A multilayered stress‐response circuit: The mammalian mitochondrial UPR

Abstract

Mitochondrial proteotoxic stress activates the mammalian UPR^mt through a multilayered mechanistic architecture rather than a linear pathway. At its core lies an import‐gated sensing logic: reduced preprotein import and mito–nuclear stoichiometric imbalance activates the integrated stress response (ISR) toward the translation of ATF4, CHOP, and the mitochondria‐targeted transcription factor ATF5. These factors cooperatively reprogram transcription to expand the chaperone–protease capacity while transiently reducing the nuclear‐encoded OXPHOS load. Parallel translational mechanisms that include eIF2α‐dependent repression, stress‐granule triage, and miRNA‐driven selective silencing reduce the mitochondrial precursor import and maintain proteostatic symmetry between the cytosol and mitochondria. Within the organelle, LONP1‐ and CLPP‐dependent proteolysis, mitoribosome pausing, and tRNA‐processing checkpoints further dampen nascent chain pressure. Epigenetic licensing by demethylases and acetyltransferases links metabolic and bioenergetic status to promoter accessibility at UPR^mt loci. Together, these import‐gated, translational, and epigenetic control layers form a coherent mechanistic circuit ensuring that mitochondrial recovery is matched to folding, assembly, and metabolic capacity. We propose a unified framework explaining how these layers cooperate to determine adaptive versus maladaptive outcomes.

Mitochondrial stress signals converge at three proximal hubs—import‐gated Δψ/MPIS, ER–mitochondria contact sites, and lysosomal mTORC1, feeding an ISR gate (ATF4/ATF5/CHOP) that shapes a mitochondria‐skewed UPR^mt program. The output reinforces proteostasis (HSP60/HSP10, CLPP/LONP1), routes damaged organelles to PINK1–Parkin/FUNDC1 mitophagy, or commits to apoptosis, aligning translation, import capacity, and metabolic state with cellular and systemic demands across the interconnected stress network, via redox, Ca²⁺ exchange, membrane remodeling, endocrine mitokines FGF21/GDF15 signaling.

Abbreviations

AGO2

argonaute

ATF

activating transcription factor

ATFS 1

activating transcription factor associated with stress–1

CDDO

2‐cyano‐3,12‐dioxo‐oleana‐1,9(11)‐dien‐28‐oic acid

CHOP

C/EBP‐homologous protein

CLPP/CLPXP

caseinolytic mitochondrial matrix peptidase (proteolytic subunit/chaperone subunit X)

COX

cytochrome c oxidase

CREBH

cyclic AMP‐responsive element‐binding protein H

DELE1

death associated protein 3 (Dap3) binding cell death enhancer 1

DRP1

dynamin‐related protein 1

eIF2α

eukaryotic translation initiation factor 2A

ER

endoplasmic reticulum

FUNDC1

FUN14 domain‐containing protein 1

G3BP

stress‐granule assembly factor

GCN2

general control non‐derepressible 2

H3K

histone trimethylation

HAF 1

homolog of ABC protein family‐1

HDAC

histone deacetylase

HIF

hypoxia‐inducible factor

HIMPA

hypoxia‐induced mitochondrial protein aggregation

HRI

heme‐regulated inhibitor

HSP

heat shock protein

HtrA2

high temperature requirement protein A2

IMS

intermembrane space

IP3R

inositol 1,4,5‐trisphosphate receptors

ISR

integrated stress response

JmjC

JumonjiC

JMJD

lysine‐specific demethylase

LIN

lineage‐defective protein family

LONP1

mitochondrial ATP‐dependent protease Lon

m AAA/i AAA

matrix/intermembrane ATPases associated with diverse cellular activities

mito(CPR)

compromised protein import response

MPIS

mitochondrial protein import stress

mPOS

mitochondrial precursor overaccumulation stress

MRPP

mitochondrial RNase P protein

mtDNA

mitochondrial DNA

mTORC1

mechanistic target of rapamycin complex 1

MTS

mitochondrial targeting sequence

MTS

mitochondrial targeting sequence

NLS

nuclear localization signal

NRF

nuclear respiratory factor

OMA1

overlapping with the M‐AAA zinc metallopeptidase

Parkin

Parkin RBR E3 ubiquitin protein ligase

PERK

protein kinase R‐like ER kinase

PINK1

phosphatase and tensin homolog (PTEN)‐induced kinase 1

RISC

microRNA‐induced silencing complex

SATB2

special AT‐rich sequence‐binding protein 2

SG

stress granules

SPG7

spastic paraplegia 7

TCA cycle

tricarboxylic acid cycle

TIM

translocase of the inner mitochondrial membrane

TOM

translocase of the outer mitochondrial membrane

uORF

upstream open reading frame

UPRam

unfolded protein response activated by mistargeting

UPR^mt

mitochondrial unfolded protein response

V‐ATPase

vacuolar H + ‐ATPase

YME1L

ATP‐Dependent Zinc Metalloprotease

ΔOTC

ornithine transcarbamylase

Δψ

membrane potential

Introduction

Mitochondria are multifunctional organelles that integrate metabolism, signaling and quality control pathways that safeguard mitochondrial proteostasis and are critical for cellular homeostasis and organismal physiology [1, 2, 3, 4, 5]. Their endosymbiotic origin and the dual genome of 13 hydrophobic OXPHOS subunits encoded by mtDNA together with their ~1000 nuclear‐encoded imported proteins render mitochondria particularly vulnerable to proteostatic imbalance [6, 7, 8, 9, 10, 11, 12, 13, 14, 15]. Miscoordination of translational flux and assembly, persistent exposure to reactive oxygen species (ROS), proteotoxic stress, non‐imported precursors, orphan OXPHOS subunits, and stalled nascent chains all require sensing and correcting [16, 17, 18, 19].

A central component necessary for mitochondrial homeostasis is the mitochondrial unfolded protein response (UPR^mt), a conserved stress‐response program activated by misfolded proteins, imbalanced subunit stoichiometry, or import defects in the organelle (Fig. 1) [16, 17, 18, 19, 20, 21, 22]. The UPR^mt induces chaperones and proteases, enhances antioxidant defenses, and modulates mitochondrial translation and import to restore proteostasis. If stress persists, UPR^mt signaling promotes mitophagy to selectively eliminate irreversibly damaged mitochondria [16, 17, 18]. In mammals, transcriptional control of UPR^mt effectors converges with the integrated stress response (ISR) that provides an important translational layer of control [23, 24, 25, 26]. This integration provides a flexible, rapidly inducible mechanism for stress adaptation but complicates efforts to delineate mitochondria‐specific signaling nodes. For example, ISR activation can arise from diverse perturbations including endoplasmic reticulum (ER) stress, amino‐acid deprivation, and oxidative insult [23]. Beyond this transcription factor control, epigenetic licensing, via metabolite‐sensitive demethylases and acetylation dynamics, likely pre‐condition the UPR^mt loci while restraining non‐essential transcription during energetic stress [22, 27, 28].

Fig. 1.

Import‐gated sensing architecture of the mammalian UPR^mt. Mitochondrial import stress activates an eIF2α‐P–dependent translational checkpoint, reducing global protein synthesis while enabling preferential translation of ATF4, CHOP, and ATF5, which together initiate the UPR^mt nuclear programme. This transcriptional output induces mitochondrial chaperones and proteases (HSP60/HSP10, CLPP, LONP1) to expand proteostasis capacity, while simultaneously promoting import reinforcement by upregulating TOM/TIM components and supporting MPP/OCT1 processing and mtHsp70–PAM function. Parallel import‐surveillance cues (TOM/TIM flux, precursor load) help adjust import capacity during stress. Mitochondrial stress can also release DELE1 to the cytosol, where it contributes to ISR engagement. Downstream, UPR^mt signaling coordinates mitochondrial remodeling, including fusion–fission balance and mitophagy, to restore organellar homeostasis.

Although the UPR^mt is widely invoked as a key mediator of mitochondrial proteostasis, the molecular principles that define mammalian UPR^mt sensing and signal relay remain poorly understood. Outstanding questions include how mitochondrial proteotoxic stress is detected, how stress‐derived cues such as peptides, ROS, metabolites or mitochondrial RNA species are exported, and how they are interpreted by the nucleus to produce a mitochondria‐biased transcriptional output. Moreover, mitochondrial stress responses intersect with other major cellular pathways, including the unfolded protein response of the ER (UPR^ER), hypoxia signaling and metabolic rewiring, and this creates an intricate, multilayered regulatory network. Furthermore, lysosome‐centered nutrient sensing and cell‐non‐autonomous responses can also bias the ISR output and systemic adaptation [29, 30]. Understanding how the cell resolves these competing or cooperative signals is central to defining the mechanisms involved in mitochondrial homeostasis.

These issues have substantial implications for human physiology and disease. Aging, neurodegeneration, cardiomyopathies, and cancer all feature mitochondrial proteotoxic stress, and in several contexts, notably tumor progression and metastasis, cancer cells appear to co‐opt UPR^mt circuits to maintain mitochondrial function under hostile microenvironmental conditions [20, 31, 32, 33, 34, 35]. Deciphering the regulatory mechanisms of mammalian UPR^mt signaling therefore offer opportunities for understanding the fundamental principles of stress‐response integration and for identifying the therapeutic opportunities.

In this Review, we describe a unified mechanistic framework for the mammalian UPR^mt through integrating three core regulatory networks: (1) the import‐gated sensing that biases the ISR; (2) the selective translational control across cytosolic and mitochondrial compartments; and (3) the metabolite‐driven epigenetic licensing of the stress‐responsive loci (Fig. 2). In this context, we frame the UPR^mt as a mitochondria‐centric, multilayered control architecture, incorporating ER stress pathways, hypoxia, and systemic signaling only when they impinge on UPR^mt‐specific sensing, ISR tuning, and functional outcomes.

Fig. 2.

Three‐layer control of the mammalian UPR^mt. The mammalian UPR^mt is coordinated through three mechanistic layers. Translational gating (ISR): eIF2α phosphorylation reduces global translation while promoting selective synthesis of ATF4, CHOP, and ATF5, supported by stress‐granule sequestration and miRNA‐mediated tuning of mitochondrial mRNA translation. Transcriptional execution: ATF4 regulates metabolic rewiring, CHOP induces chaperone and protease networks, and ATF5 provides mitochondria‐proximal specificity; an ERα‐dependent branch contributes to NRF2 and IMS‐specific signaling. Epigenetic licensing: chromatin accessibility at UPR^mt loci is shaped by metabolite‐sensitive demethylases and acetyltransferases, enabling selective transcriptional activation during mitochondrial stress.

Notably, mitochondrial stress associated with the UPR^mt does not reflect generic energetic failure but a restricted set of perturbations that disrupt the balance between mitochondrial protein load, folding capacity, and import competence. Operationally, UPR^mt‐relevant stress arises from impaired matrix assembly, defective co‐translational insertion, reduced membrane‐potential–dependent import, or mito–nuclear stoichiometric imbalance of OXPHOS subunits; secondary redox or metabolic effects are common, but the defining feature is accumulation of non‐imported, misfolded, or orphan mitochondrial proteins. Although these insults often converge on the integrated stress response, ISR activation alone does not specify a mitochondrial output. Thus, mitochondrial stress reconfigures ISR context to yield outputs distinct from other cellular stress responses.

Mitochondrial proteostasis

Because most mitochondrial proteins in human cells are encoded by nuclear genes and translated in the cytosol, mitochondrial biogenesis and function depend on the post‐translational import of unfolded precursors and their precise trafficking to the outer membrane, intermembrane space, inner membrane and the matrix [36, 37]. This reliance has driven the evolution of specialized import machineries that ensure accurate targeting, sorting, and assembly.

Mitochondrial respiratory chain complexes I, III, and IV, as well as ATP synthase, exemplify the complexity of mitochondrial proteostasis, as they consist of subunits encoded by both mitochondrial and nuclear genomes. Mitochondrially encoded proteins typically form the catalytic cores of these complexes and associate with numerous nuclear‐encoded subunits, requiring many dedicated assembly factors and tight coordination of mitochondrial and nuclear gene expression [38, 39, 40, 41, 42]. Failure of this coordination results in the accumulation of unassembled subunits or incomplete complexes, promoting ROS production and cellular toxicity [43, 44, 45], and is strongly associated with human disease, including neurodegenerative disorders and cardiomyopathies [46, 47, 48, 49].

Nuclear‐encoded mitochondrial proteins are synthesized as precursors on cytosolic ribosomes, where chaperones maintain them in an unfolded state and deliver them to receptors of the translocase of the outer mitochondrial membrane (TOM complex) [50, 51, 52]. The TOM channel mediates translocation across the outer membrane, after which distinct translocase systems direct proteins to their appropriate mitochondrial subcompartments [53, 54, 55, 56]. Matrix‐targeted proteins typically carry an N‐terminal mitochondrial targeting sequence (MTS) that consists of a positively charged amphipathic helix of ~ 25–35 amino acids that engages TOM and drives membrane potential (Δψ)‐dependent translocation through translocase of the inner mitochondrial membrane 23 (TIM23) [57, 58, 59, 60]. Complete import requires the ATP‐dependent presequence translocase‐associated motor (PAM), whose core component is mitochondrial heat shock protein 70 (mtHsp70) [53, 61]. In parallel, carriers lacking presequences use the TIM22 pathway for inner membrane insertion [62]. Also many of the cysteine‐motif intermembrane space (IMS) proteins undergo oxidative folding via the mitochondrial intermembrane space assembly/coiled‐coil‐helix‐coiled‐coil‐helix domain containing 4 (MIA/CHCHD4) system [63], and β‐barrel outer membrane (OM) proteins are sorted by mitochondrial sorting and assembly machinery (SAM) with TOM [64, 65], and these routes impose compartment‐specific proteostatic demands.

mtHsp70 supports both protein translocation and folding in the matrix. Following import, the MTS is cleaved by the mitochondrial processing peptidase (MPP), and folding is completed by the Hsp60–Hsp10 chaperonin system, which provides a protected environment and limits aggregation, particularly under stress [66, 67, 68]. Secondary N‐terminal trimming by octapeptidyl aminopeptidase 1/intermediate cleaving peptidase (OCT1/Icp55) (mitochondrial intermediate peptidase) often stabilizes processed precursors, linking processing fidelity to folding efficiency [69]. Proteostasis is further maintained by ATP‐dependent proteases, including mitochondrial ATP‐dependent protease (LON) and caseinolytic mitochondrial peptidase (ClpXP) in the matrix and paraplegin and ATP‐dependent zinc metalloprotease (YME1L) at or near the inner membrane, while antioxidant enzymes limit ROS‐induced damage [70, 71, 72].

Protein import efficiency is a tightly regulated node responsive to cellular and metabolic conditions. Phosphorylation of targeting sequences or TOM subunits, as well as cytosolic binding partners or metabolites, can modulate import competence by altering precursor folding or TOM activity [36, 37, 73, 74, 75]. Because import depends on the inner membrane potential and chaperone availability, it functions as a sensitive sensor of mitochondrial homeostasis. Severe impairment triggers mitophagy, whereas partial defects activate the mitochondrial unfolded protein response to restore function [76, 77].

In parallel, mitochondrially encoded proteins are synthesized by mitoribosomes that associate with the inner membrane to enable co‐translational insertion of highly hydrophobic respiratory chain subunits [78, 79, 80]. Despite the absence of extended 5′ untranslated regions on human mitochondrial mRNAs, translation is regulated through co‐translational checkpoints that couple protein synthesis to assembly [81]. At complex IV, this translational plasticity is mediated by cytochrome c oxidase 14 (COX14), which stalls COX1 translation at the mitoribosome, while mitochondrial translation regulation assembly intermediate of the cytochrome c oxidase/cytochrome c oxidase assembly factor 3 (MITRAC12/COA3) stabilizes the nascent chain until incorporation of the first nuclear‐encoded subunit, COX4‐1, permits translation to resume [82, 83, 84, 85, 86, 87]. This mechanism limits the accumulation of unassembled hydrophobic subunits and links mitochondrial translation to assembly status.

By coupling import capacity and folding demand and co‐translational stall–resume checkpoints, human mitochondria generate proteostatic cues, including stalled nascent chains, unassembled subunits, and altered peptide or metabolite flux, that are sensed and relayed to the nucleus by the UPR^mt to recalibrate chaperone, protease, and import programs and, when required, engage mitophagy [88].

Furthermore, in yeast, disruptions of mitochondrial proteostasis lead to the cytosolic accumulation of unimported mitochondrial precursor proteins, engaging both mitochondrial precursor overaccumulation stress (mPOS) and the mitochondrial compromised protein import response (mitoCPR) [89, 90, 91]. mPOS activates the unfolded protein response activated by mistargeting (UPRam), an ISR‐linked pathway that reduces the influx of newly synthesized mitochondrial proteins, thereby alleviating pressure on the import machinery. In parallel, mitoCPR restores mitochondrial function by recruiting factors that clear stalled precursors from import channels [89, 90, 92]. This surveillance network may be complemented by the formation of cytosolic granules termed “MitoStores,” which transiently sequester a specific subset of mitochondrial precursor proteins until stress is resolved [93]. Taken together, these pathways intersect with the UPR^mt to protect both cytosolic and mitochondrial proteostasis, although their significance in mammalian systems remains largely undefined.

Beyond local protein‐level quality control, mitochondria form a dynamic network that undergoes continuous remodeling through fusion and fission [94]. Fusion, mediated by mitofusin 1/2 (MFN1/2) and optic atrophy 1 (OPA1), enables functional complementation and dilution of damaged components, whereas fission, driven by dynamin‐related protein 1 (DRP1) and its adaptors mitochondrial fission factor (MFF) and mitochondrial dynamics protein 49/51 (MiD49/51), segregates dysfunctional regions for selective removal by mitophagy, notably via the PTEN‐induced kinase 1 ‐ Parkin RBR E3 ubiquitin protein ligase (PINK1–Parkin pathway) [95, 96, 97, 98, 99, 100]. These processes operate hierarchically to repair, isolate or eliminate damaged proteins or entire organellar units, with apoptosis engaged when damage is irreparable [101].

Because most mitochondrial quality control factors are nuclear‐encoded, mitochondrial homeostasis must be communicated to the nucleus to adjust proteostatic capacity. Retrograde signaling integrates proteotoxic, metabolic and redox cues [102, 103, 104], enabling the UPR^mt to recalibrate folding capacity, protease activity and import efficiency. How these cues are converted into a transcriptional program is discussed next, with emphasis on the activating transcription factor 4 (ATF4)‐ and ATF5‐dependent arms of the mammalian UPR^mt.

In addition to intramitochondrial quality control mechanisms, mitochondrial protein import stress (MPIS) and iron‐dependent regulation of import have emerged as key triggers of mammalian mitochondrial stress signaling. Central to this pathway is death‐associated protein 3 (Dap3) binding cell death enhancer 1 (DELE1), a short‐lived mitochondrial protein whose import and stability depend on membrane potential and iron availability. Under acute mitochondrial stress, including loss of Δψ or OXPHOS inhibition, the inner mitochondrial membrane protease OMA1 cleaves DELE1, releasing a cytosolic fragment (DELE1‐S) that binds and activates the ISR kinase heme‐regulated inhibitor (HRI), promoting selective translation of ATF4/ATF5 [105, 106]. During iron deficiency, mitochondrial import arrest stabilizes full‐length DELE1 on the organelle surface, enabling OMA1‐independent HRI activation and ISR induction [107]. Furthermore, iron mitochondrial ATP binding cassette (ABC) transporters, ABCB7/ABCB8, modulate this axis, linking iron–sulfur/heme metabolism to ISR output [107]. MPIS now appears to be the overarching upstream signal sensed by DELE1: endogenous DELE1 can be processed into multiple fragments (DELE1‐S; DELE1‐VS) by distinct proteases in a cell‐type–specific manner [107], including high temperature requirement protein A2 (HtrA2), whose Parkinson's disease–associated variant shows reduced DELE1 processing [108]. Disruption of inner membrane scaffolds prohibitins likewise impairs matrix import and engages DELE1–HRI–dependent ISR activation [109]. Collectively, these findings position the DELE1–HRI axis as a mammalian surveillance system that integrates import efficiency, iron availability, and proteotoxic stress to drive ISR‐biased nuclear programs and, when repair fails, mitophagy.

The mitochondrial unfolded protein response

Mitochondrial dysfunction provokes a wide range of cellular responses, complicating efforts to define a mitochondria‐specific stress program. Here, we use the term mitochondrial unfolded protein response to denote a transcriptionally coordinated adaptive pathway activated by defined signaling mechanisms when mitochondrial proteostasis or respiratory function becomes compromised. In this framework, mitochondrial dysfunction encompasses diverse insults—including misfolded protein accumulation, defects in mitochondrial translation, mtDNA perturbation, loss of membrane potential (Δψ), and OXPHOS inhibition—with the unifying principle that these stresses disrupt the balance between mitochondrial protein load and folding capacity [110]. This working definition prioritizes regulatory architecture rather than individual downstream effectors and will continue to evolve as additional signaling layers are uncovered.

The UPR^mt was first detected in mammalian cells, where mtDNA depletion or expression of a terminally misfolded, mitochondria‐targeted protein induced robust expression of mitochondrial chaperones and quality control proteases [111]. Genetic studies in Caenorhabditis elegans subsequently delineated a mitochondria‐to‐nucleus communication axis that integrates import‐sensitive transcription factors and chromatin regulators, and the identification of homologous regulators in mammals confirmed evolutionary conservation of this adaptive program [112, 113, 114]. Notably, mammalian signaling is not organized as a single linear cascade but as a multilayered, partially redundant network in which multiple upstream sensors and transcriptional executors operate independently or cooperatively to sustain mitochondrial proteostasis. Conceptually, the UPR^mt is a mitochondria‐to‐nucleus pathway initiated when misfolded or unassembled proteins accumulate within the organelle; its activation restores proteostasis by enhancing folding capacity, accelerating degradation of damaged proteins and coordinating transcription, translation and import of nuclear‐encoded mitochondrial proteins [18]. If damage persists, prolonged signaling promotes mitophagy to remove irreversibly compromised organelles and preserve cellular fitness.

A broad spectrum of perturbations engages this response [115]. Among the strongest triggers are disruptions of mitochondrial chaperones or proteases, which overwhelm folding capacity and cause precursor accumulation [116]. These diverse stimuli, including import inhibition, mitoribosome blockers, and respiratory chain poisons are summarized in Table 1 with their sources and modes of action. Direct inhibition of protein import, via TIM23 depletion or dissipation of membrane potential, likewise produces a robust signal [117]. Mitochondrial protein import has recently been shown to modulate the toxic aggregation of polyQ, amyloid‐β, and α‐synuclein proteins implicated in neurodegeneration and aging [118, 119]. Furthermore, the pro‐apoptotic sphingolipid intermediate [trans‐2‐hexadecenal (t‐2‐hex)] inhibits mitochondrial import, triggering a cytosolic proteostasis response [120]. Inhibition of mitochondrial translation by RNA interference or mitoribosome‐targeting antibiotics such as doxycycline or chloramphenicol induces the pathway by destabilizing respiratory‐chain assembly [116, 121, 122]. UPR^mt activation is not restricted to proteostatic defects: mtDNA depletion, pharmacological inhibition of respiratory complexes, and perturbation of TCA‐cycle enzymes such as fumarate hydratase all activate the pathway, indicating integration of proteotoxic and metabolic inputs [123, 124, 125]. Selective mitochondrial perturbations, including knockdown of mitochondrial proteases (SPG7, LONP1, ClpP) or chaperones (mtHSP70, mtHSP60), activate mitochondrial stress responses without inducing cytosolic or ER chaperone pathways, underscoring signaling specificity [126, 127, 128, 129]. Pharmacological tools such as ClpP inhibitors and mitochondrial‐targeted Hsp90 inhibitors [130], including gamitrinib triphenylphosphonium [131, 132], similarly engage UPR^mt‐associated transcription. Despite this diversity, these perturbations converge on a recurring defect signature: non‐imported precursors, stalled nascent chains, unassembled respiratory‐chain subunits and elevated ROS. Importantly, in mammalian cells, it was revealed that the UPR^mt functions during two distinct phases of proteostasis stress: a protection phase that limits damage, followed by a repair phase that restores proteostasis [133].

Table 1.

Defects that activate the mitochondrial unfolded protein response.

Perturbation impairing mitochondrial homoeostasis	Stress source	Mode of action and/or resulting defects	Ref.
Mitochondrial translation and protein synthesis perturbation	Doxycycline	Mitochondrial ribosome inhibitor	[21, 122]
	Chloramphenicol	Mitochondrial ribosome inhibitor	[21, 121]
	Rapamycin as inhibitor of cytosolic translation	Accumulation of excess of mitochondrial‐encoded subunits	[21, 136]
	Pharmacological interventions enhancing mitochondrial biogenesis, e.g. sirtuin activator, resveratrol	Perturbation in mitochondrial subunit stoichiometry reinforcing its imbalance	[137, 138]
	Knockdown of multiple mitochondrial ribosomal components	Defective ribosomes	[21, 116]
	Muscle‐specific Crif1 knockout (mouse)	Ribosomal subunit; defective ribosome assembly	[241]
	Heart‐muscle‐specific Dars2 knockout (mouse)	Mitochondrial aspartyl‐tRNA synthetase; reduced mitochondrial translation	[238]
OXPHOS impairment	Rotenone	Complex I inhibitor	[123, 125, 275]
	Oligomycin	Complex V inhibitor	[123, 125, 176]
	Antimycin	Complex III inhibitor	[123, 125, 176, 275, 276]
	Surf1 knockout (mouse)	Complex IV assembly factor	[277]
	isp‐1 (qm150) mutant (C. elegans)	Reduced complex III activity	[278]
	clk‐1 (qm30) knockout (C. elegans)	Impaired ubiquinone biosynthesis	[278]
	cco‐1 knockdown (C. elegans)	Complex IV component; impaired complex assembly	[135]
Mitochondrial proteostasis perturbations	Knockdown of chaperones hsp‐6 or hsp‐60 (C. elegans) or proteases spg‐7, lonp1, clpp	Mitochondrial proteases and chaperones; accumulation of misfolded protein in the mitochondrial matrix	[116, 126, 127, 128, 129]
	gamitrinib triphenylphosphonium	Inhibitor of the mitochondrial HSP90 chaperone TRAP1	[131, 132, 232]
	ΔOTC (C. elegans and mammalian cells)	Protein truncation; leads to misfolding and the accumulation of terminally misfolded protein in the mitochondrial matrix	[112, 127]
	CDDO (2‐cyano‐3,12‐dioxo‐oleana‐1,9(11)‐dien‐28‐oic acid)	LON protease inhibitor	[232]
	EndoG N174A expression in mammalian cells	Accumulation of terminally misfolded protein in the mitochondrial intermembrane space	[187]
Metabolic perturbation	Missense variant (mouse) and knockdown (C. elegans) of FH1	perturbation of TCA cycle enzymes such as FH1 catalyzes the hydration of fumarate to malate	[123, 124]
mtDNA defects	mtDNA lesion uaDf5 (C. elegans)	3.1 kb mtDNA deletion of four essential OXPHOS genes	[279, 280]
	Deletor mice	Mutant for the mitochondrial helicase Twinkle	[281]
	Ethidium bromide	Reduced mtDNA synthesis	[116, 282]
ROS	Paraquat	Generation of excessive ROS	[112, 117, 275]
Mitochondrial protein import machinery impairment	tim‐17, tim‐23 or tom‐40 knockdown (C. elegans)	Lack of essential components of the import translocases	[117, 283]
	Pro‐apoptotic sphingolipid intermediate [trans‐2‐hexadecenal (t‐2‐hex)]	Inhibits mitochondrial import, triggering a cytosolic proteostasis response	[284]

A key conceptual advance is that mito–nuclear protein imbalance is a dominant trigger. Impaired mitochondrial translation—caused by loss of mitoribosomal proteins, mtDNA depletion or mitoribosome inhibition—reduces synthesis of mitochondrial‐encoded complexes subunits and leads to accumulation of orphaned nuclear‐encoded partners, strongly activating the response [134, 135]. Conversely, inhibition of cytosolic translation, for example by rapamycin, generates excess mitochondrial‐encoded subunits and likewise activates the program [21, 136]. Interventions that enhance mitochondrial biogenesis, such as sirtuin activation or resveratrol treatment, can likewise trigger activation by perturbing subunit stoichiometry, reinforcing that imbalanced stoichiometry ‐ not the absolute protein abundance ‐ is the predominant initiating signal [137, 138]. A practical challenge follows from the breadth of commonly used stimuli: respiratory‐chain inhibitors, depolarizing agents, and mitoribosome antibiotics concurrently activate the ISR, UPR^ER, redox signaling and metabolic rewiring, making it difficult to distinguish mitochondria‐proximal signals from secondary consequences. This caveat is particularly relevant when interpreting transcriptional and epigenetic changes, where generalized energetic collapse can mimic or mask bona fide UPR^mt signatures. There is a consequent need for refined models that selectively perturb mitochondrial proteostasis—by manipulating import capacity, matrix chaperones or proteases, or mitochondrial assembly checkpoints—to achieve mitochondria‐biased activation with minimal engagement of confounding pathways.

A defining feature of UPR^mt activation is its tight association with compromised mitochondrial protein import [139]. Because matrix‐directed import through TOM–TIM23 depends on membrane potential and mtHSP70‐driven ATP cycling, import is exquisitely sensitive to OXPHOS impairment, translational stress, and proteotoxic load [140]. Thus, diverse insults converge on reduced import efficiency and the accumulation of unimported or misassembled precursors. In C. elegans [117], import decline prevents mitochondrial degradation of activating transcription factor associated with stress 1 (ATFS‐1) and enables its nuclear accumulation; in mammals, import impairment biases the ISR toward preferential translation and nuclear accumulation of ATF4, ATF5 and C/EBP‐homologous protein (CHOP), thereby gating the strength of the UPR^mt to the degree of proteostatic stress [16]. Yeast data underscore the conservation of this logic. In Saccharomyces cerevisiae, non‐imported mitochondrial precursors mislocalize to the cytosol and are toxic unless rapidly cleared by proteasomes, necessitating activation of compensatory quality control pathways [89, 90, 141].

Once activated, the UPR^mt initiates a coordinated recovery program that restores proteostasis, restrains oxidative damage, and rebalances metabolism [123, 133]. Mitochondrial chaperones such as HSP60 and mtHSP70 expand folding capacity and suppress aggregation [142], while matrix and intermembrane ATPases associated with diverse cellular activities (m‐AAA, i‐AAA) proteases eliminate damaged or misassembled proteins [143]. Antioxidant enzymes (notably superoxide dismutase isoforms) limit ROS accumulation [144, 145, 146]. In parallel, induction of iron–sulfur cluster assembly and ubiquinone‐biosynthetic pathways [147, 148] supports respiratory complex maturation [140, 141], and increased expression of the fission GTPase DRP1 links organelle renewal to segregation and removal of damaged regions [149, 150]. Because assembly of OXPHOS complexes is energetically costly and stoichiometry‐sensitive, the program transiently downregulates nuclear‐encoded OXPHOS gene expression to reduce proteostatic burden, while induction of glycolytic genes supports ATP production during recovery [151, 152]. Altogether, these mechanisms enable stressed mitochondria to re‐establish proteostasis and restore organellar function. Reflecting its deployment across species and tissues, the UPR^mt operates as a distributed, partially redundant signaling system that maintains robustness even when individual arms are impaired [18].

Transcriptional layer of the mammalian UPR^mt

Although the mitochondrial unfolded protein response (UPR^mt) was first identified in mammalian cells, its transcriptional architecture is considerably more complex than the streamlined ATFS‐1‐centered pathway characterized in C. elegans. In mammals, no single transcription factor serves as a dedicated mitochondria‐to‐nucleus messenger. Instead, UPR^mt gene expression emerges from the intersection of mitochondria‐proximal stress sensing, ISR‐driven translational control, and combinatorial transcription factor activity. This multilayered design reflects the diversity of mitochondrial insults that activate the pathway and ensures transcriptional specificity despite extensive crosstalk with broader stress‐response systems. Deciphering mammalian UPR^mt transcription therefore requires integrating conserved import‐gated mechanisms, best defined in nematodes, with ISR‐dependent regulation of ATF4, ATF5, and CHOP.

Insights from C. elegans defined the core logic of an import‐gated transcriptional switch. ATFS‐1 carries both a mitochondrial targeting sequence (MTS) and a nuclear localization signal (NLS). Under basal conditions, efficient import results in ATFS‐1 degradation by the matrix protease LONP1; under stress, import efficiency declines, allowing ATFS‐1 to accumulate in the cytosol and translocate to the nucleus to activate the UPR^mt transcriptional program [117, 153]. This mechanism is supported by genetic evidence: weakening or removal of the MTS yields constitutive nuclear localization and chronic pathway activation [154]. ATFS‐1 does not act alone. The transcription factor DVE‐1 and the ubiquitin‐like protein UBL‐5 remodel chromatin at UPR^mt target promoters, such as hsp‐60, thereby facilitating ATFS‐1 binding [155, 156, 157]. Upstream of ATFS‐1, CLPP‐1‐dependent peptide generation and homolog of ABC protein family‐1 (HAF‐1)–mediated peptide‐export couple mitochondrial proteolysis to nuclear transcription [116, 156, 157, 158]. Beyond acute signaling, ATFS‐1 also contributes to mitochondrial genome maintenance by promoting mtDNA polymerase‐γ (POLG) loading and modulating transcriptional initiation during mtDNA repair [159], illustrating how import‐gated transcription factors link mitochondrial proteostasis to mtDNA homeostasis.

In contrast, S. cerevisiae lacks a dedicated mitochondria‐to‐nucleus transcription factor analogous to ATFS‐1 [160], yet retains the underlying logic of import‐linked stress detection. Mitochondrial proteotoxic stress activates multiple surveillance pathways. Accumulation of non‐imported precursors induces the unfolded protein response activated by mistargeting (UPRam), which enhances proteasomal degradation of cytosolic mitochondrial clients [141]. mPOS represses cap‐dependent translation while promoting selective survival programs, a shift reinforced by ROS‐mediated ribosomal modification [161, 162]. At the organelle surface, mitoCPR mobilizes pleiotropic drug resistance protein 3 (Pdr3)‐dependent citrinin resistance protein (CIS1) and recruits outer mitochondrial transmembrane helix translocase (Msp1) to extract stalled precursors at TOM [89, 139]. Although the downstream transcriptional machinery differs across phyla, these pathways underscore a principle shared with metazoans: Impaired mitochondrial protein import is a primary stress signal that must be detected and mitigated.

The mammalian UPR^mt retains this import‐gated logic but embeds it within a more elaborate regulatory framework [133]. The functional analogue of ATFS‐1 is ATF5, which also contains an MTS and an NLS. Under non‐stress conditions, ATF5 is imported into mitochondria and degraded; when import declines, ATF5 accumulates in the cytosol and translocates to the nucleus, where it induces mitochondrial chaperones such as HSP60 and represses select mitochondrial proteins, thereby reducing proteostatic load [112, 117]. Consistent with its dual localization, ATF5 also contributes to basal mitochondrial function: it supports respiratory capacity under non‐stress conditions, and ATF5 depletion reduces both basal and maximal respiration [112, 117]. This dual mitochondrial–nuclear role mirrors that of ATFS‐1 and highlights deep conservation of the core regulatory logic.

In mammals, however, ATF5 acts within the broader ISR. Expression of ATF4, CHOP and ATF5 depends on eukaryotic translation initiation factor 2A (eIF2α) phosphorylation, which reduces global translation while selectively enhancing translation of upstream open reading frame (uORF)‐containing stress‐responsive mRNAs [163]. CHOP (DDIT3), one of the earliest identified components of the mammalian UPR^mt, is induced following mtDNA depletion or ΔOTC (ornithine transcarbamylase) expression, and together with C/EBPβ, CHOP binds promoters of UPR^mt targets, including HSP60, HSP10, mitochondrial isoform of DnaJ (mtDnaJ), CLPP and YME1L1 [111, 127]. These promoters contain CHOP‐binding elements flanked by mitochondrial UPR elements (MUREs), which appear to confer selectivity for mitochondrial stress over canonical ER stress [164, 165]. CHOP and C/EBPβ also harbor activator protein‐1 (AP‐1) motifs bound by c‐Jun [166], linking mitochondrial dysfunction to JNK signaling and revealing additional integration with cellular stress pathways.

Subsequent studies established that mitochondrial insults activate eIF2α kinases such as general control non‐derepressible 2 (GCN2) and protein kinase R‐like ER kinase (PERK), positioning eIF2α phosphorylation as an upstream gate for mammalian UPR^mt gene expression [27, 167, 168]. ISR activation alone, however, is insufficient to define a mitochondria‐specific output. ATF4 and CHOP are induced by diverse stresses, including ER stress, amino‐acid deprivation and viral infection [169, 170, 171], raising the central question of how mitochondrial dysfunction biases ISR output toward a UPR^mt transcriptional program.

Current evidence supports a cooperative, non‐redundant interplay between ATF4, CHOP and ATF5. ATF4 primarily regulates metabolic and redox pathways, including amino‐acid, one‐carbon and antioxidant metabolism [172, 173, 174, 175, 176]. In models of mtDNA depletion or respiratory inhibition, ATF4‐driven remodeling enhances transsulfuration, glutathione synthesis, lipid droplets formation and nucleotide biosynthesis [113, 124, 171, 174, 177, 178, 179]. CHOP, often acting with C/EBP family partners, induces mitochondrial chaperones and proteases and contributes to cell‐fate decisions under persistent stress [180, 181]. ATF5, by contrast, provides mitochondria‐proximal specificity: Its activity is directly modulated by import efficiency, allowing mitochondrial functional status to bias ISR output toward mitochondrial proteostasis [112, 182]. ATF5 can rescue UPR^mt activation in C. elegans lacking ATFS‐1 [112], emphasizing functional conservation despite divergent network architecture. In this hierarchical structure, ATF4 and CHOP establish a permissive transcriptional and metabolic state, while ATF5 couples transcriptional execution to mitochondrial import capacity. ATF5 transcriptionally upregulates mitochondrial chaperones such as HSP60 and HSP10 and quality control proteases including CLPP and YME1L [112, 183]. Transcription of ATF5 is regulated by ATF4 and CHOP [184], whereas ATF5 protein stability is further gated post‐translationally by mitochondrial import [185, 186]. Together, ATF4, CHOP, and ATF5 drive the core UPR^mt recovery program, thereby linking ISR‐primed transcriptional activation directly to the restoration of mitochondrial proteostasis.

Submitochondrial localization introduces additional specificity. Perturbations affecting the intermembrane space elicit a distinct transcriptional branch requiring estrogen receptor α (ERα) in addition to ATF4, CHOP and ATF5 [187]. ERα‐dependent signaling enhances proteasome activity, induces the IMS protease HTRA2 and activates nuclear respiratory factor 1 (NRF1) to promote mitochondrial biogenesis and clearance of mislocalized IMS proteins [187]. Mitochondrial stress also engages lysosomal nutrient‐sensing pathways [188], augmenting its UPR^mt‐specific transcriptional activity and revealing mitochondria–lysosome–nucleus crosstalk that is not present in invertebrate models.

Collectively, these observations support a multilayered model in which mitochondrial stress is first translated into a permissive ISR‐dependent translational state and then resolved into mitochondria‐specific transcriptional outputs. Diverse mitochondrial perturbations, converge on activation of eIF2α kinases, enabling selective translation of ATF4, CHOP, and ATF5. These factors cooperatively activate overlapping but distinct gene sets encompassing chaperones and proteases, protein import and assembly factors, antioxidant defenses and metabolic pathways that support survival during mitochondrial dysfunction. In this architecture, eIF2α phosphorylation acts as a permissive gate, whereas mitochondrial import efficiency biases ISR output toward a bona fide UPR^mt transcriptional program, ensuring that nuclear gene expression is precisely matched to mitochondrial needs.

Importantly, although the UPR^mt relies on a conserved principle of import‐sensitive stress detection, the organization of this response varies markedly between organisms. In C. elegans, UPR^mt regulation is largely encoded in a mitochondria‐proximal, single‐factor circuit centered on ATFS‐1, whose nuclear availability is dictated by mitochondrial import competence. Yeast, by contrast, lack a dedicated ATFS‐1–like transcriptional arm and instead mitigate mitochondrial proteotoxic stress through cytosolic quality control pathways, including mPOS, mitoCPR, and UPR^am. Mammalian systems retain import‐based sensing but distribute it across a layered regulatory architecture that combines ISR‐driven translational control, cooperative action of ATF4, CHOP, and ATF5, and extensive epigenetic and organelle‐level integration. Together, these distinctions illustrate that while stress sensing is evolutionarily conserved, mammalian UPR^mt has expanded structurally to accommodate tissue diversity, regulatory flexibility, and whole‐organism coordination, precluding direct extrapolation from yeast or nematode models.

Translational regulation as a core component of the UPR^mt

Mitochondrial stress elicits profound changes in cellular translation that constitute a central arm of the mitochondrial unfolded protein response. In metazoans—and most prominently in mammals—these changes are largely coordinated by the integrated stress response, which couples mitochondrial dysfunction to global translational repression while permitting selective synthesis of stress‐adaptive factors. ISR activation is initiated by phosphorylation of the α subunit of eukaryotic initiation factor 2 (eIF2α) by stress‐responsive kinases, resulting in broad attenuation of protein synthesis that alleviates proteostatic pressure on cytosolic and mitochondrial folding systems [189]. At the same time, eIF2α phosphorylation selectively enhances translation of a restricted set of mRNAs containing upstream open reading frames, most notably those encoding the basic leucine zipper (bZIP) transcription factors ATF4, CHOP and ATF5, which together orchestrate the transcriptional and metabolic arms of the UPR^mt [23, 24, 25, 26, 189].

Multiple ISR kinases can be engaged during mitochondrial dysfunction, reflecting the diversity of stress signals emanating from impaired organelles. GCN2 responds to amino acid limitation and uncharged tRNAs and is also activated by ribosome stalling, reactive oxygen species, and broader bioenergetic stress [190, 191, 192, 193]. PERK, although classically associated with endoplasmic reticulum stress, is frequently engaged during mitochondrial perturbations owing to ER–mitochondria crosstalk involving calcium flux, lipid imbalance, and oxidative stress [189, 194]. HRI senses heme insufficiency and redox imbalance—a particularly relevant axis given that heme biosynthesis and iron–sulfur cluster metabolism are mitochondrial processes and is additionally activated when mitochondrial protein import stress or iron‐dependent import arrest stabilizes DELE1 on the organelle surface, or when OMA1‐ or HtrA2‐mediated processing of DELE1 generates cytosolic signaling fragments that engage the ISR [105, 106, 107, 179, 195, 196]. Protein kinase R (PKR) can be activated by double‐stranded RNA and innate immune signaling [197, 198, 199], which may intersect with mitochondrial damage through release of mitochondrial RNA or DNA species. Regardless of the initiating kinase, eIF2α phosphorylation depresses ternary‐complex recycling, thereby reducing bulk initiation while biasing re‐initiation toward uORF‐containing transcripts. The net effect is a selective rise in stress effectors despite an overall translational lull. The ensuing transcriptional program in the UPR^mt is dominated by the cooperative yet non‐redundant activities of ATF4, CHOP and ATF5 [133, 180] described in previous section. In addition, mitochondrial stress engages lysosomal nutrient sensing: activation of the vacuolar H⁺‐ATPase—mechanistic target of rapamycin complex 1 (v‐ATPase–mTORC1) on lysosomes leads to ATF4 phosphorylation [188], which licenses a UPR^mt‐biased transcriptional output and helps explain how mitochondrial ISR signaling achieves specificity relative to other stress contexts.

Translational control during the UPR^mt is not confined to the cytosol but extends into mitochondria, enforcing compartmental symmetry. Mitochondrial dysfunction reduces intramitochondrial protein synthesis in part through protease‐dependent mechanisms. A well‐defined example is LONP1‐dependent turnover of mitochondrial RNase P protein 3 (MRPP3), a catalytic subunit of the mitochondrial RNase P complex. Loss of MRPP3 impairs tRNA 5′‐end processing and limits mitoribosomal throughput [200, 201]. Functionally, this mirrors cytosolic ISR control: both reduce nascent polypeptide pressure, cytosolic via eIF2α phosphorylation and mitochondrial via tRNA biogenesis control, thereby minimizing accumulation of orphan subunits and preventing mito–nuclear stoichiometric imbalance. Additional mitochondrial checkpoints are likely engaged, including mitoribosome pausing on hydrophobic ORFs when membrane insertion or assembly is delayed, feedback from respiratory‐chain assembly checkpoints (for example at COX1/complex IV), and metabolite‐dependent tuning of RNA processing and elongation [202, 203].

Beyond the ISR, translational repression during mitochondrial stress is further refined by the formation of stress granules (SGs), dynamic cytoplasmic ribonucleoprotein assemblies that provide a reversible and selective layer of translational control [204]. SGs nucleate when initiation is limited and ribosomes disengage from mRNAs, exposing transcripts that are captured by intrinsically disordered RNA‐binding proteins, like stress‐granule assembly factor (G3BP1/2), via liquid–liquid phase separation [205]. Although SG formation is facilitated by eIF2α phosphorylation, it can also occur independently through alternative initiation blocks or ribosome collisions [206]. Recent mechanistic work indicates that SGs along with their nucleators G3BP1/2, are not passive sequestration sites, but active regulators of the stress translation landscape [207]. SGs are enriched for stress‐resistant mRNAs whose translation persists during global repression, and G3BP1/2 are necessary and sufficient to prioritize translation of SG‐resident transcripts while suppressing global cytosolic translation [207]. Thus, SGs operate as mesoscale organizers that reinforce the translational program of the stress response. In the context of the UPR^mt, SGs are well positioned to fine‐tune translation of nuclear‐encoded mitochondrial proteins [204]. Because mitochondrial clients constitute a minority of total protein synthesis, indiscriminate translational shutdown can be excessive or maladaptive during recovery. Indeed, SGs form rapidly upon induction of UPR^mt to be disassembled upon EIF2 dephosphorylation during recovery [204]. Together, SG‐mediated sequestration provides a mechanism to transiently promote ISR programs, while limiting synthesis of import‐dependent mitochondrial clients and assembly factors and preserving these mRNAs for rapid re‐engagement once proteostasis is restored. This SG selectivity is likely especially important during reversible mitochondrial stress. Interestingly, in C. elegans, hypoxia as well as loss of LONP‐1 leads to constitutive formation of mitochondrial stress granules (mitoSGs), that precede overt protein aggregation and represent an early mitochondrial pathology [208], however the exact molecular role of these phenomenon remains unknown.

Post‐transcriptional control by microRNAs (miRNAs) constitutes a distinct, intersecting layer. Rather than enforcing a blanket repression, miRNAs provide targeted, context‐dependent tuning via Argonaute protein ‐ microRNA‐induced silencing complex (AGO2–RISC)‐mediated inhibition of initiation of translation or promotion of deadenylation/decapping [209, 210]. The magnitude and direction of repression depend on competition with RNA‐binding proteins at AU‐rich, as well as on partitioning of mRNAs into SGs and P‐bodies, which co‐enrich AGO2 to favor reversible silencing rather than decay [209, 211]. Consistent with this selectivity, several miRNAs previously linked to ATF5 activity (for example, miR‐141‐3p [212], miR‐520b‐3p [213] in cancer stress, and miR‐134‐5p in infection [214]) do not map cleanly onto canonical UPR^mt outputs, underscoring that ATF5–miRNA interactions are highly context‐specific and not uniformly coupled to mitochondrial stress. By contrast, miRNA deregulation can directly engage the UPR^mt: silencing miR‐382‐5p in muscle triggers a robust UPR^mt with mito–nuclear protein imbalance and coordinated downregulation of mitochondrial ribosomal protein mRNAs [215], implicating miRNA control of translational fidelity as an upstream node in mitochondrial stress signaling. More broadly, curated miRNA sets modulate mitochondrial metabolism, respiratory capacity and biogenesis across tissues [216, 217], positioning miRNAs as active modulators of mitochondrial adaptation. These observations support a model in which UPR^mt‐regulated miRNAs may operate as negative‐feedback regulators, transiently reducing synthesis/import of nuclear‐encoded mitochondrial proteins during proteostatic overload to alleviate chaperone and protease burden and avoid excessive response amplification. Mechanistically, overlap in cis‐elements (AU‐rich elements, miRNA motifs) and shared SG/P‐body niches fosters cooperation between SG sequestration and miRNA repression: SG residency lowers ribosome engagement and increases RISC access [218], whereas miRNA occupancy promotes SG association by further diminishing initiation [219]. In concert with ISR‐driven attenuation, these post‐transcriptional mechanisms provide a rapid, reversible, and transcript‐selective layer that matches mitochondrial protein flux to folding, assembly, and import capacity during the UPR^mt, while preserving the potential for swift translational re‐engagement during recovery.

Together, ISR‐driven translational attenuation, stress‐granule triage, and miRNA‐mediated tuning create a multilayered system that matches mitochondrial protein flux to folding and import capacity, stabilizing the organelle during acute stress and enabling rapid recovery once homeostasis is restored.

Epigenetic regulation and chromatin remodeling as parallel inputs to UPR^mt activation

In addition to the discussed above transcriptional and translational layers of control, epigenetic mechanisms make a substantial contribution to UPR^mt activation by reorganizing nuclear chromatin during mitochondrial stress.

Mechanistic principles were first delineated in C. elegans, where mitochondrial stress provokes nuclear accumulation of LIN‐65 together with the H3K9 methyltransferase MET‐2 (the orthologue of mammalian SETDB1) [157, 220, 221]. This module enforces genome‐wide chromatin compaction and transcriptional dampening while preserving competence at UPR^mt targets through the homeobox factor DVE‐1, which safeguards accessibility at stress‐responsive regions [222]. Functional counterparts of these regulators exist in mammals (special AT‐rich sequence‐binding protein 2 (SATB2) as the DVE‐1 orthologue and UBL‐5 as the conserved UBL‐5 homolog) supporting the conservation of this chromatin‐licensing architecture across metazoans [223, 224]. Notably, nuclear entry of LIN‐65 requires the matrix protease CLPP‐1 [157], placing mitochondrial proteolysis upstream of chromatin reorganization and suggesting that mitochondria‐derived cues instruct nuclear epigenetic responses.

Parallel pathways in C. elegans reveal a second mode of chromatin activation. The JumonjiC (JmjC) demethylases JMJD‐1.2 (PHF8) and JMJD‐3.1 erase repressive H3K27 methylation, creating permissive promoter states at UPR^mt genes, and when overexpressed, are sufficient to drive UPR^mt activation and longevity—directly linking cofactor‐dependent demethylation (α‐ketoglutarate, Fe²⁺) to stress competence [157, 225]. Worm CBP‐1, the orthologue of mammalian CBP/p300, not only acetylates histones at UPR^mt loci but can also acetylate ATFS‐1 itself [226], illustrating how chromatin enzymes modulate both chromatin structure and transcription factor function.

The analogous epigenetic control operates in mammals. Functional homologs of nematode UPR^mt demethylases, PHF8 (KDM7B) and JMJD3 (KDM6B), increase chromatin accessibility at mitochondrial chaperone and proteostasis loci during mitochondrial stress [157, 227], while CBP/p300‐mediated H3K18ac/H3K27ac correlates with induction of UPR^mt‐associated genes [226]. Conversely, inhibition of histone deacetylase (HDAC1/2) attenuates a subset of these responses, underscoring a methylation–acetylation balance that shapes UPR^mt transcription [223]. As in worms, mammalian HDAC1/2 cooperate with SATB2 to modulate mitochondrial homeostasis [223], reinforcing conservation of a chromatin‐remodeling module across species. Given the reliance of JmjC demethylases on α‐ketoglutarate and iron [228], and the sensitivity of acetyltransferase/deacetylase activities to acetyl‐CoA and NAD⁺ [229], epigenetic enzyme output is naturally coupled to mitochondrial metabolism.

Matrix AAA⁺ proteases position mitochondrial proteostasis upstream of chromatin readiness through two convergent routes. First, LONP1 and CLPP degrade misfolded or unassembled matrix proteins, shaping pools of mitochondrial‐derived peptides and metabolites that can function as retrograde signals [156, 230, 231]. In C. elegans, CLPP‐1–dependent peptide generation and HAF‐1–mediated export lie upstream of nuclear UPR^mt activation [156, 158]; while a direct peptide‐export conduit has not been fully delineated in mammals. Perturbation of CLPP or LONP1 alters the amplitude of UPR^mt gene induction, consistent with a role in priming chromatin at stress loci [110, 232]. Secondly, these proteases recalibrate metabolic fluxes that feed chromatin enzymes. By governing proteostasis of clients that influence mtDNA transcription/replication and respiratory capacity, LONP1, in particular, modulates TCA throughput, redox balance and iron handling, thereby determining the availability of α‐ketoglutarate, acetyl‐CoA and NAD⁺ [233, 234]. Tuning these cofactors alters the effective activities of PHF8/JMJD3 and CBP/p300/HDACs and biases promoter–enhancer states at UPR^mt loci toward accessibility or repression [70, 71]. Through these routes, mitochondrial proteolysis is directly coupled to nuclear chromatin readiness, linking matrix proteostasis to epigenetic control.

Although a SETDB1‐directed H3K9 compaction module precisely analogous to MET‐2/LIN‐65 has not yet been mapped in mammals, the regulatory logic appears conserved: broad transcriptional restraint at non‐essential genes coupled with locus‐specific competence at UPR^mt effectors. This competence is reinforced when ISR‐induced transcription factors, particularly ATF4 and ATF5, engage enhancers pre‐conditioned by demethylation and acetylation [170, 235], favoring efficient recruitment of the transcriptional machinery.

Together, these observations support a two‐input model for mammalian UPR^mt gene activation. Epigenetic licensing, driven by LONP1/CLPP‐dependent proteolysis and metabolite provision, establishes accessible chromatin at UPR^mt loci while globally restraining transcription under energetic stress. Transcription factor execution, principally via ATF4, CHOP, and ATF5, then converts this chromatin competence into productive, mitochondria‐biased transcriptional output. In this way, epigenetic remodeling operates in parallel with mitochondrial stress sensing and translational reprogramming to ensure that UPR^mt activation is both selective and proportional to mitochondrial dysfunction.

Taken together, these epigenetic mechanisms establish a chromatin landscape that is at once selectively permissive and globally restrained, ensuring that UPR^mt loci remain accessible while non‐essential transcription is limited under energy stress. Yet chromatin readiness alone cannot dictate the magnitude or specificity of the response. A defining feature of the mammalian UPR^mt is that transcriptional activation is preceded—and in many cases dictated—by translational reprogramming through the integrated stress response (ISR). This translational gating shapes the availability of the very transcription factors that act on licensed chromatin and thus represents a crucial interface linking mitochondrial dysfunction to nuclear gene expression and organellar recovery. The next section examines how mitochondrial stress rewires cytosolic and mitochondrial protein synthesis, how eIF2α phosphorylation biases translation toward ATF4, ATF5, and CHOP, and how selective translational control integrates with chromatin remodeling to produce a mitochondria‐adapted transcriptional state.

Yet epigenetic, transcriptional and translational control circuits represent only one dimension of how cells adapt to mitochondrial dysfunction. In physiological settings, mitochondrial stress rarely occurs in isolation; instead, it unfolds within a densely interconnected signaling landscape that also includes the endoplasmic reticulum, lysosomes, cytosolic quality control networks, and whole‐body metabolic circuits. These pathways share regulatory nodes—including eIF2α kinases, mTORC1 activity, redox state, Ca²⁺ flux, and organelle contact sites—and their interactions collectively determine whether mitochondrial dysfunction is contained, compensated or propagated to neighboring cells and tissues. The next section examines how the UPR^mt operates within this broader stress network: how ER stress, hypoxia, nutrient sensing, and mitophagy intersect with mitochondrial quality control; how organelle contact sites coordinate signaling across compartments; and how cell‐non‐autonomous responses such as mitokine secretion integrate mitochondrial status with systemic metabolic adaptation.

UPR^mt in an interconnected stress landscape: Organelle crosstalk and systemic adaptation

Although UPR^mt activation is generally adaptive, its deployment can be insufficient, chronic, or repurposed, producing disease‐ and tissue‐specific failure modes [16, 77]. In age associated disorders, including neurodegenerative and cardiovascular disease contexts, mitochondrial dysfunction is repeatedly linked to UPR^mt engagement, yet prolonged activation can become detrimental at the organismal level [16, 77]. Mechanistically, mitochondrial perturbations frequently engage a mitochondria‐to‐cytosol ISR relay through the OMA1–DELE1–HRI axis, providing a direct route by which defects in mitochondrial import or presequence processing bias ISR output toward sustained signaling [105, 106]. Neurodegeneration‐relevant perturbations including impaired PITRM1/MPP‐dependent processing, Parkinson's disease‐associated HtrA2 variants, and CHCHD10‐driven inner membrane proteotoxic stress, converge on DELE1 licensing and ISR activation, directly coupling mitochondrial proteostasis failure to chronic translational repression in neurons [108, 236, 237]. Consistent with this framework, ATF4 was identified as a dominant driver of the mammalian mitochondrial stress program, while canonical ATF5‐mediated UPR^mt gene sets are not uniformly engaged across stressors, underscoring that UPR^mt outputs differ substantially across disease‐relevant perturbations [16, 77, 113]. In the heart, mitochondrial proteostasis and translation defects can elicit strong, tissue biased stress pathways, notably, loss of DARS2 drives pronounced stress responses predominantly in cardiomyocytes and induces FGF21, linking mitochondrial proteostasis impairment to endocrine/systemic adaptation [238]. Moreover, under mild‐to‐moderate cardiac OXPHOS dysfunction, FGF21 can modulate a portion of the mitochondrial ISR program, highlighting dose dependence and the risk that chronic signaling becomes maladaptive as energetic reserve is eroded [239]. In metabolic disease settings, mitochondrial stress responses frequently operate in a constitutively low‐grade mode and associate with mitokine programs (FGF21/GDF15) that mediate systemic adaptation downstream of mitochondrial ISR signaling [240, 241]. However, persistent nutrient excess or insulin resistance uncouples UPR^mt signaling from recovery, favoring chronic ISR activation without effective restoration of mitochondrial proteostasis, thereby promoting inflammation, altered lipid handling, and endocrine stress signaling rather than organellar repair [240, 241]. Cancer provides a contrasting paradigm in which UPR^mt and mitochondrial stress pathways are often co‐opted: tumor cells operate under hypoxia, ROS, and proteotoxic pressure, and UPR^mt linked regulators are viewed as contributors to tumor fitness. Tumor cells exploit import‐gated ISR biasing and ATF4/ATF5 activity to sustain mitochondrial function under hypoxia, nutrient limitation and therapeutic stress. Rather than resolving proteotoxicity, UPR^mt pathways are tuned to maintain minimal mitochondrial competence, suppress apoptosis, and promote metabolic plasticity. In this setting, the UPR^mt is not failed but rewired, with prolonged signaling supporting survival, drug resistance, and metastatic potential [110, 182]. Accordingly, UPR^mt dysfunction in disease rarely reflects simple pathway “loss”; instead, pathology emerges from context‐dependent miscalibration of signaling amplitude, duration, and coupling to downstream recovery mechanisms, within broader organelle and systemic stress networks [77, 110]. These patterns underscore that UPR^mt signaling is deeply embedded in broader stress networks, particularly those emanating from the ER.

Furthermore, although the core logic of the UPR^mt is conserved, downstream outcomes differ markedly between post‐mitotic and proliferative tissues [77, 242]. In long‐lived post‐mitotic cells such as neurons, cardiomyocytes and skeletal muscle, mitochondrial stress is often chronic and recovery is constrained by limited or absent cell renewal, biasing UPR^mt activation toward sustained proteostasis through chaperone induction, translational restraint and organellar quality control [243, 244, 245]. Under these conditions, prolonged UPR^mt engagement can become maladaptive, as extended translational attenuation and suppression of OXPHOS gene expression impair synaptic transmission or contractile performance. By contrast, proliferative cells, including epithelial progenitors, immune cells, and many cancers, tolerate dynamic UPR^mt signaling more effectively, as transient translational repression and metabolic rewiring can be offset by cell division, mitochondrial biogenesis and metabolic plasticity [246]. In these contexts, UPR^mt outputs are preferentially channeled toward ATF4‐dependent signaling supporting amino‐acid metabolism, redox balance and nucleotide biosynthesis rather than sustained proteostasis enforcement, a divergence further amplified in cancer to preserve mitochondrial function while suppressing differentiation or apoptosis [242, 247]. Thus, UPR^mt outcomes reflect not only stress magnitude but also proliferative capacity, underscoring the need for tissue‐ and cell‐state‐specific interpretation rather than a uniform stress paradigm.

The pathways highlighted below in this section are not intended as standalone stress responses; instead, they are discussed as regulatory influences that tune UPR^mt signaling outcomes via shared control axes encompassing mitochondrial import dynamics, ISR‐biased translational regulation, and chromatin permissiveness. Mitochondrial proteotoxic stress rarely arises in isolation but unfolds within a densely interconnected stress landscape that includes oxidative stress, hypoxia and ER dysfunction. These pathways converge on shared regulatory nodes, most prominently translational control, organelle contact sites, and metabolic–epigenetic coupling, that collectively determine whether mitochondrial stress is resolved through adaptive recovery or progresses toward maladaptive degeneration. Within this network, the mitochondrial unfolded protein response functions not as an autonomous pathway but as a mitochondria‐biased module embedded in a broader, multicompartmental triage system operating across cells and tissues (Fig. 3).

Fig. 3.

UPR^mt within an interconnected stress‐response landscape. The mammalian UPR^mt operates within a broader, interconnected stress‐response network that integrates local organellar signals with cellular and systemic adaptation. Mitochondrial proteotoxic stress intersects with ER stress pathways, including PERK‐driven eIF2α phosphorylation, IRE1 signaling, and ATF6 activation, which reshape translation, mRNA levels, and ER homeostasis. This crosstalk extends to post‐transcriptional regulation through stress granules (SGs) and miRNA. Impaired oxidative phosphorylation, lipid and redox imbalance, and defects in iron–sulfur cluster homeostasis further engage antioxidant and metabolic programs. ER–mitochondria contact sites coordinate Ca²⁺ exchange, membrane remodeling, and bidirectional signaling that modulates mitochondrial resilience. Under hypoxia, HIF1/HIF2 and FUNDC1‐dependent mitophagy link oxygen sensing to mitochondrial turnover. Endocrine mediators such as FGF21 and GDF15 extend this response to the organismal level, enabling systemic metabolic adaptation. Together, these pathways illustrate how the UPR^mt is embedded within a multicompartmental signaling architecture that aligns organellar crosstalk, translational control, and metabolic reprogramming during mitochondrial stress.

Defects in oxidative phosphorylation elevate mitochondrial ROS, activating NRF2‐dependent antioxidant programs and perturbing iron–sulfur cluster biogenesis, lipid redox balance and heme metabolism [248, 249]. In this section, hypoxia is framed as a physiological modifier of mitochondrial proteostasis that impacts UPR^mt deployment through changes in import efficiency, redox state, and mitophagy regulation. Hypoxia further reshapes mitochondrial function through stabilization of hypoxia‐inducible factors (HIFs), which reprogram metabolism and electron transport and indirectly influence mitochondrial proteostasis [161, 250, 251]. Genetic and physiological analyses in C. elegans indicate that mitochondrial dysfunction is an early and causative event in hypoxic pathology: sublethal hypoxia rapidly disrupts mitochondrial proteostasis, induces the UPR^mt and elicits hypoxia‐induced mitochondrial protein aggregation (HIMPA), which precedes mitochondrial fragmentation and organismal death [208, 252, 253, 254]. Systematic genetic analyses in C. elegans further reveal a dissociation between mitochondrial proteostasis and hypoxia survival: resistance to hypoxic injury invariably suppresses HIMPA [253], yet hypoxia resistance does not require UPR^mt activation. These data position the UPR^mt as a selective protector against early mitochondrial proteotoxic pathology rather, than a universal hypoxia‐survival program, with response amplitude graded by import efficiency and ISR‐driven transcription factor activation. Hypoxia engages mitochondrial quality control through selective mitophagy. HIF‐dependent regulation of receptors such as FUN14 domain‐containing protein 1 (FUNDC1) directly links oxygen sensing to mitochondrial turnover as reversible phosphorylation of FUNDC1's LC3‐interacting region integrates hypoxic signaling with mitochondrial fission and autophagosome recruitment [255, 256, 257, 258]. In C. elegans, the FUNDC1 orthologue fundc‐1 is transcriptionally induced by hypoxia through HIF‐1 binding sites, and loss of fundc‐1 confers protection against hypoxic injury and death in an ATFS‐1‐dependent manner, indicating that disruption of mitophagy can secondarily elicit adaptive mitochondrial stress signaling [259].

The unfolded protein response emerges as a central organizing layer in this integrated network. Activation of the UPR^ER through PERK, IRE1, and ATF6 imposes global translational attenuation, reducing the burden of nuclear‐encoded mitochondrial precursors during periods of compromised folding and import [260]. PERK‐dependent phosphorylation of eIF2α imposes global translational attenuation, reducing the burden of nuclear‐encoded mitochondrial precursors and thereby lowering mitochondrial import pressure during UPR^mt activation; in parallel, selective translation of ATF4, CHOP and ATF5 reinforces transcriptional programs that overlap with and potentiate UPR^mt outputs [194]. IRE1, enriched at mitochondria‐associated membranes (MAMs), remodels the mRNA landscape through XBP1 splicing and regulated IRE1‐dependent decay (RIDD), plausibly restricting cytosolic pools of mitochondrial‐destined transcripts, while ATF6 indirectly modulates mitochondrial homeostasis by adjusting ER folding capacity, lipid biosynthesis and membrane composition [261, 262, 263, 264, 265]. These ER‐derived inputs converge with intrinsic mitochondrial throttles that limit protein synthesis and assembly during stress.

Beyond canonical UPR^ER signaling, ER‐proximal transcriptional regulators confer tissue‐specific control over mitochondrial proteostasis. In the liver, the ER‐tethered transcription factor cyclic AMP‐responsive element‐binding protein H (CREBH) acts as a physiological activator of the UPR^mt in response to fasting and circadian cues, inducing mitochondrial chaperones, proteases and metabolic regulators via ATF4 and ATF5, independently of classical UPR^ER arms [208]. CREBH activity couples to expansion of ER–mitochondria contact sites, and its loss causes accumulation of unfolded mitochondrial proteins, impaired membrane potential, and redox imbalance [208], underscoring a direct ER–mitochondria axis for maintaining mitochondrial homeostasis. ER stress also remodels mitochondrial architecture: acute PERK activation promotes stress‐induced mitochondrial hyperfusion downstream of translational attenuation [266, 267], providing a transient protective response that preserves bioenergetic competence and suppresses pathological fragmentation. At ER–mitochondria contact sites, Ca²⁺ signaling provides an additional rapid regulatory layer; redox‐sensitive modulation of sarcoplasmic reticulum Ca²⁺‐ATPase (SERCA) pumps, IP3R and ryanodine receptors shapes mitochondrial Ca²⁺ uptake, TCA enzyme activity and, when excessive or prolonged, apoptotic signaling via permeability transition [268].

Mitochondrial stress responses extend beyond the originating cell to engage a cell‐non‐autonomous layer of UPR^mt signaling. In this mode, mitochondrial dysfunction in one tissue triggers adaptive UPR^mt activation in distant tissues through secreted cues, functioning as a mitohormetic mechanism that integrates local mitochondrial stress with systemic metabolic and proteostatic adaptation. In mammals, mild mitochondrial perturbations in discrete tissues, particularly skeletal muscle, activate an ATF4 program that promotes secretion of the endocrine mitokines fibroblast growth factor 21 (FGF21) and growth differentiation factor 15 (GDF15) [240, 269]. These factors remodel whole‐body metabolism by suppressing anabolic input, promoting lipid utilization and enhancing insulin sensitivity, thereby reducing nutrient pressure and proteostatic burden [239, 240, 270]. Mitokine secretion thus reflects a downstream, organism‐level adaptation that arises from sustained UPR^mt–ISR engagement, rather than a canonical transcriptional arm of UPR^mt signaling itself. This mode of cell‐non‐autonomous communication was first defined in C. elegans, where mitochondrial stress originating in neurons or the intestine elicits UPR^mt activation in distal tissues and promotes lifespan extension [271, 272, 273, 274].

Taken together, however, ER stress signaling, hypoxia, and systemic mitokine responses do not supplant the core UPR^mt machinery, but instead shape its deployment by modulating import‐gated sensing, ISR‐dependent translational control, and epigenetic competence across tissues.

In mammals, population genetics and experimental models indicate that modest impairment of mitochondrial translation, for example, reduced expression of mitochondrial ribosomal proteins activates the UPR^mt and correlates with increased lifespan [21]. Crucially, the benefits of cell‐non‐autonomous signaling are dose, timing and tissue dependent: low‐grade, transient stress is adaptive and longevity‐promoting, whereas chronic or severe mitochondrial dysfunction overwhelms proteostatic capacity and promotes metabolic or degenerative pathology.

These observations position the UPR^mt as a modular component of a multimodal stress‐response network operating across subcellular, cellular and systemic scales. ER stress and ER–mitochondria contact sites emerge as pivotal organizers of this network, coordinating translational control, organellar crosstalk and endocrine signaling to align mitochondrial capacity with cellular and organismal demand. We propose that mitochondrial stress signals are funneled through three proximal hubs that together set UPR^mt amplitude and fate outcomes: (i) import‐gated proteostasis sensors (reduced Δψ, MPIS) that activate DELE1–HRI and bias the ISR toward ATF4/ATF5, restraining global translation while preserving UPR^mt executors; (ii) ER–mitochondria contact sites where PERK–eIF2α–dependent attenuation, redox and Ca²⁺ exchange coordinate with mitochondrial dynamics to stabilize bioenergetics under load; and (iii) lysosomal mTORC1–ATF4 phosphorylation, which licenses a mitochondria‐skewed transcriptional output and interfaces with mitophagy routing. In this scheme, low‐to‐moderate import stress favors ATF4/ATF5‐driven proteostasis, antioxidant and metabolic remodeling; sustained import failure or severe depolarization diverts to PINK1–Parkin or FUNDC1 mitophagy modules depending on network state (fusion–fission, hypoxia), whereas unresolved proteotoxicity converges on apoptosis (Fig. 4).

Fig. 4.

Multi‐hub integration of mitochondrial stress signals shaping UPR^mt outcomes. Mitochondrial stress signals are integrated across three proximal hubs that converge on the ISR to determine the amplitude and fate of the mammalian UPR^mt. Import‐gated cues– including reduced Δψ/MPIS, precursor accumulation, and assembly stress– signal import overload, activating translational attenuation through eIF2α phosphorylation. ER–mitochondria contact sites (MERCS) contribute additional regulatory inputs via PERK–eIF2α signaling, Ca²⁺ and redox exchange, and membrane‐associated control of mitochondrial dynamics. In parallel, lysosomal nutrient‐sensing pathways, particularly the mTORC1–ATF4 axis, modulate transcriptional bias toward a mitochondria‐focused program. These hubs converge on the ISR (ATF4/ATF5/CHOP), which shapes the UPR^mt transcriptional output and channels cells toward distinct outcomes: proteostasis reinforcement via mitochondrial chaperones and proteases, mitophagy through pathways such as PINK1–Parkin and FUNDC1, or apoptotic elimination when damage is terminal. Together, these interconnected pathways coordinate mitochondrial recovery, quality control, and cellular fate decisions.

Conclusions

The mammalian UPR^mt constitutes a distributed, multilayered stress‐response circuit that aligns proteostatic demand with folding, assembly, and metabolic capacity. Central to this architecture is import‐gated sensing, whereby reduced preprotein import and mito–nuclear stoichiometric imbalance bias the integrated stress response toward the selective production and nuclear activity of ATF4, CHOP, and ATF5. The resulting transcriptional program expands chaperone–protease networks and transiently throttles the nuclear‐encoded OXPHOS load to reduce proteostatic burden. A second layer, selective translation, enforces proportionality. eIF2α phosphorylation, stress‐granule triage, and miRNA‐mediated tuning suppress import‐dependent mitochondrial cargo while preserving synthesis of stress‐adaptive effectors. Within mitochondria, protease‐ and RNA‐processing checkpoints impose an additional layer of control, restraining nascent chains and safeguarding the assembly of OXPHOS complexes. A third layer, epigenetic licensing, links metabolite availability to chromatin accessibility at UPR^mt loci, thereby defining activation thresholds, response amplitude, and transcriptional decay kinetics.

This mechanistic core is embedded within a broader, interconnected stress landscape that includes ER–mitochondria contact sites, lysosomal nutrient signaling, hypoxia, and mitophagy. Such integration not only supports adaptive recovery but also clarifies failure modes that drive persistent proteotoxicity, inflammatory signaling, or metabolic collapse. The layered architecture of the UPR^mt therefore determines recovery versus degeneration, shaping when and how adaptation becomes maladaptive. Elucidating how these regulatory layers coordinate—and how they fail in disease—will be essential for defining therapeutic strategies to modulate mitochondrial stress, resilience, and organismal aging.

Therapeutically, this framework highlights “tunable” regulatory nodes—import efficiency, ISR kinases and their phosphatases, SG dynamics, CLPP/LONP1‐dependent proteostasis and mitophagy effectors—that can be leveraged to rebalance mitochondrial proteostasis across aging, neurodegeneration, cardiometabolic disease and cancer. By integrating import‐gated logic with translational and epigenetic control, the UPR^mt emerges as a coherent regulatory system whose calibrated modulation may realign cellular energetics and resilience with clinical benefit.

Author contributions

Writing – original draft, R.B., J.F.C.; Writing – review and editing, R.B., J.F.C., P. C., A.W.‐W. Revising original draft, P.C., A.W.‐W., S.K., J.D.

Conflicts of interest

The authors declare no conflict of interest.