The mitochondrial unfolded protein response: acting near and far

1 Introduction

Protein homeostasis (proteostasis) is constantly undermined by extrinsic stressors and potentially harmful by-products of cellular metabolism, such as reactive oxygen species (ROS). The paramount importance of maintaining a healthy protein pool is highlighted by the significant fraction of the proteome that is devoted to protein surveillance across species (Muller et al. 2020). An extensive plexus of chaperones and the proteolytic degradation machinery, coordinated by stress response pathways, collectively referred to as the proteostasis network (PN), safeguards proteostasis (Hipp et al. 2019). Notably, the efficacy of the PN declines with age, leading to the accumulation of misfolded proteins, toxic oligomers and protein aggregates, culminating in proteotoxicity. Post-mitotic cells, such as neurons, are particularly susceptible to protein aggregation and PN dysfunction. Intensive scientific efforts have been focused on slowing PN decline to mitigate late-onset neurological disorders.

Mitochondria are the result of endosymbiotic events between ancestral eukaryotic cells and free-living proteobacteria. They are central to cellular metabolism, producing ATP via oxidative phosphorylation (OXPHOS) and being involved in processes such as the TCA cycle and the beta-oxidation of fatty acids, but are also critical for the production of essential cofactors and regulatory metabolites (Suomalainen and Nunnari 2024). Mitochondrial dysfunction is a key hallmark of aging and is associated with the manifestation of a wide spectrum of human pathologies affecting the muscular, neuronal and immune systems (Lopez-Otin et al. 2023). Sophisticated quality control and protein turnover mechanisms (i.e. chaperones, proteases, mitochondrial-associated degradation) maintain protein integrity in various mitochondrial compartments, while others ensure that irreversibly damaged or superfluous mitochondria are removed by autophagic degradation (i.e. mitophagy) (Ng et al. 2021). The vast majority (more than 99 %) of mitochondrial proteins are encoded by the nuclear genome, translated by cytosolic ribosomes and then imported into the mitochondria. Therefore, any change in the mitochondrial status should be communicated to the nucleus so that the mitochondrial network can successfully adapts to ever-changing physiological demands and functionally recover from stress. Here, we review the mechanisms of mitochondrial unfolded protein response and describe paradigms of cell-nonautonomous induction of the mitochondrial stress response.

2 The mitochondrial unfolded protein response

The mitochondrial unfolded protein response (UPR^mt) is a stress response pathway, in which mitochondrial stress is sensed by the nucleus, and induces the expression of mitochondrial proteases and chaperones to restore mitochondrial proteostasis and promote cell recovery and survival. A variety of defects, such as OXPHOS dysfunction, inhibition of mitochondrial translation, dissipation of mitochondrial membrane potential, as well as blockade of mitochondrial protein import and mitochondrial protein misfolding have been shown to activate the UPR^mt. This adaptive response was first described in mammalian cells, where depletion of mitochondrial DNA (mtDNA) following treatment with ethidium bromide or overexpression of a mutant, aggregation-prone variant of the matrix mitochondrial protein ornithine transcarbamylase was shown to induce the expression of mitochondrial chaperones encoded by the nuclear genome (Martinus et al. 1996; Zhao et al. 2002). UPR^mt was later characterized and studied in the model nematode Caenorhabditis elegans, and key factors involved in UPR^mt regulation were elucidated.

Central to UPR^mt activation is the basic leucine zipper (bZIP)-activating transcription factor associated with stress-1 (ATFS-1). The ATFS-1 protein sequence contains both a mitochondrial localization signal (MLS) and a nuclear localization signal (NLS). This allows ATFS-1 to shuttle between the mitochondria and the nucleus, with its abundance in these two compartments serving as an indicator of mitochondrial proteostasis. Under basal conditions, ATFS-1 is imported into mitochondria and degraded by the ATP-dependent protease LONP-1 (LONP1 in mammals) (Nargund et al. 2012). However, when mitochondrial import efficiency is decreased by stress, ATFS-1 cannot enter mitochondria as efficiently due to its weak MLS. This leads to its stabilization and subsequent accumulation of a significant fraction of ATFS-1 in the nucleus. The MLS of ATFS-1 acts as a sensor that responds to changes in mitochondrial protein import efficiency. Substitution of the endogenous MLS of ATFS-1 with a counterpart of a higher net charge resulted in the inability of the chimeric protein to fully induce the expression of nuclear-encoded mitochondrial chaperones following mitochondrial stress (Rolland et al. 2019).

Once in the nucleus, ATFS-1 orchestrates a multifaceted transcriptional response program. Upon binding to its consensus regulatory sequence found in numerous promoters, it activates transcription of genes encoding mitochondrial chaperones, detoxification enzymes, mitochondrial nuclear protein import components, and innate immune proteins, among others (Nargund et al. 2012, 2015; Pellegrino et al. 2014). Several glycolytic enzymes are also upregulated, indicating a switch from OXPHOS to anaerobic glycolysis to meet cellular energy demands. In contrast, other genes, such as those encoding components of the electron transport chain (ETC), are downregulated by ATFS-1, limiting the accumulation of nuclear-encoded OXPHOS proteins. Within the nucleus, ATFS-1 cooperates with additional factors to induce the UPR^mt. In particular, the homeodomain-containing transcription factor DVE-1, which interacts with the small ubiquitin-like protein UBL-5, binds to the promoters of genes encoding mitochondrial chaperones upon mitochondrial stress (Haynes et al. 2007). Epigenetic regulators such as the histone lysine demethylases JMJD-1.2 (PHF8 in mammals) and JMJD-3.1 (JMJD3 in mammals) and the histone acetyltransferase CBP/p300 are involved in chromatin remodeling to induce transcription of UPR^mt-associated genes (Li et al. 2021; Merkwirth et al. 2016). In addition, the key UPR^mt transcription factors ATFS-1 and DVE-1 are post-translationally modified by covalent attachment of the small ubiquitin-like modifier (SUMO), which regulates their stability and subcellular localization, respectively (Gao et al. 2019). Removal of SUMO marks by the SUMO-specific peptidase ULP-4 is essential for full UPR^mt activation (Gao et al. 2019). Another mode of regulation is exerted by ZIP-3, which is a direct transcriptional target of ATFS-1 but acts as its negative regulator. ZIP-3 probably heterodimerizes with ATFS-1 via its bZIP domain, thereby repressing its transcriptional activity and limiting UPR^mt activation (Deng et al. 2019).

Interestingly, CHIP sequencing data revealed that ATFS-1 also binds to the non-coding region (D loop) of mtDNA during mitochondrial stress (Nargund et al. 2015). Consistent with its nuclear role, ATFS-1 binding to mtDNA limits the expression of mtDNA-encoded OXPHOS proteins. Given that a stoichiometric imbalance between nuclear and mitochondrial encoded OXPHOS proteins, a condition referred to as mitonuclear imbalance, triggers the UPR^mt, coordinated regulation of the nuclear and mitochondrial genome by ATFS-1 ensures the balanced assembly of newly formed ETC complexes and full recovery of mitochondrial function (Houtkooper et al. 2013). Importantly, the finding that ATFS-1 binds directly to mtDNA suggests that its mitochondrial import is not limited to baseline conditions solely for degradation. Two recent studies have provided compelling insights into the pleiotropic functions of ATFS-1 within mitochondria. Yang and colleagues showed that OXPHOS deficiency triggers the accumulation of ATFS-1 in mitochondria, which promotes the binding of DNA polymerase γ (POLG) to mtDNAs, leading to a subsequent increase in mtDNA content (Yang et al. 2022). Interestingly, LONP-1 also binds to the non-coding region of mtDNA and the binding sites of ATFS-1 and LONP-1 overlap. LONP-1 knockdown increased both ATFS-1 and POLG binding to mtDNA, suggesting that LONP-1 antagonizes ATFS-1 for mtDNA binding. The above highlight that LONP-1 not only determines the abundance of ATFS-1 through its degradative function, but also its availability for mtDNA binding. In the second study, Dai et al. developed a system to introduce double strand breaks (mtDSBs) into mtDNA in vivo. They observed that mitochondria-targeted ATFS-1 could alleviate the deleterious mtDSB-associated phenotypes (Dai et al. 2023). ATFS-1 has been proposed to cooperate with components of base excision repair (BER), the major functional DNA repair mechanism within mitochondria, to limit the accumulation of DNA damage in the mitochondrial genome. Specifically, the authors showed that ATFS-1 physically interacts with HMG-5, the worm homolog of mitochondrial transcription factor A (TFAM), a key component of the mitochondrial nucleoid required for initiation of mtDNA transcription, thereby preventing assembly of the mitochondrial pre-initiation complex and shifting the balance from mtDNA transcription to repair. This finding may also explain the previous observation that ATFS-1 limits the accumulation of mitochondria-encoded OXPHOS proteins during mitochondrial stress (Nargund et al. 2015). Importantly, this novel function of ATFS-1 appears to be evolutionarily conserved, as overexpression of ATF5, a putative mammalian ATFS-1 homolog (see below), also prevented the accumulation of mtDNA damage in a mitochondrial localization-dependent manner.

More complex mechanisms of mitonuclear communication have been described in mammals, where mitochondrial stress induces extensive metabolic rewiring characterized by altered one-carbon metabolism (Bao et al. 2016) and amino acid biosynthesis (Forsstrom et al. 2019; Quiros et al. 2017). ATF5 has been proposed as a functional mammalian ATFS-1 homolog because it is required for cell recovery after mitochondrial stress and its heterologous expression could rescue atfs-1 deficiency in the nematode. Notably, this property was specific for ATF5 but not for ATF4, which also shares sequence similarity with the nematode ATFS-1 (Fiorese et al. 2016). Like ATFS-1, ATF5 harbors both an MLS and an NLS, localizes to mitochondria under basal conditions, and is nuclearized upon mitochondrial stress, inducing the expression of mitochondrial chaperones and proteases encoded by the nuclear genome, thereby enhancing mitochondrial proteostasis. However, a subsequent study using a multi-omics approach identified ATF4, which is activated in the context of the integrated stress response (ISR), as a key coordinator of the retrograde mitochondrial stress response in mammalian cells (Quiros et al. 2017). The ATF4-mediated cytoprotective response to mitochondrial stress in mammalian cells involves extensive metabolic reprogramming and differs from the canonical UPR^mt transcriptional program described in invertebrate models. ISR activation leads to a decrease in global protein synthesis, while transcription factors with upstream ORFs (uORFs) in their 5′UTR, including ATF4 and ATF5, are preferentially translated (Costa-Mattioli and Walter 2020). A central event for ISR activation is the phosphorylation of eukaryotic initiation factor 2α (eIF2α) in response to various types of stress, which are mainly sensed by four kinases: GCN2 (amino acid starvation), PERK (ER stress), PKR (double-stranded RNAs), and HRI (reduced heme levels) (Taniuchi et al. 2016). However, in this initial report, single knockdown of the aforementioned kinases failed to abolish the transcriptional induction of the ATF4 target genes upon administration of mitochondrial stress (Quiros et al. 2017); thus the nature of the upstream signals driving eIF2α phosphorylation in response to mitochondrial stress remained elusive. An important advance in this field was the discovery that the mitochondrial protein DAP3 binding cell death enhancer 1 (DELE1) is proteolytically processed by the mitochondrial stress-activated zinc metallopeptidase OMA1 upon its entry into mitochondria. This leads to the accumulation of a short DELE1 fragment in the cytosol, where it oligomerizes and activates the HRI kinase (Fessler et al. 2020; Guo et al. 2020). Subsequently, HRI phosphorylates eIF2α, thereby transducing the signal from dysfunctional mitochondria to ISR activation and ATF4 translation (Figure 1). Recent evidence suggests that the silencing factor of the integrated stress response (SIFI), an E3 ubiquitin ligase complex targets cytosolic DELE1, HRI and mitochondrial protein precursors for proteasomal degradation, terminating the ISR (Haakonsen et al. 2024). Although eIF2α phosphorylation also occurs in response to mitochondrial stress in C. elegans, a direct link to the ATFS-1-mediated UPR^mt transcriptional program has not been established. ISR activation is thought to alleviate mitochondrial stress by reducing global protein synthesis, thereby preventing the excessive accumulation of mitochondrial precursors when the organelle’s protein-folding capacity is compromised (Baker et al. 2012).

Figure 1:

Overview of mitochondrial unfolded protein response signaling. (A) In C. elegans, mitochondria-nucleus communication is established by the transcription factor ATFS-1. In the absence of stress, ATFS-1 is imported into mitochondria and degraded by the mitochondria-resident protease LONP-1. When the efficiency of electrochemical potential-dependent mitochondrial import is compromised, ATFS-1 is not imported into mitochondria as efficiently, but enters the nucleus where it activates a broad transcriptional program to promote recovery of mitochondrial function. Nuclear ATFS-1 cooperates with additional transcription and epigenetic factors to activate or repress the transcription of numerous genes. Recent evidence suggests that a portion of ATFS-1 localizes to mitochondria under stress conditions, where it binds to mitochondrial DNA (mtDNA), and restricts the expression of mitochondrial-derived ETC components. (B) In mammals, ATF5 has been proposed as a functional ATFS-1 homologue. However, activation of the integrated stress response (ISR) and its downstream effector ATF4 is also critical. DELE1 processing by the mitochondrial protease OMA1 results in the release of a short DELE1 fragment into the cytosol. This binds and activates the kinase HRI, which then phosphorylates the translation initiation factor eIF2α, thereby attenuating global protein synthesis, a hallmark of ISR. However, ATF4, which has an upstream open reading frame, can evade the protein synthesis inhibition, is translated and then imported into the nucleus. HSF1 also translocates to the nucleus upon mitochondrial misfolding stress to activate the transcription of UPR^mt genes.

ATF4 and ATF5 are not the only mammalian effectors activated by mitochondrial stress. The C/EBP homologous protein (CHOP) is a key transcription factor identified in early reports describing the UPR^mt and is robustly induced upon mitochondrial stress (Zhao et al. 2002). Recent evidence suggests that CHOP acts as a rheostat, preventing excessive activation of the ATF4-regulated ISR transcriptional program, which can otherwise have deleterious consequences for cellular homeostasis (Kaspar et al. 2021). The notion that UPR^mt is merely embedded in the ISR in mammals was challenged by a thorough transcriptomic analysis, which revealed that the expression of UPR^mt-associated genes can be induced independently of the two main ISR effectors (i.e. ATF4 and CHOP) (Sutandy et al. 2023). Importantly, this study highlighted the essential role of heat shock factor 1 (HSF1), widely known as the master regulator of the heat shock response, in UPR^mt gene activation in mammalian cells. Specifically, mitochondrial misfolding stress triggers the release of mitochondrial ROS (mtROS) into the cytosol, which oxidizes DNAJA1, a co-chaperone of cytosolic HSP70. Oxidized DNAJA1 enhances the recruitment of HSP70 to cytosolic protein precursors, releasing HSF1, the binding partner of HSP70, to translocate to the nucleus and activate the transcription of UPR^mt genes (Sutandy et al. 2023) (Figure 1). These observations are consistent with a previous study showing that HSF1 occupancy at the promoters of mitochondrial chaperones is increased upon mitochondrial stress (Katiyar et al. 2020). Future research is needed to delineate the interplay between ATF4, ATF5, CHOP and HSF1 in different physiological and pathological contexts.

3 Organismal mitochondrial unfolded protein response signaling

A key feature of UPR^mt is that it can be activated systemically in response to mitochondrial stress detected in distal tissues. A seminal publication showed that knockdown of ETC components specifically in C. elegans neurons could induce the expression of nuclear-encoded chaperones in the intestine (i.e. gut), representing a bona fide paradigm of cell-nonautonomous signaling and inter-tissue communication (Durieux et al. 2011). Interestingly, this is also true for other types of neuronal mitochondrial stress. For example, expression of a polyglutamine tract of 40 amino acids (Q40), which associates with mitochondria and inhibits their function, and the mitochondria-targeted KillerRed, which generates high levels of ROS, or perturbation of neuronal mitochondrial dynamics by knocking down mitofusin, which promotes mitochondrial fusion, elicited effects similar to those of knockdown of ETC components, formulating the “mitokine” hypothesis (Berendzen et al. 2016; Chen et al. 2021; Shao et al. 2016). A thorough analysis shed light on the neuronal circuitry involved in the perception of the mitochondrial challenge and its transduction to the C. elegans intestine. CRISPR-mediated knockout of the mitochondrial protease spg-7 in three specific sensory neurons, namely ASK, AWA and AWC, and the interneuron AIA was sufficient to induce mitochondrial chaperone expression in the gut (Shao et al. 2016). This study also demonstrated that a neuroendocrine signal, specifically the neuropeptide FLP-2 released by the interneuron AIA, is necessary and sufficient for the cell-nonautonomous induction of the UPR^mt in response to neuronal mitochondrial stress. FLP-2 acts in concert with additional secreted factors to mediate neuron-intestinal communication and systemic UPR^mt induction. A genetic screen showed that a functional retromer complex, which mediates the retrieval of the Wnt secretion factor MIG-14 from the cell surface, is indispensable for cell-nonautonomous UPR^mt induction (Zhang et al. 2018). Of note, secretion of the Wnt ligand EGL-20 from serotonergic neurons is necessary and sufficient for UPR^mt induction. EGL-20 binding to the Wnt-Frizzled receptor MIG-1 in the intestine leads to the activation of the canonical Wnt pathway, via β-catenin translocation to the nucleus and the TCF/LEF transcription factor. The systemic effects of EGL-20 on UPR^mt induction could be abolished when serotonin biosynthesis or secretion was inhibited (Zhang et al. 2018). Taken together, the above studies suggest that distinct neuron-derived signals coordinate the systemic activation of the UPR^mt in response to mitochondrial stress in neurons (Figure 2).

Figure 2:

Systemic mitochondrial stress signaling. (A) In the nematode C. elegans, secretion of the neurotransmitter serotonin and the WNT family ligand EGL-20 from serotonergic neurons in response to different types of neuronal mitochondrial stress induces UPR^mt in the intestine. In addition, secretion of the neuropeptide FLP-2 from the AIA interneuron is also necessary and sufficient for cell-nonautonomous UPR^mt induction. An intact germline is essential for full somatic UPR^mt induction, whereas germline-derived hedgehog signals dampen UPR^mt during adulthood. (B) In the fruit fly D. melanogaster, mild mitochondrial stress in muscle triggers the release of the insulin antagonist peptide (ImpL2), which systemically suppresses insulin signaling. Stressed muscles also secrete activin-β (Actβ), a ligand of the TGF-β family, which binds to its cognate receptor in fat body cells, triggering mitochondrial function and lipid accumulation in the fat body. (C) In the mouse M. musculus, mitochondrial stress in muscle is associated with the secretion of the hormone FGF21, which acts in a paracrine manner to increase serine biosynthesis and the transsulfuration pathway, and in an endocrine manner to induce glucose uptake and alter mitochondrial biogenesis in the CA2 hippocampal region of the brain.

The organism-wide effects associated with mitochondrial dysfunction have also been reported in other species. In Drosophila melanogaster, mild mitochondrial stress via ETC complex I inhibition preserved muscle function during aging and extended lifespan. These beneficial effects were compromised when UPR^mt genes, such as mitochondrial chaperones or proteases, were inhibited (Owusu-Ansah et al. 2013). Intriguingly, like the nematode model, mitochondrial stress in fly muscle exerts systemic effects by inducing secretion of the insulin antagonist peptide (ImpL2), which systemically suppresses insulin signaling. It also triggers mitochondrial dysfunction and lipid accumulation in a distant tissue, the fat body, in an ImpL2-independent manner. This muscle-fat body communication axis is established by the release of the TGF-β ligand activin-β (Actβ) from muscle. The binding of Actβ to the baboon TGF-β receptor, and the concomitant activation of the Drosophila Smad2 (dSmad2) transcription factor impairs mitochondrial function and triggers lipid accumulation in the fat body (Song et al. 2017). A careful analysis of Deletor mice and human patients, which accumulate mtDNA deletions due to the ubiquitous expression of a Twinkle mtDNA helicase mutant, showed that the induction of the UPR^mt and the development of mitochondrial myopathy in the skeletal muscle and heart are the conserved consequences (Forsstrom et al. 2019). The Deletor mice show no evidence of mitochondrial dysfunction inducing UPR^mt in non-affected organs, such as the liver, arguing against the existence of a mammalian “mitokine” analog. However, this study has shown that muscle-derived fibroblast growth factor 21 (FGF21), an early and robustly upregulated metabolic hormone with autocrine and endocrine functions, promotes body weight loss and glucose uptake and remodels metabolism in favor of the transsulfuration pathway and de novo serine biosynthesis. Interestingly, the effects of FGF21 extend far beyond muscle, fostering glucose uptake and affecting mitochondrial biogenesis in the hippocampal CA2 region of the brain. Taken together, these studies demonstrate that mitochondrial dysfunction can be transmitted to distant organs via long-range secreted signals (Figure 2).

Emerging evidence indicates that the germline, the unique immortal cell lineage, is a central hub for sensing mitochondrial stress and for organism-wide activation of the mitochondrial stress response. A recently published study from our laboratory showed that a cohort of germline mutants that exhibit reproductive defects fail to induce the full UPR^mt when challenged with mitochondrial stressors (Charmpilas et al. 2024). Specifically, depletion of gametes (sperm or oocytes) impaired the animals’ ability to mount efficient UPR^mt. The use of genetic tools that allow the knockdown of genes of interest exclusively in the germline provided new insights into the interplay between germline and soma. Germline-specific knockdown of several ETC components was sufficient to induce UPR^mt in somatic tissues (Charmpilas et al. 2024; Lan et al. 2019). Interestingly, an intact germline is also essential for cell-nonautonomous UPR^mt activation in response to neuronal Q40 mitochondrial stress (Shen et al. 2024). These findings suggest that reproductive capacity is linked to the ability of somatic cells to respond to mitochondrial stress and are consistent with the disposable soma theory of aging, which predicts a trade-off between somatic maintenance and reproductive investment. Notably, germline signals affect not only the inducibility of UPR^mt in somatic tissues but also its decline with age, as germline-derived Hedgehog signals regulated by piRNAs suppress somatic UPR^mt in adult nematodes (Zhou et al. 2024). These studies highlight that, in addition to its traditional role in transmitting genetic material to offspring through sexual reproduction, the germline also releases signals that regulate the mitochondrial stress response in somatic cells. It would be interesting to test whether similar mechanisms of germline-somatic communication fine-tune mitochondrial stress signaling in mammals.

4 Future directions

The fact that a fraction of ATFS-1 binds to the mtDNA upon mitochondrial stress, thereby regulating the mitonuclear balance, suggests that the degree of mitochondrial stress or import disruption may lead to different consequences. Similar, but poorly described mechanisms that maintain coordinated expression of the nuclear and mitochondrial genomes under stress may exist in mammals.

Retrograde mitochondrial signaling allows cells to cope with and recover from mitochondrial misfolding stress. Its excessive activation can be detrimental to cellular homeostasis and has been associated with cancer cell populations with high metastatic capacity (Kenny et al. 2019). The recent identification of the silencing factor of the integrated stress response (SIFI), a specialized E3 ubiquitin ligase complex that targets HRI and DELE1 for proteasomal degradation, underscores the importance of tight control of the ISR (Haakonsen et al. 2024). Notably, mutations in UBR4, a core SIFI component, have been associated with early-onset dementia, making ISR regulation a promising therapeutic target. Other mechanisms that control the extent and duration of UPR^mt or ISR would be important for further investigation.

Although cell-nonautonomous regulation of the UPR^mt has been demonstrated in different species, the mechanisms that mediate systemic stress responses or metabolic remodeling appear to be different in worms, flies and mice. Whether there are evolutionarily conserved mechanisms for the systemic regulation of UPR^mt would be interesting for further studies.