Targeting Adaptive Cellular Responses To Mitochondrial Bioenergetic Deficiencies In Human Disease

Abstract

Mitochondrial dysfunction is increasingly appreciated as a central contributor to human disease. Oxidative metabolism at the mitochondrial respiratory chain produces ATP and is intricately tied to redox homeostasis and biosynthetic pathways. Metabolic stress arising from genetic mutations in mitochondrial genes and environmental factors such as malnutrition or overnutrition is perceived by the cell and leads to adaptive and maladaptive responses that can underlie pathology. Here, we will outline cellular sensors that react to alterations in energy production, organellar redox, and metabolites stemming from mitochondrial disease (MD) mutations. MD is a heterogenous group of disorders primarily defined by defects in mitochondrial oxidative phosphorylation from nuclear or mitochondrial-encoded gene mutations. Pre-clinical therapies that improve fitness of MD mouse models have been recently identified. Targeting metabolic/energetic deficiencies, maladaptive signaling processes, and hyper-oxygenation of tissues are all strategies aside from direct genetic approaches that hold therapeutic promise. A further mechanistic understanding of these curative processes as well as the identification of novel targets will significantly impact mitochondrial biology and disease research.

Graphical Abstract

Mutations in mitochondrial genes, from either mitochondrial or nuclear genomes, result in aberrant mitochondrial function. As the hub of metabolism and bioenergetics, dysfunctional mitochondria result in perturbed cellular homeostasis. Molecular sensors of mitochondrial bioenergetics regulate cell survival, proliferation, and death processes that modulate tissue pathology. Hence, understanding the adaptive cellular responses to mitochondrial defects provides therapeutic opportunities for these rare but detrimental mitochondrial diseases.

Introduction

Cellular metabolism involves a series of biochemical processes that produce or consume energy sustaining cellular function and organismal health. The breakdown fats, carbohydrates, and protein generate energy and the metabolic intermediates necessary for enzymatic reactions. When cells experience metabolic stress arising from nutrient limitation, distinct sensing mechanisms are engaged to modulate central transducer and effector proteins promoting cell adaptation and survival. Often this entails activation or repression of cell signaling pathways and cellular processes that control energy metabolism, redox homeostasis, and survival. Here, we will describe metabolic-protein networks perturbed by mitochondrial gene mutations and discuss promising therapeutic interventions that target them. As mitochondrial dysfunction is recognized as a key contributor to human pathologies including cancer, neurodegenerative disease, and mitochondrial disease (MD), understanding the principles of this cellular response to abnormal metabolic states is needed for rational and effective drug design.

Mitochondria are central organelles of metabolism and cellular physiology. There are approximately 1,100 proteins that function within human mitochondria to support numerous metabolic reactions that produce energy, co-factors, metabolites, and maintain redox homeostasis [1–3]. As the metabolic hub in the cell, mitochondria coordinate with other intracellular compartments to sustain specific cellular processes and physiology. The classical view of mitochondria recognizes their role in energy metabolism via two fundamental metabolic processes: the production of intermediary metabolites through the tricarboxylic acid (TCA) cycle and ATP through the respiratory chain. However, there is a wealth of literature establishing mitochondria as principal players in cellular communication relaying metabolic signals to the cytosol and other organelles [4]. This transfer of information and signals permit an active interplay between mitochondrial metabolic status and cellular function. The bulk of mitochondrial signaling cascades occur via the release of metabolites, proteins, or reactive oxygen species (ROS) or the establishment of signaling platforms on the outer membrane [4]. This is manifested by pathological mitochondrial gene mutations that stimulate aberrant signaling responses, highlighting the importance of understanding and targeting pathway components for therapy.

Mitochondrial gene mutations cause disease

The most common inherited genetic disorders are MDs, which are a clinically heterogenous group of disorders characterized by mitochondrial gene mutations in mitochondrial or nuclear DNA. MDs cause significant morbidity in patients usually affecting multiple organ systems and present substantial medical and socioeconomic costs [5]. About 1 in 5,000 newborns [6] and 1 in 4,300 adults suffer from MD [7]. Unfortunately, treatments for MDs are non-existent due to a lack of promising drug targets. This is partly due to the infancy of mitochondrial biology research, insufficient funds for MD research owing to its classification as a rare disorder, and insufficient pharmaceutical industry engagement [8]. Current therapies mainly address secondary complications rather than underlying biological determinants of disease. For example, management of diet and lifestyle such as supplementation with specific metabolites like Coenzyme Q10 (CoQ10), creatine, arginine, carnitine, or vitamins that mitigate oxidative stress or specific metabolic deficiencies [8, 9], while avoiding compounds with potential adverse effects on mitochondrial function are the primary methods to relieve MD symptoms. Aside from targeted therapies where a specific metabolic deficiency is known, there is little proven efficacy behind these treatment regimens. However, basic science research is beginning to bridge the gap between mitochondrial gene mutations, beneficial and maladaptive cellular responses, and therapeutic targets.

Currently, mutations in more than 250 mitochondrial genes cause disease [10]. Pathogenic mutations primarily impact respiratory chain function either directly or indirectly as many affected genes control mitochondrial processes outside the respiratory chain that ultimately impinge upon it. Functional categories of gene mutants encompass respiratory chain subunits/assembly factors/electron carriers, mitochondrial DNA (mtDNA) maintenance, mitochondrial dynamics, and metabolism of substrates, cofactors, or toxic compounds [11]. The classic example of diagnosed mutations is in the mtDNA characterized by mitochondrial inheritance patterns from mother to children. In contrast, mutations in nuclear-encoded genes inherited in an autosomal recessive, autosomal dominant, or X-linked fashion are also appreciated with conservative estimates of 5.9 in 100,000 individuals at risk [7]. As next-generation sequencing technologies continue to be applied to patients with MD phenotypes, the number of known disease-causing mutations are rising [12]. Despite the presence of known pathological mutations, there is a high degree of phenotypic variance among MD patients, which complicates diagnosis as individuals do not fall within one disease category. MDs usually affect multiple organ systems with prominent phenotypic features in muscle and brain. Few mutations result in phenotypes isolated to one organ system such as the eye in Leber hereditary optic neuropathy (LHON) or ear in mitochondrial non-syndromic hearing loss. The phenotypic consequences of an identified mutation are often variable, especially for mtDNA mutations. Mutation burden differs between individuals and even for a given individual, large differences exist at both the tissue and cell level (defined as heteroplasmy). In mitochondrial encephalomyopathy, lactic acidosis, and stroke-like episodes (MELAS) patients with A3243A>G mutations, a poor correlation exists between phenotype and mutation load in blood cells, especially in symptomatic patients where negative selection occurs over time [13]. However, a stronger correlation exists in muscle, indicating that diagnostic measurements in pathological tissue can help predict at-risk individuals or families [14, 15].

Adding to phenotypic complexity at the cellular level, distinct levels of heteroplasmy influence downstream signaling responses associated with different degrees of bioenergetic deficiency. Progressive increases of the MELAS A3243A>G mutation in cybrid cell lines cause distinct nuclear DNA (nDNA) and mtDNA expression changes including those encoding glycolytic genes, signal transduction components, epigenetic regulators, and chaperones in either the cytosol, ER, or mitochondria [16]. Heteroplasmy also correlates with acetyl-CoA and α-ketoglutarate (α-KG) levels and NAD⁺/NADH redox state, which lead to specific histone modification patterns [17]. Other metabolic consequences dependent on respiratory chain deficiency include the relative flow of α-KG towards oxidative TCA flux or reductive carboxylation [18]. Nuclear background and mutations can also interact with mtDNA mutations to influence downstream cellular responses with potential for detrimental phenotypes [19–23]. In Drosophila, transcriptional responses to hypoxia treatment are dependent on mitochondrial haplotypes [24] and in mice, different nDNA:mtDNA combinations altered metabolic parameters and adipose tissue transcriptional profiles [25]. This emphasizes that even outside MD, nDNA and mtDNA mutations can impact health and environmental stress tolerances of an organism. Furthermore, nuclear mutations in genes associated with mitochondrial encephalopathy, sensorineural hearing loss, diabetes, epilepsy, seizure and cardiomyopathy were identified in a subset of MELAS patients, suggesting nuclear mutations cooperate with mtDNA mutations to control disease severity and clinical symptoms [23]. In summary, the relationship between nuclear and mitochondrial genome mutations, variability in heteroplasmic burden across tissues and individuals, and mutation specific pathophysiologies contribute to the complexity of MD that may prevent the general application of therapies designed for mitochondria dysfunction in a particular context.

Metabolic substrates drive respiratory metabolism

One must consider the inputs and outputs of mitochondrial bioenergetics to better understand targetable metabolic stress mechanisms such as occurring in MD mutations. Cells consume carbohydrates, amino acids, and lipids to generate mitochondrial substrates pyruvate, glutamine, and fatty acids. Enzymatic reactions oxidize these substrates and produce ATP and redox metabolites NADH, NADPH, and FADH₂ that control the rate of anabolic and catabolic reactions that employ these cofactors. Mitochondrial bioenergetics largely depends on the availability and utilization of metabolic substrates discussed below that vary under different cellular and environmental stimuli.

Pyruvate Metabolism

Pyruvate is a fundamental driver for TCA carbon flux and mitochondrial ATP synthesis. The majority of pyruvate is generated from glycolytic metabolism. In certain cellular contexts and tissue types, lactate can also contribute to the pyruvate pool, where it is imported into the cell through monocarboxylate transporters and converted to pyruvate via lactate dehydrogenase [26, 27]. Cytosolic pyruvate, like other polar metabolites, diffuses through the outer mitochondrial membrane (OMM) via porins [28] and is transported into the mitochondrial matrix through the mitochondrial pyruvate carrier (MPC) [29]. From there, it assimilates into the TCA either through the pyruvate dehydrogenase complex (PDH) or pyruvate carboxylase (PC) where it generates acetyl-CoA or oxaloacetate, respectively. Due to its important role in mitochondrial physiology, aberrant pyruvate metabolism in the form of PDH complex deficiency manifests in tissues and presents with lactic acidosis, neurological and muscular defects, and can be fatal in children [30]. This preference for glycolysis occurs despite potential for PC driven anaplerosis, indicating that PDH deficiency is sufficient to shift cellular metabolism away from mitochondrial respiration [30]. This metabolic phenotype is a hallmark for mitochondrial respiratory deficiencies and observed in a Leigh syndrome mouse model, where pyruvate, lactate, and other glycolytic intermediates accumulate in the brain [31].

Glutamine Metabolism

Another major source of carbon, nitrogen, and energy for the cell is the amino acid glutamine. The mitochondrial enzyme glutaminase converts glutamine into glutamate, which can enter the TCA as α-KG following oxidative deamination by glutamate dehydrogenase. This latter enzymatic step is coupled to NADH production, supporting complex I mediated respiration. Glutamine is a major source of anaplerosis, replenishing TCA intermediates consumed by biosynthetic pathways for amino acids, nucleotides, and lipids. Aside from this, glutamine is used for synthesis of neurotransmitters and glutathione for redox homeostasis. Unsurprisingly, altered glutamine metabolism is observed in diseases with mitochondrial etiology [32, 33]. Patient-derived cytoplasmic hybrid cell lines harboring MD mutations were found to exhibit reprogrammed glutamine metabolism dependent on the severity of respiratory defect [18]. Under severe oxidative phosphorylation (OXPHOS) dysfunction, and similar to hypoxia, cells preferentially utilize α-KG to fuel reductive carboxylation to generate citrate and support lipid synthesis. In contrast, oxidative glutamine metabolism predominates in cells with moderate respiratory defects, especially since pyruvate is preferentially converted into lactate rather acetyl-CoA. These observations were further supported in a skeletal muscle COX10 KO mouse model of mitochondrial myopathy where elevated glutamate-derived α-KG metabolism triggered muscle protein breakdown [18].

Fatty Acid Metabolism

Long-chain fatty acids (e.g. palmitate) are oxidized in the mitochondrial matrix in a series of catabolic reactions termed fatty-acid β-oxidation (FAO). Due to their high level of stored energy compared to other substrates, fatty acids are a major source of energy under nutrient stress. Uptake of fatty acids into the mitochondria is the rate limiting step of FAO, where first carnitine palmitoyl transferase (CPT1) on the OMM produces acylcarnitines from FAs, which are subsequently shuttled into the matrix via inner mitochondrial membrane carnitine-acyl carnitine translocase (SLC25A20) [34]. CPT2, also localized to the inner mitochondrial membrane, releases carnitine in the matrix to initiate FAO, where fatty acids first undergo dehydrogenation by acetyl-CoA dehydrogenase, yielding one equivalent of FADH₂. Subsequent dehydration by enoyl-CoA hydratase and dehydrogenation by 3-hydroxyacyl-CoA dehydrogenase provides additional NADH [34]. Finally, cleavage by thiolase produces acetyl CoA. This series of reactions is repeated, and FADH₂, NADH, and acetyl-CoA are produced proportional to the carbon-chain length. FAO in the mitochondria sustains redox potential where NADH and FADH₂ enter the electron transport chain (ETC), and carbon equivalents via acetyl-CoA combine with citrate in the TCA cycle. As expected, disorders associated with mitochondrial fatty-acid oxidation are caused by defects in a variety of β-oxidation enzymes including SLC25A20, acyl-CoA dehydrogenase, carnitine palmitoyltransferase-2, and many others [35]. Mitochondrial respiratory chain deficiencies also correlate with aberrant fatty acid metabolism. Leigh syndrome mice exhibit a progressive loss in bodyfat, liver fat droplets, and brain free fatty acids [31], while accumulating fatty acids in serum driving inflammatory responses [36]. In patients with Leigh syndrome French Canadian variant, saturated even chain acylcarnitines accumulate in circulation reflecting disrupted mitochondrial fatty acid β-oxidation [37]. Fibroblasts from these patients are also sensitive to exogenous palmitate, indicating defects in fatty acid handling [38].

Other CoQ Donors And Electron Transfer Reactions

Beyond pyruvate, glutamine, and fatty acids, there are other important enzymatic reactions that utilize mitochondrial substrates to fuel electron transfer and support respiratory processes such as choline dehydrogenase, electron transfer flavoprotein dehydrogenase, sulfide:quinone oxidoreductase, proline dehydrogenase (PRODH), glycerol phosphate dehydrogenase, and dihydroorotate dehydrogenase (DHODH). These reactions, along with Complex I and II, contribute to electron transport by reducing the CoQ pool to fuel complex III activity. DHODH, an enzyme found on the inner mitochondrial membrane, catalyzes the oxidation of dihydroorotate to orotate, and is a critical step in pyrimidine de novo synthesis [34]. DHODH activity is central to nucleotide abundance, DNA/RNA homeostasis, respiratory capacity, and cellular proliferation [39]. Intriguingly, inhibition of DHODH augments respirasome assembly in proliferative cells to support cellular proliferation and survival [39], demonstrating an interplay between DHODH and complex I-driven respiratory reactions. Mitochondrial PRODH functions in the catabolism of proline to pyrroline-5-carboxylate, facilitating the proline cycle that concomitantly shuttles NADPH/NADP⁺ between the mitochondria and cytosol [40]. PRODH associates directly with CoQ1 on the inner membrane supporting the function of complex III and IV [40]. Direct binding of PRODH to ETC subunits links its activity to the production of mitochondrial ROS (mtROS) and signaling responses [41] including control of cell fate through PPARγ, FOXO1, AMP-activated protein kinase (AMPK), and others [40]. PRODH is also crucial to mitochondrial function and redox homeostasis in cardiomyocytes under hypoxic conditions [42].

Another mitochondrial enzyme, NADH-NADPH transhydrogenase (NNT), is a major generator of mitochondrial NADPH that is coupled to the TCA activity. Present in the inner mitochondrial membrane, NNT functions to transfer hydride ions between NADH and NADP⁺ which is coupled to proton translocation across the inner membrane [43]. NNT overexpression and altered NADP+/NADPH ratios stimulate glutamine oxidation and reductive carboxylation with concomitant decreases in glycolysis [44]. Specifically, high levels of NADPH under these conditions allow for the reduction of glutamine-derived α-KG to citrate via isocitrate dehydrogenase [44]. Conversely, NNT dysfunction favors the oxidation of α-KG to succinate [44] and results in high levels of ROS and oxidative stress [45] that leads to metabolic abnormalities in mice [46].

As the above section demonstrates, numerous biochemical factors that originate in the mitochondria control cellular energy, redox, and biosynthesis to sustain cellular homeostasis. Further, defects in these biochemical pathways result in a pathological reprogramming of metabolism and redox resulting in a variety of different diseases. In the following section, we will discuss mechanisms by which the cell senses and signals changes in mitochondrial metabolism, demonstrate how defects in these signaling pathways may sustain pathology, and illustrate how these signaling pathways hold promise for therapeutic intervention.

Outputs of mitochondrial bioenergetics regulate signaling mechanisms

The metabolic network in the mitochondria evolved multiple regulatory points that enable energy-sensing mechanisms to communicate with the cell. Alterations in the proton and ion gradient across the mitochondrial inner membrane, energy status reflected by chemical equivalents of ATP, redox homeostasis, or even certain metabolites coordinate downstream signaling pathways that reprogram metabolism and gene expression. These pathways are vital in normal energy homeostasis and implicated in MD. In the following section, we will describe communication mechanisms between the mitochondria and the cell that mediate compensatory responses to altered mitochondrial bioenergetics.

Mitochondrial Reactive Oxygen Species

Beyond substrate oxidation and electron transport, mitochondria are major producers of ROS. These species include superoxide anions (O₂⁻), hydrogen peroxide (H₂O₂), and hydroxyl radicals (OH⋅) and possess distinct chemical properties [47, 48]. ROS can cause oxidative damage to cellular components such as DNA, lipids, and protein, but also serve as potent signaling molecules that modify cysteine residues within proteins [47, 48]. At physiological pH, cysteines are primarily thiolate anions (Cys-S) and are vulnerable to oxidation by H₂O₂ to create its sulfenic form (Cys-SOH) [47, 48]. Certain proteins such as thioredoxin (Trx) and glutaredoxin can then return the cysteine to its initial state leading to reversible signal transduction mechanisms [47, 48]. Mitochondrial complex I and III are responsible for mtROS, as electron leakage in combination with iron-sulfur (Fe-S) clusters within these complexes react with molecular oxygen. Specifically, the leakage of one electron to O₂ produces O₂⁻ that is detoxified in the mitochondrial matrix by superoxide dismutase 2, and in the intermembrane space and cytosol by superoxide dismutase 1, ultimately forming H₂O₂ [48]. Further processing by peroxiredoxins, glutathione peroxidases, and catalase reduce H₂O₂ to H₂O at varying affinities [49], highlighting a role of these enzymes in balancing cellular damage and signaling depending on intracellular ROS concentrations [48].

When elevated above a certain threshold, mtROS acts as a messenger of mitochondrial dysfunction through activation of downstream molecular sensors through oxidation of their thiol residues (Figure 1). Tyrosine phosphatases characterized by conserved cysteines in their active site are especially susceptible to oxidation [48]. Phosphatase and tensin homolog deleted on chromosome 10 (PTEN) is a repressor of PI3K/AKT/mTOR signaling pathways [50]. Mechanistically, PTEN negatively regulates PI3K/AKT signaling by dephosphorylating phosphatidylinositol 3,4,5-trisphosphate (PIP₃), a downstream signaling molecule that activates AKT [51]. PTEN is modulated by phosphorylation and oxidation events. Oxidation of Cys 124 by H₂O₂ results in a disulfide linkage with Cys 71 resulting in its inactivation [52]. Upstream effectors such as epidermal growth factors, platelet derived growth factors, and insulin further increase H₂O₂ production, PTEN oxidation, and PIP₃ stabilization [51], amplifying PI3K signaling. For example, during muscle differentiation, mtROS-derived H₂O₂ inactivates PTEN and hyperactivates PI3K and mechanistic target of rapamycin 1 (mTORC1) to enhance myogenesis-specific autophagy [50]. Other phosphatases that are sensitive to oxidation include PTP1B, cdc25, low Mr PTP [53], and others involved in B-cell receptor signaling [54] and TNFα mediated regulation of JNK [55].

Figure 1. Signaling mechanisms responsive to mitochondrial bioenergetics.

Biosynthetic and energetic processes of the mitochondria enable energy-sensing mechanisms to communicate with the cell. Pyruvate, glutamine, and fatty acids supply the TCA cycle of oxidative substrates that fuel the ETC, providing energy and redox equivalents. Further, proton pumps across the ETC in coordination with ions to sustain the MMP driving electron transfer, ATP synthesis, and electron leak. Mitochondrial energy production and redox, MMP, and mtROS communicate mitochondrial integrity to the cell and is sensed by a variety of pathways such as mTORC1, ISR, MAPK, hypoxia, and mitophagy signaling.

The Mitogen-activated protein kinases (MAPKs) are also sensitive to ROS (Figure 1). MAPKs are a widely studied class of serine/threonine kinases that largely converge on three central molecular players: ERKs, c-Jun N-terminal kinases (JNKs), and p38 MAP kinases [56]. Stress inputs from inflammatory cytokines, radiation, heat, or osmotic shock, along with growth factors initiate this signaling cascade to the nucleus. The apoptosis-signaling kinase 1 (ASK1), activates both JNK and p38 MAPK pathways, and can be stimulated by H₂O₂ [56, 57]. ASK1 forms higher order assemblies with itself and other protein intermediates that directly sense ROS in the cytoplasm. The protein antioxidant Trx inhibits ASK1 kinase activity through disulfide-exchange mechanisms in its reduced state and dissociates from ASK1 when oxidized, permitting ASK1 activation [57–59]. Peroxiredoxins also respond to H₂O₂ levels and regulate ASK1 in a similar fashion [60]. In complex I MD cell lines, ASK1 and downstream p38/JNK MAPK signaling are elevated under nutrient stress conditions from disruptions to cytosolic redox associated with elevated endoplasmic reticulum (ER) stress [61, 62]. Activation of p38/JNK signaling in complex I mutant cells requires the ER-resident type I transmembrane serine/threonine protein kinase IRE1 [61]. This presumably occurs through IRE1-TRAF2-ASK1 complex formation, which induces p38/JNK signaling and cell death [63, 64]. Accordingly, treatment with chemical chaperones 4-PBA and TUDCA ameliorated ER stress and p38/JNK signaling in complex I mutant cells [61], while pharmacological inhibition of IRE1/p38 restores redox homeostasis, suppresses inflammation, and promotes cell survival [61, 65]. Overexpression of malic enzyme 1 (ME1), which generates NADPH, or supplementation with reduced glutathione (GSH) also rescued MAPK signaling and cell death in these cells [62], indicating ROS is an upstream mediator of these responses. ASK1/MAPK signaling from ER stress and cytosolic ROS accumulation illustrates an important effector that dictates cell fate in in vitro MD models worth exploring in vivo.

mtROS directly regulates activation of the hypoxia pathway through stabilization of the transcriptional factor hypoxia-inducible factor 1α (HIF-1α) (Figure 1). HIF-1α is normally degraded in normoxia after hydroxylation by prolyl hydroxylase enzymes (PHDs) and ubiquitination by the E3 ubiquitin ligase von Hippel Lindau (VHL). In hypoxia, PHD activity decreases leading to HIF-1α stability and activity. Intriguingly, even in normoxia, ROS generated from complex III stabilizes HIF-1α [66–70] to initiate downstream transcriptional responses targeting glycolytic metabolism [71]. This metabolic shift is of particular importance during mitochondrial dysfunction, as cells unable to respire at normal rates require glycolysis for ATP production. Accordingly, forced activation of the hypoxia pathway is protective in cancer cells treated with the complex III inhibitor antimycin [72]. HIF-1 stabilization from mitochondrial stress occurs at the organismal level as well. Mitochondrial mutations or sublethal ROS doses induces HIF-1 transcriptional activity and lifespan extension in C. elegans [73, 74]. Continous hypoxia treatment (11% O₂ compared to ambient 21% O₂) is also protective in Leigh syndrome mice, while hyperoxia (55% O₂) is detrimental [72, 75, 76], as discussed later in this review.

The nutrient sensors of the cell: mTORC1, AMP-Activated Protein Kinase, and PGC-1α

The nutrient sensing mTORC1 is involved in numerous control mechanisms of anabolic and catabolic states of the cell (Figure 1) [77]. This Ser/Thr kinase complex responds to multiple cues including growth factors and amino acids and plays important roles in cell growth by controlling protein synthesis, metabolism, and autophagy. Direct targets of mTORC1 include the eukaryotic translation initiation factor 4E (eIF4E)-binding protein 1 (4E-BP1) and ribosomal protein S6 kinase beta-1 (S6K1) [77]. mTORC1 phosphorylates S6K1 which in turn phosphorylates and activates eIF4B promoting translation of mRNAs [78]. 4E-BP1 phosphorylation by mTORC1 further enhances translation by disrupting 4E-BP1 association with mRNA cap-binding protein eIF4E [79, 80]. Aside from protein synthesis, mTORC1 is a master regulator of autophagy through phosphorylation-dependent inhibition of ULK1/2 [81–83] and the VPS34 autophagy complexes [84] and the lysosomal biogenesis transcription factor EB [85–87]. Interestingtly, mTORC1 signaling is altered in response to mitochondrial dysfunction in a context dependent fashion. In HEK293T cells, mTORC1 is inhibited through activation of heme-regulated eIF2α kinase (HRI) and AMPK [88]. In models of MD arising from mitochondrial gene mutations, mTORC1 is hyperactivated through an unknown mechanism [31, 89–93] and responsible for pathology in mice [31, 90, 92].

AMPK is a key regulator of energy status through sensing low ATP levels (AMP/ATP ratios) that is correlated with dysfunctional mitochondrial respiration and ATP synthesis (Figure 1). Adenine nucleotides activate AMPK to phosphorylate a variety of targets to enhance catabolism and stimulate mitochondrial functions such as fission and fusion, mitophagy, and mitochondrial biogenesis [94]. AMPK also coordinately regulates autophagy with mTORC1 through the differential phosphorylation of the autophagy-activating kinase ULK1 [95]. Under nutrient deprivation, cells activate autophagy through AMPK-mediated phosphorylation of ULK1 at Ser 317 and Ser 777 [95]. In contrast, when nutrients are available, mTORC1 inhibits autophagy through phosphorylation of ULK1 at Ser 757, which blocks ULK1 interaction with AMPK [95].

Since mTORC1 inhibition or AMPK activation promote turnover of damaged mitochondria and energy homeostasis, these interventions are linked to MD therapeutics. mTORC1 inhibitors are beneficial in multiple MD models [31, 89, 91–93, 96–98], but underlying mechanisms remain elusive. With implications in mitochondrial turnover, the mTORC1 inhibitor rapamycin rescued mitochondrial myopathy in a muscle specific Cox15^−/− mouse model and correlated with restored autophagy and lysosomal biogenesis [98]. Rapamycin also enhances the removal of heteroplasmic mtDNA mutations in a human cybrid cell line through the activation of mitophagy [99], a mitochondrial-specific form of autophagy. The efficacy of AMP-dependent activators of AMPK like metformin in MD models is often variable owing to inhibition of the ETC [100]. An alternative approach involving allosteric activation of AMPK through an AMP-independent mechanism holds promise. A chemical screen in fibroblasts derived from patients with MD found that the direct AMPK activator PT1 enhanced mitochondrial performance and cellular redox homeostasis [100].

AMPK and mTOR also control the mitochondrial biogenesis program of the peroxisome proliferator-activated receptor gamma coactivator 1-α (PGC-1α or Ppargc1a) (Figure 1). Specifically, mTORC1 stimulates mitochondrial oxidative metabolism through phosphorylation of YY1, which drives interaction with PGC-1α to enable its targeting to mitochondrial gene promoters [101, 102], while AMPK directly phosphorylates two sites (Thr177 and Ser538) on PGC-1α to drive PGC-1α expression in a feed-forward loop [103]. Positive modulation of PGC-1α is beneficial in models of neuronal inflammation and MD. Neuronal Ppargc1a expression elevates mitochondrial content, bioenergetic capacity, and calcium handling in neurons, buffering them from inflammation [104]. Accordingly, a neural progenitor human model of Leigh syndrome with SURF1 mutation displayed altered metabolism and hindered neuronal morphogenesis capacity that could be rescued through pharmacological activation or overexpression of PGC-1α [105]. Since PGC-1α is a difficult factor to target pharmacologically, certain modifiers of PGC-1α function are attractive target strategies. Through a high-throughput chemical and CRISPR loss-of-function screens, our group identified bromodomain-containing protein 4 (BRD4) inhibition as a method to drive the PGC-1α gene expression program, mitochondrial content, and rescue NADH-ubiquinone oxidoreductase chain 1 protein (ND1) mutant cybrid cells from galactose-induced cell death [106]. BRD4 is a transcriptional and epigenetic regulator that is a member of the BET (bromodomain and extra terminal domain) family that binds acetylated lysine residues in chromatin [107]. BRD4 competes for binding with PGC-1α at genetic promoters and negatively regulates transcription of mitochondrial genes [106]. As such, BRD4 loss or bromodomain inhibitor I-BET 525762A treatment promoted assembly and activity of complex II and IV, bypassing complex I defects [106].

Mitophagy

One mechanism of mitochondrial protein quality control involves mitophagy, a process that selectively degrades damaged mitochondria with decreased mitochondrial membrane potential (MMP) through bulk autophagy (Figure 1). This occurs through stabilization of the serine-threonine kinase PTEN Induced Kinase 1 (PINK1) on the OMM of damaged mitochondria [108]. In healthy mitochondria, PINK1 is normally degraded by mitochondrial proteases Lon [109], AFG3L2, and PARL [110–112]. When MMP is decreased, PINK1 accumulates at the outer membrane and recruits a cytosolic E3-ubiquitin ligase Parkin to ubiquitinate mitochondrial proteins. The accumulation of the ubiquitin signal activates downstream p62 and LC3 mediated degradation by autophagosomes [113]. Although the PINK1/Parkin pathway is important in sensing and facilitating mitophagy, numerous other parkin-independent mechanisms exist including receptor-mediated, lipid-mediated, and ubiquitin-mediated pathways that are extensively reviewed elsewhere [28, 114]. In brief, there are LIR (LC3-interacting region) receptor proteins or lipid species such as cardiolipin that localize to the mitochondrial outer membrane and interact with LC3 proteins associated with the phagophore, targeting autophagy vesicles to the mitochondria [28, 114]. Ubiquitin ligases aside from Parkin such as MUL1, SIAH1, and ARIH1 also ubiquitinate mitochondrial proteins and promote mitochondrial recycling [115–118]. Thus, a complex network of effector proteins and lipids control mitophagy induction in mammalian cells.

mtDNA mutations often persist in a state of heteroplasmy, coexisting with healthy organelles. Fractions of these enduring mutant mitochondria that ultimately lead to disease can be controlled by mitophagy. Drosophila S2 cell line transiently transfected with a mitochondrial-targeted restriction enzyme AflIII results in heteroplasmic cell populations where mutants can be degraded by induction of autophagy [119]. Specifically, overexpression of PINK1 and Parkin result in drastic decreases in mutant mitochondrial populations [119]. This observation is further substantiated in C. elegans where deletion of PDR-1 (Parkin in mammals) resulted in an increase in mtDNA mutant populations [120]. These data support the potential of modulating mitophagy to reduce mutant mtDNA pools, especially in the situation of severe mutations that compromise MMP.

Degradation of potentially pathogenic mitochondria sustains cellular health, and mutations in mitophagy genes can lead to disease. Parkinson’s disease is a neurodegenerative disorder that is a result of tissue damage in the substantia nigra. This post-mitotic tissue is a site of high levels of mtDNA mutations [121–123] owing to the unique cytoarchitecture and mitochondrial requirements of dopaminergic neurons within this brain region [124]. Mutations in PINK1 and Parkin cause autosomal recessive forms of Parkinson’s disease that selectively damages the substantia nigra [108, 125] and correlate with deficiencies in PINK1/parkin mitochondrial quality control pathways including mitophagy, fission/fusion dynamics, mitochondrial biogenesis, and regulation of localized translation of mitochondrial genes [124]. For example, PARIS (ZNF746) is a KRAB zinc-finger protein that is a substrate for PINK1 and Parkin [126, 127], which is elevated in sporadic and familial PD brains and drives DA neuron loss in Parkin loss-of-function mouse models [127–130]. PARIS accumulation contributes to pathology through binding to insulin response sequences in the PGC-1α promoter and repressing its transcription [127]. Mitophagy proteins thus function beyond turnover of damaged mitochondria, and in a concerted fashion, stimulate mitochondrial biogenesis. Methylmalonic acidemia, a mitochondrial disorder caused by mutations in mitochondrial methylmalonyl-coenzyme A mutase, results in aberrant processing and toxic accumulation of amino acid byproducts in the mitochondria. Disease progression has been linked to dysfunctional PINK1/Parkin processing, ultimately leading to disease manifestation in the kidneys [131]. Mitophagy proteins PINK/Parkin thereby promote mitochondrial quality control through multiple mechanisms essential to mitochondrial homeostasis and tissue health.

The Integrated Stress Response

The integrated stress response (ISR) is a coordinated signaling pathway that responds to diverse stress stimuli to reestablish cellular and metabolic homeostasis. Numerous intrinsic and extrinsic factors including hypoxia, deprivation of glucose and amino acids, viral infection, along with ER and mitochondrial stress are examples of ISR-sensitive stressors [92, 132, 133]. Upon stress, eIF2α kinases GCN2, PKR, PERK, and HRI phosphorylate eIF2α inhibiting the eIF2B-mediated formation of the 43S pre-initiation complex. This results in global attenuation of 5’ cap-dependent translation, and specific translation of mRNAs with upstream open reading frames that code for genes involved in cell survival including ATF4, ATF5, CHOP, and GADD34 [132, 134, 135]. Emerging evidence illustrates that GCN2 and HRI play important roles in responding to specific consequences of mitochondrial stress such as decreased aspartate/asparagine synthesis, altered NADH/NAD⁺ ratios, and hyperpolarization of membrane potential [132, 133] (Figure 1), while PERK reacts to ER stress by increasing mitochondrial fusion and respiratory capacity [136, 137]. The protective aspects of the ISR in mitochondrial health are discussed below, but prolonged activation of the ISR can also be maladaptive, triggering apoptosis and pro-inflammatory events. Therefore, targeting the ISR in mitochondrial disorders is complicated and should be approached in context dependent manner.

GCN2 communicates various mitochondrial stressors to the cytosol in worms and in mammals [138]. In C. elegans, GCN-2 is protective during mitochondrial stress and acts in a complimentary pathway to the mitochondrial unfolded protein response (UPR^mt) [139], which upregulates mitochondrial-specific chaperones and proteases, ROS detoxification enzymes, and protein import components [140–143]. GCN-2 knockdown in the context of mitochondrial stress exacerbates developmental defects, triggers mitochondrial fragmentation, and hyperactivates the UPR^mt [139]. In mammals, however, the ISR is central to UPR^mt activation through the bZIP transcription factor ATF5 [144]. HeLa cells treated with the mitochondrial translation inhibitor doxycycline and gastrointestinal cancer cells treated with oligomycin induce the ISR in a GCN2 dependent fashion [145, 146], indicating that diverse mitochondrial stressors can elicit the response. The relationship between mitochondrial dysfunction and the ISR also depends on cellular state such as differentiation status. Impairment of NADH oxidation as a result of ETC inhibition decreased cellular asparagine and activated the ISR through GCN2 specifically in myoblasts but not myotubes [133]. Accordingly, normalization of NADH/NAD⁺ ratios in myoblasts through expression of NADH oxidase from Lactobacillus brevis (LbNOX) in the cytosol or mitochondria ablated ISR activation [133].

HRI is another eIF2α kinase that is sensitive to multiple mitochondrial stressors including inhibition of ATP synthase, mitochondrial protein synthesis, or MMP. Under these conditions, mitochondrial membrane protease OMA1 cleaves the mitochondrial protein DELE1 that translocates to the cytosol to bind and activate HRI, leading to increased translation of ATF4 [147, 148]. Importantly, the OMA1-DELE1-HRI pathway is not generally protective for cells undergoing mitochondrial stress and is even detrimental to proliferation in the presence of oligomycin treatment or mitoribosomal inhibition [147].

Communication between the ER and mitochondria is vital for cellular homeostasis by regulating calcium, mitochondrial fission, the inflammasome, and lipid biosynthesis and trafficking [149]. The eIF2α kinase PERK is located at the ER and is activated by numerous ER stressors but is also intimately linked to mitochondrial function. PERK activation coincides with its dimerization and autophosphorylation leading to phosphorylation of eIF2α [150] and ATF4 translation. In cybrid cell lines under nutrient stress, PERK-eIF2α-ATF4 induces survival through the upregulation of super-complex assembly factor 1 (SCAF1), with concomitant increases in mitochondrial super-complexes and respiratory capacity [137]. PERK also stimulates mitochondrial hyperfusion during ER stress conditions through translational attenuation to prevent pathological mitochondrial fragmentation and metabolic decline [136]. Independent of the ISR, PERK regulates mitochondrial respiratory capacity through formation of mitochondrial cristae, a platform for respiratory complex activity [137, 151]. A post-translational modification cascade involving PERK phosphorylation of O-linked N-acetyl glucosamine transferase (OGT) and OGT glycosylation of TOM70, drives MIC19 import, MICOS assembly, and cristae biogenesis [151].

NADPH and 1C Metabolism

The mitochondria play an important role in NADPH homeostasis and one-carbon (1C) metabolism (Figure 1). Compartmentalized both in the cytosol and mitochondria, 1C metabolism is mediated by the folate co-factor, enzyme methylenetetrahydrofolate dehydrogenase (MTHFD), and facilitated by oxidative reactions and NAD(P)⁺/NAD(P)H. Mitochondrial 1C metabolism is required for numerous biosynthetic processes including the metabolism of choline, purines, and histidine, the conversion of glycine and serine, and the synthesis of formate [152]. Additionally, folate is required for the methylation of mitochondrial tRNAs for mitochondrial protein synthesis [153]. In cells or tissues burdened with redox but not ATP synthesis or membrane potential defects arising from a stalled respiratory chain, the ISR activates 1C metabolism gene expression [154–157], indicating an important link between specific types of mitochondrial dysfunction and metabolic reprogramming. This is consistent with NADH redox defects from complex I inhibition compromising mitochondrial 1C [62, 156], clarifying the celluar rationale for 1C gene expression.

The largest contributor to NADPH production is assumed to be the oxidative pentose phosphate pathway (PPP), however, the maintenance of 1C metabolism in the mitochondria is a near equivalent contributor, as is ME [158]. Interplay between NADPH producing pathways is evident in complex I mutant cells that are sensitive to glucose deprivation due to an inability to compensate reduced PPP activity with mitochondrial 1C metabolism [62]. A CRISPR activation screen for genetic modifiers of ND1 cybrid cell survival under glucose restriction conditions identified cytosolic ME1 [62]. Overexpression of ME1 improved NADPH/NADP⁺ ratios and survival in ND1 cybrids during PPP inhibition through glucose withdrawal or 6-aminonicotinamide treatment [62]. This effect could be mimicked through supplementation with reduced glutathione (GSH), arguing that NADPH redox homeostasis is compromised in mitochondrial mutant cells and leads to MAPK activation and cellular dysfunction [62]. The mitochondrial 1C enzyme MTHFD2/L uses NAD⁺ as a cofactor and is predicted to be sensitive to a reduced NAD⁺/NADH ratios from respiratory chain inhibition. Consequently, complex I inhibition leads to an accumulation of upstream metabolite methylene-THF quantified by serine isotopic scrambling [156] and reduces mitochondrial formate production when serine is supplemented to isolated mitochondria [62, 156]. This latter effect is rescued by expression of NADH oxidase from Lactobacillus brevis (LbNOX) that restores NAD⁺ levels from complex I inhibition and leads to cell death rescue in mutant cells [62]. In contrast, a recent report argues that MTHFD2/L is rather insensitive to NAD⁺/NADH ratios relative to other NADH producing reactions in the mitochondria and consequently contributes to the NADH pool at a high level under mitochondrial respiratory inhibition, driving cellular dysfunction [159]. This does not imply that mitochondrial 1C metabolic flux is higher during respiratory chain inhibition, but rather other mitochondrial NADH producing pathways are heavily restrained [159]. Complicating the authors model that mitochondrial 1C contributes to pathology in MD, treatment of Leigh syndrome mice with a SHMT1/2 inhibitor did not improve their survival [159]. The benefits or potential harmful effects of one carbon metabolism in mitochondrial dysfunction need to be further addressed, especially with consideration of different media conditions for in vitro studies (e.g., proliferation studies in high-glucose versus survival studies in low-glucose or supplementation with pyruvate) and mutational contexts for in vivo studies.

The therapeutic potential of compounds that normalize not only NADH, but NADPH redox, yield promise for treatment of MD. In fact, across multiple mitochondrial inhibitors synthetic lethality was observed with mutation of the PPP enzyme G6PD [160]. Small molecule activators of ME1 that boost NADPH production when NADH redox is compromised may be a novel approach from the aforementioned studies [62]. In a chemical screen to identify compounds that activate respiration in mitochondrial defective cell lines, small molecule phosphofrucktokinase-1 inhibitor tryptolinamide alleviated disease phenotypes by shunting carbons from glycolysis into the PPP and enhancing NADPH production [161]. This buffered cells from oxidative stress and enabled AMPK activated fatty-acid oxidation and enhanced OXPHOS function [161]. Thus, potential exists for pharmacological approaches that reprogram metabolite/redox regulation and nutrient signaling to prevent cellular dysfunction in the MD setting.

Therapeutic interventions for mitochondrial disease

Chronic unresolved stress from nuclear- and mitochondrial-encoded mitochondrial gene mutations causes pathological conditions. Bioenergetic decline propagates signals that coordinate gene programs to promote cell survival or cell death. Hence, targeting maladaptive signaling pathways stimulated by mitochondrial dysfunction or engaging cytoprotective responses holds promise in treating these debilitating diseases. In the following sections, we will highlight emerging therapies for MD mutations with preclinical validation in animal models or humans and provide insights on conserved responses from lower eukaryotes. One limitation to many of MD mouse studies is the heavy reliance on the NADH dehydrogenase [ubiquinone] iron-sulfur protein 4 (Ndufs4) nuclear-encoded gene mutant mouse model, highlighting the need to generate and translate these findings to other mouse models, especially those carrying mtDNA mutations.

Rapamycin and mTOR inhibitors

mTORC1 inhibition is a promising therapeutic target in neurodegenerative disease models [162] and MDs are no exception [31, 89, 91–93, 96–98]. The first inklings of this connection emanated from studies in S. cerevisiae exploring the genetic response to dietary restriction [31, 163]. Decreasing glucose concentrations in growth media associated with decreased cytoplasmic mRNA translation and mTORC1 signaling increased the replicative lifespan of yeast mitochondrial gene mutants [31, 163]. This protective effect was also mimicked genetically through deletion of sch9 (S6K1 homolog) and ribosomal protein rpl20b or pharmacologically through cycloheximide treatment [163]. Similar beneficial effects of reduced cytoplasmic protein synthesis were reported in C. elegans and human cell lines presenting mitochondrial dysfunction [90, 93].

Extending these findings to the Ndufs4^−/− mouse model of Leigh syndrome, the pharmacological dietary restriction mimetic rapamycin improved animal lifespan and healthspan [31]. Ndufs4^−/− mice exhibit progressive neurodegeneration marked by vacuolation and gliosis in the olfactory bulb, cerebellum, and vestibular nuclei [164, 165]. The neurological phenotypes manifest in behavior changes including severe ataxia, lethargy, hypothermic events, and bodyweight loss leading to euthanasia criteria around P55 [164, 165]. Remarkably, rapamycin increased rotarod performance, delayed clasping, and decreased neuroinflammation in brains of Ndufs4^−/− mice [31]. In addition, rapamycin caused a metabolic shift in Ndufs4^−/− brains alleviating the buildup of glycolytic intermediates and increasing levels of free fatty acids [31, 166].

Although the protective mechanisms downstream of mTORC1 inhibition remain elusive, there are a number of correlated phenotypes. Inappropriate mTORC1 signaling in the context of mitochondrial dysfunction is undoubtedly a driving force behind disease pathology. A prominent feature of respiratory distress in many cell types and animal models is aberrant activation of mTORC1 signaling [31, 89–93] and downstream ATF4 activity [92]. Consequently, pharmacological attenuation of hyperactive mTORC1 is beneficial in these MD models, suggesting that increased protein synthesis and anabolic metabolism is maladaptive and exacerbates existing energetic deficits. In an iPSC-based disease model of maternally inherited Leigh syndrome caused by a T8993G mtDNA mutation in the ATP6 gene, differentiation of cells into neurons led to an ATP deficit and degenerative phenotypes that were alleviated by rapamycin or cycloheximide [93]. Similarly, we find that mTORC1 inhibition increases ATP levels of several MD cell lines including ND1 and MELAS cybrids under glucose restriction [61, 65].

Aside from energy conservation, mTORC1 inhibition affects other metabolic phenotypes downstream of mitochondrial dysfunction [31, 92, 97, 166]. In Ndufs4^−/− mice, glycolytic intermediates accumulate, supporting a shift towards glycolytic metabolism, while glutamine/glutamate/α-KG metabolites are depleted [31, 166]. These metabolic signatures are attenuated with rapamycin treatment, suggesting a beneficial effect of normalizing glycolytic and glutamine metabolism in MD [31, 92, 97, 166]. Another phenotype associated with mTORC1 inhibition is diminished ISR/ATF4-dependent metabolic rewiring [92]. In a mouse model of mitochondrial myopathy caused by a mutation in the mtDNA helicase Twinkle, rapamycin significantly reduced ATF4-dependent metabolic signatures [92] defined by increased serine biosynthesis, mitochondrial folate cycle, 1C metabolism, and associated purine and glutathione synthesis [156, 167, 168]. The dramatic reversal of metabolic derangement and disease phenotype by rapamycin in this MD model [92] and others [31, 97], suggests that chronic anabolism drives pathogenesis in MD.

The protective effects of rapamycin in MD are not limited to reduced mTORC1 activity. A recent proteomic analysis of Ndufs4^−/− mice treated with rapamycin revealed decreased phosphorylation of mTORC2 targets PKC and AKT [169]. Although rapamycin is fairly specific for mTORC1, chronic treatment in mice is known to reduce mTORC2 signaling [169]. PKC is regulated negatively through phosphorylation, which leads to its instability and degradation, and positively by calcium [169]. All PKC isoforms (PKC-α, PKC-β and PKC-γ) were affected with rapamycin treatment, however, the most dramatic effects were with PKC-β [169]. In particular, this isoform stimulates the pro-inflammatory NF-κB pathway via IKK-α and IκB phosphorylation [170]. The PKC-β inhibitor ruboxistaurin decreased glial fibrillary acidic protein levels in Ndufs4^−/− mouse brains, delayed neurological decline, and prolonged lifespan without the typical bodyweight decline associated with rapamycin [169]. These data support a model where rapamycin rescues lifespan of Ndufs4^−/− partly through suppression of PKC-driven neuroinflammation.

As highlighted, mTORC1 inhibition mediated rescue of mitochondrial mutants is well conserved across organisms yeast to human cells. The translation of mTORC1 inhibitors to MD human patients is also proving successful. Kidney transplant recipients suffering from MELAS/MIDD syndrome switched from calcineurin to mTOR inhibitors for immunosuppression exhibited clinical improvements [91]. In a separate study, mTOR inhibition failed to rescue symptoms of a patient with early-onset MELAS, but ameliorated symptoms of a patient with Leigh syndrome [96]. Continued clinical studies with focus on mutational context and disease severity at the onset of treatment will better address the efficacy of mTOR inhibitors in human MD.

Hypoxia

For two decades, hypoxia signaling was known to be stimulated by mitochondrial-derived ROS that in many cases is exacerbated by mitochondrial mutations. In C. elegans, mitochondrial dysfunction from mitochondrial gene mutations stabilizes HIF-1 to promote longevity in certain conditions [73, 74]. Paired with this, animals with mitochondrial mutations are sensitive to oxygen in their environment. The short-lived gas-1 complex I mutants, originally identified for their anesthetic sensitivity [171, 172], are vulnerable to changes in oxygen tension [173]. In normoxia, gas-1 animals are short-lived at standard temperatures (11.1 days compared to 14.8 days), a phenotype that is greatly exacerbated at continuous 60% oxygen tension (1.7 days compared to 10.9 of controls) [173]. Embryogenesis in gas-1 animals follows a similar pattern: <25% hatching at ambient O₂ (versus >95% for WT), 0.5% hatching at 60% O₂ (versus >95% for WT), and >75% hatching at 2% O₂ [173]. These data are consistent with the benefits of hypoxia and lethality of hyperoxia in animals with mitochondrial mutations.

Not surprisingly, a CRISPR loss-of-function screen for proliferation in cancer cells treated with complex III inhibitor antimycin identified VHL mutations, which stabilizes HIF-1α independent of oxygen tension [72]. This positive effect of HIF-1α stabilization was confirmed in antimycin treated zebrafish [72]. Highlighting the importance of this response across evolution, Ndufs4^−/− mice treated with continuous 11% oxygen, a dose experienced in high altitude environments, led to a remarkable improvement in survival and ameliorated disease symptoms [72, 75]. Locomotor and brain pathological phenotypes were dramatically rescued [72]. Even at 250 days of age, long-term hypoxia treated KO animals did not exhibit signs of brain lesions or microglial activation [75]. Rather, death was proposed to result from cardiac abnormalities due to the observation of decreased left ventrical ejection fraction at 200 days [75]. Unfortunately, the beneficial effects of hypoxia were absent at continuous normobaric 17% oxygen or under intermittent hypoxia (10 hours 11% oxygen, 14 hours normoxia) [174]. In addition, Ndufs4^−/− mice exposed to continuous 11% oxygen and then placed under normoxia deteriorated rapidly [174], indicating that translation of hypoxia treatment in humans may be complicated.

Similar to rapamycin, despite strong therapeutic rescue, mechanisms downstream of hypoxia responsible for attenuating pathology are not entirely clear. As mentioned above, canonical HIF-1α signaling is beneficial to C. elegans, D. rerio, and mammalian cell lines suffering from mitochondrial stress [72–74]. In contrast, the story in mice is quite different. Genetic mutations in PHD1/2/3 or VHL to stabilize HIF-1α were not sufficient to rescue lifespan in Ndufs4^−/− mice, and in some situations were harmful [76]. Intriguingly, measuring oxygen consumption in mice revealed that there was an age-dependent decrease in whole-body oxygen consumption simultaneous with an increase in brain PO₂ [76]. Establishing brain hyperoxia as a driver of pathogenesis, administration of carbon monoxide or an iron-deficient ‘anemia’ diet to Ndufs4^−/− mice improved lifespan markedly [76]. Thus, oxygen toxicity from decreased complex I function appears to drive neurological degeneration in Leigh syndrome mice. Whether a similar pathological basis occurs in other eukaryotes is not yet known.

Whether hypoxia treatment can benefit animal models with other mitochondrial gene mutations is an area of future interest. In support of this possibility, growth deficits from mitochondrial mutations in complex I, PDH complex, MPC, type II fatty acid synthesis, CoQ biosynthesis, and Fe-S cluster biosynthesis are all buffered by hypoxia treatment in cell culture [175], while continuous hypoxia improves neurological symptoms in a mouse model of Friedreich’s ataxia caused by mutation in frataxin, an essential gene in Fe-S cluster biosynthesis [176]. These data argue specificity to hypoxia-mediated rescue as deficiencies in complex II-V subunits are not adequately buffered by hypoxia in mammalian cell culture [175], a point that should be heavily considered in translational studies. On the opposite spectrum, unsaturated phospholipids derived from peroxisomal ether phospholipid biosynthesis genes were required for survival of proliferative cells under continuous 1% and 5% hypoxia conditions [175], arguing that shielding cells against saturated lipid toxicity or decreased membrane fluidity is protective, at least in cell culture. Another interpretation of this data is that ether lipids promote more efficient mitochondrial respiration through assembly of mitochondrial SCs [39], which is advantageous under hypoxic conditions [177, 178]. In sum, hypoxia therapy is a promising treatment for MD caused by certain gene mutations [175], but may not be appropriate for mitochondrial abnormalities present in peroxisomal disorders [179].

Tetracycline antibiotics

Our group recently uncovered multiple cellular mechanisms possibly valuable in the treatment of MD [61, 62, 65, 106, 137]. One paradoxical finding was that mitochondrial protein synthesis inhibitors could improve the fitness of MD models including Ndufs4^−/− mice [65], challenging the notion mitochondrial toxins should be avoided by MD patients. We screened a library of biologically active compounds in MELAS cybrids undergoing cell death from glucose deprivation and found that tetracycline antibiotics robustly rescued cell survival [65]. This effect was not specific to the MELAS tRNA^{Leu (UUR)} mutation, but general across MD cell lines and even pharmacological treatment with mitochondrial inhibitors [65]. Screening a library of tetracycline analogs, we found that inflammatory gene expression correlated with mitochondrial ribosomal inhibition and cell survival [65]. The underpinnings to this inflammatory response modulated by mitoribosomal inhibition associated with glycolysis, PPP and nucleotide synthesis metabolites [65]. NADPH/NADP⁺ ratios were increased with doxycycline prior to other metabolic alterations, indicating mitoribosomal inhibition improves redox homeostasis through an unknown mechanism, preventing inflammatory gene expression and cell death [65].

Doxycycline alleviated neuromuscular decline, neuroinflammation and neuroimmune signaling, and lifespan deficits in Ndufs4^−/− mice [65]. At the brain proteomic level, doxycycline decreased microglial and astrocyte activation, complement, and the interferon response proteins [65] thereby providing an anti-inflammatory signature consistent with other interventions [31, 72, 169]. In brain tissue, doxycycline triggered metabolic shifts in NADPH/glutathione redox, polyamine biosynthesis, and itaconic acid metabolites, and similar to rapamycin [166], rescued glutamine and glutamate levels [65].

Despite in vitro evidence of mitoribosomal inhibition from tetracyclines, the mechanism of rescue in mice may differ. Accumulation of doxycycline in the liver or gut – acting as an antibiotic and altering the microbiome – could release of neuroprotective factors into the bloodstream [180]. Doxycycline does buildup in the brain of Ndufs4^−/− mice at doses similar to in vitro studies, indicating that direct effects on brain mitoribosomes are also entirely possible. The antibiotic linezolid was shown to prevent pathogenic type 17 helper T-cell driven autoimmune disease through cell intrinsic effects on the mitochondrial translation machinery [181], providing supplementary evidence for in vivo effects of antibiotics on mitochondria.

These findings support additional investigation of tetracycline antibiotics as therapy in MD patients. Although fairly safe for the general population, concerns with aminoglycosides, another class of antibiotics that can target the mitoribosome, in MD patients do exist [182]. Follow-up studies in humans should be approached with caution, but the ease of administration, low cost, and well-studied effects of tetracyclines makes them an attractive therapeutic candidate. Another feasible path forward is to disentangle beneficial effects of tetracyclines in MD models from the gut microbiota [180]. If the gut microbiota is not required for improvements in MD in vivo models, one could design tetracycline analogs with enhanced specificity to the mitoribosome to offset side effects. Alternatively, identification and targeting of cellular mechanisms stimulated by tetracyclines independent of the mitoribosome or gut are promising.

NADH and NADPH redox

Mitochondrial respiratory inhibition leads to an accumulation of NADH, hindering TCA cycling and other NAD⁺ sensitive enzymes. As a result, glucose-derived pyruvate accumulates and is converted to lactate by lactate dehydrogenase (LDH) to re-oxidize NADH [183], causing lactate acidemia in patients. Aside from metabolic roles, NAD⁺ is a critical cofactor for multiple enzymes such as polyADP ribose polymerase, cyclic ADP ribose synthetases, and sirtuin deacetylases. One function of the sirtuin SIRT1 is to increase PGC-1α activity through deacetylation and drive the mitochondrial biogenesis transcriptional program [184]. A consequence of improving NADH homeostasis in MD models is augmented mitochondrial mass to compensate for existing respiratory defects. In agreement, PGC-1α overexpression ameliorates complex IV assembly in Surf1 KO mice [185] and improves skeletal muscle and cardiac function in mtDNA mutator mouse [186].

A promising therapeutic strategy in MD models is supplementation with NAD⁺ or targeting enzymes involved in its biosynthesis. This targets at least two aspects of mitochondrial biology, the mitochondrial biogenesis program through activation of PGC-1α as previously discussed, and NADH accumulation in the mitochondria. In Leigh syndrome complex I deficient cell lines, supplementation of cells in galactose media with NAD⁺ rescues cell survival and ATP levels [187]. Oral administration of nicotinamide riboside (NR), a NAD⁺ precursor, in mitochondrial myopathy Deletor mice delayed disease progression, increased mitochondrial biogenesis in certain tissues, and prevented other mitochondrial abnormalities [188]. NR supplementation improved muscular performance and increased mitochondrial gene expression of the Sco2 knockout/knockin (Sco2^KOKI) mouse with impaired complex IV function [189]. Similarly, treatment of Ndufs4^−/− full body KO mice with the NAD⁺ precursor nicotinamide mononucleotide increased lifespan, decreased lactic acidosis, and affected NAD⁺/NADH homeostasis primarily in the muscle, but not the brain [190]. Genetic restoration of NAD⁺ in Ndufs4^−/− mice also prevents pathology. NDI1 is a single subunit NADH dehydrogenase that oxidizes NADH to NAD⁺ without concomitant proton pumping and ATP generation [191, 192]. Expression of yeast NDI1 in Ndufs4^−/− brain-specific KO mice dramatically extended lifespan, prevented neuroinflammation, but failed to improve motor function [193], establishing NADH accumulation as a primary driver of brain pathology in MD. In aggregate, these studies firmly establish that treatments that effectively reestablish NADH redox in MD models are strong candidates for clinical trials.

Reestablishing NADPH redox homeostasis ameliorates cell death of MD cell culture models [62], but whether this translates in vivo is not completely clear. Complex I mutant cybrids are susceptible to cell death upon loss of PPP activity since NADPH production from mitochondrial 1C metabolism is reduced [62]. Forced expression of cytosolic ME1 buffers cells from oxidative damage through NADPH production and improves their survival [62]. Hallmarks of redox stress linked to NADPH dysregulation in vitro such as reduced GSH/GSSG ratios, JNK MAPK activation, and inflammatory markers are also observed in brains of Ndufs4^−/− mice [62], indicating that similar responses occur in vivo. It is conceivable that small-molecule strategies to boost NADPH production such as ME1 activators could be therapeutically beneficial in MD. The fact that PGC-1α prevents pathology in MD models also supports this view since PGC-1α is a major regulator of mtROS and oxidative damage [194–196]. Preclinical MD models show improvement with antioxidants like N-acetylcysteine and vitamin E [197], but such strategies are yet to translate in patients. Modulating metabolic enzyme activity or regulators that directly affect cellular redox instead of relying on antioxidant supplementation, with its many caveats, is a more logical and convincing strategy.

Antioxidants

The majority of ROS species are derived from the mitochondria at respiratory complexes I and III, where escaped electrons interact with molecular oxygen [198]. The reduction of O₂ by a single electron creates O₂^•−, which is highly reactive and oxidizes cellular components in proximity to the site of production such as lipids, proteins, and nucleic acids. If this occurs at a high enough level, it can lead to breakdown of mitochondrial respiratory function and ATP generation. To protect mitochondria from oxidative damage and prevent disease pathogenesis, chemical antioxidants have been developed and even entered into clinical trials. KH176, a vitamin E derivative, and the mitochondrial-targeted ubiquinone derivative MitoQ are being explored as treatments for MD. Idebenone, an analog of CoQ10 is also in trials for patients with LHON. Unfortunately, antioxidant approaches are not proving efficacious in clinical trials. In patients with m.3243A>G mutations, KH176 failed to reach its primary endpoint of improving motoric abnormalities and movement in 28 days [199]. A more long-term follow-on study is now in place. Similarly, the phase II clinical trial RHODOS concluded that idebenone did not improve primary outcomes in patients [200].

Mitochondrial-targeted peptides are another therapeutic avenue for mitochondrial disorders. The cell-permeant cationic tetrapeptide Elamipretide sustains mitochondrial function and prevents cellular deterioration in pre-clinical models of mitochondrial dysfunction reviewed elsewhere [201]. Originally designated as an antioxidant or ROS scavenger due to correlation with reduced ROS levels from pathological mitochondria [202, 203], it is now understood to interact with cardiolipin (Birk et al., 2013), improving inner membrane structure and respiratory function during mitochondrial dysfunction [204]. Unfortunately, MMPOWER-3, a recent phase III clinical trial, reported that Elamipretide did not improve the primary outcome in patients with mitochondrial myopathies [205], complicating its future use in MD.

The lack of antioxidant efficacy in clinical trials possibly stems from the inherent benefit of ROS in intracellular signaling and adaptation to mitochondrial dysfunction. Recent studies demonstrate that expression of alternative oxidase (AOX) in models of mitochondrial deficiencies reduce lifespan and healthspan [206]. AOX is membrane bound enzyme present in plants and lower eukaryotes encoded by a single gene product that transfers electrons from CoQH₂ to O₂, bypassing complex III and IV [207]. It restores CoQ redox in the mitochondria when the pool becomes overly reduced from defects in complex III or IV [207]. In this situation, AOX expression prevents reverse electron transport and complex I-driven ROS production [206]. In a mouse model of complex IV-defective mitochondrial myopathy, expression of AOX enhanced degenerative phenotypes [206]. This correlated with decreased AMPK/PGC-1α and NFE2L2 signaling and PAX7/MYOD-dependent muscle regeneration [206]. Similarly, AOX expression prevented reverse electron transport-induced longevity in Drosophila [208]. In constrast, AOX expression is beneficial in certain mutational contexts such as in the complex III deficient GRACILE mouse model, where it rescued cardiomyopathy, kidney atrophy, respiratory activity, and decreased metabolic stress responses [209]. Phenotypic improvements in GRACILE mice were largely independent of ROS [209], suggesting that restoration of mitochondrial electron transport may be more relevant. Thus, targeting mtROS production can be therapeutically productive in certain pathological contexts, but perhaps not MDs.

Genetic technologies

Maternally inherited MD is inherently complex due to mtDNA heteroplasmy. However, the existence of wild-type genome copies provides a therapeutic advantage. Shifting heteroplasmy away from mutant mtDNA is possible through enhancing mitophagy or more feasibly, genome editing or degradation technologies. Zinc-finger nucleases (ZFNs), transcription activator-like effector nuclease (TALENS), and CRISPR-Cas9 possess the ability to recognize mutant mtDNA and degrade it, but the requirement of RNA import currently precludes CRISPR-Cas9. For these technologies to function, enzymes are imported into the mitochondrial matrix to access mtDNA, excluded from the nucleus, and designed to specifically recognize mutant mtDNA. Inclusion of a mitochondrial targeting sequence at the N-terminus of the ZFN or TALEN traffics these enzymes to the mitochondria [210]. To ensure degradation specificity, a dual system for ZFN or TALENS is employed where each copy is fused to a FokI nuclease that requires dimerization for activity [210]. One ZFN/TALEN recognizes wild-type sequence proximal to the mutation, whereas the second binds the mutation, enabling dimerization of FokI only at mutant genomes. Using these strategies, ZFN and TALEN technologies delivered via Adeno-associated virus (AAV) vectors successfully shifted heteroplasmy towards wild-type in the pathogenic mt-tRNA^Ala mouse model containing the m.5024C>T mutation [211, 212]. Another approach that builds upon these technologies and circumvents the destruction of mtDNA molecules is the mitochondrial base editor [213]. Recently, an interbacterial toxin that deaminates cytidines within dsDNA termed Ddda was identified [213]. An engineered split version of this enzyme targeted to the mitochondria and fused to TALE proteins and a uracil glycosylase inhibitor catalyzed C•G-to-T•A conversion in mtDNA with efficiencies around 5–50% [213]. Intriguingly, this occurs without CRISPR-Cas9 requirements such as a protospacer adjacent motif, guide RNA, or double-stranded breaks [213]. Overall, genetic approaches hold considerable promise for correcting pathological mtDNA variants, but with existing barriers to gene therapy, it will take time to reach MD patients compared to the pharmacological treatments mentioned above.

Concluding remarks and future directions

While progress in MD therapy has been slow owing to the rarity of the disease, phenotypic complexity, and lack of standardized measures and biomarkers for clinical success, there are considerable reasons to be hopeful. As outlined in this review, our understanding of how specific aspects of mitochondrial function are sensed by the cell and addressed through compensatory stress responses has never been greater. Basic science inquiries into MD models over the past decade has revealed causes of pathogenesis and the pharmacological, genetic, and environmental tools that ameliorate phenotypic manifestations of disease. We are at a point where we can start to envision effective therapies with solid scientific foundations. This contrasts many therapeutics currently in trial based off anecdotal reports or that ignore the wealth of literature surrounding the benefits of ROS for normal physiology. As the technical hurdles of gene therapy such as delivery, tissue targeting, safety, among others are passed, editing/removal of nuclear or mitochondrial encoded gene mutations will become a realistic goal. Furthermore, screening technologies will continue to uncover novel drug targets and mechanistic information regarding current pharmacological approaches outlined in this review, allowing better on-target specificity. The application of personalized medicine to differentiate patients by their mutational status and metabolic and energetic defects will also benefit future clinical trials. Overall, these points paired with advances in clinical trial design [9] and understanding of the cellular and organismal responses to mitochondrial dysfunction, including mouse pre-clinical models, will enable rationale therapeutic design for mitochondrial disorders.

Abbreviations

4E-BP1

eukaryotic translation initiation factor 4E (eIF4E)-binding protein 1

AMPK

AMP-activated protein kinase

AOX

alternative oxidase

ASK1

apoptosis-signaling kinase 1

BRD4

bromodomain-containing protein 4

CoQ10

Coenzyme Q10

CPT1

carnitine palmitoyl transferase

DHODH

dihydroorate dehydrogenase

ER

endoplasmic reticulum

ETC

electron transport chain

FAO

fatty-acid β-oxidation

Fe-S

iron-sulfur

H₂O₂

hydrogen peroxide

HIF-1α

hypoxia-inducible factor 1α

HRI

heme-regulated eIF2α kinase

ISR

integrated stress response

LHON

Leber hereditary optic neuropathy

MAPKs

mitogen-activated protein kinases

MD

mitochondrial disease

ME1

malic enzyme 1

MELAS

mitochondrial encephalomyopathy, lactic acidosis, and stroke-like episodes

MMP

mitochondrial membrane potential

MPC

mitochondrial pyruvate carrier

mtDNA

mitochondrial DNA

mtROS

mitochondrial ROS

ND1

NADH-ubiquinone oxidoreductase chain 1 protein

nDNA

nuclear DNA

NDUFS4

NADH dehydrogenase [ubiquinone] iron-sulfur protein 4

NNT

NADH-NADPH transhydrogenase

O2⁻

superoxide anions

OMM

outer mitochondrial membrane

OXPHOS

oxidative phosphorylation

PC

pyruvate carboxylase

PDH

pyruvate dehydrogenase complex

PGC-1α or Ppargc1a

peroxisome proliferator-activated receptor gamma coactivator 1-α

PHDs

prolyl hydroxylases

PINK1

PTEN Induced Kinase 1

PRODH

proline dehydrogenase

PTEN

phosphatase and tensin homolog deleted on chromosome 10

ROS

reactive oxygen species

S6K1

ribosomal protein S6 kinase beta-1

TCA

tricarboxylic acid cycle

Trx

thioredoxin

UPR^mt

mitochondrial unfolded protein response

VHL

E3 ubiquitin ligase von Hippel Lindau

α-KG

α-ketoglutarate