Mitochondrial homeodynamics in ageing: mechanisms, resilience, and interventions

Abstract

Mitochondria integrate bioenergetics, redox signalling, calcium handling, biosynthesis, apoptosis, and stress responses. Their contribution to ageing depends less on any single pathway than on the ability to sustain these functions through continuous maintenance, remodelling, and inter-organelle communication. This review proposes mitochondrial homeodynamics as a systems-level framework for that ability, which rests not on static preservation but on three linked capacities. Maintenance safeguards mitochondrial genome, proteome, and membrane integrity. Adaptation adjusts metabolism and remodels network and cristae architecture to match changing demand. Recovery restores function and reserve after challenge. These capacities emerge from mitochondrial quality control, network and cristae remodelling, biogenesis, mitophagy, retrograde stress signalling, and inter-organelle communication. So defined, mitochondrial dysfunction becomes a measurable loss of capacity rather than a descriptive category. Ageing erodes these capacities in tissue- and context-specific ways, which reduces physiological reserve, slows recovery after stress, and amplifies sterile inflammation. The mechanisms underlying these capacities, the biomarkers that report them, and the interventions proposed to preserve them are evaluated in turn. Exercise provides the strongest human evidence for coordinated mitochondrial and functional adaptation, whereas evidence for energy restriction, NAD+ precursors, mitophagy-supporting compounds, and mitochondria-targeted agents remains heterogeneous and endpoint-specific. No mitochondrial intervention has been shown to slow ageing or extend lifespan in healthy humans, and movement of a biomarker towards a younger reference value does not establish rejuvenation. Progress will require dynamic measures of maintenance, adaptation, and recovery, obtained in defined tissues and interpreted alongside clinically meaningful outcomes.

Introduction

Age is the major risk factor for chronic disease and geriatric syndromes, but biological ageing varies across individuals, organs, cell types, and molecular systems (Partridge et al. 2018; López-Otín et al. 2023). Geroscience seeks modifiable processes that contribute to functional decline and multiple age-related diseases rather than treating each disorder as an isolated endpoint (Kennedy et al. 2014; Kritchevsky and Cummings 2025). A useful framework must therefore account for both progressive damage and differences in the capacity to maintain function during repeated metabolic, inflammatory, and environmental challenges.

Mitochondria integrate oxidative metabolism with redox signalling, calcium homeostasis, biosynthesis, apoptosis, innate immunity, and inter-organelle communication (Spinelli and Haigis 2018; Mottis et al. 2019). Their ageing phenotype is not a uniform decline in abundance or respiration, but a context-dependent combination of altered genome maintenance, proteostasis, network architecture, turnover, spatial distribution, metabolic flexibility, and stress responsiveness. Blood-cell respiration, for example, can increase with age and differs by sex (Phang et al. 2026). Mitochondrial homeodynamics is defined here as the systems property through which mitochondrial networks maintain integrity, adapt to demand, and recover after challenge. The term has been used previously for a nutrient-responsive switch between mitochondrial states during development (Zhang et al. 2025). It is used here in a related but distinct sense, for the capacity to sustain and restore such transitions across the lifespan. The concept emphasizes regulated flux, remodelling, and adaptive range rather than a fixed set point, and it follows the biogerontological view that ageing contracts the homeodynamic space despite persistent maintenance, repair, and defence (Rattan 2014). This formulation extends earlier accounts of mitochondrial resilience and stress adaptation (Picard and McEwen 2018; Sharma et al. 2019), and it organizes mitochondrial capacity into three interacting domains: maintenance, adaptation, and recovery. It does not designate a newly discovered pathway or a master mechanism of ageing.

This review examines the systems that support these domains, their alteration during ageing and inflammageing, and the evidence for candidate interventions. It proposes that precision geroscience should assess dynamic reserve and recovery alongside static biomarkers, without equating target engagement or partial normalization of mitochondrial function with slower ageing (Rattan 2020, 2026).

Mechanisms underlying mitochondrial homeodynamic capacity

Mitochondrial homeodynamic capacity depends on coordinated surveillance, renewal, and signalling. Maintenance rests on quality-control systems that preserve genome and proteome integrity and sustain organelle distribution. Adaptation depends on the capacity of mitochondrial networks to remodel their structure and metabolism, activate transcriptional responses, and communicate with other cellular compartments during stress. Recovery requires restoration of adequate function without persistent inflammatory or cell-death signalling. Ageing can reduce this adaptive range despite continued maintenance and repair, even when basal measurements remain relatively preserved (Rattan 2014; Kauppila et al. 2017; Song et al. 2021).

Mitochondrial genome maintenance and proteostasis

Mitochondrial genome integrity is sustained through replication, repair, nucleoid organization, selective turnover, and population-level buffering across many mtDNA copies. Heteroplasmic variants can remain functionally silent until their proportion exceeds mutation- and tissue-specific biochemical thresholds (Stewart and Chinnery 2015). Ageing is associated with the accumulation and clonal expansion of some mtDNA mutations, which arise predominantly from replication error rather than from oxidative damage (Kauppila et al. 2017), although mutation load alone does not determine mitochondrial failure. The phenotypic effect depends strongly on cell lineage, mutation type, energetic demand, and the efficiency with which dysfunctional genomes and organelles are constrained or removed. Recent work further links mitochondrial genome instability to inflammatory signalling, which extends the consequences of defective mtDNA maintenance beyond impaired oxidative phosphorylation (Bahat et al. 2025).

Proteostasis is equally important because most mitochondrial proteins are encoded in the nucleus, synthesized in the cytosol, imported into mitochondria, assembled into multiprotein complexes, and continuously surveyed. The human mitochondrial proteome exceeds one thousand proteins and is organized across subcompartments with distinct import, folding, and degradation requirements (Rath et al. 2021). Chaperones, ATP-dependent proteases, mitochondrial ribosome quality control, the mitochondrial unfolded protein response, and mitophagy form an integrated proteostasis network (Jensen and Jasper 2014; Suomalainen and Battersby 2018; Moehle et al. 2019; Song et al. 2021; Borlepawar et al. 2026). This network is particularly important in long-lived postmitotic cells, where cumulative proteotoxic stress cannot be diluted efficiently by cell division (Coleman et al. 2026). In primary human T cells, age-related loss of mitoribosome abundance and organization has recently been linked to impaired mitochondrial biogenesis and reduced cellular function, illustrating how organelle translation can become a cell-type-specific limiting node in human ageing (Chen et al. 2026).

Network dynamics and spatial organization

Mitochondrial fusion and fission regulate content mixing, segregation of damaged regions, organelle distribution, and adaptation to energetic demand (Friedman and Nunnari 2014; Chan 2020). Neither fragmentation nor elongation is intrinsically beneficial. Functional relevance depends on the capacity to remodel appropriately. Experimental manipulation of either fission or fusion can extend lifespan under specific conditions in model organisms, which argues against treating one morphology as a universal marker of mitochondrial health (Sharma et al. 2019; Traa et al. 2024). The more informative property is network plasticity, which links morphology to turnover, bioenergetics, and stress recovery.

Studies of human skeletal muscle support the importance of structural organization. Mitochondrial fragmentation has been associated with lower physical capacity in older adults, and three-dimensional imaging has identified age-related changes in mitochondrial architecture that correlate with muscle characteristics (Goulding et al. 2025; Scudese et al. 2025). Changes in mitochondria-endoplasmic reticulum contacts can also occur early during age-related muscle dysfunction, with experimental evidence that regular exercise preserves aspects of contact-site organization (Allen et al. 2025). These observations do not establish that morphology is a causal biomarker of organismal ageing, but they show that mitochondrial structure and inter-organelle topology provide information that is not captured by abundance or resting respiration alone.

Spatial organization is especially critical in neurons and other highly polarized cells. Axonal and dendritic transport places mitochondria near sites of high ATP demand and calcium flux, while local quality-control pathways determine whether damaged organelles are repaired, returned to the soma, or removed (Misgeld and Schwarz 2017). With ageing, defects in transport or positioning can therefore produce local energetic failure before whole-cell mitochondrial measurements show major abnormalities. Similar principles apply to cardiomyocytes, skeletal muscle fibres, and stem-cell niches, where mitochondrial position is constrained by cellular architecture and functional demand (Ravindran and Gustafsson 2025). In the brain, mitochondrial dysfunction intersects with synaptic maintenance, proteostasis, inflammatory signalling, and neuronal vulnerability (Bartman et al. 2024), which makes regional and cell-type resolution particularly important when interpreting age-related mitochondrial phenotypes.

Cristae organization and bioenergetic reserve

Mitochondrial architecture also extends to the inner membrane. Cristae provide a spatially organized environment for the respiratory-chain complexes and ATP synthase, and they support metabolite exchange and local electrochemical gradients (Spinelli and Haigis 2018; Chan 2020). Age-related changes in inner-membrane organization can therefore alter respiratory efficiency even when mitochondrial content is preserved. However, cristae density or shape should not be interpreted in isolation because structural remodelling can represent adaptation, compensation, or damage depending on tissue demand and the accompanying metabolic state. The relevant question is whether membrane organization supports sufficient reserve when ATP demand rises.

This distinction is important for human ageing studies. Bioenergetic reserve emerges from the interaction among membrane organization, calcium-sensitive metabolic activation, substrate availability, and network topology, as indicated by three-dimensional analyses of aged skeletal muscle and by evidence that impaired mitochondrial calcium uptake constrains muscle performance (Gherardi et al. 2025; Scudese et al. 2025). No single morphological feature can therefore serve as a universal readout of mitochondrial age. A stronger phenotype combines ultrastructure with respiratory reserve, metabolite handling, and tissue-level performance.

Mitophagy, biogenesis, and mitochondrial turnover

Mitophagy removes damaged or superfluous mitochondria through PINK1-Parkin-dependent and receptor-mediated pathways, while basal mitophagy can proceed independently of PINK1 in several metabolically active tissues (McWilliams et al. 2018; Pickles et al. 2018). Mitophagy is not an isolated degradation process. It is coupled to mitochondrial fission, lysosomal competence, cellular energy status, and compensatory biogenesis. Experimental work in primary cells indicates that suppression of basal mitophagy can promote cellular ageing phenotypes, which supports a causal role for inadequate mitochondrial clearance in some contexts (Kelly et al. 2024).

Removal must be balanced by replacement. Mitochondrial biogenesis requires coordinated nuclear transcription, mtDNA replication, protein import, and organelle assembly, with PGC-1 coactivators and TFAM among its central regulators (Scarpulla 2008; Abu Shelbayeh et al. 2023). Work in Caenorhabditis elegans showed that coordination between mitophagy and biogenesis is necessary for organismal adaptation during ageing (Palikaras et al. 2015; Roussos et al. 2025). In mammalian tissues, the key translational issue is therefore turnover quality rather than maximal turnover rate. Excessive elimination without adequate replacement can reduce energetic capacity, whereas biogenesis without effective quality control can expand a dysfunctional pool.

The link between mitophagy and inflammatory control has become clearer. Impaired mitochondrial clearance can increase cytosolic mtDNA and activate cGAS-STING signalling, whereas enhanced mitophagy can restrain this inflammatory axis in ageing models (Jiménez-Loygorri et al. 2024). Skeletal muscle is a useful example because ageing, exercise, and disuse alter mitochondrial turnover in different directions and at different stages of the remodelling response (Kuznyetsova and Hood 2025). These findings support a dynamic interpretation in which mitophagy should be assessed together with lysosomal flux, biogenesis, and tissue function.

Stress adaptation and inter-organelle communication

Mitochondria communicate continuously with the nucleus and with other organelles. Endoplasmic reticulum-mitochondria contact sites coordinate calcium transfer, lipid synthesis, organelle dynamics, and stress signalling. Lysosome-mitochondria contacts link nutrient sensing and organelle turnover, while contacts with peroxisomes and lipid droplets contribute to fatty-acid handling and redox balance (Murley and Nunnari 2016; Prinz et al. 2020; Wong et al. 2018). Metabolic kinases integrate these contact-site functions with cellular nutrient and stress signals, making inter-organelle communication an important component of ageing biology rather than a structural curiosity (Chowdhury et al. 2026).

Mitochondrial proteotoxic or bioenergetic stress can activate retrograde pathways, including the integrated stress response, which is relayed from the organelle by DELE1 (Fessler et al. 2020). At controlled levels, mitochondrial reactive oxygen species act as signals rather than solely as damaging by-products (Shadel and Horvath 2015; Sies and Jones 2020). This principle underlies mitohormesis, in which transient stress induces adaptation that increases subsequent resistance (Yun and Finkel 2014). The benefit of such signalling depends on stress intensity and duration, tissue, age, and baseline reserve.

Calcium handling is another homeodynamic node because mitochondrial calcium uptake links cellular work to oxidative metabolism. Recent mechanistic evidence indicates that mitochondrial calcium uptake can decline with ageing in skeletal muscle and contribute to reduced energetic performance (Gherardi et al. 2025). This observation fits the broader framework in which ageing can impair the ability to match mitochondrial output to fluctuating demand even when basal respiration is not uniformly depressed.

Mitochondrial metabolites and nuclear signalling

Mitochondrial metabolism influences ageing biology not only via ATP production but also through metabolites that act as substrates, cofactors, or signals for nuclear and cytosolic enzymes. Changes in the NAD+/NADH ratio can alter redox reactions and NAD+-dependent signalling, while acetyl-CoA availability can influence protein acetylation and chromatin state. Tricarboxylic-acid-cycle intermediates such as alpha-ketoglutarate, succinate, and fumarate can affect dioxygenases and other regulatory enzymes. These relationships provide a mechanistic route by which mitochondrial state can reshape gene regulation, stress responses, and cell identity (Aon and Cortassa 2026; Covarrubias et al. 2021; Mottis et al. 2019).

The translational implication is more nuanced than the idea that one metabolite can reset biological age. Metabolite concentrations depend on compartment, flux, substrate use, oxygen availability, and cellular composition. The same metabolite can have different consequences across tissues or disease states. Thus, metabolomic and epigenomic changes can help define mitochondrial adaptation, but causal inference requires perturbation studies that show whether changing a specific mitochondrial process alters downstream gene regulation and ultimately preserves tissue function.

Mitochondrial homeodynamics as an integrated systems property

The preceding mechanisms can be organized into three overlapping domains (Fig. 1). Maintenance comprises genome and proteome quality control, membrane integrity, turnover, and repair. Adaptation comprises metabolic reprogramming, network and cristae remodelling, and stress signalling. Recovery comprises the processes that restore function and reserve after challenge. Mitochondrial homeodynamic capacity is therefore a systems-level construct, not a single pathway, score, or youthful target state. Its value depends on reproducible measurement and association with functional outcomes. So defined, mitochondrial dysfunction is a measurable loss of capacity in one or more domains rather than a descriptive category.

Fig. 1.

Mitochondrial homeodynamic capacity and its proposed relationship to tissue resilience. The central panel shows the three overlapping domains that together constitute mitochondrial homeodynamic capacity. Maintenance comprises genome and proteome quality control, membrane integrity, organelle turnover, and repair. Adaptation comprises metabolic reprogramming, network and cristae remodelling, and stress signalling. Recovery comprises the processes that restore function and reserve after challenge. The domains overlap because individual molecular processes contribute to more than one of them. The six boxes above list the intervention classes evaluated in this review, which differ in the domains they engage and in the strength of the supporting human evidence. The dashed arrow denotes a hypothesized relationship: whether preserving homeodynamic capacity slows functional decline or extends healthspan has not been tested in humans

The concept should not become a label for every favourable mitochondrial change. A valid homeodynamic interpretation must retain context, trade-offs, and the distinction between adaptive acute responses and harmful chronic activation (Rattan 2014). The safeguard against such circularity is to state the framework as predictions that can fail.

Three predictions follow. First, among individuals matched for age, sex, and resting mitochondrial capacity in a given tissue, the amplitude and rate of recovery after a standardized physiological challenge should predict subsequent functional decline, whereas resting measures alone should predict it less well or not at all. Second, an intervention that improves resting mitochondrial biomarkers without improving adaptation or recovery should not improve the functional outcomes attributed to its mitochondrial action, whereas an intervention that improves recovery kinetics should improve function even when resting measures are unchanged. Third, because the limiting domain differs across tissues and individuals, the response to a domain-specific intervention should be predictable from which domain is limiting at baseline, rather than idiosyncratic.

These predictions are testable with methods already applied in human skeletal muscle and blood, although standardized challenge protocols and validated recovery endpoints remain to be established (Ferrucci et al. 2020; Grevendonk et al. 2021). If measures of adaptation and recovery add no predictive information beyond resting mitochondrial phenotypes, and if intervention responses track biomarker change rather than restored adaptive range, then the homeodynamic framing carries no explanatory advantage over conventional accounts of mitochondrial decline and should be abandoned.

Mitochondrial homeodynamics in ageing

Ageing does not produce a universal mitochondrial phenotype. It instead increases the probability that maintenance, adaptation, or recovery becomes limiting in a particular cell or tissue. Defects in mitophagy, proteostasis, biogenesis, network architecture, organelle communication, and metabolic flexibility vary in magnitude (Sun et al. 2016; Kauppila et al. 2017). Reduced reserve may therefore emerge only during exercise, fasting, infection, injury, or another challenge that requires rapid mitochondrial reconfiguration.

Declining quality control and dynamic reserve

Long-lived cells are particularly sensitive to cumulative defects in quality control. In neurons, cardiomyocytes, and skeletal muscle, impaired proteostasis and organelle turnover can progressively increase the fraction of mitochondria that are structurally intact but functionally constrained (Affourtit and Carré 2024; Miwa et al. 2022; Moehle et al. 2019; Ravindran and Gustafsson 2025). Sarcopenia illustrates the interaction between mitochondrial decline, denervation, reduced physical activity, inflammation, impaired regeneration, and altered amino-acid metabolism (Zuo et al. 2025). At the same time, cross-sectional human studies show that mitochondrial phenotypes are not uniform across tissues, sexes, or age strata (Goulding et al. 2025; Phang et al. 2026). This variation should be treated as biological information rather than as measurement noise.

Stem cells provide a complementary example. Quiescent and activated stem cells occupy different metabolic states, and lineage commitment requires coordinated changes in mitochondrial dynamics, biogenesis, and oxidative metabolism (Khacho et al. 2016; Mohrin et al. 2015). Ageing-associated defects in these transitions can impair self-renewal and regenerative responses even when the baseline mitochondrial content of quiescent cells is low. Recent work in aged bone marrow further shows that mitochondrial abundance can identify haematopoietic stem-cell states with distinct self-renewal properties rather than functioning as a simple damage marker (Totani et al. 2025). The relevant phenotype is therefore the ability to transition between states rather than maximal mitochondrial output in one state.

Tissue, sex, and life-course heterogeneity

A major limitation of broad claims about mitochondrial ageing is that different tissues face different constraints. Skeletal muscle must match large changes in energetic demand and preserve contractile function, the heart requires continuous oxidative output, neurons depend on long-distance organelle transport and local calcium buffering, immune cells undergo rapid metabolic reprogramming, and haematopoietic stem cells must balance quiescence with regenerative activation (Affourtit and Carré 2024; Chen et al. 2026; Coleman et al. 2026; Ravindran and Gustafsson 2025; Totani et al. 2025). A process that is limiting in one lineage may therefore be compensable in another. This heterogeneity argues against a single mitochondrial ageing score unless its components are explicitly tissue-aware.

Sex and life stage can also modify mitochondrial phenotypes and intervention responses. Human blood-cell respiration differs by sex across age, and exercise-associated adaptations depend on baseline activity, tissue, and physiological context (Bishop et al. 2025; Phang et al. 2026). Cross-sectional differences should be interpreted cautiously because they mix ageing effects with survivor selection, cohort effects, medication exposure, disease burden, and differences in cellular composition. Longitudinal designs and repeated challenge-response measurements are better suited to identifying trajectories of declining reserve. This point is especially relevant for precision geroscience, where biological heterogeneity is the object of study rather than a source of variance to be averaged away.

The mitochondrial-inflammatory axis

Mitochondrial damage can amplify sterile inflammation when mitochondrial nucleic acids or other danger-associated signals enter inappropriate cellular compartments (Lin et al. 2022; Jindal et al. 2026). Cytosolic mtDNA can activate innate immune pathways, including cGAS-STING and the NLRP3 inflammasome (West et al. 2015; West and Shadel 2017; Zhong et al. 2018). In senescent cells, permeabilization of a minority of mitochondria can release mtDNA and reinforce the senescence-associated secretory phenotype (Victorelli et al. 2023). cGAS-STING signalling has also been implicated causally in age-related inflammation and neurodegeneration in experimental models (Gulen et al. 2023).

Recent evidence extends this axis beyond mtDNA. Mitochondrial genome instability caused by aberrant ribonucleotide incorporation can drive inflammatory signalling (Bahat et al. 2025), and mitochondrial RNA can accumulate in the cytosol of senescent cells and activate RNA-sensing pathways that sustain inflammatory secretion (Victorelli et al. 2025). Taken together, these observations support a broader concept in which defective mitochondrial compartmentalization and clearance can convert organelle stress into chronic innate immune activation. Mitophagy can interrupt this cycle by limiting the persistence and cytosolic exposure of damaged mitochondrial material (Jiménez-Loygorri et al. 2024).

Inflammation also acts in the opposite direction. Chronic cytokine exposure, altered nutrient signalling, and oxidative stress can impair mitochondrial biogenesis, dynamics, and quality-control pathways, which creates a self-reinforcing loop between mitochondrial dysfunction and inflammageing (Bektas et al. 2018; Franceschi et al. 2018; Miwa et al. 2022). This bidirectionality is important therapeutically because suppression of one inflammatory mediator does not necessarily restore the underlying mitochondrial maintenance defect, and correction of one mitochondrial parameter may not extinguish a sustained inflammatory programme.

Intercellular and systemic mitochondrial signals

Mitochondrial quality control also extends beyond the boundaries of a single organelle. Mitochondria-derived vesicles can transport selected proteins, lipids, and other cargo to lysosomes, peroxisomes, or extracellular compartments, providing an additional route for organelle surveillance and intercellular communication (Ferrucci et al. 2024; König and McBride 2024). The significance of these vesicles in human ageing is still being defined, but they broaden the concept of mitochondrial homeodynamics from intracellular turnover to systemic signalling. Experimental approaches that transfer intact mitochondria or mitochondrial material are also being explored in regenerative medicine, although delivery, integration, immune compatibility, durability, and regulatory issues remain major barriers to clinical translation (Kubat et al. 2025).

Mitochondrial stress is also communicated systemically through secreted proteins. Activation of the integrated stress response (Fessler et al. 2020) induces GDF15 and FGF21, both of which circulate and are used as diagnostic markers of primary mitochondrial disease (Lehtonen et al. 2016; Yatsuga et al. 2015). Circulating GDF15 increases with age and has been associated with mortality, frailty, and functional decline in older cohorts (Conte et al. 2022; Wiklund et al. 2010). These proteins are currently the most accessible human readouts of mitochondrial stress signalling. They are not mitochondria-specific, however, because concentrations also respond to inflammation, renal function, malignancy, and common medications including metformin (Coll et al. 2020). They therefore report activation of a stress pathway rather than the capacity of the underlying domains, and they should be interpreted alongside tissue-level and challenge-based measures rather than as substitutes for them.

Preserving and modulating mitochondrial homeodynamic capacity

Mitochondrial phenotypes are modifiable, but support for individual interventions is unequal. An intervention may change a molecular marker, improve an organ-specific function, or reduce a disease risk factor without slowing biological ageing. The sections below therefore distinguish mechanistic rationale, human functional evidence, and direct evidence for modification of ageing biology (Justice et al. 2016; Kritchevsky and Cummings 2025). They can also be read against the three domains. Mitophagy activators and mitochondria-targeted redox or membrane compounds act principally on maintenance, pharmacological nutrient-sensing modulators act principally on adaptation, and exercise and energy restriction, which are repeated physiological challenges, engage all three. Almost none of these interventions has been evaluated for effects on recovery after a standardized challenge, which is the gap the predictions above are intended to expose.

Physical exercise

Exercise provides the strongest human evidence for coordinated improvement in mitochondrial function and physical capacity. In skeletal muscle, training activates AMPK- and calcium-sensitive signalling, PGC-1-coordinated transcription, mitochondrial protein synthesis, antioxidant defences, and turnover (Bishop et al. 2025; Grevendonk et al. 2021). Exercise is a repeated multidimensional challenge followed by recovery, not a single-target therapy. Its benefits reflect integrated adaptation across tissues and cannot be inferred from one downstream pathway.

Responses vary by modality, intensity, baseline health, age, and tissue. Endurance training generally increases oxidative capacity, while resistance training supports muscle mass and strength. Human studies indicate that exercise can preserve mitochondrial network architecture and organelle contact sites (Allen et al. 2025; Scudese et al. 2025), and physical activity can modify immune-cell metabolism in older adults (Hazeldine et al. 2025). These effects support functional adaptation but do not establish reversal of mitochondrial ageing.

Frailty, pain, disability, and multimorbidity can limit structured exercise in older adults, supporting the development of adjunctive strategies. However, reproducing one exercise-responsive pathway cannot substitute for the integrated mechanical, metabolic, vascular, immune, neural, and behavioural stimulus. Candidate interventions should therefore be judged by safe functional benefit rather than resemblance to selected molecular effects of exercise (Rattan 2020).

Dietary and nutrient-sensing interventions

Energy and nutrient restriction engage AMPK, mTOR, insulin-IGF signalling, autophagy, and stress responses (de Cabo and Mattson 2019; Longo and Anderson 2022). Caloric restriction extends lifespan in several model organisms. In the CALERIE trial, sustained caloric restriction improved cardiometabolic risk factors and induced metabolic adaptation, but the study did not test human lifespan (Kraus et al. 2019; Redman et al. 2018). Fasting-mimicking and time-restricted approaches can modify metabolic markers, although evidence for durable age-related clinical outcomes remains incomplete (Wei et al. 2017).

Nutrient composition can influence ageing pathways independently of energy intake. Restriction of specific amino acids or protein alters metabolism, frailty, and lifespan in rodents, with effects modified by sex and life stage (Green et al. 2023; Richardson et al. 2021; Yeh et al. 2024). These findings should not be extrapolated directly to older adults because frailty, sarcopenia, and malnutrition can make sustained restriction harmful. Nutritional strategies must preserve lean tissue, bone health, and adequacy.

NAD+ metabolism and mitochondrial signalling

NAD+ links redox metabolism with sirtuin activity, DNA repair, and stress signalling, and its metabolism changes with age (Covarrubias et al. 2021). Nicotinamide riboside and nicotinamide mononucleotide increase NAD+-related metabolites in humans, but effects on metabolic, vascular, inflammatory, and functional outcomes are inconsistent (Elhassan et al. 2019; Yoshino et al. 2021). Biochemical target engagement establishes biological activity rather than geroprotection, and NAD+ precursors remain unproven as interventions that slow human ageing.

Mitophagy-supporting and autophagy-related strategies

Urolithin A induces mitophagy in experimental systems and produces molecular evidence of target engagement in humans (Andreux et al. 2019; Ryu et al. 2016). A randomized trial in older adults reported improvements in selected muscle-endurance outcomes with mitochondrial biomarker changes, but the magnitude and durability of clinical benefit remain uncertain (Liu et al. 2022). These findings support further study, not claims of organismal rejuvenation.

Spermidine promotes autophagy and longevity in model organisms, and higher dietary intake is associated with lower mortality (Eisenberg et al. 2009; Kiechl et al. 2018). A randomized trial of spermidine supplementation in older adults with subjective cognitive decline did not improve its primary cognitive endpoint, and no reproducible geroprotective effect has been established in humans (Schwarz et al. 2022). Translation of mitophagy modulation requires evidence of tissue benefit without maladaptive mitochondrial depletion or other long-term effects.

Mitochondria-targeted redox, membrane, and metabolic approaches

Mitochondria-targeted compounds test whether selected redox or bioenergetic defects are modifiable in humans. MitoQ improved endothelial function in a small study of healthy older adults (Rossman et al. 2018). Elamipretide acutely increased skeletal-muscle ATP production, but a longer heart-failure trial did not improve its primary endpoint (Butler et al. 2020; Roshanravan et al. 2021). These results show why short-term target engagement cannot be generalized to organismal ageing, particularly when reactive species also mediate adaptive signalling.

Metformin and mTOR inhibition affect nutrient sensing rather than mitochondria alone. Both have extensive mechanistic support, but neither is a validated anti-ageing therapy for healthy humans. Metformin attenuated some exercise-induced mitochondrial adaptations in older adults (Konopka et al. 2019). mTOR inhibition has improved selected immune outcomes, although results vary across endpoints (Kell et al. 2026; Mannick et al. 2021). The metformin finding illustrates a problem that a homeodynamic reading makes explicit. An agent can engage a plausible target while reducing the adaptive response to a stimulus of established benefit, so combinations of interventions require testing for interference rather than assumed additivity.

Senescence, inflammageing, and indirect mitochondrial rescue

Senescent cells can alter neighbouring tissues through the senescence-associated secretory phenotype and mitochondrial nucleic-acid signalling (Gorgoulis et al. 2019; Victorelli et al. 2023). Senolytic and senomorphic strategies may therefore reduce inflammatory pressure on mitochondrial maintenance (Kirkland and Tchkonia 2020). Early human studies show feasibility in selected diseases but not anti-ageing efficacy (Justice et al. 2019). Effects will probably be tissue specific because senescent cells also contribute to wound repair and tumour suppression in some contexts.

Translation, safety, and intervention windows

Interventions must be evaluated in the physiological context of older adults. Frailty, multimorbidity, polypharmacy, altered pharmacokinetics, reduced renal or hepatic reserve, and nutritional vulnerability can alter both benefit and risk. Suppression of growth signalling, changes in substrate availability, or increased organelle turnover may have different consequences in middle age than in late life with sarcopenia or acute illness. Preclinical lifespan extension therefore cannot substitute for human safety and functional evidence (Justice et al. 2016; Rattan 2020).

Intervention timing also matters. Some mechanisms may be most modifiable before extensive tissue damage, whereas pathways governing inflammation, immunity, or recovery may remain actionable later. Human trials show that pathways can be engaged in older adults, but organ-specific effects cannot be generalized to whole-body ageing (Farr et al. 2024; Mannick et al. 2021). Development should proceed from target engagement to organ function and validated healthspan outcomes, with safety assessed at each stage.

Mitochondrial homeodynamics in precision geroscience

Precision geroscience requires measures that separate chronological age from variation in reserve, vulnerability, and recovery (Ahadi et al. 2020; Ferrucci et al. 2020). Multi-omic and plasma-proteomic studies show that ageing trajectories differ among individuals and organs (Oh et al. 2023). Cross-species analyses can help determine whether mitochondrial phenotypes are upstream, downstream, or parallel to broader ageing programmes (Tyshkovskiy et al. 2026). This heterogeneity argues against assuming a universal mitochondrial age or intervention response.

Static measures such as resting respiration, mitochondrial copy number, or a circulating marker are informative but cannot fully capture a response property. Standardized challenge-based phenotyping could quantify the amplitude and kinetics of adaptation and recovery, using measures such as respiratory reserve, substrate switching, mitophagic flux, redox recovery, and functional performance. Such measures require validation for reproducibility, tissue specificity, sex effects, and prediction of clinical outcomes (Grevendonk et al. 2021; Phang et al. 2026).

Tissue access remains a major limitation. Muscle biopsy permits detailed phenotyping but is invasive, while blood-cell assays may not represent brain, heart, or muscle. Circulating mtDNA, extracellular vesicles, proteomic signatures, metabolites, and stress-induced mitokines such as GDF15 are influenced by inflammation, exercise, infection, injury, and pre-analytical handling. Multimodal frameworks should therefore integrate molecular signals with organ function and longitudinal trajectories rather than rely on one surrogate (Ferrucci et al. 2020; Oh et al. 2023).

Operationally, maintenance can be assessed through genome and proteome integrity, turnover, and organelle quality. Adaptation can be assessed through metabolic reprogramming, stress signalling, and network remodelling, and recovery through the time required to restore function after challenge. This is a hypothesis-generating framework, not a validated composite biomarker. Its value should be tested by whether it improves prediction of resilience, frailty, mobility loss, or multimorbidity beyond established measures (Ferrucci et al. 2020; Kritchevsky and Cummings 2025). Testing the predictions set out above depends on precisely these measurements, which is at present the main practical obstacle to evaluating the framework.

Causal attribution in geroscience interventions

The central challenge in geroscience is causal attribution. If an intervention improves walking speed, glucose control, respiration, and inflammatory markers, the changes may arise from a shared upstream mechanism, parallel mechanisms, or secondary effects of increased activity (Kennedy et al. 2014; Kritchevsky and Cummings 2025).

Mediation studies are therefore needed to determine whether mitochondrial changes contribute to clinical benefit or merely accompany it, especially for pleiotropic interventions such as exercise, energy restriction, and nutrient-sensing modulation. The term geroscience intervention should describe a hypothesis, not confer therapeutic status in advance (Rattan 2020).

Exceptional longevity is consistent with preserved resilience but does not establish mitochondrial homeodynamics as its cause (Andersen et al. 2012). Likewise, mitochondrial biomarker improvement is not rejuvenation without durable evidence of better function or reduced vulnerability. Even healthspan requires an explicit operational definition (Rattan 2026). Conservative terminology improves testability and credibility.

Cell-state manipulation, organelle transfer, and mitochondrial engineering can test constraints inaccessible to lifestyle interventions, but remain predominantly preclinical or early translational (Kubat et al. 2025). Development requires tissue targeting, durable integration, immune compatibility, genomic and epigenomic stability, assessment of tumour risk, and benefit beyond short-term molecular endpoints. These requirements determine whether an experimental manipulation can credibly preserve human function.

Trial design: from target engagement to healthspan

Clinical trials should state the evidentiary step being tested. Early studies can establish feasibility, safety, target engagement, and short-term molecular or physiological effects, but not modification of ageing. Subsequent trials should determine whether target engagement produces durable changes in organ function, resilience, or validated outcomes across more than one age-related domain (Justice et al. 2016; Kritchevsky and Cummings 2025). The interpretation should not exceed the endpoint measured.

Dynamic outcomes may be especially informative because mitochondrial reserve often becomes apparent only under challenge. Recovery after exercise, metabolic stress, vaccination, or another standardized challenge could complement resting biomarkers. These outcomes should be paired with established measures of mobility, cognition, immune function, cardiometabolic health, frailty, and adverse events and analysed within a prespecified causal framework (Ferrucci et al. 2020).

Because lifespan trials are rarely feasible initially, longer studies may use incident multimorbidity, disability-free survival, major age-related disease events, or validated measures of intrinsic capacity and frailty. Molecular biomarkers can add mechanistic information, but should serve as primary or co-primary outcomes only after analytical validity, biological interpretation, and responsiveness to intervention have been established (Kritchevsky and Cummings 2025).

More mitochondrial activity is not invariably beneficial. Excessive substrate oxidation, persistent stress signalling, uncontrolled biogenesis, or defective removal can be maladaptive. The objective is regulated flexibility rather than maximization. Mitochondrial homeodynamics therefore describes the capacity to change appropriately and recover sufficiently, not a mandate to maximize respiration, NAD+, mitophagy, biogenesis, or any other single parameter.

Research priorities and unresolved questions

Five questions should define the next phase of mitochondrial geroscience:

• (1) Which measurements reproducibly capture mitochondrial maintenance, adaptation, and recovery in humans?

• (2) Do these measurements predict functional decline independently of chronological age and established risk factors?

• (3) Which evidentiary claim is a given trial designed to support: target engagement, organ-specific functional improvement, modification of an ageing-related risk process, or reduction in clinical events? These outcomes are related but not equivalent.

• (4) How do sex, ancestry, tissue, baseline fitness, nutritional status, and comorbidity modify mitochondrial phenotypes and intervention responses (Ahadi et al. 2020; Phang et al. 2026)

• (5) What do negative results and failed extrapolations reveal about the validity and limits of the proposed model (Rattan 2024)

Experimental design must also distinguish restoration from compensation. Function may recover through alternative substrates, increased mitochondrial mass, altered calcium handling, or changes in cell composition without correction of the initiating lesion. Conversely, a molecular defect may improve while another component continues to limit organ function. Studies should therefore combine molecular assays with physiology and clinically meaningful outcomes (Ferrucci et al. 2020; Justice et al. 2016).

Conclusions

Mitochondrial ageing is better described as erosion of coordinated maintenance, adaptation, and recovery than as uniform loss of ATP production or a simple increase in oxidative stress. Genome maintenance, proteostasis, network dynamics, turnover, calcium handling, stress signalling, and inter-organelle communication jointly determine mitochondrial responses to demand. Their failure can reduce tissue reserve and amplify inflammageing, but mitochondrial changes may be initiating, amplifying, compensatory, or consequential depending on tissue and context. Exercise provides the strongest human evidence that mitochondrial and physiological reserve remain modifiable in later life. Nutritional, NAD+-directed, mitophagy-supporting, mitochondria-targeted, and broader geroscience interventions have plausible targets but less consistent evidence for durable clinical benefit. None has established extended human lifespan or generalized slowing of ageing. Mitochondrial homeodynamics will be useful only if operationalized through dynamic, tissue-aware measures that improve prediction of resilience, frailty, and multimorbidity. The translational aim is to preserve adaptive capacity and function without disrupting compensatory biology, not to restore every mitochondrial feature to a youthful value.

Author contributions

P.P.C. wrote the entire manuscript text and prepared the Fig. 1.

Data availability

No datasets were generated or analysed during the current study.