Step 8: Creatine Phosphate Shuttle
1 Normal Function and ATP Accounting
The creatine phosphate (PCr) shuttle provides spatial and temporal buffering between mitochondrial ATP production and cytoplasmic ATP consumption. In brain and muscle, creatine kinase at the mitochondrial outer membrane converts mitochondrial ATP into phosphocreatine, which diffuses to the myofibril or synapse where the cytoplasmic CK isoform instantly regenerates ATP on demand. Direct ATP yield is zero (the shuttle is a kinetic buffer, not a net producer); failure is kinetic (ATP not available fast enough at the demand site) rather than stoichiometric (ATP count unchanged per cycle).
2 Documented ME/CFS Findings and Failure Modes
The PCr shuttle has additional significance in the context of ANT dysfunction (Step 7: ATP Export - The Adenine Nucleotide Translocator (ANT)): if ANT is partially blocked, each ATP molecule that does exit the matrix is immediately captured by mitochondrial CK and shuttled to the cytosol as PCr, preventing product inhibition at the ANT exit site. Creatine supplementation ensures the shuttle substrate is not limiting, effectively functioning as a partial bypass of an ANT bottleneck. This is one possible mechanistic rationale — among several, including simple energy buffering — for why creatine might benefit ME/CFS patients; the ANT-bypass explanation is post-hoc and would require demonstration of ANT dysfunction before it can be preferred over simpler accounts.
In a feasibility study (\(n=11\) completers), Godlewska et al. found that 6 weeks of creatine monohydrate (16 g/day) significantly increased brain creatine in the pregenual anterior cingulate cortex (+8.3%, \(p=0.004\)) and dorsolateral prefrontal cortex (+2.9%, \(p=0.012\)), measured by 7T MRS (Godlewska et al. 2024). Secondary outcomes showed decreased fatigue scores and improved Stroop reaction time and hand-grip strength. This is the first intervention evidence that brain creatine deficiency in ME/CFS — previously documented by 7T MRS cross-sectionally — is pharmacologically correctable. Study: (open-label feasibility, no placebo arm; certainty: 0.45 — requires blinded RCT before clinical recommendation). Note: the 16 g/day dose is 3.2\(\\times\) the standard loading dose and substantially above the 6 g/day used in the positive Long COVID RCT (Dos Santos 2026); at this dose, GI distress (bloating, diarrhea, cramping) is common and may be functionally debilitating, particularly for ME/CFS patients with comorbid IBS.
Godlewska 2024 is an open-label feasibility study without a placebo control. The observed improvements in fatigue and cognition cannot be distinguished from placebo effects at this stage. A blinded RCT is required.
Converging evidence from related conditions: Three lines of evidence from outside ME/CFS constrain expectations for a future creatine RCT:
- CFS GAA study (creatine precursor): Ostojic et al. conducted the only placebo-controlled RCT of a creatine-loading compound in CFS proper — a double-blind crossover trial (\(n = 21\) women, CDC 1994 criteria) of guanidinoacetic acid (GAA) 2.4 g/day for 3 months . GAA successfully increased muscular creatine (+36.3% vs +2.4%, \(p < 0.01\)) and improved muscular strength and aerobic power (\(p < 0.05\)), but had no effect on general fatigue or musculoskeletal pain. This dissociation — muscle creatine loading without fatigue benefit — suggests that the fatigue-relevant compartment in ME/CFS may be the brain rather than skeletal muscle, consistent with Godlewska’s brain-focused MRS findings.
A critical pharmacological distinction: GAA is converted to creatine by guanidinoacetate N-methyltransferase (GAMT), which consumes S-adenosylmethionine (SAM) as the methyl donor. Creatine synthesis is the single largest consumer of labile methyl groups in human metabolism — accounting for approximately 40% of all SAM-derived methylation (the commonly cited ~70% figure does not account for methyl-group recycling) (Brosnan, Silva, and Brosnan 2011). In ME/CFS patients with methylation pathway impairment (MTHFR C677T polymorphism reduces enzyme activity and is hypothesized as a vulnerability factor — see Homocysteine-Mediated ETC Disruption as a Contributing Mechanism; homocysteine may be elevated; B12/folate cycling may be disrupted), GAA supplementation may worsen the methylation bottleneck by diverting SAM toward creatine synthesis at the expense of DNA methylation, neurotransmitter synthesis (catecholamines, melatonin), phospholipid methylation, and other SAM-dependent processes. Direct creatine monohydrate supplementation bypasses GAMT entirely, eliminating the methylation cost. If methylation impairment is present, the Ostojic 2016 null result for fatigue may partly reflect a pharmacological confound (GAA-induced SAM depletion opposing any benefit from creatine loading) rather than a clean negative for the creatine hypothesis. This does not rescue the creatine hypothesis from the null — alternative explanations include genuine absence of benefit from muscle creatine loading regardless of compound choice. Importantly, the Ostojic trial ran for 3 months without reporting methylation-related adverse effects (elevated homocysteine, clinical deterioration), which argues against acute GAA-induced methylation toxicity at the dose used — though subclinical methylation diversion may not produce detectable adverse events within that timeframe. The methylation concern is theoretical, not demonstrated. Nonetheless, a future ME/CFS creatine RCT should use creatine monohydrate rather than GAA to avoid the methylation confound.
Long COVID RCTs: Dos Santos et al. 2026 conducted the largest placebo-controlled creatine trial in post-viral fatigue (\(n = 67\) randomized, 58 completers; single-blind): 6 g/day creatine for 4 weeks significantly reduced fatigue (Piper Fatigue Scale change \(-2.05\), \(p = 0.005\)) and improved handgrip strength (\(+4.40\) kgf, \(p = 0.037\)), while 18 g/day showed no superiority (Santos et al. 2026). Smaller long COVID trials (\(n = 12\)–$ 15$) by the Ostojic group also reported benefit, including increased tissue creatine on MRS and improved clinical features (effect sizes \(d = 0.80\)–$ 1.33$) (Slankamenac et al. 2023). These results are encouraging but carry a conflict-of-interest note: the Ostojic group has received creatine-industry funding.
Fibromyalgia RCT: Alves et al. conducted a 16-week double-blind placebo-controlled RCT of creatine in fibromyalgia (Alves et al. 2013). Creatine dramatically increased muscle phosphocreatine (\(+80.3\)% vs \(-2.7\)%, \(p = 0.04\)) and improved muscle strength across multiple measures, but had no effect on pain, fatigue, aerobic conditioning, cognitive function, sleep quality, or quality of life. This is a cautionary precedent: creatine can successfully load muscle without resolving fatigue in a comorbid condition.
Pattern: Muscle creatine loading alone does not resolve patient-reported general fatigue in either CFS (GAA trial; caveat: GAA is a creatine precursor, not creatine monohydrate, so pharmacological differences cannot be excluded) or fibromyalgia (creatine monohydrate RCT), despite improving muscular strength and aerobic power in both. The Godlewska signal is brain-specific — brain creatine correlated with cognitive improvement, suggesting the fatigue-relevant deficit may be central (CNS energy buffering) rather than peripheral (muscle energy stores). A future ME/CFS creatine RCT should therefore be designed with brain MRS as the primary mechanistic outcome and should measure brain-specific endpoints (cognition, brain fog) alongside whole-body fatigue, explicitly distinguishing central and peripheral fatigue components.
Creatine supplementation routinely elevates serum creatinine by approximately 20–30 μmol/L due to non-enzymatic cyclization of the creatine excess. This does not reflect renal impairment but can trigger unnecessary concern (or further testing) on routine labs if the prescribing clinician is unaware of the supplementation. Document creatine use clearly before any renal function panel.
NOT STUDIED: The creatine transporter (SLC6A8) and the synthesis enzymes (AGAT, GAMT) have not been characterized in ME/CFS. Whether brain creatine deficiency reflects impaired synthesis, impaired transport into cells, or increased consumption (e.g., by activated astrocytes or microglia) is unknown.
3 Research Gaps at Step 8
Four gaps at Step 8 revolve around creatine-phosphate-shuttle components: (G29) creatine transporter SLC6A8 expression; (G30) creatine synthesis enzymes AGAT (l-arginine:glycine amidinotransferase) and GAMT (guanidinoacetate N-methyltransferase); (G31) mitochondrial vs cytosolic creatine kinase isoform balance; (G32) a placebo-controlled creatine RCT to confirm the Godlewska 2024 feasibility signal (Godlewska et al. 2024).
| Gap | Essentiality | Worst-case impact | Tractability | Therapeutic leverage | Measurement | Priority |
|---|---|---|---|---|---|---|
| G29: SLC6A8 (creatine transporter) | Medium — sole route for creatine uptake into muscle and brain, but creatine can be partially synthesized in situ | Kinetic, not stoichiometric; affects burst capacity not steady-state ATP | Moderate — serum creatinine tracking plus MRS of brain creatine | Immediate — creatine supplementation is OTC | Blood draw (serum creatinine) + MRS | 2 |
| G30: AGAT + GAMT (synthesis enzymes) | Low — dietary creatine compensates | Minor in presence of diet | Hard — specialist LC-MS or biopsy | Immediate — creatine supplementation bypasses both | Muscle biopsy (specialist LC-MS) | 3 |
| G31: Mitochondrial vs cytosolic CK isoform balance | High if disrupted — mito-CK is the direct coupler of matrix ATP to cytosolic demand | Kinetic; up to 100% of demand-response capacity | Hard — muscle biopsy with isoform fractionation | Long-term | Muscle biopsy (isoform fractionation) | 2 |
| G32: Placebo-controlled creatine RCT with brain MRS | N/A — confirmatory intervention; GAA loaded CFS muscle without fatigue benefit (caveat: GAA \(\neq\) creatine monohydrate); FM creatine loaded muscle without fatigue benefit (Alves et al. 2013) (caveat: FM \(\neq\) ME/CFS) | Determines whether the brain-specific Godlewska signal is real; Long COVID RCTs encouraging (Santos et al. 2026) | Easy — RCT design is simple; brain MRS adds cost but is essential for mechanism | Immediate — creatine is OTC | Blood draw + brain MRS (RCT) | 1 |
G32 remains priority 1 but the landscape has shifted. A creatine-precursor RCT in CFS proper (Ostojic 2016, GAA, \(n = 21\)) successfully loaded muscle creatine but did not reduce fatigue ; a fibromyalgia creatine RCT similarly improved muscle strength without affecting fatigue or cognition (Alves et al. 2013). Long COVID creatine RCTs show promise (\(n = 58\) completers, fatigue \(p = 0.005\) (Santos et al. 2026)), but the Godlewska signal is brain-specific: brain creatine correlated with cognitive improvement. A future ME/CFS creatine RCT should include brain MRS as a mechanistic outcome. The muscle-loading question is not definitively answered — GAA is pharmacologically distinct from creatine monohydrate, and the fibromyalgia population differs from ME/CFS — but the available evidence shifts the prior toward brain-specific mechanisms. Step 8 is not individually essential to ATP production (creatine shuttles are kinetic buffers) but its failure mode is consistent with the demand-response failure pattern observed in PEM, making it a candidate mechanism for the selective-dysfunction hypothesis (Selective Energy Dysfunction).
4 Brain Creatine and the Selective Energy Dysfunction Hypothesis
The brain-specific creatine signal deserves particular emphasis because the PCr shuttle plays a uniquely critical role in CNS energy metabolism. Unlike skeletal muscle, where glycolytic ATP generation can partially substitute during demand spikes, synaptic transmission and action potential propagation require ATP delivery on a millisecond timescale where the PCr shuttle is the primary buffering system. Mitochondrial CK at the synapse pre-phosphorylates creatine using locally generated ATP; when a burst of synaptic vesicle release or ion-pump activity occurs, cytosolic CK instantly regenerates ATP from the local PCr pool without waiting for mitochondrial oxidative phosphorylation to ramp up. Glycolytic enzymes physically associated with ion pumps and synaptic vesicle machinery provide a secondary ATP source, but glycolysis operates on a seconds timescale and cannot match PCr-mediated buffering for transient demand spikes.
The brain-to-muscle dissociation in the creatine data — brain creatine decreased while muscle creatine loads normally with supplementation — maps directly onto the selective energy dysfunction pattern described in Selective Energy Dysfunction. Cognitive processes (working memory, executive function, sustained attention) place high burst demands on the PCr shuttle in prefrontal and cingulate cortex — precisely the regions where Godlewska found creatine depletion. Autonomous peripheral processes (resting cardiac output, hepatic metabolism, basal skeletal muscle tone) operate at steady-state demand levels that do not depend critically on PCr temporal buffering. If brain creatine is selectively depleted while muscle creatine is normal, the clinical result would be preserved resting peripheral function alongside impaired cognition, impaired demand-responsive coordination, and exercise intolerance driven by central rather than peripheral fatigue — which is the ME/CFS phenotype.
Certainty: 0.30. The brain-specific creatine depletion documented by Godlewska 2025 may represent the molecular substrate of the selective energy dysfunction hypothesis (Selective Energy Dysfunction). If the PCr shuttle — the sole millisecond-timescale ATP delivery system — is depleted in brain but not muscle, the resulting phenotype would show precisely the CNS-dependent, demand-responsive failures characteristic of ME/CFS (cognitive impairment, orthostatic intolerance via autonomic coordination failure, exercise intolerance from central motor drive dysfunction) alongside preserved autonomous peripheral functions (resting cardiac output, hepatic metabolism).
Testable prediction: The brain-creatine-to-muscle-creatine ratio (measurable by comparing brain 1H-MRS with muscle 31P-MRS in the same participants) will be lower in ME/CFS patients than healthy controls. This ratio — if validated — could serve as a non-invasive index of selective energy dysfunction severity. Patients with the lowest brain/muscle creatine ratio should show the greatest cognitive impairment relative to their physical capacity.
Limitation: This speculation rests on a single open-label feasibility study (\(n = 11\)) for brain creatine and a small RCT (\(n = 21\)) for muscle creatine loading. The brain-vs-muscle comparison has never been performed within the same ME/CFS cohort — the apparent dissociation is a cross-study inference from different populations, compounds, and outcome measures. The cause of brain-selective depletion (impaired SLC6A8 transport, astrocyte consumption, or reduced local synthesis) is unknown. A simpler alternative: brain creatine depletion may be a downstream consequence of chronic deconditioning and reduced physical activity rather than a primary pathological mechanism — exercise increases brain creatine in healthy subjects, and the Godlewska study did not include activity-matched deconditioned controls.
Certainty: 0.25. Activated astrocytes undergo metabolic reprogramming toward glycolysis (a Warburg-like shift documented in neuroinflammation) and have markedly increased creatine kinase activity and PCr turnover — they consume creatine at an accelerated rate to fuel immune signaling and glutamate recycling. If neuroinflammation is present in ME/CFS (supported by TSPO-PET elevation in Nakatomi et al. and Mueller et al. (Mueller et al. 2020)), reactive astrocytes may create a creatine sink in neuroinflamed tissue, depleting the neuronal PCr pool.
This provides a mechanistic explanation for why brain creatine is decreased while muscle creatine loads normally : astrocytes in neuroinflamed regions sequester creatine, starving neurons of their primary rapid-response ATP buffer. The resulting neuronal energy deficit forces increased glycolytic reliance during demand spikes, producing the brain lactate elevation documented by Godlewska et al. 2025 .
Testable prediction: In brain regions with higher neuroinflammation (measured by TSPO-PET or elevated myo-inositol on MRS), creatine depletion should be more severe. A combined TSPO-PET + 1H-MRS study would test this directly.
Therapeutic implication (research-stage only): If confirmed, anti-neuroinflammatory interventions might restore brain creatine levels by reducing the astrocyte sink, while creatine supplementation alone might require perpetual high-dose maintenance. These are mechanistic inferences from a 0.25-certainty speculation, not clinical recommendations — no intervention study has tested this framework.
Limitation: The neuroinflammation premise itself is contested in ME/CFS. The TSPO-PET evidence (Nakatomi et al. 2014, \(n = 9\); Mueller et al. 2020) uses small samples with a radiotracer (PK11195) that has known specificity limitations, and replication attempts have produced mixed results. Astrocyte creatine kinase upregulation during neuroinflammation is characterized in rodent models; human in vivo confirmation is lacking. If neuroinflammation is absent or mild in ME/CFS, the entire astrocyte-sink mechanism collapses and brain creatine depletion would require a different explanation (e.g., SLC6A8 transport impairment or reduced local synthesis).
Certainty: 0.35. Godlewska/Syan 2025 reported decreased brain creatine and elevated brain lactate as separate findings , but they are mechanistically linked and may represent a single underlying failure. The PCr shuttle normally absorbs transient ATP demand within milliseconds; without it, glycolysis must ramp up within seconds, producing lactate as a byproduct. The seesaw model: brain [Cr] down → PCr buffering capacity down → glycolytic demand up during demand spikes → lactate up. This is a direct kinetic consequence (PCr depletion forces reliance on the slower glycolytic backup), not merely a correlation.
Testable prediction (falsifiable): In a creatine supplementation RCT, measure both brain creatine (should increase) and brain lactate (should decrease) with MRS pre- and post-supplementation. Plot \(\Delta\)-brain-creatine against \(\Delta\)-brain-lactate: the seesaw model predicts a negative correlation (\(r < -0.5\)). If \(r \approx 0\), the phenomena are independent and the lactate elevation has a separate cause (e.g., mitochondrial dysfunction independent of creatine). This provides an elegant dual-endpoint for the creatine RCT design proposed at G32.
Competing explanation: Brain lactate may be elevated due to impaired mitochondrial oxidation (Complex I dysfunction, PDC inhibition) rather than PCr buffer depletion. Under this model, creatine supplementation would raise brain creatine but not reduce lactate, because the lactate source is upstream of the PCr system.
Limitation: The seesaw model treats brain creatine and brain lactate as primarily linked through the PCr buffer mechanism, but both are influenced by multiple additional factors (glucose delivery, astrocyte-neuron lactate shuttle, monocarboxylate transporter activity).
Certainty: 0.30. Physical PEM is extensively discussed in this chapter, but cognitive PEM (brain fog worsening after mental exertion) is equally debilitating and less mechanistically explained. The PCr shuttle model offers a specific mechanism:
- Mental effort depletes brain PCr in activated regions (prefrontal cortex, cingulate — the regions where Godlewska found creatine depletion ).
- In healthy individuals, mitochondrial ATP production rapidly regenerates the PCr pool via mitochondrial creatine kinase.
- In ME/CFS with impaired mitochondrial function and a depleted baseline creatine pool, recovery is doubly impaired: the starting pool is smaller (longer to deplete) and the regeneration rate is slower (impaired oxidative phosphorylation).
Cognitive PEM is therefore a depletion + repletion failure event: the brain’s creatine buffer is already running near-empty, so moderate cognitive demand causes a transient ATP crisis in activated brain regions. The characteristic delay in cognitive PEM (symptoms worsening hours after mental exertion) reflects the time course of PCr pool recovery failure, analogous to the delayed PCr recovery kinetics documented in Long COVID muscle by Finnigan et al. (\(\tau_\text{PCr}\) 92.5 vs 51.9 s (Finnigan et al. 2024)).
Testable prediction: Brain 31P-MRS immediately after a standardized cognitive task (e.g., 30-min Stroop test) should show greater PCr depletion in ME/CFS vs controls, with slower recovery kinetics. This is technically feasible with 7T 31P-MRS.
Competing explanations: Cognitive PEM could equally reflect neuroinflammatory flare (microglial activation triggered by mental exertion), autonomic dysregulation (cerebral blood flow instability), or astrocyte-neuron metabolic uncoupling — none of which require PCr depletion. The PCr model is one of several candidate mechanisms and does not exclude the others.
Limitation: 31P-MRS of brain during cognitive tasks is technically challenging (low SNR, motion artefacts). The analogy between muscle PCr recovery kinetics and brain PCr recovery kinetics is assumed but not established.