Mitochondria and Longevity

Forget the motivational-poster powerhouse line. Mitochondria are 1.5-billion-year-old endosymbiotic bacteria that moved into our cells and stayed. They kept their own circular genome. And they make roughly 90 percent of cellular ATP through a four-complex electron transport chain (ETC, the assembly line that turns food into usable energy).

The bioenergetic core. Electrons enter the ETC from NADH at Complex I, or from FADH₂ at Complex II (succinate dehydrogenase). Coenzyme Q10 (ubiquinone) carries them through the lipid bilayer to Complex III. Cytochrome c then hands them to Complex IV (cytochrome c oxidase), which turns oxygen into water. Complexes I, III, and IV pump protons against their gradient, building a 150 to 180 mV electrochemical potential across the inner membrane. ATP synthase (Complex V) lets those protons back in and uses the kinetic energy to turn ADP into ATP. Burn one glucose molecule fully and you get about 30 to 32 ATP. Palmitate β-oxidation gives you about 106.

Reactive oxygen species (ROS) leak is unavoidable. Two sites make most of the physiological superoxide: Complex I's ubiquinone-binding pocket (site I_Q, especially during reverse electron transport) and Complex III's Q_o site. Brand's group at Cambridge has mapped at least 11 distinct mitochondrial superoxide-producing sites in mammalian tissues. Ubiquinol (the reduced form of CoQ10) does double duty as the inner membrane's main lipophilic antioxidant.

Mitochondrial DNA is tiny and exposed. Human mtDNA is a 16,569 bp circular molecule with just 37 genes: 13 OXPHOS subunits (7 of Complex I, 1 of Complex III, 3 of Complex IV, 2 of Complex V), 22 tRNAs, and 2 rRNAs. No histones, no introns. It sits tethered to the inner membrane, meters from the ROS source on a cellular scale. Its mutation rate is 10 to 20x higher than nuclear DNA. mtDNA is maternally inherited, present at hundreds to thousands of copies per cell, and it has a biochemical threshold effect: clinical symptoms usually only show up above 60 to 90 percent mutant load.

Dynamics are quality control. Mitochondria constantly fuse (MFN1/MFN2 outer membrane, OPA1 inner membrane) and divide (DRP1 + FIS1/MFF). When a mitochondrion loses membrane potential, PINK1 stabilizes on its outer membrane, calls in the E3 ligase Parkin, tags outer-membrane proteins with ubiquitin, and the damaged organelle gets cleared through mitophagy (the cell's trash-and-recycle program built just for mitochondria). The 2008 work from Youle's lab (Narendra, J Cell Biol) first showed Parkin homing in on damaged mitochondria. Break any of these proteins and you get real disease: MFN2 becomes Charcot-Marie-Tooth 2A; OPA1 becomes dominant optic atrophy; PINK1/PRKN becomes recessive early-onset Parkinson's.

Biogenesis is on demand. The master regulator is PGC-1α (PPARGC1A), cloned by Spiegelman's group in 1998 (Puigserver, Cell) from a cold-exposed brown-fat library. PGC-1α coactivates NRF1 and NRF2, which switch on nuclear-encoded OXPHOS genes and TFAM, the mitochondrial transcription factor. Upstream activators: AMPK (the cellular fuel-low sensor, which fires when ATP runs low) and SIRT1 (an NAD⁺-dependent deacetylase). The physiological triggers with the best evidence are exercise, cold, fasting, and caloric restriction. None of them come in a capsule.

Mitochondrial function fades with age in basically every tissue studied. The pattern is consistent. The cause is still argued over.

The classic free radical theory of aging (Harman 1956, expanded in 1972 to point the finger at mitochondrial ROS specifically) said that ROS damage to mtDNA, lipids, and proteins drove aging. For three decades it shaped the field. In its strong form, it is now dead.

What killed it. The mutator mouse (two landmark papers, Trifunovic 2004 Nature and Kujoth 2005 Science) carries a proofreading-deficient mitochondrial DNA polymerase (PolgA D257A). These mice pile up 3 to 5x more mtDNA mutations than wild type and age early: graying, kyphosis, sarcopenia, cardiac hypertrophy, anemia, median lifespan around 12 months versus about 30 in controls. Here is the twist. Kujoth's group found no global rise in oxidative-stress markers in these animals. You can break mtDNA, speed up aging, and never see the ROS signature the theory predicted. A 2009 Biochimica et Biophysica Acta study from the Van Remmen and Richardson labs (Pérez et al.) tested 18 antioxidant gene manipulations in mice; only one (Sod1 deletion) shortened lifespan. Cochrane meta-analyses of human antioxidant supplementation show no mortality benefit and possible harm at high doses.

What survives. Ristow's mitohormesis framework (Schulz 2007 Cell Metabolism; later reviewed by Ristow and Schmeisser in 2014) says this: low, brief, compartmentalized mitochondrial ROS, especially during exercise, fasting, and caloric restriction, trigger adaptive responses (NRF2-driven antioxidant transcription, mitochondrial biogenesis, autophagy) that improve healthspan. Chronic, high, dysregulated ROS is a result of mitochondrial dysfunction and feeds disease. But it is not the prime mover. Wallace's 2018 mitochondrial-aging theory (J Clin Invest) reframes the picture: heteroplasmy buildup, bioenergetic decline, and signaling failures, not raw ROS damage, are the drivers.

Specific markers of age-related decline. Tissue CoQ10 in the human heart drops roughly 50 percent between age 20 and 80 (Kalén 1989). NAD⁺ in human skin falls about 50 percent between age 20 and 60 (Massudi 2012). You see similar curves in adipose, muscle, and cerebellum. More and more, this looks like a consumption problem rather than a synthesis problem: CD38 (an NAD⁺-consuming enzyme) climbs with age, pushed by senescence-associated inflammation (Camacho-Pereira 2016 Cell Metabolism; a 2020 Nature Metabolism paper from the Verdin lab, Covarrubias et al., showed senescent cells activate CD38⁺ macrophages). Plasma taurine declines with age in mice and monkeys. The human picture is now contested (more below).

Practical takeaway. Mitochondrial dysfunction with age is real. Calling it the cause of aging is not justified by current evidence. That matters for treatment: broad antioxidants are a dead end. The things that improve mitochondrial quality control (exercise, fasting, cold, NAD⁺ precursors in some contexts, sleep) have far better evidence behind them.

A 2013 Cell paper from López-Otín and colleagues organized the modern aging field around nine hallmarks. The 2023 update ("Hallmarks of Aging: An Expanding Universe," also in Cell) added three more (disabled macroautophagy, chronic inflammation, and dysbiosis) for a total of twelve.

Mitochondrial dysfunction is one of them. And it touches most of the others. It drives proteostatic stress, feeds "inflammaging" through mtDNA release and mtROS-induced NLRP3 activation, sits downstream of nutrient sensing (mTOR, AMPK, SIRT1), and builds up alongside stem-cell exhaustion.

A hallmark is not a cause. López-Otín's framework asks for three things: the phenomenon must show up with age, it must speed up aging when made worse, and slow aging when fixed. "Mitochondrial dysfunction with age" passes all three. "Mitochondrial dysfunction is the cause of aging" does not. It is one of twelve correlated phenomena, all of which lean on each other.

The mitochondrial unfolded protein response (UPR^mt) is an active longevity-research target. In C. elegans, the transcription factor ATFS-1 senses how well proteins import into mitochondria: when import works, ATFS-1 gets degraded; when import stalls (stress), it builds up in the cytosol and heads to the nucleus instead, switching on mitochondrial chaperones and quality-control proteases. A 2013 Nature paper from the Auwerx lab (Houtkooper et al.) showed that a mild knockdown of mitochondrial ribosomal proteins triggers "mitonuclear protein imbalance," fires up UPR^mt, and extends C. elegans lifespan by 50 to 60 percent. NAD⁺-boosting interventions copy a slice of this in worms and mice. Human translation is plausible but unproven.

Tissue-specific dependencies. Heart, brain, slow-twitch skeletal muscle, and liver pack the highest mitochondrial density. Primary mitochondrial diseases (MELAS with m.3243A>G in mt-tRNA-Leu, MERRF with m.8344A>G in mt-tRNA-Lys, LHON with Complex I subunit mutations) wreck these tissues first. Here is the aging link: the same tissues that fail in mitochondrial disease are the ones that fail with age. Neurons and cardiomyocytes are post-mitotic or slow-turnover. They collect damaged mitochondria and heteroplasmy (a mix of healthy and mutant mtDNA inside the same cell) over decades.

Clinically significant mtDNA disease hits roughly 1 in 5,000 people. About 1 in 200 healthy individuals carries a pathogenic mtDNA variant at sub-threshold heteroplasmy (Gorman 2016, Nature Reviews Disease Primers). Mitochondrial biology is not a rare-disease footnote. It sits under a large chunk of neurology and cardiology.

No supplement matches what training, sleep, and dietary structure do for mitochondrial function. The PGC-1α / AMPK / SIRT1 axis (the master regulator of mitochondrial biogenesis) answers to muscle contraction and energy stress. Capsules barely touch it.

Zone 2 endurance. Holloszy's 1967 Journal of Biological Chemistry paper is the founding study. Rats ran 2 h/day, five days a week, for 12 weeks, and roughly doubled their muscle mitochondrial protein and respiratory enzyme activity. This is the original exercise-induced biogenesis result. A 2014 Biochimica et Biophysica Acta review (Bishop and Granata) boiled down four decades of follow-up: training volume drives mitochondrial content (density, citrate synthase activity, OXPHOS protein abundance); training intensity drives mitochondrial function (respiration per unit mitochondria). You need both. A 2018 Sports Medicine study (San Millán and Brooks) put elite cyclists next to metabolic-syndrome patients and found the elite athletes kept burning fat at much higher absolute power outputs before lactate climbed. "Metabolic flexibility" is mostly a story about mitochondrial mass and quality.

Practical zone 2 prescription: 3 to 4 sessions per week, 30 to 60 minutes each, RPE 5 to 6 on a 10-point scale ("can talk but not sing"), aiming for a heart rate just below the first lactate threshold (~2 mmol/L lactate, often ~60 to 70 percent HRmax in trained adults). The reason to favor zone 2 is volume tolerability, not magic. Harder efforts produce comparable signals per minute, but you cannot repeat them as often.

HIIT. A 2017 Journal of Physiology review (MacInnis and Gibala) is the go-to reference. Sprint interval training and high-intensity intervals raise citrate synthase, COX, and total mitochondrial content about as well as traditional endurance training, in less time. The "10 × 1 minute at ~90 percent HRmax with 1-minute recovery" protocol fires up AMPK and p38 MAPK harder per minute than steady moderate work. HIIT adds to zone 2. It does not replace it. Two HIIT sessions per week is a sensible ceiling for most adults.

Resistance training. A 2017 Cell Metabolism study (Robinson et al.) compared HIIT, resistance, and combined training in young and older adults. HIIT produced the broadest mitochondrial gene-transcription response and, in older adults, pushed the mitochondrial proteome back toward a younger profile. Resistance training boosted muscle protein synthesis but did not match HIIT for mitochondrial signals. Translation: lift weights for sarcopenia and strength; do HIIT and zone 2 for mitochondria. Concurrent training is fine if you split the modalities by day.

Time-restricted eating and fasting. A 2018 Cell Stem Cell study (Mihaylova et al.) showed a 24-hour fast switches on a PPAR-driven fatty-acid oxidation program in intestinal stem cells through CPT1A. A 2018 Cell Metabolism trial (Sutton et al.) tested a 6-hour eating window (last meal before 3 PM) for 5 weeks in prediabetic men and improved insulin sensitivity, β-cell responsiveness, blood pressure, and oxidative stress, without weight loss. The TREAT trial (Lowe 2020, JAMA Internal Medicine) is the negative counterweight: 16:8 with a noon-to-8-PM eating window produced only ~1 percent weight loss over 12 weeks and worrying lean-mass loss. The signal favors an early-window TRE kept up for months. Late-window 16:8 on its own is unimpressive.

Caloric restriction. The CALERIE 2 trial (Kraus 2019, Lancet Diabetes Endocrinology) hit ~12 percent CR for 2 years in healthy non-obese adults. Every cardiometabolic risk factor improved beyond what weight loss alone would predict. A 2023 Nature Aging re-analysis of CALERIE 2 with epigenetic clocks (Belsky's group) found the DunedinPACE pace-of-aging clock slowed 2 to 3 percent in the CR arm. That is comparable, in projected mortality reduction, to quitting smoking.

Cold exposure. A 2012 J Clin Invest study (Ouellet et al.) established that adult humans have working brown adipose tissue (BAT) that switches on with cold; UCP1 uncouples the ETC from ATP synthesis to make heat instead. Van der Lans 2013 confirmed 10 days of mild cold acclimation raises BAT activity and non-shivering thermogenesis. The cold-plunge culture oversells the mitochondrial story: a 2021 Cell Reports Medicine study (Søberg et al.) found habituated winter swimmers showed thermoregulatory efficiency adaptations, not raw biogenesis expansion. Cold helps at the margins. It is not a stand-in for zone 2.

Sleep. A 2018 Science Advances study (Cedernaes et al.) showed that one night of sleep deprivation changes DNA methylation in clock genes and dials down oxidative phosphorylation transcripts in human skeletal muscle. A 2022 Frontiers in Endocrinology study (Saner et al.) found 5 nights of sleep restriction cut muscle mitochondrial respiration and protein synthesis. Sleep is the cheapest mitochondrial intervention you can give yourself.

Avoiding obvious damagers. Smoking directly blocks Complex IV through cyanide and carbon monoxide binding. Alcohol suppresses mitochondrial protein synthesis, triggers electron leak from Complex III, and shifts the NAD⁺/NADH ratio in liver and muscle. Ultra-processed food intake links to dysbiosis, LPS translocation, systemic inflammation, and downstream mitochondrial dysfunction at the population level. The cellular-level certainty here is inferred more than proven.

The honest prescription for a healthy adult. Zone 2 + HIIT + enough protein + 7 to 9 hours of sleep + occasional fasting will shift every meaningful mitochondrial marker (VO₂ max, lactate threshold, citrate synthase activity, OXPHOS protein content) more than any supplement stack on the market. Anyone selling you a mitochondrial supplement stack as a replacement is selling you the weaker intervention at the higher price.

Say you wanted to anchor a mitochondrial-focus supplement framework on biochemistry alone. Four molecules sit at structurally privileged spots in the ETC. The clinical evidence varies by indication. The mechanistic case is unusually strong.

CoQ10 (Ubiquinone / Ubiquinol)

CoQ10 is the obligate electron carrier between Complexes I/II and III. There is no biochemical substitute. It is also the inner membrane's main lipophilic antioxidant. Synthesis from tyrosine and the mevalonate pathway gets blocked by statins, which inhibit HMG-CoA reductase upstream; statins cut circulating CoQ10 by 16 to 54 percent across trials. Tissue levels in the human heart fall roughly 50 percent from age 20 to 80 (Kalén 1989).

The flagship trial is Q-SYMBIO (Mortensen 2014, JACC Heart Failure, n=420 NYHA III to IV chronic heart failure, 300 mg/day for 2 years). Major adverse cardiac events fell by half (HR 0.50, 95% CI 0.32 to 0.80). Two-year all-cause mortality: 10 percent on CoQ10 versus 18 percent on placebo. Cardiovascular mortality: 9 versus 16 percent. Symptoms (NYHA class) and hospitalizations improved too. KiSel-10 (Alehagen 2013, with 12-year follow-up published in 2018) showed lasting cardiovascular mortality reduction in elderly Swedes on selenium plus CoQ10. Confounded by selenium, but consistent in direction.

Where evidence is mixed: statin-induced myopathy (a 2015 Mayo Clinic Proceedings meta-analysis from Banach was equivocal on CK and pain endpoints), male sub-fertility (Lafuente's 2013 meta-analysis found better sperm motility, no pregnancy benefit), migraine prophylaxis (Sándor 2005, Neurology: 50-percent-responder rate 48 percent versus 14 percent on placebo, NNT 3). Where the evidence is clearly negative: Parkinson's disease. The NIH QE3 trial (Beal 2014, JAMA Neurology, 600 patients on 1200 or 2400 mg/day) was stopped for futility, and trended worse on active treatment. Do not claim CoQ10 helps Parkinson's.

Ubiquinone versus ubiquinol: plasma runs about 95 percent ubiquinol in healthy adults. The "ubiquinol is 2 to 3x more bioavailable" framing rests largely on manufacturer-funded studies. A 2019 Nutrition independent comparison (López-Lluch) showed formulation matters more than redox form. A well-made ubiquinone soft-gel can match a poorly-made ubiquinol capsule. See the dedicated CoQ10: Ubiquinol vs Ubiquinone guide for the full funding map and DACH product detail.

Dose: 100 to 300 mg/day with a fat-containing meal, split above 200 mg (absorption saturates). Q-SYMBIO used 300 mg/day. Tolerated up to 1200 to 2400 mg/day in trials. Drug interactions: monitor INR with warfarin above 100 mg/day.

EU food-claims status: EFSA reviewed CoQ10 health-claim dossiers in 2010 (EFSA Journal, 2010) and rejected all six categories (energy metabolism, blood pressure, antioxidant protection, cognitive function, cholesterol, endurance). No authorised health claim exists in the EU. The honest editorial line is to describe what studies measured (Q-SYMBIO, statin myopathy trials, EFSA's rejection itself) and skip label-style "supports cardiovascular function" wording.

Riboflavin (Vitamin B2): The Unglamorous Heavyweight

Riboflavin makes FMN and FAD inside cells. FMN is the cofactor at Complex I's electron entry point (NDUFV1 subunit). FAD is covalently bound to Complex II's SDHA subunit. The ETF / ETFDH bridge, which all fatty-acid β-oxidation dehydrogenases use to feed electrons into the chain, needs FAD. So riboflavin sits at three load-bearing nodes of mitochondrial metabolism and acts as a prosthetic group for roughly 90 human flavoproteins.

Riboflavin-responsive mitochondrial disorders. This is one of the cleanest "vitamin treats a mitochondrial disease" stories in medicine. Late-onset multiple acyl-CoA dehydrogenase deficiency (RR-MADD), caused by ETFDH mutations, responds to oral riboflavin 100 to 400 mg/day under specialist neurology care. A 2007 Brain study (Olsen et al.) pinned down the genotype; later cohorts (Wen 2010, Xi 2014, Hong 2022) report response rates near 98 percent in type-3 MADD. Brown-Vialetto-Van Laere syndrome (caused by SLC52A3/A2 riboflavin-transporter mutations) can be life-saving on high-dose oral riboflavin under specialist neurology care (Foley 2014, Brain; a 2017 Practical Neurology case from Bashford described an adult patient who went from tetraparetic and ventilated to independent living on 1200 mg/day).

Migraine prophylaxis is the best non-rare-disease RCT evidence. Schoenen 1998, Neurology: 55 patients, 400 mg/day for 3 months, 50-percent-responder rate 59 percent versus 15 percent on placebo (p=0.002), NNT 2.3. Pediatric trials at 200 mg/day are negative (MacLennan 2008; Bruijn 2010). AAN/AHS rate riboflavin Level B for adult migraine prevention.

Cardiovascular niche. A 2013 Hypertension study (Wilson et al.) showed 1.6 mg/day riboflavin for 16 weeks dropped systolic blood pressure by 5.6 mmHg in MTHFR 677TT homozygotes specifically. It is a genotype-specific effect, because the thermolabile TT variant loses its FAD cofactor and gets stabilized by extra riboflavin.

Dose, safety, EU compliance: maximum single-dose absorption is ~27 mg (Zempleni 1996), so 400 mg/day works best split up. No EFSA tolerable upper intake level. No toxicity threshold has ever been set. Excess turns your urine a fluorescent yellow, which is harmless and just confirms absorption.

Riboflavin has the strongest EU food-claims portfolio of any nutrient on this list: nine authorised Article 13(1) claims (Regulation 432/2012) including "contributes to normal energy-yielding metabolism" and "contributes to the reduction of tiredness and fatigue."

NAD⁺ Precursors: Nicotinamide Riboside and NMN

NAD⁺ is the redox cofactor for hundreds of dehydrogenases and the obligate co-substrate for sirtuins (NAD⁺-dependent deacetylases), PARPs (DNA repair), and CD38 (the dominant NAD⁺-consuming enzyme in aged tissues). NAD⁺ falls with age in multiple tissues. In human skin, the Massudi 2012 cross-sectional series reported a more than 50 percent drop from newborn to young adult and a further ~60 percent decline from young adult to middle age. Similar curves show up in adipose, muscle, and cerebellum. More and more, this looks like a consumption problem driven by CD38 upregulation in senescence-associated inflammation, not a synthesis failure.

Mechanism. The longevity story rests on SIRT1 deacetylating PGC-1α to drive mitochondrial biogenesis, plus SIRT3 deacetylating ETC components and SOD2 inside mitochondria. The mechanism is solid molecular biology. The translation to human outcomes is more modest.

Human RCTs for NR: Trammell 2016, Nature Communications: dose-dependent blood NAD⁺ rose up to 2.7-fold after a single 1000 mg dose in a single-subject pilot (the cohort-level n=12 dose-response curve was shallower; the 2.7x ceiling describes one participant, not a group average). Martens 2018, Nature Communications (1000 mg/day × 6 weeks): PBMC NAD⁺ up ~60 percent; trends toward lower systolic BP and aortic stiffness in stage-1 hypertensives. Dollerup 2018, AJCN (2000 mg/day × 12 weeks in obese insulin-resistant men) was clearly negative on metabolic endpoints despite NAD⁺ elevation. The NADPARK trial (Brakedal 2022, Cell Metabolism, Parkinson's, 1000 mg/day × 30 days): cerebral NAD⁺ rose; clinical improvement was small and exploratory.

Human RCTs for NMN: Yoshino 2021, Science (250 mg/day × 10 weeks in postmenopausal prediabetic women) improved muscle insulin sensitivity by clamp. Yi 2023, GeroScience (300 to 900 mg/day × 60 days) improved 6-minute walk distance dose-dependently with a plateau at 600 mg. Igarashi 2022, NPJ Aging (250 mg/day × 12 weeks) showed blood NAD⁺ elevation and nominal gait-speed improvement.

The 2PY/4PY caveat. Ferrell 2024, Nature Medicine: this study analyzed 4,000+ cardiac patients across three cohorts and reported that terminal niacin metabolites (N1-methyl-2-pyridone-5-carboxamide and N1-methyl-4-pyridone-3-carboxamide) in the top quartile carried 1.6 to 2x the 3-year MACE risk (major adverse cardiovascular event composite) of the bottom quartile, independent of traditional risk factors. 4PY induces endothelial VCAM-1. All NAD⁺ precursors (NR, NMN, NA, NAM) funnel through the same NAM → NNMT → MeNAM → 2PY/4PY pathway. Whether supplement-dose NR or NMN pushes these metabolites into a problematic range is not yet measured in trial cohorts.

EU regulatory status. NR (ChromaDex Niagen™) is authorised as a Novel Food at up to 300 mg/day for adults (Regulation EU 2017/2470 amendments; extended in 2022/1160). NMN is not yet listed on the EU Union list of novel foods. EFSA issued a positive safety opinion in May 2026 at up to 300 mg/day, but Commission Implementing Regulation has not authorised it. Sale in DACH is currently borderline. Neither has an authorised health claim.

Methyl-group pairing: TMG. NR consumption raises methylation demand through NNMT, which converts NAM to MeNAM using SAM. Trimethylglycine (betaine) regenerates SAM through BHMT, topping up the methyl pool. This is real biochemistry. The clinical case for routine TMG pairing is mechanistic rather than RCT-proven. TMG has its own authorised EU claim, "contributes to normal homocysteine metabolism" at ≥500 mg per portion, 1.5 g/day intake. That is the methyl-buffer mechanism, not a mitochondrial claim.

Honest framing: NR and NMN reliably raise NAD⁺ in humans. Health benefits in already-healthy adults are modest, with the cleanest signals showing up in older or compromised populations. The SIRT1/PGC-1α biogenesis story is mechanism, not outcome. The 2PY/4PY caution is real and unresolved.

Taurine: The Underappreciated mt-tRNA Story

Taurine is covalently installed onto the wobble position (U34) of mitochondrial tRNAs to form 5-taurinomethyluridine (τm⁵U), most importantly on mt-tRNA-Leu(UUR) and mt-tRNA-Lys. Without this modification, codon recognition goes fuzzy at UUG (Leucine) and AAG (Lysine), and translation of the 13 mtDNA-encoded ETC subunits, especially the UUG-rich Complex I subunit ND6, fails.

MELAS is the strongest mechanistic anchor. The m.3243A>G mutation in mt-tRNA-Leu(UUR) (the cause of ~80 percent of MELAS) blocks τm⁵U formation, producing systemic energy failure with stroke-like episodes, lactic acidosis, hearing loss, diabetes, and cardiomyopathy. A 2019 Journal of Neurology, Neurosurgery & Psychiatry phase III open-label trial (Ohsawa et al.) gave 9 to 12 g/day oral taurine for 52 weeks to 10 MELAS patients. The annual stroke-like-episode relapse rate fell from 2.22 to 0.72 (p=0.001). The τm⁵U modification rate in peripheral leukocytes rose. Taurine was approved in Japan for MELAS in 2019. Outside Japan it remains off-label but is standard in specialist mitochondrial-disease centers.

Cardiac evidence: taurine transporter (SLC6A6) knockout mice develop cardiac atrophy and dilated cardiomyopathy with reduced exercise capacity (Ito 2008, J Mol Cell Cardiol). The mechanism is mitochondrial. Defective τm⁵U on mt-tRNAs leads to impaired translation, then ETC failure.

The Singh 2023 Science paper and its 2025 rebuttal. Singh et al. (2023, Science) reported that plasma taurine falls with age across species and that 1 g/kg/day oral taurine in middle-aged mice extended median lifespan by ~10 to 12 percent. The paper got wide coverage as a likely longevity intervention. Then a 2025 Science paper from the de Cabo group (Fernandez et al.) used data from the Baltimore Longitudinal Study of Aging and other cohorts to challenge the central claim, finding that taurine actually rose or stayed stable with age in most human populations sampled. The NIH news release on the rebuttal concluded taurine is "unlikely to be a good aging biomarker." The MELAS biology and the TauT-knockout cardiomyopathy story are untouched. The "refill what you've lost" narrative that drove the 2023 to 2024 supplement wave has not held up.

Dose, safety, EU status: Trials use 1 to 6 g/day; MELAS uses 9 to 12 g/day. Observed Safe Level 6 g/day (Shao and Hathcock 2008). Taurine is a permitted ingredient under Regulation 1170/2009 with no authorised health claim. EFSA rejected all proposed claims in 2009 and 2011 (EFSA Journal, 2009 and 2011). It cannot be marketed with cardiac, mitochondrial, energy, or anti-fatigue claims in the EU.

The Core Four: Honest Summary

CoQ10 is mechanistically central and has the best single positive primary RCT (Q-SYMBIO). It is not a longevity drug. Riboflavin is the cheapest defensible mitochondrial cofactor, with the strongest EU claims portfolio and a real role in mitochondrial-disease therapy. NAD⁺ precursors reliably raise NAD⁺, but the human longevity case is preclinical and the 2PY/4PY question is open. Taurine has a beautiful mt-tRNA mechanism and a phase III trial in MELAS; the broader longevity case took a serious hit in 2025.

These compounds have real mitochondrial biology behind them and weaker human evidence. Use them with calibrated expectations.

Magnesium (Bisglycinate or Equivalent)

Magnesium is a Mg-ATP cofactor and a structural ion for ETC Complexes I and V. EFSA UL: 250 mg supplemental per day; NIH UL 350 mg. Forms matter for tolerance more than for bioavailability as such: oxide is laxative and ~14 to 23 percent absorbed; citrate is mid-tier; bisglycinate is gentle on the GI tract, and its glycine carrier delivers a sleep-active molecule along with the magnesium (see the deep-sleep guide). L-threonate (Magtein) is premium-priced on the strength of one rat brain-Mg study (Slutsky 2010, Neuron) and an industry-funded Oura-ring trial in humans (Hausenblas 2024). The value proposition is not matched by independent evidence.

Magnesium has six authorised EU claims including "contributes to normal energy-yielding metabolism," "normal muscle function," and "reduction of tiredness and fatigue."

Creatine Monohydrate

Creatine is not a mitochondrial-biogenesis nutrient. It is an ATP buffering system through the phosphocreatine / creatine kinase shuttle. Phosphocreatine quickly regenerates ATP where demand spikes: myofibers during contraction, neurons during activity. The strongest evidence is in muscle performance (Kreider 2017, ISSN position stand). A 2018 Experimental Gerontology meta-analysis (Avgerinos) found cognitive benefits in older adults, sleep-deprived adults, and vegetarians (who start with lower baseline creatine). Gordji-Nejad 2024, Scientific Reports: a single 0.35 g/kg dose during 21-hour sleep deprivation improved cognitive performance and changed cerebral phosphocreatine on ³¹P-MRS. Kidney safety worries in healthy adults have been debunked again and again.

Dose: 3 to 5 g/day for muscle; 5 to 10 g/day if cognitive endpoints are the goal. Monohydrate is the only form with solid evidence. The EU has two authorised claims: "increases physical performance in successive bursts of short-term, high-intensity exercise" (≥3 g/day) and "can enhance the effect of resistance training on muscle strength in adults over the age of 55" (≥3 g/day plus resistance training, Regulation EU 2017/672). A cognitive claim was rejected by EFSA in 2024.

Methylcobalamin (Vitamin B12)

B12 is the cofactor for methylmalonyl-CoA mutase, a mitochondrial-matrix enzyme that converts methylmalonyl-CoA to succinyl-CoA, feeding propionate-derived carbons into the TCA cycle. Deficiency raises methylmalonic acid (MMA), a direct, measurable functional mitochondrial defect. Subclinical B12 deficiency is common in older adults (10 to 20+ percent over age 60), metformin users (de Jager 2010, BMJ), PPI users (Lam 2013, JAMA), and vegans (up to 50+ percent without supplementation).

Forms: methylcobalamin (methylation), adenosylcobalamin (the mitochondrial form, used by methylmalonyl-CoA mutase), hydroxocobalamin (IM standard), cyanocobalamin (cheapest, fine for most). Practical reality: form matters far less than dose. The EU has eight authorised claims including energy metabolism, nervous system, homocysteine metabolism, and reduction of tiredness and fatigue.

Curcumin

Mechanism includes Nrf2 activation, AMPK, and indirect PGC-1α. But the human outcome evidence is dominated by anti-inflammatory effects, not direct mitochondrial endpoints. Bioavailability is the central problem: raw curcumin is ~1 percent absorbed. Bioenhanced forms (Theracurmin, Meriva, Longvida, CAVACURMIN, C3 Reduct) produce 27 to 65x higher plasma AUC. Best clinical data: osteoarthritis pain (Daily 2016, J Med Food meta), depression adjunct (Ng 2017, J Am Med Dir Assoc meta), metabolic syndrome / lipids (Qin 2017, Nutrition Journal meta). Mitochondria-specific human endpoints are thin.

Safety: rare but documented hepatotoxicity signals with high-bioavailability formulations (Italian, Australian, US DILI registry entries since ~2018). Watch in patients on statins or other hepatotoxic medications. No authorised EU health claim. The "contributes to normal joint function" application was rejected by EFSA in 2017.

PQQ (Pyrroloquinoline Quinone)

Mechanism: CREB phosphorylation drives PGC-1α expression and mitochondrial biogenesis in cell-culture and rodent models (Chowanadisai 2010, J Biol Chem). Human evidence is thin: one small crossover trial (Harris 2013, J Nutr Biochem, n=10) showing biomarker shifts (CRP, IL-6, methylated amines). No human RCT measuring hard clinical endpoints at standard supplemental doses. EU: novel-food authorised at ≤20 mg/day (Regulation 2018/1122); no authorised health claim. The "mitochondrial biogenesis nutrient" marketing language is not supported by human trial evidence. Treat as preclinical to weak.

These compounds got swept into mitochondrial-supplement marketing, but in the human evidence base they are not primary mitochondrial agents. Each has a legitimate role. None of those roles is "directly improves mitochondrial function in healthy adults."

Citicoline (CDP-choline) supports neuronal membrane phospholipid synthesis and acetylcholine availability. The ICTUS trial (Dávalos 2012, Lancet) was negative in stroke and stopped early for futility. It touches mitochondria only through cardiolipin (an inner-membrane phospholipid) after injury. Novel-food authorised in the EU at ≤500 mg/day; no authorised health claim. Peripheral, not a mitochondrial ingredient for healthy users.

Rhodiola rosea has weak AMPK/PGC-1α signals in rodent muscle in vitro (salidroside). No human study has shown a Rhodiola-driven change in mitochondrial respiration, biogenesis, or PGC-1α expression in tissue. A 2011 Phytomedicine review (Hung et al.) concluded efficacy "not convincingly demonstrated." EFSA rejected the "reduction of mental fatigue" claim; EMA classifies Rhodiola as a traditional herbal medicinal product based on traditional use, not modern clinical evidence. An adaptogen with mito framing extrapolated from preclinical pathway data.

P5P (pyridoxal-5'-phosphate, the active form of B6) is a cofactor for ~150 enzymes, including mitochondrial ones (ALAS for heme synthesis; SHMT2 in the mitochondrial folate cycle; mitochondrial transaminases). It is permissive. Supplementing it in non-deficient people does not drive mitochondrial output. EFSA's 2023 reassessment dropped the upper limit to 12 mg/day (from 25 mg). Long-term high B6 causes sensory neuropathy. Vitamin B6 has ten authorised EU claims; form choice (P5P versus pyridoxine HCl) matters more for neuropathy risk at chronic high doses than for clinical benefit at typical supplemental levels.

5-MTHF (L-methylfolate, Metafolin/Quatrefolic) gets around the MTHFR 677TT polymorphism limitation. The mitochondrial folate cycle is real (SHMT2, MTHFD2, MTHFD1L generate formate in mitochondria for export to the cytosol; Ducker and Rabinowitz 2017, Cell Metabolism), but 5-MTHF supplementation acts mainly on cytosolic methylation, with mitochondrial effects downstream and indirect. Best human outcome: a 2012 American Journal of Psychiatry study (Papakostas) showed 15 mg L-MTHF + SSRI in SSRI-resistant depression produced a 32 percent response versus 15 percent placebo. Multiple authorised EU claims for folate including "normal homocysteine metabolism."

Betaine (TMG, trimethylglycine) is a methyl donor (BHMT). Cytosolic, hepatic, not mitochondrial. Its rational pairing is with NAD⁺ precursors: NR and NMN consumption raises NAM, which is methylated by NNMT, draining the SAM pool; TMG regenerates SAM. The authorised EU claim is "contributes to normal homocysteine metabolism" (≥500 mg per portion, 1.5 g/day intake). This is methyl-pool buffer biology, not mitochondrial. Best as a co-administration partner for NR/NMN, never sold solo as mitochondrial.

Myo-inositol is a second messenger (IP3) and a substrate for PIP2/PIP3 in insulin signaling. Strongest evidence: PCOS (Unfer 2017, Endocrine Connections meta) and possibly gestational diabetes (a 2023 Cochrane review, Crawford et al.). Not a mitochondrial agent. An insulin sensitizer at best.

Piperine (BioPerine, black pepper extract) is a CYP3A4 / P-glycoprotein inhibitor. Sole purpose: raise the bioavailability of co-administered substrates such as curcumin (Shoba 1998, Planta Medica: ~2000 percent AUC increase for curcumin co-dosed with 20 mg piperine). No direct mitochondrial action. Piperine carries non-trivial drug-interaction risk. It raises plasma levels of tacrolimus, cyclosporine, simvastatin, midazolam, calcium channel blockers, theophylline, phenytoin, and others. Cap at ≤5 mg/day in supplement contexts; not in patients on interacting medications.

The honest editorial framing: these belong in a "support cast" bracket. Useful in specific contexts (MTHFR polymorphism, NR/NMN methyl-pool support, curcumin bioavailability, PCOS comorbidity), not headline mitochondrial ingredients. If marketing copy calls any of them a "mitochondrial booster," push back.

Food beats single-compound pills for almost everything mitochondrial. Three categories have evidence worth knowing.

Urolithin A (Mitopure)

Urolithin A is a postbiotic. Your gut bacteria make it from dietary ellagitannins in pomegranate, walnuts, and berries. Roughly half of Western adults are "metabotype 0," meaning they make little or no urolithin on their own no matter how much pomegranate juice they drink. Direct supplementation skips the microbiome lottery.

The mechanism is unusually specific: direct induction of mitophagy (the cell's targeted recycling program for damaged mitochondria) through the PINK1/Parkin pathway. Singh and colleagues at EPFL spotted UA in a screen of >1,500 compounds for mitophagy induction. Three Amazentis-sponsored trials are the human evidence base:

• Andreux 2019, Nature Metabolism: first-in-human PK and safety, healthy elderly, dose-dependent plasma exposure and modulation of skeletal-muscle mitochondrial gene expression and plasma acylcarnitines.
• Liu 2022, JAMA Network Open: 66 adults aged 65 to 90, 1000 mg/day for 4 months. Missed the primary 6-minute walk endpoint but improved muscle endurance (~17 to 26 percent in hand and leg testing) and lowered inflammatory and acylcarnitine biomarkers.
• Singh 2022, Cell Reports Medicine: 88 middle-aged adults, 500 or 1000 mg/day for 4 months. Missed the peak-power primary endpoint but produced ~12 percent muscle strength gain, improved VO₂peak and 6MWD, and lowered acylcarnitines and CRP.

Honest reading: this is the most distinctive recent mitochondrial-supplement story with human RCTs published in tier-1 journals. All three trials are Amazentis-funded and missed their primary endpoints. The case rests on secondary endpoints and biomarkers. Effect sizes are modest. The mechanism is unusually clean for a supplement.

Dose: 500 to 1000 mg/day. EU: authorised as a Novel Food (Mitopure synthetic UA); no health claim.

Omega-3 EPA and DHA

This is the one entry on this list with EU-authorised health claims at a single specified intake. The mitochondrial mechanism runs through cardiolipin, the inner-membrane phospholipid that holds ETC supercomplex assembly together. DHA in particular gets built into cardiolipin and reshapes supercomplex architecture (Watson 2018, J Lipid Res).

Human outcome evidence:

• McBurney 2021, AJCN: Framingham Offspring cohort, ~2,200 adults, 11-year follow-up. Red-blood-cell Omega-3 Index inversely linked with all-cause mortality, with an effect-size magnitude comparable to smoking status. Observational.
• REDUCE-IT (Bhatt 2019, NEJM): 4 g/day icosapent ethyl (pure EPA) in statin-treated patients with high triglycerides cut major CV events by 25 percent.
• VITAL (Manson 2019, NEJM): 1 g/day EPA+DHA in primary prevention was null overall, with signals in fish-poor subgroups.
• STRENGTH (Nicholls 2020, JAMA): 4 g/day EPA+DHA carboxylic acid was null. Some read REDUCE-IT's positive result as partly mineral-oil-placebo-driven.

EU authorised claims (Regulation 432/2012):

• "EPA and DHA contribute to the normal function of the heart" at a daily intake of 250 mg combined.
• "DHA contributes to maintenance of normal brain function" and "normal vision" at a daily intake of 250 mg DHA.
• At higher intakes (2 to 3 g/day): blood pressure and triglyceride claims.

Safety: up to 5 g/day combined is safe (EFSA 2012). Modest atrial-fibrillation signal at gram doses across REDUCE-IT, STRENGTH, OMEMI.

Polyphenols: Drink the Tea, Eat the Berries

Resveratrol has plenty of hype and modest human evidence. The original SIRT1-activation story (Lagouge 2006, Cell, in mice) took a hit when a 2010 J Biol Chem paper (Pacholec et al.) showed much of the in-vitro SIRT1 activation was a fluorophore artifact. Timmers 2011, Cell Metabolism (150 mg/day in obese men): small gains in mitochondrial respiration and liver fat; later RCTs are mixed-to-null. Bioavailability is poor (<1 percent free) and the compound inhibits CYP3A4, with interaction risk against statins, anticoagulants, and immunosuppressants. EU Novel Food at ≤150 mg/day; no authorised claim.

EGCG (green tea) activates Nrf2 and AMPK in preclinical models. Human cardiometabolic effects are small. EFSA's 2018 risk assessment flags hepatotoxicity risk for green tea extracts above 800 mg EGCG/day, especially on an empty stomach. Drink the tea (no DILI from brewed tea); be careful with concentrated extracts.

Fisetin is a senolytic (Yousefzadeh 2018, EBioMedicine: extended median and maximum lifespan in mice). Mayo Clinic Phase 2 trials at 20 mg/kg/day for 2 days monthly are ongoing. No positive clinical primary endpoints have been published as of May 2026. Don't pay for "senolytic" branding ahead of the RCT readouts.

Sulforaphane (broccoli sprouts) is the strongest dietary Nrf2 inducer known. Human RCTs in autism spectrum disorder (Singh K. 2014, PNAS) and biochemically recurrent prostate cancer (Alumkal 2015) are encouraging but not mitochondria-specific. A daily handful of broccoli sprouts (with a myrosinase source like mustard powder) delivers ~30 to 60 mg sulforaphane equivalent.

Mediterranean Pattern: A-Grade Whole Diet

PREDIMED (Estruch 2018, NEJM corrected reanalysis) randomized ~7,400 high-CV-risk adults to a Mediterranean diet with extra-virgin olive oil or mixed nuts versus a low-fat control. Major cardiovascular events fell ~30 percent over 5 years. The polyphenol-dense Mediterranean pattern beats any single-compound supplement in long-horizon outcome data. The mechanism mix (fiber, monounsaturates, marine omega-3, polyphenols, lower glycemic load, social co-effects) cannot be cleanly pinned on one compound. Which is the whole point.

This is an editorial guide, not a supplement label. The distinction matters legally. Regulation (EC) No 1924/2006 governs health claims on commercial communications (product packaging, advertising, sponsored content). Editorial journalism is excluded by Article 1(2) and Recital 4, provided the writing stays scientific rather than promotional.

The safe editorial pattern: describe what studies measured. "An RCT (Mortensen 2014) reported that 300 mg/day CoQ10 reduced major adverse cardiac events by half in moderate-to-severe heart failure." That is reporting on a trial. It is not a claim that CoQ10 reduces your heart-failure risk.

The unsafe pattern: "CoQ10 supports cardiovascular function" or "take 300 mg of CoQ10 to support your heart." That reads like a label.

Authorised Article 13(1) claims (Regulation 432/2012) for nutrients discussed in this guide:

• Magnesium: "contributes to normal energy-yielding metabolism," "to a reduction of tiredness and fatigue," "to electrolyte balance," "to normal functioning of the nervous system," "to normal muscle function," "to normal protein synthesis," "to normal psychological function."
• Zinc: claims on metabolism, cognitive function, immune function, cell division, protection from oxidative stress, and more.
• Riboflavin (B2): nine claims including normal energy-yielding metabolism, normal functioning of the nervous system, reduction of tiredness and fatigue, protection of cells from oxidative stress.
• Niacin (B3), Pantothenic acid (B5), Vitamin B6, Vitamin B12, Folate: each carries energy-metabolism and fatigue-reduction claims; B6 was downgraded to UL 12 mg/day in 2023.
• Vitamin C: energy metabolism, immune function, collagen, iron absorption, oxidative-stress protection.
• Vitamin D: bone, muscle function, immune system, calcium absorption.
• EPA + DHA: "contribute to the normal function of the heart" at 250 mg/day combined. DHA alone has claims for brain function and vision at 250 mg/day.
• Creatine: "increases physical performance in successive bursts of short-term, high-intensity exercise" at ≥3 g/day; "can enhance the effect of resistance training on muscle strength in adults over 55" at ≥3 g/day plus resistance training.
• Betaine: "contributes to normal homocysteine metabolism" at ≥500 mg per portion, 1.5 g/day intake.

No authorised claim, in the EU, for:

• CoQ10 (EFSA rejected all six categories in 2010).
• NR (Niagen has Novel Food authorisation at ≤300 mg/day; no health claim).
• NMN (EFSA positive safety opinion May 2026; Commission Implementing Regulation has not authorised; sale in DACH is borderline).
• Taurine (EFSA rejected all proposed claims).
• Resveratrol (Novel Food ≤150 mg/day; no claim).
• Curcumin (food additive E100; "joint function" claim rejected by EFSA 2017).
• L-theanine, glycine: no authorised claims.
• Urolithin A (Mitopure): Novel Food authorised; no health claim.
• PQQ (Novel Food ≤20 mg/day; no claim).
• Citicoline (Novel Food ≤500 mg/day; no claim).
• Fisetin: uncertain Novel Food status; no claim.

EU Tolerable Upper Intake Levels (the EFSA figures are EU-binding):

• Magnesium 250 mg/day supplemental (NIH: 350)
• Zinc 25 mg/day total (NIH: 40)
• Vitamin B6 12 mg/day (revised down 2023; NIH: 100)
• Vitamin D 4000 IU/day
• Selenium 255 µg/day (revised down 2023)
• EPA+DHA 5 g/day combined is safe
• Riboflavin: no UL (no toxicity threshold)
• B12: no UL

DACH-specific reality. Germany applies the Bundesgerichtshof's "doppelte Zweckbestimmung" doctrine. A food supplement labeled or advertised with disease-related effects can be re-classified as a functional medicinal product under §2(1) AMG. BfArM is the relevant regulator. For melatonin specifically (relevant to the sleep guide companion), OLG Koblenz 9 U 1947/22 (May 2023) softened earlier strictness on ≤1 mg products; BfR's 17 September 2024 statement still reminds that melatonin is "not a gentle sleep aid." Practical line: ≤1 mg is food-supplement-acceptable in Germany; above 1 mg drifts toward medicinal-product classification.

The pattern we follow on this site: discuss the science transparently, cite the trials, cite the regulators (including their rejection notices), state safety thresholds whenever doses appear, and never write "take X to do Y" in a product-recommending voice.

If a reader walked into this guide asking "what should I take for mitochondria," the honest answer, in order of evidence-to-effort ratio, is this.

Before the tier list, one ground rule. The Tier 2 to 5 doses below are the doses commonly studied or sold over the counter. They are not a personal recommendation. Talk to your Hausarzt before starting a stack, especially if you take medication or have a chronic condition.

Tier 0: Foundation (not supplements). Zone 2 endurance 3 to 4 sessions per week. HIIT 1 to 2 sessions. Resistance training 2 to 3 sessions. Sleep 7 to 9 hours. Protein at least 1.6 g/kg body weight. Time-restricted eating with an earlier window if it fits your life. Don't smoke. Moderate alcohol or none. Limit ultra-processed food. This single bullet point produces measurable gains in VO₂ max, lactate threshold, mitochondrial protein content, and OXPHOS gene expression that no supplement matches. Skip this tier and the supplement tiers below will not rescue it.

Tier 1: Foods with the best evidence.

• A Mediterranean-pattern diet: vegetables, legumes, whole grains, fish, extra-virgin olive oil, nuts, modest red wine if you drink at all. PREDIMED-grade evidence.
• Fatty fish 2 to 3x per week or omega-3 EPA+DHA ≥250 mg/day (EU-authorised heart claim).
• Berries, walnuts, pomegranate for ellagitannin substrate. Half of you won't make urolithin from it; the other half will.
• Broccoli sprouts (with mustard powder for myrosinase) for sulforaphane.
• Green tea (brewed, not concentrated extracts above 800 mg EGCG).

Tier 2: Defensible cofactor coverage. Especially relevant for older adults, vegans, metformin or PPI users, statin users, or anyone whose diet does not consistently cover the basics.

• A multivitamin with riboflavin, B12, folate, and the rest of the B-complex at near-RDI levels. Riboflavin has the strongest EU claim portfolio and the cleanest mitochondrial cofactor story.
• Magnesium bisglycinate: typical consumer dose 200 to 400 mg elemental in the evening. EU authorised claims for muscle, nervous system, and fatigue.
• Vitamin D: dose individualized to keep 25(OH)D in the 30 to 50 ng/mL range. Test, don't guess.
• Methylcobalamin or cyanocobalamin: studied dose 500 to 1000 µg sublingual or oral in deficiency-risk groups (vegan, metformin, PPI, age 60+). MMA and homocysteine are the functional markers for a quantitative read.

Tier 3: Discretionary mitochondria-leaning additions.

• Creatine monohydrate: studied dose 5 g/day for muscle endpoints, with 5 to 10 g/day used in cognitive or sleep-deprivation studies. Two authorised EU claims. ATP buffer, not biogenesis.
• CoQ10: studied dose 100 to 200 mg/day taken with fat. The case is strongest for statin users, adults over 60, and anyone with cardiovascular risk. No authorised EU claim. This is evidence-based personal choice territory, not a prescribed mitochondrial booster.
• Omega-3 EPA+DHA: typical consumer dose 1 to 2 g/day in people who do not eat fatty fish twice a week.

Tier 4: Higher-cost, higher-uncertainty experiments.

• NR: typical consumer dose 300 mg/day for testing the NAD⁺ story (EU Novel Food authorised; only Niagen / ChromaDex). Usually paired with TMG 500 to 1000 mg to cover the methyl-pool tax.
• Urolithin A (Mitopure): studied dose 500 mg/day, relevant population is over-60 adults with sarcopenia concerns and the budget. Three RCTs missed their primary endpoints but produced reasonable secondary signals.
• Curcumin in a bioenhanced form (Meriva, Theracurmin, Longvida, CAVACURMIN): typical consumer dose 500 to 1000 mg/day for joint or anti-inflammatory use cases. Not a primary mitochondrial agent. Watch for hepatotoxicity.
• Taurine: studied dose 1 to 3 g/day, cheap and safe at that range. The MELAS evidence is real; the longevity case took a hit in 2025. Plausible adjunct, not a headline ingredient.

Tier 5: Where the evidence is preclinical or contested.

• NMN. EFSA opinion suggests up to 300 mg/day is safe; not yet on the EU Novel Food list; sale in DACH is regulatorily borderline. Comparable mechanistic case to NR but weaker EU legal footing.
• PQQ: typical consumer dose 10 to 20 mg/day. Marketed as the biogenesis nutrient; human evidence is one n=10 biomarker study.
• Fisetin. Skip until Mayo Clinic Phase 2 results publish.
• Resveratrol. The SIRT1 story has largely unwound. Skip unless you're keen on polyphenol mechanism research.

If you only do one thing from this guide: do Tier 0. If you only buy one thing: a quality multivitamin covering riboflavin, B12, folate, and magnesium bisglycinate. Everything else is fine-tuning. And fine-tuning never beats foundations.

A Note on Marketing Versus Mechanism

The mitochondria-supplement space is among the most aggressively marketed corners of consumer health. The mechanism stories are real. CoQ10 really is the obligate electron carrier, riboflavin really is the FMN/FAD source for Complex I and II, taurine really does modify mt-tRNAs, NAD⁺ really declines with age. But the mechanism stories are not the same as longevity outcome data. Every time a label or an influencer compresses "X is required for the ETC" into "X extends your healthspan," they're skipping the part where a randomized controlled trial in humans showed up.

The best you can buy with money is high-quality food, time to exercise, time to sleep, and competent medical care. The cofactor supplements above are second-order optimizations. PQQ is not going to add a decade to your life. Lifting weights twice a week and getting eight hours of sleep might.