Step 6: Key Cofactors as Cross-Cutting Failure Modes
1 Normal Function and ATP Accounting
Several cofactors are required by multiple steps in the chain simultaneously. Their depletion therefore impairs energy metabolism at several nodes at once. Direct yield from this step is zero (cofactors are not consumed stoichiometrically in catalysis), but gated yield is up to ~25 ATP (Steps 3, 5, and the ETC combined), because a single cofactor failure can disable multiple enzymatic steps. This makes cofactor depletion uniquely high-leverage for therapy and uniquely hard to diagnose (the metabolomic signature is indistinguishable from direct enzymatic damage).
2 Documented ME/CFS Findings and Failure Modes
Several cofactors are required by multiple steps in the chain simultaneously. Their depletion therefore impairs energy metabolism at several nodes at once.
Coenzyme Q10 (CoQ10/ubiquinone). CoQ10 is the mobile electron carrier between Complexes I/II and Complex III, and also a lipid-soluble antioxidant in the inner mitochondrial membrane. Reduced CoQ10 is among the most replicated findings in ME/CFS: Holden et al.’s systematic review of 19 studies identified it in every study that measured it (Holden et al. 2020). The inverse correlation between CoQ10 levels and fatigue/autonomic symptom severity has been documented , and a randomised controlled trial of CoQ10 combined with NADH showed improved heart rate recovery . CoQ10 deficiency thus simultaneously impairs electron transfer (reducing ATP yield), increases electron leakage at Complexes I and III (increasing ROS), and reduces membrane antioxidant protection.
The Maes 2009 case-control study (n=58 ME/CFS, 22 controls) demonstrated that 44.8% of ME/CFS patients have plasma CoQ10 below the lowest control value (490 micro g/L), with lower levels correlating with fatigue, autonomic, and neurocognitive symptoms (Maes et al. 2009). A 2022 meta-analysis of 13 RCTs (n=1,126) found CoQ10 significantly reduces fatigue (Hedges’ g = -0.398, p = 0.001) with dose-dependent (coefficient = -0.0017 per mg, p < 0.001) and duration-dependent (coefficient = -0.0042 per day, p = 0.007) effects (Tsai et al. 2022). Novel cocrystal ubiquinol formulations show 2.01-3.43x higher bioavailability vs standard ubiquinone (Mei et al. 2026).
Cardiolipin. Cardiolipin is a phospholipid unique to the inner mitochondrial membrane, required for the structural stability of all five ETC complexes, ANT, and the cristae architecture that concentrates the proton gradient. Its oxidation by ROS impairs all IMM-dependent functions simultaneously. Historical data report anti-cardiolipin antibodies in ME/CFS , and Molnar et al. highlight cardiolipin-targeting elamipretide (the SS-31 peptide, currently in trials for Barth syndrome and heart failure) as an uninvestigated therapeutic candidate for Long COVID mitochondrial dysfunction — with equivalent theoretical rationale in ME/CFS.
NAD+/NADH ratio. NAD+ is required by the E3 subunits of both PDC and KGDHC, by isocitrate dehydrogenase (TCA step 3), by malate dehydrogenase (TCA step 8), by Complex I, and by the sirtuin family of metabolic regulators. Two independent mechanisms threaten NAD+ availability in ME/CFS: the kynurenine pathway diverts tryptophan away from NAD+ synthesis, and PARP overactivation (responding to oxidative DNA damage) consumes NAD+ as a substrate. Germain et al. found nicotinate and nicotinamide metabolism — the NAD+ precursor pathway — significantly disrupted at baseline in female ME/CFS plasma (KEGG \(p = 0.0006\), \(q = 0.04\); \(n = 60 + 45\) controls), even before exercise challenge (Germain et al. 2022). Groven et al. independently demonstrated that kynurenine metabolite profiles distinguish CFS from fibromyalgia and controls (\(n = 49/57/54\), all female): the neuroprotective KA/QA ratio was specifically reduced in CFS (\(p = 0.016\)), suggesting condition-specific diversion of kynurenine toward the neurotoxic/NAD+-synthetic branch (Groven et al. 2021). Neither PARP activity (poly-ADP-ribose quantification) nor CD38 expression has been directly measured in ME/CFS — both depletion mechanisms remain inferred from indirect evidence.
The picture is complicated by the only direct PBMC NAD+ measurement in ME/CFS. Heng et al. (\(n = 61 + 61\) controls) found NAD+ elevated in ME/CFS white blood cells (300.3 nM vs 257.7 nM), while the ATP/ADP ratio was simultaneously reduced (Heng et al. 2025). This pattern — substrate accumulation upstream of a processing block — is consistent with Complex I dysfunction: NAD+ is present but cannot be efficiently reduced to NADH because the ETC cannot accept electrons. It does not refute systemic NAD+ insufficiency: in energy-intensive tissues (skeletal muscle, neurons), PARP overactivation and kynurenine diversion could produce genuine depletion even if PBMCs show substrate backup. No muscle biopsy NAD+ data exist for ME/CFS. An earlier, lower-resolution study found serum NAD(P)H reduced in CFS (8.0 vs 10.8 nmol/mL, \(n = 44 + 30\) controls) by fluorescence emission, but the method does not distinguish individual nicotinamide adenine dinucleotides and has not been replicated (Mikirova, Casciari, and Hunninghake 2012).
Downstream of NAD+ availability, Missailidis et al. reported SIRT3 and SIRT6 downregulation in ME/CFS lymphoblast cell lines (transcriptomics; \(n = 23\) vs 17 controls), with SIRT2 upregulated (Missailidis et al. 2021). SIRT3 is the principal mitochondrial deacetylase regulating oxidative metabolism; its downregulation is consistent with either reduced NAD+ flux through the mitochondrial compartment or a compensatory transcriptional response to impaired electron transport. The EBV-immortalization caveat applies.
Therapeutically, the largest NAD+ precursor trial to date — nicotinamide 1000 mg/day (\(n = 900\), double-blind RCT in acute COVID-19) — accelerated clinical recovery and, in a post-hoc 6-month follow-up, reduced the burden of post-COVID symptoms (median symptom score 6.5 vs 10.5, \(p = 0.010\)) (Schreiber et al. 2025). In established Long COVID, nicotinamide riboside (NR) 2000 mg/day (\(n = 58\), double-blind RCT) raised blood NAD+ 2.6–3.1-fold within 5 weeks but did not significantly improve the primary cognitive endpoint; secondary outcomes (fatigue, sleep, depression) improved after \(≥\) 10 weeks on NR . No NR or NMN trial has been conducted specifically in ME/CFS. A mechanistic bridge between NAD+ depletion and the persistent interferon-like immune signature observed in ME/CFS was provided by Chini et al., who demonstrated in vitro that chronic NAD depletion triggers mitochondrial DNA leakage into the cytoplasm via VDAC1, activating the cGAS-STING pathway and inducing a Type I interferon response indistinguishable from viral infection (Chini et al. 2025) — a candidate explanation for why post-viral ME/CFS may sustain an innate immune activation state even after pathogen clearance.
Riboflavin (FAD precursor). FAD is required by Complex I (as FMN), Complex II, the E3 subunits of both PDC and KGDHC, all four acyl-CoA dehydrogenases in beta-oxidation, ETF, ETF-QO, and glutathione reductase. Three independent lines of evidence now converge on riboflavin pathway disruption in ME/CFS. First, Naviaux et al. identified riboflavin pathway dysregulation among the 20 abnormal metabolic pathways in their plasma metabolomics study, with FAD decreased in both male and female CFS patients . Second, Baraniuk found riboflavin and FMN levels lower in ME/CFS cerebrospinal fluid than in sedentary controls (\(n = 45 + 20\)), with Riboflavin Metabolism enriched among significantly different CSF metabolite pathways — extending the deficit from plasma to the central nervous system (Baraniuk 2025). Third, Heap et al. measured erythrocyte glutathione reductase (GTR) basal activity in CFS (\(n = 12 + 18\) controls) and found it significantly reduced (6.13 vs 7.42 U/g Hb, \(p < 0.04\)) (Heap, Peters, and Wessely 1999) — however, they measured basal enzyme activity only, not the activation coefficient ratio (EGRAC), so the result cannot distinguish between reduced enzyme protein mass and functional FAD depletion. The EGRAC assay (ratio of GTR activity with and without exogenous FAD) remains the gold-standard functional riboflavin biomarker and has never been performed in ME/CFS (G23).
Despite this converging evidence, riboflavin as a standalone intervention has never been formally trialled in ME/CFS — an absence notable given three features that would make a pilot trial low-risk and potentially informative. First, riboflavin-responsive forms of multiple acyl-CoA dehydrogenase deficiency (RR-MADD, typically ETFDH mutations) produce a clinically similar syndrome of fatigue, exercise intolerance, and proximal myopathy that responds to riboflavin 100–400 mg/day; a fatigue-dominant MADD phenotype without overt myopathy has been reported in ETFDH c.1130T>C carriers. Second, riboflavin 400 mg/day is an evidence-based migraine prophylactic (59% responder rate vs 15% placebo, NNT = 2.3 (Schoenen, Jacquy, and Lenaerts 1998)), and migraine is among the most common ME/CFS comorbidities — the proposed mechanism (mitochondrial energy reserve enhancement via FAD-dependent Complex I/II) is identical to the energy-deficit hypothesis in ME/CFS. Third, riboflavin is inexpensive, widely available, and has an excellent safety profile even at high doses.
Glutathione reductase (GR) is FAD-dependent — it recycles GSSG back to GSH. Under the chronic oxidative stress documented in ME/CFS (elevated F2-isoprostanes (Kennedy et al. 2005) (Robinson et al. 2010), depleted GSH, elevated 8-OHdG), GR turns over at abnormally high rates to maintain the GSH/GSSG ratio. This constitutes a potential “FAD sink”: riboflavin may not be deficient in dietary supply but may be consumed faster than it can be replenished by the very redox-cycling machinery that defines oxidative stress. If so, EGRAC (G23) will correlate with F2-isoprostane and GSSG/GSH ratio: patients with the highest oxidative stress burden will have the worst functional riboflavin status. The therapeutic implication is that supplementation dose calculations must account for the consumption rate, and that co-administration of antioxidants (NAC, CoQ10) with riboflavin may be more effective than riboflavin alone — because reducing oxidative demand simultaneously reduces FAD consumption. This reframes the riboflavin question from “do ME/CFS patients get enough riboflavin?” to “do ME/CFS patients use up riboflavin faster than they can replace it?”
Thiamine (TPP precursor). Thiamine pyrophosphate is the cofactor for the E1 subunits of both PDC and KGDHC. Functional thiamine deficiency (documented by elevated erythrocyte transketolase activation ratio in historical CFS series) would simultaneously impair pyruvate entry into the TCA cycle and reduce alpha-ketoglutarate oxidation. High-dose thiamine (600–1800 mg/day) is used in thiamine-responsive metabolic disorders and in POTS-adjacent autonomic conditions. It has not been tested in ME/CFS in a formal trial . The closest formal evidence is Bager et al.’s RCT of high-dose oral thiamine (600 mg TID) in IBD patients with chronic fatigue (\(n = 40\)), which showed significant fatigue reduction versus placebo (Bager et al. 2021) — but the IBD fatigue mechanism may differ from ME/CFS. A large patient-reported outcomes survey (\(n = 3{,}925\)) found that benfotiamine and TTFD (fat-soluble thiamine derivatives) showed statistically significant differential response rates between ME/CFS and Long COVID patients, suggesting disease-specific thiamine handling (Eckey et al. 2025).
3 Research Gaps at Step 6
Cofactor gaps are inherently high-leverage because a single cofactor gates multiple steps simultaneously. Open gaps: (G22) direct NAD+ measurement in ME/CFS muscle or neuronal tissue — PBMC data now exist (Heng 2025: elevated NAD+, consistent with substrate backup rather than depletion) but no energy-intensive tissue has been assayed, and direct PARP activity and CD38 expression measurements remain absent; (G23) EGRAC functional riboflavin assay — the gold-standard activation coefficient ratio has never been computed in ME/CFS, though converging evidence from plasma FAD (Naviaux 2016), CSF riboflavin/FMN (Baraniuk 2025), and erythrocyte GTR basal activity (Heap 1999) all point to riboflavin pathway disruption; (G24) high-dose thiamine trial; (G25) lipoic acid status and supplementation; (G26) CoQ10 tissue (not plasma) measurement; (G27) magnesium status — Cox 1991 reported reduced RBC magnesium and a positive IM magnesium RCT, but three independent replications failed to confirm the finding (Clague, Edwards, and Jackson 1992) (Hinds et al. 1994) (Swanink et al. 1995); the positive finding has not been replicated.
| Gap | Essentiality | Worst-case impact | Tractability | Therapeutic leverage | Measurement | Priority |
|---|---|---|---|---|---|---|
| G22: NAD+ muscle/tissue measurement + PARP/CD38 activity | High — NAD+ gates Complex I, PDC, KGDHC, IDH, MDH, sirtuins; PBMC data exist (Heng 2025: elevated) but no energy-intensive tissue | Up to ~83% of gated yield across multiple steps | Moderate — LC-MS NADome profiling validated (Braidy 2021); muscle requires biopsy; PARP/CD38 assays standardized | Immediate — NR/NMN are OTC; NAM 1000 mg/day reduced post-COVID symptoms (Schreiber 2025, \(n = 900\)) | Muscle biopsy (LC-MS NADome); PBMC data exist | 1 |
| G23: EGRAC functional riboflavin assay | High — gates PDC E3, KGDHC E3, β-ox, Complex I, Complex II; converging evidence from plasma FAD (Naviaux 2016), CSF riboflavin/FMN (Baraniuk 2025), erythrocyte GTR (Heap 1999) | Up to ~83% | Easy — EGRAC is a simple erythrocyte assay; standardized protocol published (Powers 2023) | Immediate — riboflavin 400 mg/day; evidence-based for migraine (NNT = 2.3); RR-MADD responds to 100–400 mg/day | Blood draw (erythrocyte EGRAC) | 1 |
| G24: High-dose thiamine trial | High — gates PDC E1 and KGDHC E1; proximate evidence from IBD fatigue RCT (Bager 2021: significant benefit, \(n = 40\), but different population) (Bager et al. 2021); large patient survey shows differential benfotiamine/TTFD response between ME/CFS and Long COVID (Eckey et al. 2025) | Up to ~50% | Easy — erythrocyte transketolase activation is standardized | Immediate — thiamine is OTC and cheap | Blood draw (erythrocyte transketolase) | 1 |
| G25: Lipoic acid status and supplementation | Medium — gates PDC E2 and KGDHC E2 but cellular synthesis is normally sufficient; no ME/CFS baseline measurement exists; an unblinded, non-randomised Long COVID trial (CoQ10+ALA, \(n = 174\), Barletta 2022 (Barletta et al. 2022)) reported fatigue benefit but did not measure lipoic acid levels | Up to ~50% if fully depleted; rare | Hard — no routine assay; requires LC-MS | Immediate — ALA is OTC | Blood draw (specialist LC-MS) | 2 |
| G26: CoQ10 tissue (muscle biopsy) measurement | High — sole mobile electron carrier between Complexes I/II and III | ~50% of gated yield | Hard — muscle biopsy | Immediate — CoQ10 supplementation is widely used | Muscle biopsy | 2 |
| G27: Magnesium status (serum + RBC) | High — required by every kinase and ATPase; but Cox 1991 positive finding not replicated (3 negative: Clague 1992, Hinds 1994, Swanink 1995) | Indirect; gates ATP utilization more than production | Easy — RBC magnesium is available | Immediate — magnesium supplementation is OTC | Blood draw (RBC magnesium) | 2 |
Step 6 has the highest density of priority-1 gaps in the chapter (G22, G23, G24). This reflects the cross-cutting nature of cofactors: a single measurement resolves multiple downstream steps simultaneously. G27 (magnesium) was downgraded to priority 2 because the sole positive finding (Cox 1991) was not replicated by three independent groups (Clague, Edwards, and Jackson 1992) (Hinds et al. 1994) (Swanink et al. 1995). G23 (EGRAC) is especially high-value because the riboflavin master-rate-limiter hypothesis (Riboflavin Deficiency as a Multi-Step Master Rate-Limiter) predicts that ≥30% of ME/CFS patients will have EGRAC ≥1.3, and the assay is both cheap and immediately actionable.
Step 6 is not a distinct pipeline step in the stoichiometric sense — cofactors are consumed catalytically, not stoichiometrically, and their supply is typically upstream (diet, absorption, tissue-level handling). But the functional essentiality of Step 6 is uniquely high because a single cofactor failure cascades across multiple downstream steps. Cofactor gaps are therefore disproportionately leveraged: if any one of G22–G27 turns out positive, multiple apparent failures at Steps 3, 5, and 7 can be explained by a single upstream lesion.
Certainty: 0.30. The Heng 2025 finding — NAD+ elevated in PBMCs while ATP/ADP is reduced (Heng et al. 2025) — is a single-study observation (\(n = 61\)) requiring replication that, if confirmed, would challenge the simple depletion narrative. It suggests the conventional framing (NAD+ is depleted, supplement it) may be wrong in PBMCs, while muscle and neurons may behave differently.
- Compartment A (PBMCs): One interpretation: NAD+ accumulates because Complex I cannot accept electrons from NADH — the ETC is blocked or dysfunctional, stranding NAD+ upstream. Alternatives include upregulated salvage pathway activity or reduced NAD+ consumption in metabolically quiescent PBMCs.
- Compartment B (skeletal muscle / neurons): NAD+ is predicted to be depleted by three concurrent consumers: (i) PARP overactivation from exercise-induced DNA damage and oxidative stress, (ii) CD38 upregulation from chronic immune activation, (iii) kynurenine pathway diversion reducing de novo NAD+ synthesis . This compartment has never been measured in ME/CFS.
If both compartments are dysfunctional but in opposite directions, any blood-based NAD+ measurement (predominantly PBMCs) will systematically mislead. Supplementing NAD+ precursors would raise an already-elevated PBMC pool (useless or harmful) while potentially helping muscle/neurons (beneficial) — explaining the inconsistent clinical trial results for NR/NMN.
Testable predictions: (1) Muscle biopsy NAD+ will be reduced (opposite to PBMC NAD+) in ME/CFS vs controls. (2) The PBMC NAD+ / muscle NAD+ ratio will be \(> 1\) in ME/CFS and \(< 1\) in controls. (3) NR/NMN supplementation will improve muscle-specific outcomes (handgrip, exercise tolerance) more than cognitive outcomes if muscle is the depleted compartment. (4) PBMC NAD+ elevation will correlate with Complex I dysfunction markers (NADH/NAD+ ratio, Complex I activity assay) — confirming the backup rather than surplus interpretation.
Falsifiability: This model would be falsified if muscle NAD+ is also elevated in ME/CFS (same direction as PBMCs, eliminating the compartment dissociation); or if PBMC NAD+ elevation is absent in a larger replication cohort (Heng 2025 is a single study); or if PBMC Complex I activity is normal (removing the substrate-backup explanation). Note that the model cannot be tested by blood-based NAD+ measurements alone — it specifically predicts that such measurements are uninformative, which risks making the model unfalsifiable by the most accessible assay. The critical test requires paired PBMC + muscle tissue measurement.
Limitation: The tissue-compartment dissociation is a prediction extrapolated from a single PBMC dataset (\(n = 61\)); PBMC and muscle NAD+ have never been measured in the same ME/CFS cohort. The substrate-backup interpretation (PBMC NAD+ elevated because Complex I cannot accept electrons) is one of several possible explanations — alternatives include upregulated salvage pathway activity, reduced NAD+ consumption in quiescent PBMCs, or assay artefact. This speculation should not be used to guide NR/NMN prescribing decisions until the tissue-compartment prediction is directly tested.
Certainty: 0.25. CD38 is upregulated on senescent cells and is the dominant NAD+-consuming enzyme in aged tissues. A critical confound: CD38 expression increases with chronological age independently of disease state; if the cited studies did not rigorously age-match, the observations below could partly reflect normal ageing rather than ME/CFS-specific pathology. With that caveat, a self-amplifying loop may connect four previously separate observations in ME/CFS:
- Viral infection triggers endothelial senescence ((Nunes and Kruger 2026), hypothesis).
- Senescent endothelial cells upregulate surface CD38.
- CD38 consumes NAD+ in the local microenvironment (endothelium, perivascular tissue).
- NAD+ depletion triggers mtDNA leakage via VDAC1, activating cGAS-STING (Chini et al. 2025).
- The type I interferon response from cGAS-STING further promotes cellular senescence (IFN-\(\beta\) is a known senescence inducer).
- More senescence → more CD38 → more NAD+ depletion → more mtDNA leakage → more interferon → more senescence.
In principle, breaking the loop at any node could interrupt the cycle. Candidate intervention classes include senolytics, CD38 inhibitors (apigenin, luteolin), NAD+ precursors (NR, NMN), and cGAS-STING inhibitors — but none has been tested in ME/CFS in this context, and senolytic agents such as dasatinib carry significant toxicity (myelosuppression, pleural effusion, hepatotoxicity) that precludes use outside formal clinical trials.
Testable predictions: (1) ME/CFS patient endothelial cells will co-express senescence markers (p16, SA-\(\beta\)-gal) and CD38 at higher rates than age-matched controls. (2) In vitro CD38 inhibition (78c) will reduce cGAS-STING activation in ME/CFS patient-derived cells. (3) Senolytics will reduce CD38 expression and partially restore tissue NAD+ levels.
Critical caveat: No link in this six-step chain has been demonstrated in ME/CFS tissue. Each arrow is inferred from a different disease context: CD38-senescence from ageing biology, NAD+-VDAC1-cGAS-STING from in vitro cancer models (Chini et al. 2025), IFN-\(\beta\)-senescence from viral immunology, and endothelial senescence from a Long COVID hypothesis (Nunes 2026). The chain’s validity requires all links to hold simultaneously in ME/CFS. An alternative explanation — that NAD+ depletion, senescence, and immune activation are independent consequences of a common upstream cause (e.g., chronic viral reactivation) rather than a self-amplifying loop — would predict the same individual observations without requiring the cyclic mechanism.
Certainty: 0.35. Most cofactors are not consumed stoichiometrically — they cycle between oxidized and reduced forms. The rate of recycling depends on other cofactors, creating an interdependent recycling network:
- FAD recycling in glutathione reductase requires NADPH.
- Lipoic acid recycling on PDC-E2 and KGDHC-E2 requires NAD+ (via E3/dihydrolipoamide dehydrogenase, which also requires FAD).
- CoQ10 reduction (ubiquinone → ubiquinol) occurs at Complexes I and II; the reduced form is then oxidized at Complex III via the Q-cycle. CoQ10 pool depletion impairs electron shuttling between these complexes.
A deficit in one cofactor can impair recycling of others, creating cascading failure. Specifically: FAD depletion → impaired E3 function → impaired lipoic acid and NAD+ recycling simultaneously. FAD depletion also impairs Complex I (FMN cofactor shares the flavin pathway), Complex II, and glutathione reductase (the master antioxidant recycler). This network analysis mechanistically strengthens the riboflavin master-rate-limiter speculation (Riboflavin Deficiency as a Multi-Step Master Rate-Limiter) by identifying FAD as the cofactor that gates the most other recycling reactions — it is required by E3 (which regenerates both lipoic acid and NAD+), Complex I, Complex II, and glutathione reductase.
Testable prediction: In a simultaneous cofactor panel (G9), FAD deficiency (EGRAC \(\geq\) 1.3) will predict deficiency in the most other cofactors (highest co-deficiency count). Riboflavin supplementation will partially normalise lipoic acid function and NAD+ status even without directly supplementing those cofactors — because restoring FAD restores the E3 recycling hub that regenerates both.
Competing explanation: A simpler model where each cofactor is depleted independently by oxidative stress (Step 6: Key Cofactors as Cross-Cutting Failure Modes) does not require the recycling network. The recycling model predicts correlated depletion patterns (FAD predicts NAD+ and lipoic acid status); the independent model predicts uncorrelated depletion.
Limitation: The recycling network dependencies are established in general biochemistry but the quantitative contribution of inter-cofactor recycling (vs direct dietary supply and de novo synthesis) to steady-state cofactor pools in human tissue is poorly characterized.