Mitochondria represent the only organelles within human cells that possess their own DNA (mtDNA). mtDNA is a circular double-stranded molecule approximately 16.6 kb in length that encodes 37 genes: 13 core subunits of the OXPHOS complexes, 22 tRNAs and 2 rRNAs that are necessary for protein translation within mitochondria^[5]. However, the vast majority (> 99%) of the ~1,500 mitochondrial proteins are encoded by the nuclear genome (nDNA); these proteins are synthesized in the cytosol and are then transported into mitochondria via dedicated translocation mechanisms.

This dual genetic control dictates the diverse inheritance patterns observed in PMDs. mtDNA mutations are transmitted maternally and may occur in coding regions (directly affecting subunit structure) or in regulatory regions (impairing transcription and translation). As every cell carries hundreds to thousands of copies of mtDNA, the coexistence of mutant and wild-type mtDNA, termed heteroplasmy, is a hallmark. When the proportion of mutant mtDNA exceeds a tissue-specific bioenergetic threshold, dysfunction ensues—a phenomenon known as the threshold effect, which accounts for much of the clinical variability^[6,7]. To date, over 100 confirmed pathogenic point mutations and rearrangements (mainly deletions) of mtDNA have been identified, among which the most frequent are m.3243A>G, m.8344A>G, and m.8993T>G^[7]. Mutations in nDNA follow the Mendelian inheritance pattern, including autosomal dominant, autosomal recessive or X-linked inheritance, and exert broader effects, not only involving structural subunits of OXPHOS but also genes critical for mitochondrial biogenesis, dynamics (e.g., OPA1), protein import (e.g., TOMM40), respiratory supercomplex assembly (e.g., COX10), mtDNA replication and repair (e.g., POLG), and coenzyme Q10 biosynthesis^[7-9]. Currently, over 1,000 nuclear genes encoding proteins related to mitochondrial function have been identified, among which approximately > 400 are clearly associated with pathogenic mutations. Epidemiological data indicate that approximately 80% of childhood-onset PMDs are attributable to nDNA defects, whereas about 60% of adult-onset cases are linked to mtDNA mutations^[7].

The ultimate clinical manifestations of PMDs are remarkably tissue-specific, stemming from differential tissue reliance on adenosine triphosphate (ATP) and variable susceptibility to mitochondrial dysfunction. The brain, cardiac muscle, skeletal muscle, and retinal photoreceptors—owing to their high energy demand, low reserves, and limited regenerative capacity—are the most frequently affected systems^[10]. Furthermore, the variable heteroplasmic burden across various tissues within the same individual accounts for the striking phenotypic heterogeneity observed with identical gene mutations, ranging from Leigh syndrome to MELAS and MERRF (myoclonic epilepsy with ragged red fibers)^[11].

Mitochondrial quality control (MQC) constitutes a sophisticated interconnected network that surveils mitochondrial integrity, counteracts damage or stress, and sustains mitochondrial homeostasis. This system coordinates multiple cellular processes, such as mt-ISR, mitochondrial biogenesis, mitochondrial dynamics, mitophagy, and intercellular mitochondrial transfer^[12]. MQC serves a critical function in a large variety of human disorders, encompassing cancer, cardiovascular diseases and neurodegenerative diseases^[13].

The core pathological hallmark of PMDs is impaired oxidative phosphorylation (OXPHOS) within the mitochondrial respiratory chain, resulting in insufficient ATP production^[4]. OXPHOS is the primary aerobic energy-producing process, executed by the electron transport chain (ETC) complexes I to IV, which are embedded in the inner mitochondrial membrane, working in concert with ATP synthase (Complex V). In PMDs, a defect in any single complex disrupts this finely tuned system. For instance, Complex I deficiency impairs nicotinamide adenine dinucleotide (NADH) oxidation, reducing electron flux and proton pumping, thereby lowering Δp^[14]. Complex III or IV deficiencies directly block electron transfer to the terminal acceptor, similarly dissipating Δp. Notably, although Complex V dysfunction does not directly affect electron transfer, it reduces proton utilization, leading to a relative increase in Δp, which in turn exerts negative feedback on upstream electron transport, further impairing energy output^[15]. The dissipation of Δp not only cripples ATP production but also compromises other Δp-dependent processes, including mitochondrial protein import (requiring membrane potential), calcium uptake, and metabolite transport, thereby triggering a global cellular energetic failure^[4].

Bioenergetic impairment initiates a self-perpetuating cascade wherein ETC blockage causes upstream hyper-reduction^[16], prolonging electron-transfer intermediate lifetime at Complex I (FMN) and Complex III (Q₀) and promoting electron leakage to O₂•^-)^[17]. Subsequent dismutation to H₂O₂ and Fenton-mediated conversion to •OH^[18]. Collectively, constitute reactive oxygen species (ROS).

This ROS burst inflicts spatially targeted damage. Owing to its histone-depleted state and proximity to the ETC, mtDNA is highly vulnerable to hydroxyl radical attack, generating 8-oxoguanine (8-oxoG) and oxidative lesions that cause base mispairings, point mutations, and strand breaks during replication. This impairs respiratory chain subunit synthesis, establishing a self-amplifying cycle: ETC deficiency → electron leakage → ROS → mtDNA damage → aggravated ETC deficiency^[19]. The inner membrane, rich in cardiolipin and polyunsaturated fatty acids, undergoes lipid peroxidation upon ROS attack, altering membrane fluidity and permeability to cause proton leakage and Δp dissipation, while disrupting cardiolipin binding to ETC complexes and destabilizing respirasome assembly^[20]. Oxidative modifications of fusion proteins OPA1 and MFN2 further drive mitochondrial fragmentation and metabolic failure^[21].

When increased oxidative stress and depleted ATP accumulate beyond a critical threshold, cells execute downstream injury programs. When permeability of the outer mitochondrial membrane is increased, cytochrome c and apoptosis-inducing factor (AIF) are released into the cytosol, activating both caspase-dependent and caspase-independent apoptotic pathways^[22,23]. As an iron-dependent, lipid-peroxidation-driven form of regulated cell death, ferroptosis has been recently implicated in neuronal loss in specific PMD subtypes characterized by glutathione depletion and extensive lipid peroxidation^[24,25]. Grasping the crosstalk between apoptosis and ferroptosis is critical for developing targeted treatments for mitochondrial disorders. A combination of mitochondria-targeted antioxidants, GPX4 modulators, and mPTP inhibitors, paired with personalized diagnostic frameworks, constitutes the most promising therapeutic approach to date. Further in-depth exploration of the apoptosis-ferroptosis interaction is key to driving progress in the clinical management of mitochondrial diseases.

Upon stress or damage, mitochondria release damage-associated molecular patterns (DAMPs) including mtDNA and cardiolipin. They are recognized by microglial Toll-like receptor 9 (TLR9) and the NLRP3 inflammasome, activating innate immune responses and sustaining chronic neuroinflammation. This inflammatory state not only directly harms adjacent neurons but also exacerbates the local microenvironment through reactive astrogliosis, driving disease progression independently of the primary energetic defect. mtDNA-mediated inflammation is primarily driven by the cGAS (cyclic GMP-AMP synthase)-STING (stimulator of interferon genes) signaling pathway^[26]. In addition to the cGAS-STING pathway, the mtROS and mtDNA can activate NLRP3 inflammasome to promote IL-1β and IL-18 cleavage. A link between inflammation and mitochondrial bioenergetics has already been confirmed. Studies indicate that mitochondrial dysfunction can exacerbate inflammation, and the exacerbated inflammation in turn impairs OXPHOS and disrupts MQC^[27]. Histopathological analyses of postmortem human brain tissue and mouse models of PMDs indicate that neuroinflammation is not merely a consequence of neurodegeneration but a potential pathogenic driver of disease progression warranting further investigation. This insight may pave the way for novel therapeutic strategies for these devastating disorders, which currently lack effective treatments^[28].

In summary, the pathogenesis of PMDs is a multi-dimensional process initiated by genetic defects, centered on OXPHOS dysfunction, amplified by oxidative stress and impaired mitochondrial dynamics, and culminating in cell death and inflammatory injury. A thorough elucidation of these mechanisms not only enhances our understanding of disease heterogeneity but also provides a solid theoretical foundation for developing novel therapeutic strategies targeting mitochondrial biogenesis, ROS scavenging, inflammation modulation, and gene editing.

The mitochondrial medicine society (MMS) initially published its guidelines for the management of mitochondrial medicine therapies in 2009. In 2015, the MMS released a consensus statement that reviewed the existing literature on mitochondrial diseases and provided recommendations for the optimal diagnosis and treatment of PMDs^[30]. Subsequently, a comprehensive update in 2020 offered detailed dosage protocols for the clinical treatment of these diseases^[29], alongside an international Delphi-based consensus on drug safety in patients with PMDs^[31]. Most recently, an Australian adaptation of the MMS recommendations, titled Patient Care Standards for PMDs in Australia, was published^[32]. While multimodal therapeutic strategies (Cocktail therapy) are currently prevalent in the management of PMDs, the rapid evolution of targeted medication offers a promising new direction for future treatment paradigms.

Additionally, several other agents are in preclinical or early clinical evaluation, including resveratrol/SRT2104 (SIRT1 activators), cysteine donors (N-acetylcysteine, whey protein), among others^[1]. The table below summarizes the clinical trial status of major therapeutic agents.

Viewed thematically, the targeted therapies for mitochondrial disease cluster into three broad categories. First, cofactor pathway restoration—exemplified by riboflavin, nicotinamide, thiamine, CoQ10, and biotin—represents the most immediately translatable strategy, capitalizing on the high plasticity of vitamin-responsive enzymology. Second, gene-product-independent compensation seeks to bypass the genetic defect by supplying downstream substrates (deoxycytidine/deoxythymidine in TK2 deficiency), providing alternative fuels (L-glutamine/βHB in MPC deficiency), or activating parallel protective pathways (NRF2 activation by omaveloxolone). Third, organelle-directed biophysical stabilization (elamipretide) and dietary metabolic reprogramming (valine/fat restriction) offer complementary avenues for selected disorders. While cofactor replacement and dietary modification currently provide the broadest applicability, the approval of mechanism-based agents such as omaveloxolone, elamipretide, and deoxycytidine/deoxythymidine signals a maturing field in which therapy is increasingly dictated by molecular diagnosis rather than phenomenology alone.

Since mitochondrial diseases often involve multiple systems, clinical practice requires collaboration among multidisciplinary teams including neurology, endocrinology, cardiology, nephrology, hepatology, respiratory medicine, and gastroenterology^[32,64]. Current symptomatic management of mitochondrial diseases primarily involves targeted pharmacological or interventional therapies for affected systems or organs.

The ketogenic diet (KD) is the first-line treatment for patients with pyruvate dehydrogenase deficiency^[72]. KD may be effective for some refractory epilepsy patients with MELAS^[73]. However, individuals with a confirmed diagnosis of fatty acid oxidation disorders or pyruvate carboxylase deficiency should be categorically excluded from KD interventions, as this dietary pattern may precipitate life-threatening metabolic decompensation. For patients with certain fatty acid metabolic disorders, special dietary interventions are required, and consultation with a clinical nutrition specialist is necessary. For patients with significantly elevated lactate levels, it is recommended to restrict the intake of sugars and carbohydrates.

Individualized diets are recommended based on the specific metabolic defect—high-fat, low-carbohydrate for pyruvate dehydrogenase deficiency; fat restriction for carnitine deficiency; and high-protein, high-carbohydrate, low-fat for pyruvate carboxylase deficiency^[20]. General principles include avoiding fasting, maintaining adequate caloric intake, and avoiding high-fat, low-carbohydrate diets (except for specific indications)^[20].

Patients should receive annual assessments for potential musculoskeletal complications, including kyphoscoliosis, joint contractures, dislocations and limb deformities, with particular focus on individuals with pre-existing abnormalities in muscle tone, muscle strength or neurological function^[32]. Exercise helps improve mitochondrial function and reduce abnormal mtDNA load. Children with mitochondrial disorders are prone to exercise intolerance due to their lower maximal oxygen uptake. Therefore, a personalized rehabilitation training program should be developed under the guidance of a professional rehabilitation therapist, starting with low-intensity and short-duration sessions, and progressing gradually. However, this is not recommended during acute disease phases or when fasting.

Recombinant adeno-associated virus (rAAV)-based gene replacement therapies targeting nuclear gene defects in PMDs have been successfully validated in more than 10 distinct preclinical mouse models of PMDs, including mouse models of NDUFS3, NDUFS4, SURF1, TYMP, TK2, MPV17, AIFM1, ANT1 deficiency, et al.^[74]. This landmark progress is largely attributed to the advent of next-generation rAAV platforms with greatly enhanced tropism for therapeutically relevant organs. However, to date, none of these gene reconstitution-based therapeutic modalities have successfully progressed to formal clinical evaluation in human participants, except Friedreich Ataxia^[75]. The greatest progress of gene therapy in PMD has been made for mtDNA replacement such as LHON^[76,77].

Friedreich ataxia, an autosomal recessive and uniformly fatal monogenic disorder resulting from pathogenic variants in the FXN (frataxin) gene, is a progressive multi-system condition predominantly affecting the nervous system and heart, where cardiac complications account for the vast majority of disease-related mortality. LX2006 is an AAV-based gene therapy developed by Lexeo Therapeutics, delivered intravenously to provide a functional frataxin gene, promoting frataxin protein expression and restoring mitochondrial function in cardiomyocytes^[75]. In Phase I/II clinical trials and an investigator-initiated Phase Ia trial, LX2006 was well tolerated with no treatment-related serious adverse events, and cardiac biomarkers showed clinically meaningful improvements that increased over time. Among 8 participants followed for at least 6 months, FXN protein expression levels increased, and 3 out of 4 subjects with elevated left ventricular mass index at baseline showed a > 10% reduction at 12 months^[75].

Despite their promise, few gene reconstitution approaches have yet to enter clinical trials, largely due to three translational barriers. Firstly, poor phenotypic concordance between animal models and humans limits the reliability of efficacy predictions. Secondly, maintaining long-term transgene expression in target organs proves difficult. Thirdly, the requirement for disease-specific vectors makes this strategy unfeasible for the vast spectrum of more than 350 mitochondrial disorders.

As a rare mitochondrial disorder, LHON presents with severe bilateral sequential vision loss, caused by the irreversible selective degeneration of retinal ganglion cells (RGCs). The three canonical mtDNA point mutations—m.3460G>A in MT-ND1, m.11778G>A in MT-ND4, and m.14484T>C in MT-ND6—represent the overwhelming majority of all known LHON genetic triggers, collectively explaining ~90% of confirmed patient cohorts across global populations. LHON gene therapy is attractive due to its monogenic cause, acute onset allowing a treatment window, easy intravitreal access to RGCs, ocular immune privilege, and proven AAV safety^[76,77]. Lumevoq (lenadogene nolparvovec), is an AAV-based gene therapy delivered via intravitreal injection of the wild-type ND4 gene. Its innovative design fuses the AAV capsid VP2 with a mitochondrial targeting sequence, enabling AAV to target mitochondria and achieve allotopic expression of ND4. Relative to the visual outcomes of an externally recruited natural history control cohort, aggregated data across four phase 3/2 interventional trials (RESCUE, REVERSE, RESTORE and REFLECT) confirm that unilateral intravitreal administration of a single dose of lenadogene nolparvovec drives durable, bilateral visual function recovery in m.11778G>A LHON patients ≥ 15 years old treated within the 12-month post-onset therapeutic window^[78-80]. Despite a modest overall effect size, treated patients achieved final visual acuity outcomes that are statistically and clinically distinct from the untreated LHON natural history trajectory, with a magnitude of clinical benefit superior to that observed in m.11778G>A carriers receiving idebenone as monotherapy^[81].

Due to the extremely low permeability of the inner mitochondrial membrane to nucleic acids, the conventional CRISPR-Cas9 system cannot be used for mtDNA editing (as guide RNAs cannot be delivered into mitochondria). In recent years, RNA-independent base editing technologies have opened entirely new avenues. This base-editing platform supports seamless, lesion-free precision modification of the mitochondrial genome, in stark contrast to the undesired, broad mtDNA copy depletion initiated by targeted programmable nuclease cleavage. This unique non-destructive editing modality creates transformative opportunities for both dissecting mitochondrial genetic pathologies and advancing curative interventions for mitochondrial disorders.

Mitochondrial transplantation represents a highly promising therapeutic strategy for the clinical management of mitochondrial diseases. Professor Xingguo Liu’s team developed a significantly improved, high-efficiency mitochondrial transplantation approach by encapsulation of intact functional mitochondria within biomimetic vesicles derived from purified erythrocyte plasma membranes. This method was successfully applied to deliver mitochondria into the tissues and cells of mice and monkeys. The study demonstrates the therapeutic potential of these mitochondrial vesicles for treating PMDs and proposes a novel "organelle therapy" strategy for regenerative medicine^[91].

MRT is also a preventive reproductive strategy focusing on preventing the transmission of pathogenic mtDNA from mother to offspring. This technique involves transplanting the nucleus of a fertilized egg or oocyte carrying defective mtDNA into a donor egg cell with healthy mtDNA from which the nucleus has been removed, resulting in offspring carrying the parents' nDNA and the donor's mtDNA (> 99.8% from parents, approximately 0.1% from the donor) [Figure 2]. The world's first clinically documented live birth resulting from maternal spindle transfer (MST), the foundational MRT platform, was reported in 2016^[92]. Embryo transfer led to a successful full-term pregnancy and uncomplicated delivery of a healthy newborn, and subsequent neonatal tissue assays measured mtDNA mutation burden within the range of 2.36% to 9.23%, and he is in good health at 8 years old now. Recently, Louise Hyslop, Robert McFarland, and their colleagues used pronuclear transfer (PT) to implant healthy embryos into women in whom all copies or a large proportion of their mtDNA included mutations^[93,94], the levels of the maternal pathogenic mtDNA variant were reduced between 95% and 100% in six newborns and 77% and 88% in two others. Instead of using eggs, the third potential method of MRT relies on polar bodies, which are created during egg formation but generally do not develop into functional cells. Although polar bodies spontaneously degenerate under normal circumstances, they can be used in polar-body transfer (PBT) for MRT. Professor Lv established a spindle-protrusion-retained PB2 separation technique that yielded a near-normal proportion of euploid blastocysts while further reducing mitochondrial carryover in both mice and humans. This approach offers a valuable option for future clinical applications of mitochondrial replacement^[95]. However, further studies are warranted to validate these findings.

Comparing the three methods, spindle transfer holds a clear advantage. PT starts with two fertilized eggs—meaning two embryos would be used for one baby—which raises ethical concerns. More importantly, the embryos are activated before nuclear transfer and time window for the transfer is very narrow, and the amount of carryover cytoplasm could be higher. In contrast, spindle transfer requires only one embryo per baby. As the technique advances, the carryover rate can be reduced to less than 1%^[96]. Polar body transfer is also highly effective in minimizing carryover. However, since polar bodies are not typically involved in embryo development, their use remains uncertain. While animal studies show promise, other factors—such as DNA methylation and epigenetic modifications—should also be considered.

Although the successful application of MRT has identified one of the most effective means to block the inheritance of mtDNA mutations, despite therapeutic promise, almost all MRT approaches remain vulnerable to low-level mtDNA contamination and mutational reversion; consequently, their clinical use is outlawed in the majority of countries worldwide. Looking ahead, persistent clinical evaluation coupled with evolving ethical standards could position MRT as an essential and potentially irreplaceable remedy for such profoundly destructive conditions.

MST: The spindle-chromosome complex, which harbors the maternal nuclear genome, is extracted from the patient's oocyte and transferred into an enucleated donor egg. This reconstituted oocyte thereby retains the mother's nDNA while acquiring the donor's wild-type mitochondrial pool, and may subsequently be fertilized and implanted into the maternal uterus. PT: The maternal and donor oocytes are each fertilized with paternal sperm. Prior to the onset of cleavage, the pronuclei—containing the nuclear genome—are extracted from the maternal zygote and transferred into the enucleated donor zygote, from which the original nuclear material has previously been removed. PBT: A spindle-protrusion-retained PB2 separation technique was developed, which yielded a near-normal proportion of euploid blastocysts and further reduced mitochondrial carryover in both mice and humans.