeast, small TIM chaperones recognize and stabilize hydrophobic regions. This may not be the case in mammals (see text). (E) (1) IMS and IMM proteins with bipartite targeting signals are recognized via the presequence receptors of the TOM complex, which forms a supercomplex with TIM23. (2) After cleavage of the presequence by MPP, these precursors are released from the TIM17 channel. In yeast, this step is supported by Mgr2. Following lateral release, these proteins can either (3) remain in the IMM or (4) get released as soluble IMS proteins with the help of PARL and IMMP1/2 proteases. (F) (1) IMM proteins with internal targeting signals mostly belong to the mitochondrial carrier family (MCF). These are imported through TOM and (2) in yeast, recognized at the trans face by small Tim chaperones and membrane-inserted via the TIM22 complex. In mammals (3), tighter coupling between TOM and TIM22 may replace the need for small TIMMs to some degree (see text). (G) (1) Disulfide relay substrates require cytosolic reducing factors including glutaredoxins (GRX) and thioredoxins (TRX) to prevent misfolding. (2) The TOM complex directs precursors towards AIFM1-MIA40 complexes, which introduce (3) structural disulfide bonds, while shielding hydrophobic regions in unfolded precursors. (4) ALR is responsible for recycling reduced MIA40.

In general, mitochondrial proteins are believed to be imported post-translationally (Becker et al. 2019). However, evidence for co-translational import is accumulating (Fünfschilling and Rospert 1999; Gold et al. 2017; Lesnik et al. 2014; Lapointe et al. 2018; Tsuboi et al. 2020; Williams et al. 2014). This latter mode of protein import may be especially important for specific proteins e.g. of the inner mitochondrial membrane (IMM) and when mitochondrial biogenesis is increased. Irrespective of the coordination with the import process, localized translation near the organellar surface has been described (Devaux et al. 2010; Gao et al. 2014; Gold et al. 2017; Quenault et al. 2011; Zhang et al. 2016; Zaninello et al. 2024) and may be important to reduce the exposure of unfolded precursors in the cytosol.

The TOM complex acts as the initial entry point for most mitochondrial precursors and, due to the limited luminal diameter of its channel subunit (Ahting et al. 2001; Tucker and Park 2019), precursors are required to be in an unfolded state for their import (Araiso et al. 2022). Upon translocation through the TOM complex, different machineries sort the precursor proteins into subcompartments, directed by a variety of targeting motifs in the precursors (Figure 2A). Roughly half of all mitochondrial proteins are targeted to the matrix, another third populates the IMM, ten percent go to the outer membrane (OMM) and about five percent are localized in the IMS (Rath et al. 2021). Except for alpha-helical OMM proteins, which are introduced via the mitochondrial import machinery (MIM) in yeast (Doan et al. 2020; Papić et al. 2011) and with the help of MTCH2 in mammals (Guna et al. 2022) (Figure 2B), all mitochondrial proteins traverse the TOM channel and the IMS during their import (Figure 2C). Together with the variety of import pathways that lead to localization within or facing the compartment, this fact places the IMS at a central junction of mitochondrial protein import (Edwards et al. 2020, 2021; Wiedemann and Pfanner 2017). It further raises the question whether immature client proteins en route to their destination may constitute a threat to IMS proteostasis.

It has been recognized for almost four decades that protein translocation into mitochondria may occur through connected pores, creating a bridge across OMM, IMM and the IMS (Schleyer and Neupert 1985). Recent observations of TOM pores in a number of supercomplexes (reviewed in (Brave et al. 2024)), which form a network of import structures (termed the “mitochondrial import network / MitimNet”) provide a modern perspective on this early concept. The TOM-TIM23 supercomplex, for example, allows for direct handover of proteins destined for the mitochondrial matrix (Albrecht et al. 2006; Callegari et al. 2020). Similarly, TOM-SAM supercomplexes (Qiu et al. 2013), coupling of TOM pores to the TIM22 translocase complex (Callegari et al. 2016; Kang et al. 2016) of the IMM as well as TOM complex structures that hand over substrates to the oxidoreductase Mia40 (Gornicka et al. 2014) were described. It has not, to our knowledge, been determined to which extent import proceeds via these supercomplex assemblies or whether the import systems also operate in isolation. Some substrates employ two otherwise distinct import systems, the disulfide relay and the TIM23 complex, which might indicate that at least in some instances substrates can engage import systems in “non-canonical” combinations (Longen et al. 2014; Peker et al. 2023). In plants, for comparison, the TOC and TIC import pores of chloroplasts do form a stable supercomplex (Jin et al. 2022; Liu et al. 2023), which mediates import of the bulk of chloroplast matrix proteins (Chen and Li 2017). Such assemblies may on the one hand allow for more efficient import processes. On the other hand, the existence of these import “pipelines” may also indicate that evolution has taken preventive measures against the exposure of pre-mature proteins within the IMS. Indeed, newly imported proteins are more vulnerable to proteotoxic stress and pose a higher risk to mitochondrial proteostasis than the already correctly localized ones (Vazquez-Calvo et al. 2023).

The following paragraphs provide details on the various routes that import substrates take through the IMS, the factors that keep these proteins in a folding- and transport-competent state and challenges that these newly imported proteins may present to IMS proteostasis (overview in Figure 2C). Insight into these processes was established to a large degree by studies using the model eukaryote S. cerevisiae. Since these processes are in general well conserved, we will highlight aspects that differ, instead of providing species-specific descriptions.

Targeting of mitochondrial proteins to subcompartments is achieved via different sequence motifs, with the amphipathic, positively charged mitochondrial targeting sequence (MTS) being the most frequent signal. Most MTSs are found at the proteins’ N-termini and target them to the matrix via the presequence pathway. Substrates of this pathway are first stabilized in the cytosol by chaperones including Hsp70 and others (see e.g. Kang et al. 2018, for details), and are recognized by the Tom20, Tom22 and Tom50 subunits of TOM on the mitochondrial surface (Pfanner et al. 2019). Import proceeds through a translocation channel formed by the Tom40 subunit of TOM and by the Tim17 subunit of the major translocase of the IMM, the TIM23 complex. Substrate recognition by the translocase complexes is required to trigger the formation of the TOM-TIM23 supercomplex. The supercomplex is further stabilized by interaction of the substrate with the Tim50 subunit that guides it to the TIM23 gate as well as by Tom22 (reviewed in (Callegari et al. 2020)). The membrane potential difference across the IMM (Δψ, negative on the matrix side) attracts positively charged MTSs and supports the action of a motor complex on the matrix face of the TIM23 complex. This presequence-associated motor (PAM) unfolds and drives the substrates into the matrix, where they are further stabilized by interaction with matrix chaperones including mitochondrial HSP70 (mtHSP70), while presequences are cleaved off by the mitochondrial presequence peptidase (MPP).

Many proteins destined for the OMM have β-barrel structures. Unlike proteins translocated via the TOM-TIM23 pathway, β-barrel proteins lack a classical MTS. Instead, they contain internal sorting signals in the form of specific β-hairpin motifs (Jores et al. 2016), which are first recognized by cytosolic chaperones of the Hsp70, -90 and −40 families (Figure 2D, step 1) (Young et al. 2003; Jores et al. 2018). These chaperones in turn interact with the TOM subunit Tom70 on the OMM, while the β-hairpin motifs are bound by the Tom20 subunit. Following import through TOM into the IMS, β-barrel proteins are recognized by the Sam50 subunit of the sorting and assembly machinery (SAM) for membrane integration (Ganesan et al. 2024; Höhr et al. 2018) (Figure 2D, step 2). The transfer is assisted by proximity of the TOM- and SAM-complexes, which is induced by cytosolic regions of the subunits Tom22 and Sam35, respectively (Qiu et al. 2013) (Figure 2D, step 3). While this proximity reduces the exposure of the precursors to the IMS lumen, it was also shown that small Tim chaperones in the IMS of yeast mitochondria bind to β-barrel proteins to prevent aggregation (Figure 2D, step 4) (Diederichs et al. 2020; Wiedemann et al. 2003). We will further describe these small Tim chaperones in the next section. A few α-helical OMM proteins with large IMS domains, like Om45 and Mcp3, may also enter the IMS during their import (Pfanner et al. 2019; Wiedemann and Pfanner 2017). These proteins first utilise the presequence pathway before being released into the IMS and subsequently inserted into the OMM (Song et al. 2014; Sinzel et al. 2016; Wenz et al. 2014).

Many IMM as well as some IMS protein precursors contain a “bipartite MTS” composed of a classical MTS followed by a hydrophobic sorting signal. Such substrates are recognised by the TOM complex and transferred to TIM23 (Figure 2E, step 1). Upon reaching the TIM23 complex, the hydrophobic segment can act as a “stop-transfer signal”, halting translocation into the matrix. These proteins are then released laterally into the IMM from the “half-pipe” Tim17 subunit (Figure 2E, step 2)). After release, precursors destined for the IMS are cleaved by the inner membrane peptidase (IMP), releasing mature, soluble proteins (Figure 2E, step 3). Another IMM-localized protease that performs proteolytic maturation is Presenilin-associated rhomboid-like protein (PARL), a member of the rhomboid family of intramembrane serine proteases. It is active towards a more restricted set of substrates compared to IMP, including PINK1, the phosphatase PGAM5 (Sekine et al. 2012; Siebert et al. 2022), and the ubiquinol-transport protein STARD7, and appears to be important to control the dual distribution of these proteins between cytosol and IMS (Spinazzi and De Strooper 2016; Saita et al. 2018). Alternatively, a few proteins with transmembrane domains are imported partially into the matrix and are then integrated into the IMM via the OXA translocase (Figure 2F). This pathway has been termed “conservative sorting” due to its similarity to protein targeting to the periplasm in bacteria, and is important for biogenesis of the TIM22 translocase (Stiller et al. 2016).

The carrier pathway, also called the TIM22 pathway, imports polytopic membrane proteins, most of which serve as carriers for metabolites across the IMM (mitochondrial carrier family, MCF, Figure 2F). These proteins have multiple transmembrane segments and internal targeting signals instead of a classical MTS (Brix et al. 1999; Kreimendahl et al. 2020; Wiedemann et al. 2001). Cytosolic chaperones, such as Hsp70 and Hsp90, bind to carrier precursor proteins and transport them to mitochondria (Figure 2F, step 1). After passing through the TOM complex, carrier precursor proteins bind to small TIM chaperones in the IMS (Figure 2F, step 2), which guide them to the TIM22 complex in the IMM, a specialized translocase that inserts carrier proteins in a Δψ-dependent process. In humans, import of carrier precursors is facilitated by coupling between the TOM and TIM22 complexes by the subunit Tim29 and via interactions with the MICOS complex (Figure 2F, step 3) (Callegari et al. 2016, 2019). In yeast, a similar coupling between TOM and TIM22 is achieved by Por1 (Ellenrieder et al. 2019). Apart from ensuring a safe handover of precursors, this coupling may be important to coordinate import with energy metabolism demands (Brave et al. 2024).

The majority of soluble IMS proteins rely on the disulfide relay (Al-Habib and Ashcroft 2021; Edwards et al. 2020; Herrmann and Bykov 2023; Zarges and Riemer 2024) (Figure 2G). This pathway is specifically responsible for the import and oxidative folding of cysteine-rich proteins, which in most cases lack N-terminal MTS. These proteins often contain characteristic twin cysteine motifs (CX₃C and CX₉C) that form intramolecular disulfide bonds. Outside of mitochondria, these motifs are kept in a reduced state by the activity of the glutaredoxin and thioredoxin systems (Banci et al. 2013; Durigon et al. 2012) and possibly with the help of zinc ions (Figure 2G, step 1) (Lu and Woodburn 2005; Mesecke et al. 2008; Morgan et al. 2009; Terziyska et al. 2005). Using the MICOS subunit MIC19/CHCHD3 as a model which harbors the typical twin CX₉C motif, it was proposed that substrates of the disulfide relay in mammals are recognized by the TOM20 subunit of the TOM complex on the OMM (Marada et al. 2024). Similar to the regulation of the Tom22 receptor by PKA and casein kinase in yeast (Gerbeth et al. 2013), human TOM20 can be regulated by the cytosolic kinase DYRK1A to adjust mitochondrial protein import to cellular demands (Marada et al. 2024).

The protein mitochondrial intermembrane space import and assembly protein 40 (Mia40 in yeast, MIA40 and also CHCHD4 in humans), an oxidoreductase, chaperone holdase and import receptor, is the central enzyme of the disulfide relay in the IMS (Chacinska et al. 2004; Hofmann et al. 2005; Mesecke et al. 2005; Naoé et al. 2004; Terziyska et al. 2005). It is located at the IMM by a transmembrane helix in yeast and via its partner apoptosis inducing factor mitochondrial 1 (AIFM1) in mammals (Brosey et al. 2025; Hangen et al. 2015; Salscheider et al. 2022). It recognises and binds the incoming cysteine-rich proteins as they emerge from the Tom40 channel (Figure 2G, step 2). Notably, disulfide relay substrates and matrix-targeted substrates may pass through TOM complexes of distinct configurations (Araiso et al. 2019, 2022). The MICOS complex has been proposed to further facilitate the transfer of substrates from the TOM complex to Mia40 (von der Malsburg et al. 2011). Mia40 interacts specifically with the mitochondrial IMS sorting signal (MISS) also called IMS targeting signal (ITS) in its substrates, which is characterised by hydrophobic residues in proximity to a cysteine residue (Koch and Schmid 2014; Milenkovic et al. 2009; Sideris et al. 2009; Zarges and Riemer 2024). Following this initial, chaperone-like interaction, intermolecular disulfide bonds between Mia40 and its substrates are formed, which are released via formation of the intramolecular disulfide in the substrate (Figure 2G, step 3). The sulfhydryl oxidase essential for respiration and vegetative growth protein 1 (Erv1) in yeast and augmenter of liver regeneration (ALR) in mammals re-oxidize Mia40/MIA40, thus allowing for continuous oxidative folding of IMS proteins (Allen et al. 2005; Bien et al. 2010; Banci et al. 2011; Mesecke et al. 2005; Rissler et al. 2005) (Figure 2G, step 4). The disulfide relay may also contribute to the stabilization of proteins, which are imported by alternative pathways, as has been demonstrated for the complex I assembly factor NDUFAF8 (Peker et al. 2023) and the matrix ribosomal subunit Mrp10 (Longen et al. 2014). This mode of import has been termed the “two-step import pathway” and may be important to couple the state of disulfide import to electron transport chain (ETC) activity, as discussed further below.

In summary, the IMS serves as a critical sorting hub for proteins targeted to various mitochondrial compartments. This in turn makes this subcompartment particularly sensitive regarding the proteostatic burden caused by newly imported proteins (Vazquez-Calvo et al. 2023). We will describe in the following the challenging environmental conditions under which trafficking through and proteostasis in the IMS have to be maintained, which further underscore the importance of proteostatic mechanisms.

3 Biophysical challenges in a tiny specialized compartment compressed between two membranes

The IMS exhibits a complex morphology (Figure 3). It is embedded between the OMM and IMM. The IMM is folded into invaginations called cristae, leading to distinction between the inner boundary membrane (IBM), which runs parallel to the OMM, and the cristae membrane (CM) that invaginates towards the matrix. Cristae are connected to the IBM by narrow, pore- or slit-like structures called crista junctions (CJs). The pore created by CJs at the entrance to the cristae has been shown to act as a selective barrier, restricting the passage of proteins but also protons between the cristae lumen and the OMM-adjacent IMS, the peripheral IMS (reviewed in Rampelt et al. 2017; Giacomello et al. 2020; Caron and Bertolin 2024). This separation of cristae, which are the main site for ETC activity (Gilkerson et al. 2003) and thus subject to biophysical alterations (Figure 3), from the peripheral IMS, may be crucial to sustain proteostasis.

Figure 3:

IMS organization separates subcompartments and biophysical influences. The ultrastructure of the IMS is determined by the separation of cristae, bordering the matrix versus the peripheral region bordering the cytosol. ETC activities, along with their biophysical consequences including temperature, proton and ROS emission, are largely but not exclusively confined to cristae, whereas mitochondrial protein import traverses the peripheral IMS. The MICOS complexes induce membrane curvature and establish contacts with OMM-localized structures, forming bridges but also diffusion barriers for soluble and membrane-internal proteins. To which degree the slit junctions formed by these structures between the peripheral IMS and the cristae space are permeable for protons and small molecules remains to be established. The peripheral IMS is further characterized by its particularly restricted diameter (d), which results in a high likelihood for IMS proteins to be exposed to membrane structures.

The mitochondrial contact site and cristae organizing system (MICOS), a protein complex located at CJs in the IMM, plays a vital role for IMS ultrastructure by stabilizing membrane curvature via its subunit Mic10 (Barbot et al. 2015; Bohnert et al. 2015) and by forming contact sites between the IMM and OMM (Harner et al. 2011). Specifically, the SAM complex of the OMM interacts with MICOS to form the mitochondrial intermembrane space bridging complex (MIB) (Körner et al. 2012; Ott et al. 2015; Huynen et al. 2016). The contact is mediated by the OMM protein Sam50 and the Mic60 (Fcj1, Mitofilin) subunit of the MICOS complex (Bohnert et al. 2015; Ding et al. 2015; Ott et al. 2012; von der Malsburg et al. 2011). The organization of cristae was further found to depend on a number of proteins, which include the F₁F_O-ATP synthase inducing strong curvature in the CM (Hahn et al. 2016). Prohibitins were also implicated in cristae organization and interact in this function with the dynamin-like GTPase OPA1 (Merkwirth et al. 2008), independently of the function of OPA1 in mitochondrial fusion and fission. Further modulators of cristae morphology include the putative H⁺/Ca²⁺ antiporter LETM1, the regulator of Ca²⁺ transport MICU1 and the AAA ATPase family protein ATAD3A (reviewed in Ježek et al. 2023). Links between mitochondrial ultrastructure and Ca²⁺ homeostasis have emerged recently (Venkatraman et al. 2023) and interestingly, the mitochondrial membrane potential and the rigidity of cristae also seem to be coupled (Yoneda et al. 2022). The strain imposed by cristae curvature is relieved by an asymmetric distribution of phospholipids between the monolayer leaflets that constitute cristae membranes. Specifically, cardiolipin (CL) and phosphatidylethanolamine (PE) predominate in the negatively curved leaflet facing the cristae lumen while the opposing, positively curved, matrix-facing monolayer contains predominantly phosphatidylcholine (PC) (Ikon and Ryan 2017).

The IMS and the IMM exhibit a surprising degree of plasticity (Caron and Bertolin 2024). Early studies based on electron microscopy indicated that mitochondria could obtain so-called “condensed” or “orthodox” states, in high or low energy conditions, respectively (Hackenbrock, 1968). However, it was later shown that matrix contraction, a feature of the condensed state, might occur due to osmotic effects when mitochondria are isolated from the cellular environment (Frey et al. 2002). More recent methods that allow in situ observation of mitochondrial structure indicated that the “orthodox” state, with a large matrix volume and small IMS space is a more accurate representation of functional mitochondria (Frey et al. 2002). This implies that the diameter of the peripheral IMS is very limited – it appears roughly equal to the thickness of the IMM and OMM in cryo-electron tomography images (Gold et al. 2016, 2017), which is in the range of approximately 7 nm, similar to the size of typical globular proteins. Therefore, IMS proteins per default reside in close proximity to membranes, which may affect their folding capacities and might explain why soluble IMS proteins tend to be small and require disulfide bonds to remain stably folded. Advances in super-resolution microscopy techniques have further enhanced our understanding, revealing that cristae are actually highly dynamic structures that remodel on a timescale of seconds, as reviewed in (Kondadi et al. 2020b). These events, collectively referred to as the “cristae fission and fusion” (CriFF) model (Kondadi et al. 2020a), support the idea that cristae function as independent bioenergetic units and constitute individual subcompartments (Wolf et al. 2019). The physiological importance of cristae organization is paramount, as exemplified by the observation that alterations in this ultrastructure correlate with functional decline and ageing (Brandt et al. 2017). While cristae organization remains an active field of investigation, the dynamics of cristae organization certainly also contribute to a dynamic biophysical environment in the IMS.

The IMM has a very high concentration of proteins, estimated to exceed the mass of lipids at a ratio of 4:1 (Chrétien et al. 2018). The composition of proteins varies strongly between the CM and the IBM, with the former comprising ATP synthase, prohibitins and ETC components as major constituents, whereas the IBM features a larger number of proteins related to mitochondrial protein import as well as metabolite carriers (Iovine et al. 2021; Sukhorukov and Bereiter-Hahn 2009; Vogel et al. 2006; Wurm and Jakobs 2006; Zick et al. 2009). These IMM subdomains are separated by the MICOS complex structure (Rampelt et al. 2017; Wurm and Jakobs 2006), which limits the exchange of membrane proteins. Within the lamella-shaped cristae, ATP synthase dimers are lining the rims, consistent with their ability to induce membrane curvature (Blum et al. 2019). The localization of ETC complexes is not exclusive to the CM, as shown for example by the association of supercomplex (I)/III₂/IV with the TIM23 translocase, which facilitates protein import into the matrix by locally increasing the proton motive force and Δψ (Laan et al. 2006; Mehnert et al. 2014). Nonetheless, the bulk of proton pumping activity seems to be localized to the flat surface of the CM (Davies et al. 2011). The activity of the oxidative phosphorylation (OXPHOS) machinery results in a number of physical alterations in cristae that may need to be separated from the peripheral IMS.

The proton-pumping activity of the ETC complexes I, III and IV has led to the hypothesis that the IMS would feature a lower pH than the cytosol or the matrix. Indeed, early studies based on the pH-dependence of fluorescence intensity of proteins such as YFP found that the pH of the IMS was lower than that of the cytosol and the matrix (Porcelli et al., 2005). More recent studies based on pH-dependent ratiometric fluorescent protein sensors such as sEcGFP have further scrutinized local pH differences in mitochondria. The proton-pumping activity of ETC complexes I, III and IV and proton reflux occurring at the ATP synthase were shown to induce local pH differences within cristae (Rieger et al. 2021). Although these pH differences are mild (between 6.9 at complex IV and 7.3 at active ATP synthase), they still constitute a substantial acidification compared to the matrix, which was observed to reach pH values of up to 8 at the opposite side of the ETC complexes (Rieger et al. 2021). The extent of these pH differences yet remains to be more firmly established, as other studies indicated that dense packing of ETC complexes and kinetic coupling to ATP synthase are required to drive ATP synthase function (Toth et al. 2020). Hence, the pH differences observed may be confined to microdomains but could nonetheless contribute to proteostatic challenges in a short range. pH differences may also play a role in signaling events, for example by allowing or inhibiting the modification of thiols (which are only reactive as deprotonated, thiolate anions), or by affecting the orientation of cardiolipin, the main component of the IMM.

Another physical alteration imposed by ETC activity appears to be a considerable increase in local temperature. The Rak and Isales groups used the fluorescent dye Mito Thermo Yellow (MTY) to show that active mitochondria in a range of cell lines maintain a temperature significantly higher (approximately 10–15 °C) than their surroundings (Chrétien et al. 2018; Terzioglu et al. 2023). The Isales lab further validated these findings using mito-gTEMP, a genetically encoded ratiometric fluorescent temperature indicator targeted to mitochondria. Both groups found that inhibiting different components of the ETC leads to varying degrees of mitochondrial cooling, suggesting that heat generation is linked to specific ETC complexes. Notably, other studies using temperature-sensitive dyes and reporters found lower steady state temperatures that still increased by ca. 6 °C upon membrane potential uncoupling (Homma et al. 2015; Kiyonaka et al. 2013; Savchuk et al. 2019). Although there seems to be inconsistencies in reporting mitochondrial temperature between different studies, they commonly found spatial heterogeneities between mitochondria within a cell, and elevated mitochondrial temperatures upon membrane potential dissipation. Consistent with locally (at least temporally) elevated temperatures, thermal proteome profiling (TPP), which allows proteome-wide determination of melting points (Savitski et al. 2014), revealed that proteins associated with the ETC possess notably higher melting points than usual, averaging around 60 °C (Jarzab et al. 2020). Intriguingly, mitochondrial thermogenesis has also been shown to impact nuclear processes. Using fluorescent polymeric thermometers (FDV and ERthermoAC) and a heat shock factor 1 (HSF1)-EGFP fusion protein, Kang and colleagues demonstrated that protonophore-induced mitochondrial thermogenesis triggers the heat shock response within the nucleus (Kang et al. 2024). These studies collectively suggest that the increased temperatures near the ETC have significant implications for mitochondrial proteostasis and stress responses. Nevertheless, it is crucial to acknowledge the limitations of these studies. For example, the exact locations of probes within mitochondria have to be more clearly defined, as temperature variations may exist within the organelle itself. Moreover, many probes are also sensitive to other environmental fluctuations, such as the pH differences discussed above.

The release of ROS such as superoxide (O₂⁻) and hydrogen peroxide (H₂O₂) at the ETC has been extensively studied, revealing dual roles as signaling molecules and sources of oxidative stress (reviewed for example in (Sies et al. 2017; Santolini et al. 2019; Sies et al. 2024; Ulfig and Jakob 2024)). Within the ETC, complexes I and III are primary sites for ROS generation (Brand 2020; Dröse and Brandt 2012). Complex I generates superoxide through reverse electron transfer, especially under high proton motive force and when NADH is abundant, and releases it predominantly into the mitochondrial matrix. Complex III, specifically at the Q₀ site, produces superoxide as a byproduct of the Q-cycle and releases it into both the matrix and the IMS (Brand 2020; Chouchani et al. 2016; Murphy 2008). Mitochondrial membranes limit the diffusion of superoxide due to its charged nature. Therefore, the spatial release pattern critically determines the downstream effects. Superoxide may undergo dismutation catalyzed by superoxide dismutase (SOD) enzymes, such as SOD2 in the matrix or SOD1 in the IMS, to form H₂O₂ (Chatzi et al. 2016; Klöppel et al. 2010; Suzuki et al. 2013). In the matrix, catalase and glutathione peroxidase enzymes further detoxify H₂O₂ by reducing it to water. Similarly, peroxiredoxins, such as PRDX3 and 5 in the matrix and possibly PRDX3 in the IMS (Gomes et al. 2024), help neutralize H₂O₂. While H₂O₂ diffusion across membranes is also limited, porins in the OMM do allow some superoxide as well as H₂O₂ to exit (Han et al. 2003; Hermeling et al. 2022; Sutandy et al. 2023). The IMS thereby also benefits from the proximity to cytosolic antioxidative systems to neutralize H₂O₂ (Hoehne et al. 2022; van Soest et al. 2024).

H₂O₂ itself is not particularly damaging to proteins, since the most reactive amino acid side-chains, those of cysteines and methionines, still display low reactivity (Winterbourn and Hampton 2008). However, H₂O₂ can form more reactive hydroxyl radicals via reactions involving superoxides (Haber-Weiss reaction) or metal ions including iron (Fe^2/3+) and copper ions (Cu^1/2+) (Fenton reaction) (Jomova et al. 2023). Oxidation to form carbonyls on proline, arginine, lysine, and threonine and the introduction of non-native, intra- or intermolecular disulfide bonds are major consequences of the reaction of ROS with protein side chains, whereas base modifications like 8-oxo-guanine, and base cross-linking are major mechanisms by which ROS damage (mitochondrial) DNA. ROS also cause lipid peroxidation by attacking unsaturated fatty acids, disrupting membrane integrity and generating secondary toxic products like malondialdehyde. A comprehensive review of the chemistry involved in these modifications can be found in (Jomova et al. 2023). When detoxification fails, excess ROS can damage components of the ETC, leading to ETC dysfunction, which in turn increases the release of ROS. This vicious cycle contributes to conditions including cellular senescence, chronic diseases and ageing (Haynes and Hekimi 2022; Jomova et al. 2023; Miwa et al. 2022).

ROS as signaling molecules on the other hand are recognized within and outside of mitochondria. They elicit stress responses such as activation of redox-regulated chaperones (Ulrich 2023), mitophagy, the UPR^mt (Sutandy et al. 2023) and the NRF2 response (Ma 2013; Ngo and Duennwald 2022), as recently reviewed in (Ulfig and Jakob 2024). We will return to some of these signaling cascades in the context of IMS-related stress responses. Apart from stress responses, a multitude of signaling cascades depend on mitochondrial ROS. For example, an H₂O₂-induced signaling cascade that involves the redox-sensitive kinases Syk and Lyn, which localize to the IMS but also to the cytosol, is involved in regulation of gene expression, metabolism and the cell cycle (Patterson et al. 2015). Furthermore, ROS play important roles in the regulation of mitochondrial dynamics, by causing oligomerization of mitofusins and disulfide formation in ROS modulator 1, as reviewed in (Geldon et al. 2021). They also directly control mitochondrial protein import by an OMM-associated mechanism centered on the redox-sensitive ubiquitin ligase complex CUL2^FEM1B (McMinimy et al. 2024). Many examples illustrate the physiological importance of ROS release from mitochondria. For example, myoblasts are unable to differentiate in the presence of “reductive stress” (Rajasekaran et al. 2020) or upon over-activation of the NRF2 response (Manford et al. 2020), hypoxia-induced signaling can be suppressed by restricting ROS efflux (Sabharwal et al. 2013, 2023) and the longevity-promoting effect of peaking aerobic respiration during Caenorhabditis elegans development depends on mitochondrial H₂O₂ release (Hermeling et al. 2022). Regarding proteostasis in the IMS, it would be interesting to determine whether H₂O₂ diffuses freely from the cristae lumen towards the peripheral IMS. Here, ROS may specifically impact the folding of substrates of the disulfide relay, as will be discussed further below.

Finally, the IMS is host to the exchange of a plethora of small molecules between the mitochondrial matrix and the cytosol (ADP, ATP, NAD⁺, heme intermediates, GSH, [FeS] etc.). On the one hand, IMS compartmentalization may be important to sustain efficient and ordered metabolite exchange. For example, the recently elucidated interactions between AIFM1 and adenylate kinase 2 (AK2) may be relevant for localized ATP exchange and may also constitute an important signaling mechanism for NADH levels and the metabolic state of the cell (Rothemann et al. 2024a; Schildhauer et al. 2024). On the other hand, metabolite levels may also affect IMS organization as well as proteostasis. For example, it was shown that calcium dynamics control ETC activity and affect ROS generation as well as cristae dynamics. Calcium influx is intricately linked to mitochondrial function as well as import processes (Petrungaro et al. 2015). For space reasons, these intriguing aspects will not be further discussed here. Instead, we refer the reader to several excellent reviews (Contreras et al. 2010; Peng and Jou 2010; Pivovarova and Andrews 2010; Walkon et al. 2022).

Given these considerations on physical and chemical challenges, it is tempting to speculate that sequestration of the ETC into cristae evolved in part to avoid spreading of increases in temperature, ROS and protons to the peripheral IMS. Given the multitude of import processes traversing the (peripheral) IMS, the separation of these activities may be of high relevance for mitochondrial proteostasis. Consistently, small heat shock proteins were observed to localize specifically to the peripheral IMS region, but not to cristae (Adriaenssens et al. 2023) as will be further discussed below. In this light, alteration of cristae structures during ageing and increased release of peroxide radicals (Brandt et al. 2017) may have the harmful side-effect of challenging mitochondrial protein import fidelity and protein folding. Certainly, the link between mitochondrial subcompartment organization and proteostasis warrants further study.

4 Proteostatic safeguarding of the IMS

The dynamic and regulated ultrastructure of the IMS and its boundary membranes as well as the astoundingly complex structural organization of machineries involved in mitochondrial protein import, folding and stress responses are crucial to orchestrate processes that may affect IMS proteostasis, as described above and in several excellent recent reviews (Busch et al. 2023; Brave et al. 2024; Schulte et al. 2023). In addition, the IMS hosts a number of chaperones, proteases and stress responses, all of which are linked in a complex manner to create a robust network in charge of surveillance and preservation of proteostasis during and beyond protein import (Figure 4).

Figure 4:

Protein biogenesis and homeostasis in the IMS require specialized factors. (A) The multitude of import pathways that traverse the IMS require a dedicated set of chaperones, folding catalysts, proteases and safeguarding mechanisms (see main text). The biophysical environment challenges the fidelity of the import and, for the large set of disulfide relay substrates, can result in direct expulsion or retro-translocation from the organelle. (B) Even if import succeeds, homeostasis in the IMS faces constant challenges from ROS, elevated temperatures, and the vicinity to hydrophobic membranes, all of which may promote protein aggregation. Therefore, IMS proteostasis requires further safeguards. The ATP-dependent disaggregase CLPB collaborates with sHSPs to resolve aggregates and refold damaged proteins. When refolding fails, proteases such as HTRA2 and YME1L degrade terminally misfolded proteins, preventing toxic accumulation of soluble as well as membrane-internal proteins. For disulfide-containing proteins, proteolytic removal may require the removal of structural disulfide bonds. Whether active export from the IMS to the cytosol can occur is a matter of ongoing research (Bragoszewski et al. 2015).

Mitochondrial proteins are imported in an unfolded state. This is achieved in part by the action of cytosolic chaperones like Hsp70 or NAC, which bind mitochondrial precursors directly at the ribosome. On the other hand, the mitochondrial matrix-localized Hsp70 (mtHsp70), along with the proton motive force also exerts an unfolding force during protein import. Unfolding takes place at the TOM complex and is sufficient to denature a completely folded protein such as dihydrofolate reductase (DHFR). Interestingly, it is not sufficient to unfold methotrexate-stabilized DHFR, which has given rise to so-called “clogger” approaches, through which mitochondrial import can be blocked (Boos et al. 2019; Coyne et al. 2023; Eilers and Schatz 1986; Hsu et al. 2025). Following transport across the OMM, chaperones are still required to preserve the solubility of the imported proteins. The IMS does not contain members of the Hsp70 or Hsp90 families. However, a number of chaperones as well as the ATP-driven disaggregase CLPB (Skd3 in yeast) counteract unfolding and aggregation during and after import. In addition, IMS proteases not only process precursors but also remove misfolded proteins.

Chaperones of the IMS are small heat shock proteins (sHSPs), small TIM proteins (sTIMs) as well as MIA40. The sHSPs are a class of ATP-independent chaperones that play a critical role in maintaining proteostasis in the cytosol. Several members of this family, including HSPB1, HSPB5 and HSPB8, are imported under basal conditions into the IMS (Figure 4A and B) (Adriaenssens et al. 2023). Application of heat stress leads to additional recruitment of sHSPs, predominantly to the OMM. This may indicate that the sHSPs react to import defects that result in accumulation of import substrates at the OMM. However, their import into the IMS also increases during heat exposure, albeit to a lower degree (Adriaenssens et al. 2023). Depletion of sHSPs leads to mitochondrial swelling and reduced respiration, while aggregation of misfolding-prone substrates is countered in their presence. Regarding substrates, it was found that transmembrane proteins of the IMM (including several SLC25A family metabolite carriers) are enriched in the HSPB1 interactome (Adriaenssens et al. 2023).

The small Tim proteins are a group of highly conserved factors that shield hydrophobic mitochondrial precursor proteins from aggregation(reviewed in (Busch et al. 2023)) and in this regard, they seem to functionally overlap with sHSPs. In yeast, these proteins include Tim8, Tim9, Tim10, Tim12, and Tim13. The term “small Tims” hints to their compact size (10–13 kDa) and their involvement in the broader TIMM system of translocases, which facilitates the sorting and assembly of mitochondrial proteins. In yeast, these proteins operate as dynamic heterohexamers (Webb et al. 2006), such as Tim8-Tim13 and Tim9-Tim10, which shuttle β-barrel proteins to the SAM in the OMM and to the TIM22 translocase in the IMM (Figure 4A). In stress conditions, expression of the Tim9-Tim10 and Tim8-Tim13 complexes is increased. In humans, however, the role of TIMMs may be different. Although they do facilitate import of carrier proteins via TIM22, the complexes formed by TIMM9 and TIMM10 were reported to be tightly associated with the IMM, not present as soluble receptors (Mühlenbein et al. 2004). This may entail the requirement for a more spatially confined pathway for import of TIM22 substrates in higher eukaryotes (Callegari et al. 2019).

The proper folding and oxidative assembly of small Tims depend on the disulfide relay (Figure 4A). Mia40 recognizes incoming small Tim precursors via conserved cysteine motifs, facilitating their oxidative folding and incorporation into functional complexes. This oxidative folding process ensures that small Tims remain active and structurally stabilized to fulfill their chaperone roles. Dysfunction in Mia40 activity compromises the assembly of TIMM complexes, disrupting mitochondrial proteostasis and impairing protein import (Baker et al. 2012). The small Tims are further regulated by the mitochondrial protease Yme1/YME1L, as reviewed recently (Kan et al. 2024). Yme1 activity prevents accumulation of misfolded or unassembled small Tims (Baker et al. 2012; Spiller et al. 2015). Interestingly, cross-talk between Yme1 and small Tims also extends to regulation of the TIM23 complex. Its subunit Tim23 was reported to be a substrate for chaperoning by a Tim8–Tim9–Tim13 complex in yeast (Leuenberger et al. 1999), whereas Yme1 is involved in degradation of the TIM23 subunit Tim17, as will be further described below.

The role of Mia40 in protein oxidation for proper folding in the IMS is combined with its role as a chaperone holdase (Banci et al. 2010; Brosey et al. 2025; Koch and Schmid 2014; Peleh et al. 2016; Rothemann et al. 2024b; Weckbecker et al. 2012). During import, the recognition site for Mia40 in the substrates includes an exposed partially hydrophobic helix that typically separates the cysteine residues in the CX_3/9C motifs in these proteins (Figures 2G and 4A). Mia40 recognizes and shields these hydrophobic regions from non-native contacts (Banci et al. 2010; Longen et al. 2009). Crucially, the intermediate, intermolecular disulfide between Mia40 and its substrates is prone to reduction via alternative mechanisms, such as the activity of IMS glutaredoxins or glutathione (Habich et al. 2019a; Kojer et al. 2015). Therefore, distinct regulation of the glutathione redox potential between IMS and the cytosol is essential to allow proper oxidative folding by the disulfide relay (Hu et al. 2008; Kojer et al. 2012). Conversely, free, reduced cysteines may undergo non-native disulfide formation when exposed to oxidizing agents. The role of Mia40 as a chaperone holdase extends to proteins without cysteines including the endogenous substrate HAX1 in humans and an artificial cysteine-free variant of Atp23 (Rothemann et al. 2024b; Weckbecker et al. 2012).

Mia40 amounts are limiting for the import of cysteine-rich IMS proteins (Habich et al. 2019a; Peleh et al. 2016). The identification of small Tims among Mia40 substrates (Lutz et al. 2003; Milenkovic et al. 2007; Sideris and Tokatlidis 2007) provides an explanation for the rate-limiting role of Mia40 for the import and proteostatic surveillance of MCF proteins. This also implicates Mia40 in preventing proteotoxicity incurred by accumulation of precursors of these strongly hydrophobic proteins in the cytosol. Consistently, it was shown that overexpression of Mia40 is able to relieve the burden to cytosolic proteostasis systems (Schlagowski et al. 2021). This role may extend to other Mia40 targets since, as mentioned above, the translocation of Mia40 substrates is “reversible”. When oxidative folding is disturbed, the typically small substrates become retro-translocated into the cytosol, where they become substrates for proteasomal degradation (Bragoszewski et al. 2015; Habich et al. 2019a). The disulfide relay may also (indirectly) help to counteract defects occurring in other import pathways. Specifically, it has been proposed that the Mia40 substrate Mix23 facilitates the import of matrix proteins under stress conditions in yeast (Zöller et al. 2020). Depletion of Mix23 in unstressed cells did not cause any observable defect in growth or import of matrix proteins. However, it did show a synthetic growth defect with a temperature-sensitive mutant of the TIM23 component Tim17, suggesting that the IMS-localized Mix23 stabilizes or stimulates import via TIM23 (Zöller et al. 2020).

Stabilization of unfolded proteins during import is a main task of IMS chaperones. However, inherent imperfections of the import, leading to occasional release of misfolded proteins, necessitate further proteostatic mechanisms (Figure 4B). A particularly important factor seems to be the AAA+ domain–containing disaggregase CLPB or Skd3 in yeast. Human CLPB engages substrates as a homo-hexamer but forms dodecameric species during processing (Cupo et al. 2022; Spaulding et al. 2022; Wu et al. 2023). It has been described as the only IMS chaperone that is able to counter protein aggregation after substrate import (Cupo and Shorter 2020). However, it is only able to unfold proteins but does not itself mediate degradation. This function is likely exerted via the trimeric serine protease HTRA2 (Radke et al. 2008), which interacts with CLPB (Baker et al. 2024; Fan et al. 2022).

Interestingly, CLPB has also been observed to interact with the SLP2-PARL-YME1L (SPY) complex located at the IMM (Baker et al. 2024). This complex localizes to subdomains formed by the subunit SLP2 in a manner similar to the prohibitins that surround the matrix-oriented m-AAA+ protease (Wai et al. 2016). The precise role of this interaction is not yet clear but it appears that CLPB prevents misfolded proteins from blocking the activity of YME1L (Baker et al. 2024). It may also play an important role in maintaining the IMS-oriented part of the membrane-associated intrinsically disordered protein HAX1 in solution (Baker et al. 2024; Wu et al. 2023). Another strong dependency on CLPB function has been described for complexes I and IV of the ETC (Baker et al. 2024). Along with the link between Mia40 and ETC function, this constitutes yet another example of the intricate network that connects IMS proteostasis with energy metabolism and mitochondrial function in general.

Another highly connected IMS system comprises a subgroup of mitochondrial proteases (Deshwal et al. 2020; Lebeau et al. 2018; Ohba et al. 2020). These proteases include factors involved in structural maintenance, such as the IMS serine protease LACTB. LACTB can polymerize into stable filaments occupying the IMS. These filaments are speculated to play a role in structural maintenance (Quirós et al. 2015) and may affect metabolon organization. Potential roles in lipid metabolism and tumorigenesis have been reviewed in (Cascone et al. 2022). Another group of proteases is required for precursor maturation. IMMP1/2 and PARL were already mentioned during the discussion of the lateral release import pathway above. In addition to its import function, PARL suppresses mitophagy by cleaving PINK1, leading to PINK1’s retrotranslocation into the cytosol for degradation (Lin and Kang 2008). Conversely, failure to cleave PINK1 initiates mitophagy as further described below. Interestingly, PARL also influences the distribution between IMS and cytosol in the case of STARD7 (Deshwal et al. 2023; Saita et al. 2018). Hence, IMS accumulation of PARL substrates may burden proteostasis when cleavage by PARL is compromised.

HTRA2, also known as Omi, acts as a chaperone in steady-state conditions but displays ATP-independent serine protease activity upon heat treatment, explaining its name “high-temperature requirement A2” (Spiess et al. 1999). It plays a crucial role in mitochondrial protein quality control, apoptosis and stress response mechanisms (Bi et al. 2023; Faccio et al. 2000; Gray et al. 2000; Martins et al. 2002; Suzuki et al. 2004; Vande Walle et al. 2008). In non-stress conditions, HTRA2 is attached to the IMM. However, apoptotic stimuli induce proteolytic auto-processing of the transmembrane domain, releasing HTRA2 as a soluble protein (Suzuki et al. 2001). Its activation has further been linked to a change in the oligomerization state of the solubilized form (Aspholm et al. 2024; Toyama et al. 2021) and to divalent metal ion interactions with its PDZ domain. HTRA2 has pro-apoptotic functions when released into the cytosol, by degrading inhibitor of apoptosis (IAP) family proteins (Suzuki et al. 2004). Processing by PARL, which is assisted by HAX1 (Chao et al. 2008), and phosphorylation by PINK1 (Plun-Favreau et al. 2007) modulate the activity and balance the apoptotic role of HTRA2 with its pro-proteostatic function.

Yme1, an ATP-dependent metalloprotease belonging to the AAA family of ATPases, is involved in the turnover of defective small Tims, as mentioned above (Figure 4B). Another likely target are misfolded or non-integrated carrier proteins, since overexpression of Yme1 in yeast is able to counteract the proteotoxicity induced by overexpression of a misfolding mutant of the IMM ADP/ATP carrier protein Aac2 (Aac2-A128P) (Coyne and Chen 2019) and deletion of Yme1 leads to accumulation of TIM22 substrates (Kumar et al. 2023). It was also reported that Yme1 in yeast is required to prevent aggregation of IMS-targeted DHFR and may help to fold or degrade substrates of the disulfide relay (Schreiner et al. 2012). Mammalian YME1L was shown to play a prominent role in rewiring the mitochondrial proteome during physiological transitions. In this function, it acts downstream of mTORC1 signaling, which induces changes in the lipid composition of the IMM. In particular, mTORC1 activity decreases phosphatidylethanolamine (PE) levels by promoting LIPIN1 activity, which shifts lipid metabolism away from PE synthesis. Reduced PE levels impair YME1L activity and consequently, mTORC1 indirectly inhibits YME1L (MacVicar et al. 2019). Conversely, under stress conditions, YME1L degrades the TIM23 complex subunit TIMM17A (Kan et al. 2024; Rainbolt et al. 2013) and also the TIMM23 subunits of unoccupied IMM translocases (Hsu et al. 2025). This function of YME1L is under the control of a regulatory axis involving prohibitins, which are generally required for the stability of TIMM23, and a protein called OCIAD1. OCIAD1 blocks degradation of TIMM17A by YME1L, whereas its absence leads to a shift in TIMM23 composition towards the alternative partner TIMM17B (Elancheliyan et al. 2024). Depletion of YME1L in human cells interferes with reshaping of the mitochondrial proteome under hypoxia, and hence, during many processes including development as well as stem cell maintenance and differentiation (Ohba et al. 2020). In sum, Yme1/YME1L has a dual role in the regulation of mitochondrial protein import, via degradation of sTIMMs and TIMM17A, adding to its function as a quality control protease.

Overlapping activity with M-AAA protease 1 (OMA1) is a Zn²⁺-dependent metallo-endopeptidase in the IMM. Together with YME1L, OMA1 is involved in the regulation of the protein optic atrophy 1 (OPA1), which in turn determines mitochondrial morphology via fusion and fission mechanisms (Alavi 2021; MacVicar and Langer 2016). Through this substrate, OMA1 is crucially important for reshaping of the mitochondrial network during metabolic transitions (Kawano et al. 2023), in the context of stress responses (Gilkerson et al. 2024) and in the context of diseases including neurodegeneration (Bertholet et al. 2016). OMA1 is dormant under physiological conditions but becomes activated upon mitochondrial stress, such as loss of membrane potential or excessive ROS and in such circumstances, helps to degrade arrested import intermediates (Krakowczyk et al. 2024). OMA1 also cleaves the signaling peptide DELE1, the initiator of the OMA1-DELE1-ISR axis (Fessler et al. 2020), as further described below. There is a reciprocal inhibitory relationship between OMA1 and YME1 (Rainbolt et al. 2016), both of which process OPA1 and therefore jointly control mitochondrial dynamics. Whether this interaction also extends to the involvement of these proteases in the induction of mitochondrial stress responses has not been determined to our knowledge.

Taken together, IMS-localized or -adjacent chaperones and proteases play important roles during the import and processing of mitochondrial proteins (Figure 4A and B). They safeguard the folding of resident proteins and degrade misfolded ones, very similar to the proteostasis systems in other compartments of the cell (Figure 1). However, in contrast to the perhaps more dedicated function of the cytosolic proteostasis machinery, IMS chaperones and proteases are tightly connected to a network of factors that control mitochondrial dynamics and energy metabolism, feed back into mitochondrial protein import and regulate mitochondria-specific stress responses. The final part of this review will focus on the different principles that allow mitochondrial stress to be sensed and the models that have been used to delineate these responses, with an emphasis on proteostasis defects originating from or arising in the IMS. It will become clear that the network character of the factors noted above can serve as a connected system of sensors that elicits a common response but that it may also impede a clear differentiation between direct and indirect responses to specific perturbations.

5 Distinct principles underly a range of stress response mechanisms originating from mitochondria

A plethora of studies have been conducted to identify stress response mechanisms that sense and counteract different forms of mitochondrial functional disturbances. It is becoming increasingly clear that precise definitions of the initiating factors, the sequence and timing of the cascade of events, as well as the scope and modularity of the downstream response will be important to fully understand the complexity of cellular reactions to mitochondrial stress. Central to the idea of mitochondria-specific stress responses is that they would lead to the expression of specific factors that support proteostasis within the organelle (Figure 1, step 7). Conversely, a common response to many stresses is mediated by the integrated stress response (ISR), which attenuates cap-dependent translation via phosphorylation of the initiation factor eIF2α (Pakos-Zebrucka et al. 2016). Mitochondrial stress can trigger the ISR in different ways and it will be of particular interest to dissect the pathways leading to a specification to counteract compartment- and situation-specific challenges. Of note, we will not extensively discuss mitochondrial remodeling and mitophagy, since these have been reviewed elsewhere (Picca et al. 2023; Onishi et al. 2021; Uoselis et al. 2023; Xian and Liou 2021). We also largely neglect apoptotic processes, although MOM permeabilization leads to the engagement of pro-apoptic IMS proteins (Scarlett and Murphy, 1997; van Gurp et al., 2003; Liu et al., 2025). Instead, we focus on the mechanisms that aim to preserve IMS proteostasis and refer the reader to recent reviews on apoptotic processes involving mitochondria (Dadsena et al. 2021; Glover et al. 2024; Nguyen et al. 2023).

The central junction of many stress signaling pathways is the ability of the mitochondria to properly import proteins. This principle was first demonstrated with regard to the C. elegans UPR^mt, which is induced by deficient import of the matrix protein ATFS-1 (Nargund et al. 2012; Rolland et al. 2019). Stresses which impede the import process by various means, such as depolarization of the IMM, reduction of ATP supply to the PAM module of the TIM23 translocase complex or blocking of import channels by misfolded proteins, indeed trigger stress responses (reviewed for example in (Mukhtar et al. 2023)). Consistently, stress-sensing mechanisms were shown to employ “sentinel import substrates”, as exemplified by PINK1 (Narendra and Youle 2024) and DELE1 (Fessler et al. 2020). PINK1 senses reductions in ΔΨ, since an electrophoretic force is required to sufficiently facilitate pulling of PINK1 towards the matrix to allow PINK1 cleavage by PARL and subsequent degradation after retro-translocation to the cytosol. Hence, loss of ΔΨ induces mitophagy via stabilization of PINK1 in the OMM (Figure 5A) (Jin et al. 2010; Narendra et al. 2010; Tanaka 2020). Other pathways to activate mitophagy may sense inefficient PAM function, linking mitophagy to decreased ATP levels (Michaelis et al. 2022). Both PINK1 as well as DELE1 are expressed at very low levels, with a reported copy number of less than 500 in HEK293 cells for PINK1 (Wiśniewski et al. 2014). They also share the ability to bridge the TOM and TIM import pore. In the case of PINK1, a tethered supercomplex is established in stress conditions, in which the TIM23 channel protects the sentinel from cleavage by OMA1 (Akabane et al. 2023; Eldeeb et al. 2023).

Figure 5:

Proteotoxic stress signaling by mitochondria follows different principles. (A) The functionality of the import process is under surveillance by sentinel substrates. Defects in the import of both PINK1 and DELE1 lead to stress signaling. Both proteins can sense problems at different stages of the import process (see text), which results in sensitivity to stresses including IMM depolarization, excessive ROS production and the accumulation of unfolded or aggregated proteins. Some of the processing factors for both PINK1 and DELE1 are localized in the IMS, likely resulting in specific sensitivity to perturbations in this subcompartment. (B) As a downstream consequence of deficient import, the accumulation of mitochondrial precursors in the cytosol can induce stress responses. (1) Various perturbations were reported to result in mitochondrial precursors accumulation stress (mPOS), including inhibition of the matrix Hsp90 protein TRAP1 but also of the IMS protease HTRA2, both of lead to accumulation of matrix precursors in the cytosol, indicating defective import through TOM-TIM23. The response to mPOS, to the extent that it is mediated by HSF1, requires the release of ROS to activate the HSP40 family co-chaperone DNAJA1. (2) A similar response might be triggered by inhibition of import through the TIM22 pathway, which affects the strongly hydrophobic MCF group of IMM proteins. (3) Inactivation of Mia40 led to the accumulation of disulfide relay precursors in the cytosol and activated a response termed UPRam but did not trigger the UPR^mt (see main text). (C) Direct reactions to unfolded or aggregated proteins in mitochondria have been reported, including (1) activation of a PKR/JNK axis upon unfolding of matrix-localized OTCΔ and (2) recruitment of the transcription factor ROX1, which modulates mitochondrial translation following inhibition of MPP. (3) TRAP1 induces matrix aggregates, (4) which induce LONP1 to degrade the RNA-processing factor MRPP3. Similar to ROX1, this response modulates mitochondrial gene expression to support organellar proteostasis. A potential effect of protein unfolding stress in the matrix on IMS proteostasis results from the mosaic-like assembly process of the ETC. Specifically, aggregates may interfere with ETC assembly, thereby trapping unassembled ETC precursors in the IMM, which are partially exposed to the IMS. Whether unfolding proteins in the IMS can be sensed by similar mechanisms remains unknown.

DELE1, in contrast to PINK1, is sensitive to cleavage by stress-activated OMA1 (Guo et al. 2020) (Figure 5A). This sentinel displays slow import kinetics, apparently mediated by a stop transfer signal-like hydrophobic patch following the MTS, which provide the opportunity for cleavage in the IMS-exposed region. There may be a number of DELE1-bridged import complexes in the steady-state (Fessler et al. 2022). On the other hand, lateral release of DELE1 into the IMM is also conceivable and would release the TIMM23 channel while the C-terminal part of DELE1 still occupies a TOM complex. Interestingly, this exact configuration was recently reported for a clogger constructed with a bipartite MTS, which leads to a degradation of unoccupied TIM23 channels by YME1L (Hsu et al. 2025). Cleavage of DELE1 by OMA1 can be induced by reducing ATP supply to PAM through treatment with the ATP synthase inhibitor oligomycin A. It is also induced by dissipation of ΔΨ by treatment with the ionophore CCCP and by forcing the accumulation of cleaved presequences in the matrix (Fessler et al. 2022). How OMA1 itself is activated has not been addressed in this context but has previously been linked to IMM hyperpolarization upon oligomycin A treatment (Baker et al. 2014), oxidative stress (Garcia et al. 2018) or a combination of both (Fogo et al. 2024). After cleavage by OMA1, a short fragment of DELE1 (S-DELE1) is released into the cytosol and, following oligomerization (Yang et al. 2023), activates the ISR via binding and activation of heme-regulated kinase (HRI) (Guo et al. 2020). Other forms of DELE1 can also induce ISR activity if they fail to be imported. For example, full length DELE1 (L-DELE1) accumulates in TOM40-deficient cells and binds to HRI in the cytosol. Similarly, prolonged depolarization was also suggested to lead to cytosolic localization of L-DELE1, which induces the ISR (Fessler et al. 2022). Interestingly, DELE1 may also be cleaved by the IMS protease HTRA2 and inhibition of protein import by depletion of the TIMM23 subunit of the TIM23 complex seems to be an alternative way to activate DELE1 (Bi et al. 2023). This suggests that DELE1 signaling may be more complex and could involve yet undefined IMS-localized processes.

Deficient import has at least two further consequences that may also elicit stress signaling: First, the accumulation of mitochondrial proteins in the cytosol, often termed mPOS (mitochondrial precursor overaccumulation stress) (Figure 5B), and second, imbalances in the stoichiometry of multi-protein complexes, especially the ETC (Houtkooper et al. 2013). While responses to mPOS have been intensively investigated in recent studies (Coyne and Chen 2018; Coyne et al. 2023; Wang and Chen 2015; Weidberg and Amon 2018; Wrobel et al. 2015), the danger imposed by the loss of ETC complex stoichiometry has not been addressed directly (Eckl et al. 2021). In yeast, mitochondrial translation is tightly controlled and coordinated by cytosolic factors to coordinate ETC assembly, while in mammals, a larger excess of unassembled, nuclear-encoded ETC subunits in the mitochondrial matrix may be tolerated (Kummer and Ban 2021). Nonetheless, mitonuclear coordination of gene expression certainly is crucial for cellular and organismal health and the underlying mechanisms remain to be fully unravelled (Kim and Lee 2024; Papier et al. 2022).

mPOS was initially identified by the overexpression of a misfolding mutant of the yeast IMM ADP/ATP carrier protein Aac2 (Aac2-A128P) but is also induced by the wildtype version of Aac2 (Wang and Chen 2015). This phenomenon is not limited to substrates of TIM22. Instead, multiple perturbations that saturate or impair import pathways have been demonstrated to induce mPOS, including overexpression of IMM proteins with an α-helical stop-transfer signal or matrix-targeted precursors (Weidberg and Amon 2018). This indicates that sorting at the IMM and in the IMS are major bottlenecks of mitochondrial protein import (Eckl et al. 2021). Several factors may determine the response to mPOS, including the amount of mis-localized proteins, but also their proteotoxic potential, such as their hydrophobicity and tendency to nucleate aggregation. Many stresses may induce mPOS downstream of a perturbation of mitochondrial import. Without specification of the exact mechanism, such responses have been observed across eukaryotic kingdoms (Tran and Van Aken 2020) and are often referred to collectively as UPR^mt. However, import stress can arguably be sensed primarily by a more direct mechanism, as exemplified by the sentinel substrates PINK1 and DELE1, raising the question whether mPOS is a trigger of any independent response.

Indeed, the existence of a targeted response to mPOS has gained support recently (Figure 5B). Specifically, it was demonstrated that import precursors, which accumulate upon inhibition of the mitochondrial HSP90 chaperone TRAP1, can engage and sequester cytosolic HSP70, thereby releasing heat shock factor 1 (HSF1). This response was dependent on the release of excessive mitochondrial ROS to activate the HSP70 co-chaperone DNAJA1 and was blunted by co-treatment with antioxidants (Sutandy et al. 2023). This mechanism constitutes a redox-sensitive module that directs cytosolic proteostasis systems towards mitochondrial precursors and elicits a response that supports compartmental proteostasis by expression of mitochondrial chaperones. It will be interesting to ask whether unfolding of cytosolic proteins combined with excess ROS would elicit the same response. Conversely, is there an intrinsic specificity of DNAJA1 to recognize mitochondrial proteins? It is important to note that similar responses were elicited by the inhibition of the matrix protease LONP as well as the IMS-localized protease HTRA2 (Sutandy et al. 2023). These treatments may have caused accumulation of unfolded proteins within the matrix or IMS, respectively, although this was not directly shown. In any case, intra-mitochondrial accumulation of misfolded proteins was not required to activate HSF1. A combination of antimycin A and oligomycin A to induce ROS release and inhibit ATP synthase, respectively, induced the same response without causing mitochondrial aggregate formation (Sutandy et al. 2023). Simultaneously, intra-mitochondrial unfolding due to protease inhibition as well as excessive ROS production may well have impacted protein import, leading to activation of the DELE1-HRI axis (see above). Consistently, ISR targets ATF4 and CHOP were induced earlier than mitochondrial chaperones, even though neither ATF4 nor CHOP were required for the HSF1 response. Hence, the authors concluded that the mPOS/ROS-DNAJA1-HSF1-axis was independent of ISR activation.

Similar to mPOS, cytosolic localization of IMS precursor proteins can be triggered by interference with their import into mitochondria. This was first reported in yeast cells carrying a temperature-sensitive variant of Mia40 (mia40ts). When Wrobel and colleagues shifted these cells to the restrictive temperature, they observed increased activity and expression of the cytosolic proteasome and decreased expression of factors involved in cytosolic translation. This transcriptional and proteasomal response was termed “UPR activated by mistargeting of proteins” (UPRam) (Wrobel et al. 2015). These results support the differentiation of an import stress-induced ISR (ISR^mt), a protein misfolding and ROS-induced UPR^mt and an apparently “milder” form of overaccumulation stress resulting in the UPRam without strictly excluding other mechanisms leading to the same responses. Moreover, it is noteworthy that activation of HSF1 can be triggered by alternative mechanisms originating from mitochondria (Labbadia 2023), which collectively may be referred to as the mitochondrial heat stress response (HSR^mt) or mitoHSR (Fieler and Jae 2024).

How the ISR specifically reacts to mitochondrial stress has not been completely resolved (Anderson and Haynes 2020; Bilen et al. 2022; Mick et al. 2020). A role in this specification has been attributed to the transcription factors CHOP (Münch and Harper 2016; Quirós et al. 2015; Zhao et al. 2002) and ATF5 (Fiorese et al. 2016), based on the occurrence of binding elements in promoters of mitochondrial proteostasis factors. However, these factors are also involved in the general ISR and other factors, such as the major ISR-induced transcription factor ATF4, also regulate CHOP and ATF5 targets (Quirós et al. 2015). Hence, the question arises whether the ISR is specifically tuned to or rather augmented by specific reactions to mitochondrial stress. Certainly, the ISR^mt is important to resolve mitochondrial stress by reducing the burden to its import and proteostasis machinery. A cross-talk from the ISR to PINK1 may suppress mitophagy to allow for an attempt to rescue mitochondrial function if import efficiency can be restored (Yang et al. 2024). However, continuous ISR activity induced by mitochondrial dysfunction may be a pathological feature in diseases including amyotrophic lateral sclerosis (Landry et al. 2024).

It seems that many perturbations will result in ISR^mt as well as UPRam, UPR^mt or HSR^mt activation and it will be important to clarify precisely how import or other mitochondrial functions are affected in each case. This is complicated by the intricate connections between import and mitochondrial function, as described in the sections above. Regarding unfolded proteins in particular, it is thus far unclear how they precisely interfere with the import process. Further, it thus far remains unclear whether unfolding itself can be sensed and serve as a signaling inducer. This might be expected by analogy with proteostasis surveillance mechanisms at work in the ER, where saturation of the chaperone BiP/Grp78 triggers the UPR by releasing the BiP-dependent inhibition of IRE1, PERK and ATF6 (Preissler and Ron 2019; Hetz et al. 2020; Wiseman et al. 2022). Indications that there are pathways that specifically recognize unfolded proteins can be found in the literature. In a seminal study, the expression of a misfolding mutant of the matrix protein ornithine transcarbamylase (ΔOTC) (Zhao et al. 2002) was reported to induce UPR^mt targets via a PKR-JNK2 axis (Horibe and Hoogenraad 2007) (Figure 5C). However, this response also involved ISR activation, and it is not clear whether JNK2 activation represents a specific reaction to intramitochondrial unfolding. Subsequently, Harper and Münch observed an intra-mitrochondrial response to TRAP1 inhibition resulting in reduced mitochondrial translation via degradation of the RNA-processing factor MRRP3 (Münch and Harper 2016). The group of Vögtle employed a temperature-sensitive allele of the catalytic MPP subunit Mas1 (mas1 ^ts ) and found that this induced unfolding in the yeast mitochondrial matrix without affecting protein import. Mas1 dysfunction resulted in the recruitment and mitochondrial import of the transcription factor Rox1, which binds mtDNA and performs a TFAM-like function critical for mitochondrial transcription and translation, and matrix proteostasis (Poveda-Huertes et al. 2017).

Taken together, three principles to induce responses to mitochondrial stress emerge from these studies: (1) Sensing of the import process (Figure 5A), (2) sensing of cytosolic accumulation of mitochondrial proteins and (3) direct recognition of unfolded proteins. This categorization certainly fails to comprehensively describe all potential mechanisms. Importantly, the role of ROS signaling and sensing in these categories seems auxiliary, whereas it most likely constitutes a major factor to induce or modulate these responses. We will discuss its roles in more detail in the context of IMS-related models below. Nevertheless, the principles described above will help to better differentiate stress responses in a range of model cases, as we will see next.

6 Models to study IMS proteostasis

For the IMS, a number of model proteins have been employed to study the response to proteotoxic stress (Figure 6). One of them is endonuclease G (EndoG), an IMS protein that is involved in the induction of apoptosis. A first study with this protein indicated that the proteasome is involved in the degradation of surplus IMS proteins, regardless of their folding state (Figure 6A) (Radke et al. 2008). This is consistent with the “mild” response to Mia40 dysfunction, termed UPRam (see (Wrobel et al. 2015) and above). When the proteasome was inhibited, HTRA2/Omi cleavage of EndoG increased. Interestingly, a recent study indicated that HTRA2 is retained in the cytosol when mitochondrial import is perturbed by treatment with oligomycin A or depletion of TIM23 (Bi et al. 2023). Thus, cytosolic activity of HTRA2 might constitute another example for stress-induced degradation processes occurring at the MOM (den Brave et al. 2021) akin to the “stress variant of mitoTAD” (Mårtensson et al. 2019) - the lipid oxidation-induced recruitment of Vms1 and its partner Cdc48 in yeast (Nielson et al. 2017).

Figure 6:

IMS-specific proteostasis defects have been studied with various models. (A–C) Misfolding-prone variants of IMS-localized proteins as well as artificial constructs, which target misfolding protein domains to the IMS have been employed in a number of studies. (A) The expression of the EndoG-N174 variant in particular led to the proposition that estrogen receptor alpha (ERα) signaling is activated by unfolded proteins in the IMS and induces the expression of compartment-specific proteostasis factors including HTRA2. This pathway proceeds via the AKT kinase and requires excessive ROS production and release. (B) Exogenous CHCHD2-T61I precipitates within the IMS and co-precipitates endogenous CDCHD2 and its interaction partner CHCHD10. Collectively, this results in excessive production of ROS and induction of apoptosis. (C) The expression of misfolding domains, fused to the IMS-targeting signal of cytochrome b ₂ (Cyb ₂ SS) induces the expression of components of IMS-import pathways as well as mitochondrial translation components. (D) Deletion of the IMM-localized protease YME1L compromises the clearance of aberrant IMS proteins and simultaneously affects mitochondrial dynamics via its reciprocal interaction with OMA1. (E) Chemical inhibition of the disulfide relay, via interference with ALR function, was reported to induce the ISR^mt.

Using the misfolding protein variant EndoG-N174A, the group of Germain further aimed to detect specific responses to misfolding stress within the IMS. They surmised that estrogen receptors (ER), which were found to localize to mitochondria in breast cancer cells, might be involved in sensing unfolded proteins. Consistent with this hypothesis, they detected ROS-dependent phosphorylation of ERα by AKT upon expression of EndoG-N174A (Papa and Germain 2014). This resulted in elevated levels of the IMS protease HTRA2/Omi and of NRF1, which increases mitochondrial biogenesis. Further evidence that ERα is an important signaling factor upon IMS stress is provided by the observation that ERα-deficient cells mount a more pronounced response, involving CHOP (Papa and Germain 2014). This might be attributed to unresolved proteotoxicity in the IMS, which could interfere with the import process leading to ISR^mt activation and excess ROS release by yet unknown mechanisms.

Stress responses to ROS have been extensively studied and include stabilization of the transcription factor NRF2 after inactivation of KEAP1 (Adinolfi et al. 2023; Liu et al. 2021; Suzuki et al. 2023). In the context of intra-mitochondrial protein unfolding, ROS increase was shown to activate the deacetylase SIRT3, which subsequently elicits an antioxidant response via FOXO3A deacetylation and the induction of SOD2 and catalase (Kim et al. 2016; Münch 2018; Papa and Germain 2014). Similar to EndoG-N174A in the IMS, proteotoxic folding stress in the mitochondrial matrix also induces excess production of ROS, as discussed above in the context of TRAP1 inhibition. A range of studies investigated the relationship between disease-associated aggregators such as α-synuclein and amyloid β peptides, which are able to enter the mitochondrial matrix, and oxidative stress. However, the mechanisms by which these proteins induce ROS production are not restricted to mitochondrial processes and include factors such as dysregulated Ca²⁺ influx into mitochondria, among others (Abramov et al. 2017, 2020; Gregersen and Bross 2010). While the production of ROS in mitochondria is not limited to the ETC but includes processes such as fatty acid oxidation, the role in ROS production and the sensitivity of its assembly process seem to make the ETC an easy target for unfolding stress (Moretti-Horten et al. 2024; Needs et al. 2021). For example, unfolding of adenine nucleotide translocase 1 (ANT), the human homolog of Aac2, was reported to aggregate other IMM proteins, potentially including ETC complexes. Consistently, ETC supercomplexes are lost upon expression of Aac2 mutants in yeast (Liu et al. 2015). Furthermore, misfolded ANT could also titrate IMM chaperones away from their roles in ETC assembly, as reviewed in (Coyne and Chen 2019). However, a complete picture of how unfolded proteins in mitochondria induce ROS production seems to still be lacking.

Vice-versa, the impact of ROS on protein folding inside mitochondria is also not yet evident. For substrates of the disulfide relay in particular, the levels of ROS and the IMS redox state are an important determinant of correct folding (Habich et al. 2019b). In this context, the ability of the disulfide relay to stabilize the ETC assembly factor NDUFAF8 during its import into the matrix may be of particular relevance (Peker et al. 2023). If stabilization of NDUFAF8 fails, the assembly of complex I will be compromised. Arguably, this might constitute a “moonlighting” function of this substrate of the two-step import pathway as a sensor for the fidelity of disulfide formation and preservation.

Another model to study proteotoxic stress in the IMS employed a mutant form of CHCHD2 (Figure 6B). CHCHD2 is an IMS protein, contains a twin CX₉C motif and is imported via the disulfide relay into the IMS, where it regulates the activity of complex IV of the ETC (Straub et al. 2018). CHCHD2 forms both homodimers with itself and heterodimers with the paralogous protein CHCHD10 (Purandare et al. 2018). The formation of heterodimers is promoted in response to mitochondrial stress due to increased levels of CHCHD2 (Huang et al. 2018). Vandenberghe and colleagues have studied effects of the T61I variant of CHCHD2, which they found to precipitate inside the IMS (Cornelissen et al. 2020). Expression of this variant further induced misfolding of endogenous CHCHD2 and co-precipitated its partner CHCHD10. The accumulation of misfolded CHCHD2 T61I in the IMS led to excess ROS production and eventually to the induction of apoptosis. Interestingly, a double knockout of CHCHD2 and 10 in mice displayed mitochondrial phenotypes, with disrupted cristae structures due to OPA1-cleavage by OMA1 and activation of the ISR^mt (Liu et al. 2020). More recently, CHCHD2 and 10 were shown to directly interact with both YME1L and OMA1 and when accumulation of these proteins in the cytosol was enforced by CCCP treatment, they directly interacted with eIF2α (Ruan et al. 2022). Therefore, CHCHD2 and 10 may mediate functional connections between ETC activity, the disulfide relay and OMA1 as well as the ISR^mt. Furthermore, nuclear-localized CHCHD2 was shown to play a role in the response to hypoxia (Aras et al. 2015; Kee et al. 2021; Ikeda et al. 2022). Further mechanistic exploration of these functional connections seems promising.

In another attempt to decipher IMS-specific stress responses, the Mapa group engineered misfolding domains to the cytochrome b ₂ signal sequence (Cyb ₂ SS) to achieve IMS targeting (Narayana Rao et al. 2022) (Figure 6C). Overexpression of these constructs in yeast cells resulted in significant growth defects, indicating severe IMS stress. Through transcriptomic and proteomic surveys, they observed specific expression of IMS-specific chaperones, the yeast Mia40 partner Erv1 and TOM complex components as well as components of the mitochondrial ribosome, which were not induced upon matrix-localized unfolding of similar constructs (Narayana Rao et al. 2022). The upregulation of mitochondrial import and translation is consistent with increase of NRF1 observed upon expression of EndoG-N174A (see above) and suggests that in contrast to unfolding stress in the matrix stress, proteotoxicity in the IMS tends to upregulate compensatory mitochondrial biogenesis. However, the precise signaling events leading to this response remain obscure.

The deletion of YME1L significantly compromises the proteostatic function of the IMS (Figure 6D). Without YME1L, clearance of aberrant proteins is severely impaired, resulting in the accumulation of protein aggregates (Baker et al. 2012; Cesnekova et al. 2018; Kan et al. 2022; Potting et al. 2010; Stiburek et al. 2012; Spiller et al. 2015; Wu et al. 2017). This proteostatic imbalance triggers different compensatory mechanisms, including the activation of other proteases like OMA1, which can lead to excessive processing of proteins such as OPA1, disrupting mitochondrial dynamics (Anand et al. 2014; Ishihara et al. 2006; Ohba et al. 2020). Furthermore, the loss of YME1L’s proteolytic activity affects the mitochondria’s capacity to adapt to various stressors and to be remodeled in physiological transitions as described above. It may also affect mitochondrial lipid homeostasis (MacVicar et al. 2019), further exacerbating the organelle’s vulnerability to stress. Ultimately, the absence of YME1L’s proteostatic function renders mitochondria less resilient and more susceptible to dysfunction.

Another option to specifically interfere with protein import into the IMS is to apply chemical inhibitors of the disulfide relay pathway (Figure 6E). In this respect, the recently developed compound MitoBlock-6, which targets the MIA40 partner ALR (Dabir et al. 2013), has enabled a new type of experiment, in which the import of IMS proteins can be acutely inhibited. While more studies are expected to be conducted using this or similar compounds in the near future, it was already shown that MB6 treatment induces the DELE1-HRI axis (Bi et al. 2023).

Taken together, the models employed to elicit proteotoxic stress in the IMS – HTRA2 inhibition, overexpression of misfolding-prone proteins, YME1 KO and chemical inhibition of import processes – indicate that IMS-specific responses to proteotoxicity can be elicited. However, they have so far failed to reveal the precise sequence of events and the specific signals that dictate these responses and differentiate them from stress originating from other compartments such as the IMM and the matrix.

7 Some open questions

Many recent studies widened our knowledge on IMS proteostasis. Still numerous open questions remain. These include fundamental questions such as why so many IMS proteins contain disulfide bonds. For comparison, the introduction of disulfide bonds in the ER plays a crucial role in enhancing protein stability but also in monitoring the protein folding pathway and ensuring proper quality control before allowing secretion of proteins from the ER (Riemer et al. 2009; Oka and Bulleid 2013). The increased stability is particularly critical for proteins destined for the harsh extracellular environment, where they face various stressors such as proteases, pH changes, and oxidative conditions without having a protective proteostasis network around them. As we described here, the IMS also constitutes a biophysically challenging environment, which might explain the need for disulfides. Another indication that the IMS indeed faces severe proteostatic challenges is the extent of its proteostasis network as described above, which seems exuberant given the rather low number of around 150 soluble IMS proteins (Zarges and Riemer 2024). It will be of critical importance to delineate which classes of proteins constitute a burden to IMS homeostasis. Certainly, this may vary depending on the nature of the proteotoxic stress but some protein classes, such as hydrophobic carriers or disulfide-containing proteins may have a higher propensity to cause problems.

Moreover, for many stress-response pathways, the precise mechanisms of signaling remain unclear, including which disturbance in the IMS is precisely sensed and how the signal is conveyed from the IMS to the cytosol. An open question in this regard remains whether unfolded proteins in the IMS can be directly sensed to elicit a stress response, akin to the sensing of unfolding in the ER by the BiP chaperone. It is tempting to speculate that one of the chaperone or protease systems, e.g. CLPB or the SPY complex, might react to an accumulation of unfolded proteins. This might simply occur via saturation of its activity, which is then translated into release of an otherwise suppressed substrate. Interesting candidates for this kind of mechanism could be CLPB and a potential regulation of HTRA2, since the latter has already been implicated in the cleavage of DELE1. Further, the reciprocal regulation between YME1L and OMA1 in the cleavage of OPA1 also appears like a blueprint for a regulatory axis that might respond to YME1L saturation by unfolded IMS proteins.

The integration but also the distinction of signaling pathways responding to different stresses in different mitochondrial sub-compartments remains likewise poorly understood.