Cross-Cutting Hypotheses
The failure cartography reveals several integrative hypotheses that connect multiple steps in the chain. Four are developed here. Note that these hypotheses are not strictly additive: the riboflavin master-rate-limiter, PDK pathological/protective, MnSOD-primary, and cardiolipin-convergence models each predict a largely overlapping metabolomic signature (reduced TCA intermediates, elevated lactate, altered acylcarnitines). A design that discriminates between them — rather than retaining all four as parallel candidates — would use a combination of EGRAC (riboflavin functional status), baseline MnSOD/GPX4, plasma oxidized cardiolipin, and PBMC PDK transcript levels measured in the same individuals, so that competing predictions can be tested against shared data rather than in isolation.
Certainty: 0.45. PDK overactivation (Section Step 3: Pyruvate Dehydrogenase Complex (PDC)) is routinely framed as a pathological lesion causing energy deficit. An alternative interpretation is that PDK upregulation is a protective cellular response to a downstream failure in ROS handling. If MnSOD and PRDX3 are impaired, increasing pyruvate flux through a leaky ETC would generate additional superoxide — worsening oxidative damage to iron-sulfur clusters, cardiolipin, and ETC proteins. PDK-mediated restriction of TCA substrate is then the cell’s “dimmer switch,” trading reduced ATP production for reduced oxidative injury.
This reinterpretation has a specific clinical consequence: PDK inhibition with dichloroacetate (DCA) without simultaneous antioxidant support could worsen ME/CFS by flooding an already-compromised ETC. The small DCA studies (Section Step 3: Pyruvate Dehydrogenase Complex (PDC)) showing heterogeneous response are consistent with this: responders may be those with intact antioxidant systems; non-responders or worseners may have low MnSOD/GPX4.
Testable prediction: In a stratified DCA trial, patients with low baseline MnSOD and GPX4 activity will worsen (higher plasma F2-isoprostanes, worse PEM severity); patients with normal antioxidant status will improve (normalized lactate/pyruvate ratio, reduced PEM). DCA combined with NAC + CoQ10 + riboflavin will outperform DCA monotherapy.
Limitation: This remains entirely speculative pending direct evidence. The PDK-as-protective framing has not been tested in ME/CFS or any similar condition. The mechanism assumes a threshold beyond which additional TCA flux becomes net-harmful — a threshold that may vary widely between patients.
Certainty: 0.40. The integrated stress response (ISR) — a conserved cellular stress program governed by four eIF2α kinases (PERK, GCN2, PKR, HRI) — is activated when mitochondrial proteostasis fails (Costa-Mattioli and Walter 2020). Exertion-induced cardiolipin peroxidation, iron-sulfur cluster damage, or ETC protein misfolding triggers OMA1 cleavage of DELE1, which activates the HRI kinase, which phosphorylates eIF2α, which initiates a 24–72 hour cellular reprogramming via ATF4 (Costa-Mattioli and Walter 2020) . This reprogramming includes transient suppression of anabolic protein synthesis, amino acid redistribution, and the mitochondrial unfolded protein response (mtUPR). The characteristic 24–48 hour delay between exertion and PEM peak matches ISR kinetics precisely.
The only direct ISR-activation evidence in ME/CFS comes from the PERK arm: Wang et al. found elevated PERK markers in ME/CFS skeletal muscle, establishing that the ER-stress branch of the ISR is active in this tissue (Wang et al. 2023). PERK is one of the four eIF2α kinases; its elevation confirms ISR engagement via the ER-stress route, consistent with WASF3-mediated ER dysfunction (ER Stress–WASF3–Mitochondrial Dysfunction Pathway: Druggable Mechanism). Whether the HRI arm (mitochondrial stress), GCN2 arm (amino acid starvation), or PKR arm (viral dsRNA sensing) are also activated in ME/CFS remains unmeasured.
If PEM is mechanistically an ISR response to exertion-induced mitochondrial damage, then GDF15 and FGF21 (canonical plasma ISR biomarkers) should peak 24–48 hours post-exertion and track PEM severity (Costa-Mattioli and Walter 2020). ATF4-target gene expression in PBMCs should be elevated during PEM. This framing also predicts that protecting mitochondria from exertion-induced damage (antioxidants, cardiolipin stabilizers, pre-activity creatine loading) would attenuate PEM by reducing the ISR trigger.
Testable prediction: Plasma GDF15 and FGF21 will peak 24–48 h after a standardized exertion challenge in ME/CFS patients, correlating with symptom severity, and will not peak at comparable levels in healthy controls or in ME/CFS patients without PEM. The ISR inhibitor ISRIB (experimental) would attenuate PEM in an animal model if the mechanism is confirmed.
Limitation: Direct ISR measurement in ME/CFS is limited to PERK elevation in muscle (n=14, single study) (Wang et al. 2023). Peripheral blood stress biomarkers showed no difference between long COVID and controls in one controlled study (n=48) , suggesting ISR signatures may be tissue-confined rather than circulating. GDF15 has other causes including cardiac and inflammatory conditions. The connection between ISR kinetics and PEM timing remains mechanistic reasoning without direct empirical support.
Certainty: 0.42. FAD (flavin adenine dinucleotide, the active form of riboflavin) is an obligate prosthetic group or cofactor at eight or more distinct energy-chain steps: PDC E3 (dihydrolipoamide dehydrogenase), KGDHC E3 (same enzyme), all four acyl-CoA dehydrogenases in beta-oxidation (VLCAD, LCAD, MCAD, SCAD), the electron transfer flavoprotein (ETF) and ETF-QO that bridge beta-oxidation to the ubiquinone pool, and succinate dehydrogenase (Complex II) as FADH2 donor. A single cofactor thus gates simultaneously the carbohydrate-to-acetyl-CoA arm, the fatty-acid-to-acetyl-CoA arm, and one direct ETC entry point.
Marginal riboflavin status — common in Western diets and worsened by GI malabsorption, mast cell activation, or bile acid dysbiosis — would produce a metabolomic signature nearly indistinguishable from the combination of PDK overactivation plus beta-oxidation failure plus Complex II dysfunction: exactly the pattern reported across ME/CFS metabolomics studies . Riboflavin-responsive multiple acyl-CoA dehydrogenase deficiency (MADD) produces exercise intolerance, myopathy, and elevated acylcarnitines indistinguishable from severe ME/CFS.
Testable prediction: Erythrocyte glutathione reductase activation coefficient (EGRAC, the gold-standard functional riboflavin biomarker) will be ≥1.3 in ≥30% of ME/CFS patients vs <10% of matched controls. Patients with elevated EGRAC will show plasma C4–C18 acylcarnitine elevation (the MADD/ETF-QO signature). Riboflavin 400 mg/day for 12 weeks will normalise EGRAC and reduce lactate/pyruvate ratio in EGRAC-high responders.
Limitation: No ME/CFS study has measured EGRAC. Standard serum riboflavin may be misleading: Hypothesized impairment of riboflavin transporter (SLC52A2) or mitochondrial FAD loading means tissue-level FAD deficit could exist with normal serum riboflavin.
Certainty: 0.30 (reduced from an initial 0.40 to reflect the negative MMPOWER-3 Phase 3 result in primary mitochondrial myopathy, which provides a discouraging prior for cardiolipin-targeted intervention in adjacent mitochondrial indications). Cardiolipin (CL), the mitochondria-specific phospholipid, is uniquely vulnerable to peroxidation: it contains four polyunsaturated fatty acid chains (vs two for other phospholipids) and sits directly adjacent to the principal superoxide generators at Complexes I and III. Oxidized cardiolipin (CLox) triggers: release of cytochrome c from Complex III (disrupting electron transfer), inflammasome activation, and disassembly of ETC supercomplexes. With MnSOD decreased and PRDX3 uninvestigated, CLox accumulation is a mechanistically predicted consequence of sustained elevated ROS in ME/CFS.
Elamipretide (SS-31/bendavia) is a mitochondria-targeted tetrapeptide that selectively binds cardiolipin, stabilizes its interaction with cytochrome c, and prevents peroxidation. It has completed Phase 3 trials in Barth syndrome (CL synthesis defect) and primary mitochondrial myopathy (MMPOWER-3); the MMPOWER-3 trial did not meet its primary endpoint (6-minute walk distance) at 24 weeks. This negative result in the closest analogous indication substantially tempers expectations for ME/CFS: even if cardiolipin peroxidation is confirmed in ME/CFS, elamipretide’s clinical efficacy profile in a related mitochondrial disease is a discouraging prior, not a permissive one. Molnar et al. have specifically proposed it as a candidate for Long COVID mitochondrial dysfunction ; pursuing this rationale in ME/CFS requires first confirming CL oxidation in a ME/CFS cohort, and any eventual trial would need to pre-specify why it should succeed where MMPOWER-3 did not.
Testable prediction: Plasma oxidized cardiolipin (by LC-MS) will be >2× control in moderate-to-severe ME/CFS and will correlate with 4-HNE, F2-isoprostane, and MnSOD levels. A 12-week elamipretide open-label trial will reduce F2-isoprostanes and CLox, improve 6-minute walk distance, and reduce PEM frequency in responders with elevated baseline CLox. Non-response with normal baseline CLox would refute cardiolipin peroxidation as a driver.
Limitation: Plasma CLox may not reflect intramitochondrial CLox in muscle or brain. Elamipretide is injectable, limiting accessibility for severely ill patients.
Certainty: 0.45. Irisin, a PGC-1\(\alpha\)-dependent myokine released by exercised muscle (Boström et al. 2012), promotes mitochondrial biogenesis, oxidative phosphorylation, and glucose uptake, and signals through αV integrins (including αvβ5) in a manner requiring extracellular HSP90α (Kim et al. 2018) A et al. (2023). A 2026 cross-sectional study of 92 ME/CFS patients (CCC criteria) and 44 sedentary controls found lower baseline plasma irisin and a blunted irisin response to a standardized 90-minute mechanical stress challenge in ME/CFS (Δirisin reduced, p = 0.034) (Souma et al. 2026). In parallel, functional cellular-dielectric-spectroscopy assays showed that thrombospondin-1 (TSP-1) inhibits irisin signaling in a concentration-dependent manner, and that irisin signaling is markedly reduced by both αvβ5 blockade and HSP90α inhibition, while TSP-1 retains its inhibitory activity even when these components are disrupted (Souma et al. 2026). Elevated TSP-1 also suppresses nitric-oxide signaling and angiogenesis via CD47, a pathway with documented vascular relevance (Rogers et al. 2014) (Roberts, Kaur, and Isenberg 2017). Together these findings support a model in which elevated TSP-1 acts as a dominant antagonist of irisin-mediated metabolic signaling, contributing to the impaired adaptive response to exertion that characterises PEM. (Evidence source: plasma + Jurkat-cell in vitro — inference target: skeletal-muscle and systemic metabolic signaling. Link is indirect and requires validation in primary human muscle/PBMC systems. Severity coverage: mild and moderate-to-severe per MFI-20 stratification in the primary cohort; very-severe not separately stratified. Translation gap: in vitro → human. Not yet validated in independent ME/CFS cohorts.)
Testable prediction: In ME/CFS patients, plasma TSP-1 (or the TSP-1:irisin ratio) will rise in proportion to the blunted irisin response and to PEM severity across repeated exertional challenges; PBMC or muscle-derived cells from ME/CFS patients with elevated TSP-1 will show attenuated irisin-stimulated signaling relative to cells from patients with low TSP-1. Falsified if irisin-stimulated signaling in patient-derived cells is normal despite elevated TSP-1, or if TSP-1 reduction restores no metabolic phenotype.
Limitation: The ME/CFS evidence is a single cross-sectional cohort from one laboratory (Moreau group, CHU Sainte-Justine) with no independent replication; the functional mechanism rests on Jurkat-cell assays, not primary muscle or endothelial cells — a cell-type gap (immune cell vs the myocyte that matters for PEM); TSP-1 is also a generic activation marker, so its elevation may reflect non-specific endothelial/platelet activation rather than a specific irisin-antagonist role; plasma irisin ELISAs have known cross-reactivity and kit-variability caveats (Elsen, Raschke, and Eckel 2014); the irisin finding itself is not reproducible in adjacent conditions (unchanged in fibromyalgia (Ercan et al. 2026); poor diagnostic accuracy in sarcopenia (Lapauw et al. 2026)).
Consequence: If confirmed, this framework would reframe PEM not as simple muscle exhaustion but as a failure of a specific exercise-induced signaling pathway, pointing to a biomarker (TSP-1/irisin) and to testable targets (TSP-1 reduction, NO restoration, HSP90\(\alpha\) stabilization) — all currently hypothesis-stage with no human dosing or safety data.
In the Souma 2026 cohort, baseline irisin was an independent predictor of fatigue severity (β = 0.67, p = 0.021), and patients with moderate-to-severe fatigue showed elevated levels of both irisin and TSP-1 (Souma et al. 2026). This is paradoxical: lower group-level irisin in ME/CFS yet higher irisin in the most severely affected. The authors interpret this as a compensatory but ineffective response overridden by a concurrent rise in the antagonist TSP-1. An equally-evidenced alternative is that the severity-stratified rise reflects a biologically distinct regulatory programme (e.g. different PGC-1\(\alpha\) or FNDC5 regulation, or a sex-specific effect — the irisin–PEM correlation held in male patients, R = 0.53, p = 0.01, but not females) rather than pure compensation (Souma et al. 2026). (Evidence source: plasma ELISA; inference target: systemic metabolic signalling. Severity coverage: mild and moderate-to-severe; very-severe unknown. Cohort overlap: Moreau-group biobank — not independent of prior corpus cohort evidence. Certainty: 0.30.)
Falsifiable prediction: A longitudinal study measuring irisin and TSP-1 across PEM episodes will discriminate the two models: the compensation model predicts irisin and TSP-1 rise together before symptom recovery, whereas the distinct-biology model predicts a stable severity-associated set-point that does not track acute PEM state. Refuted if irisin and TSP-1 do not track the predicted temporal relationship.
Consequence: Resolving whether high irisin in severe ME/CFS is compensation or a distinct state determines whether TSP-1-suppressing or irisin-restoring therapies could help or could merely chase a compensatory signal — the answer is currently unknown and requires longitudinal biomarker studies.
Certainty: 0.40. Thrombospondin-1 (TSP-1) is already documented in this paper’s corpus as an endothelial-activation marker elevated in ME/CFS: it appears in the Heng 2025 biomarker panel as the endothelial-activation arm (Heng et al. 2025), and broader serum-proteome profiling documents intensified secretome vasculature interactions in ME/CFS (Hoel et al. 2026). Souma 2026 adds a second, functionally distinct role for the same molecule: TSP-1 as a dominant antagonist of irisin signaling at the HSP90\(\alpha\)/αvβ5 axis, impairing metabolic adaptation to exertion (Souma et al. 2026). The unifying hypothesis is that one upstream molecule may simultaneously drive (a) the documented vascular/endothelial dysfunction and (b) the impaired metabolic adaptation, via CD47-mediated NO suppression (Rogers et al. 2014) (Roberts, Kaur, and Isenberg 2017) and via irisin-axis antagonism respectively. If true, two currently-separate findings (vascular and metabolic) become two faces of one matricellular-signaling lesion — and explain why TSP-1 elevation spans such different symptom domains. (Evidence source: plasma + Jurkat-cell in vitro; inference target: systemic vascular and metabolic signaling. Link indirect, requires primary human muscle/endothelial validation. Severity coverage: mild and moderate-to-severe; very-severe unknown. Translation gap: in vitro → human. Origin: brainstorm.)
Testable prediction: In a single cohort, plasma TSP-1 will correlate with BOTH endothelial markers (VWF, fibronectin) AND the blunted exertional irisin response simultaneously; the top TSP-1 quartile will show both the worst perfusion/endothelial measures and the worst metabolic adaptation. Falsified if TSP-1 elevation segregates with vascular but not metabolic dysfunction (the two defects are TSP-1-independent).
Limitation: TSP-1 is a generic platelet/endothelial/immune-activation marker elevated in injury, inflammation, cancer, and cardiovascular disease (Rogers et al. 2014) (Roberts, Kaur, and Isenberg 2017); its elevation in ME/CFS may reflect non-specific activation rather than a specific irisin-antagonist role. The corpus evidence for ME/CFS elevation (Heng 2025, Hoel 2026) is not reproduced by Souma 2026 itself, which found no TSP-1 difference between ME/CFS and controls at baseline (p = 0.35) — there TSP-1 differed only by fatigue severity within patients. The convergence claim rests on two separate single-source observations (corpus vascular markers + Souma 2026), not yet demonstrated within one cohort.
Consequence: If one molecule already flagged in ME/CFS blood work as a blood-vessel-problem marker also blocks the muscle’s adaptive “exercise-good” signal, then two of the disease’s most disabling features — poor circulation and post-exertional crash — may share a single upstream cause, making it a more attractive drug target than either alone; this is currently a unifying hypothesis awaiting a combined vascular+metabolic cohort study.
The three hypotheses above jointly argue that a failure of the irisin signalling axis — a myokine pathway that normally transduces the adaptive, mitochondrial-supporting response to exertion — may contribute to the impaired metabolic adaptation and PEM of ME/CFS. Irisin Signaling Resistance via TSP-1 as a Mechanism of Impaired Metabolic Adaptation in PEM proposes that elevated TSP-1 antagonises irisin at the HSP90\(\alpha\)/αvβ5 axis, converting a physiological “exercise-good” signal into ineffective signalling; TSP-1 as a Vascular–Metabolic Convergence Node ties this to the already-documented elevation of TSP-1 as a vascular marker, so that the same molecule may underlie both poor circulation and the metabolic defect; The Irisin–TSP-1 Severity Paradox: Compensation or Distinct Biology? flags the unresolved severity paradox — high irisin in the most severely affected patients — which distinguishes a compensatory-but-ineffective response from a distinct biology. What the evidence supports: a plausible, testable mechanistic framework grounded in one direct ME/CFS cohort plus well-established irisin and TSP-1 biology. What remains speculative: that TSP-1-mediated irisin resistance is a causal driver of PEM (single cross-sectional cohort, in vitro Jurkat mechanism), that irisin-TSP-1 profiling is a valid biomarker, and any therapeutic implication. The central open question is whether reducing TSP-1 or restoring irisin signalling would actually restore metabolic adaptation in patients, or merely chase a compensatory signal.
Consequence: A single blood-measurable pair — the “exercise signal” and its blocker — may capture why exertion fails to produce recovery in ME/CFS, offering a concrete biomarker (the TSP-1/irisin pair) and a testable target (reducing TSP-1 or restoring irisin signalling), even though both remain hypothesis-stage today.