Sage — Aspirin (The Kit)
Metabolic hypothesis: aspirin as an energy-delivery and serotonin-OS intervention, not only a clot slogan. Hypothesis language.
The Kit · Coach · Surfer · Serotonin · Cardiac
*Pathways, uncoupling, serotonin, and the management of fuel and oxygen*
A working hypothesis · not medical advice
Aspirin is usually argued as a clot drug, an inflammation drug, or a cancer-prevention drug. Those frames are not wrong. They are incomplete. The molecule is a short-lived acetylating agent attached to a long-lived phenolic acid. The acetyl group edits the platelet and the vessel wall in minutes. The salicylate ion then sits in plasma for hours and, at sufficient concentration, walks protons across the inner mitochondrial membrane.
The hypothesis is that aspirin's durable physiology is an energy-delivery intervention. It changes who gets oxygen and substrate, how hard mitochondria work to hold their membrane potential, which fuel is admitted through CPT-1, and how much of the tryptophan pool is allowed to become serotonin versus NAD. Serotonin is not a mood footnote in this story. In the periphery it is a stored vasoconstrictor, a platelet amplifier, a permeability signal, and a competitor for the same amino acid that mitochondria need to make NAD. Aspirin touches every one of those jobs.
What follows is a map of the pathways, with the uncoupling chemistry and the serotonin operating system written out in enough detail to see where the dose windows actually land.
1. Two molecules, two clocks
Acetylsalicylic acid is hydrolyzed by plasma, erythrocyte, and liver esterases. The acetyl group is transferred onto nucleophilic residues --- most famously Ser530 of COX-1 --- and parent aspirin disappears from plasma in minutes (half-life roughly 15--20 minutes). Salicylate remains, with a half-life of a few hours that lengthens at high dose because of saturable glycine conjugation. After oral aspirin in the rat, peak plasma salicylate can be up to about fifty times residual aspirin; in humans the ratio is smaller at Cmax but the same pattern holds: salicylate is the durable circulating species. After a 75--80 mg antiplatelet dose, reported peaks are on the order of 6--7 µM aspirin and 15--29 µM salicylate. After anti-inflammatory gram doses, salicylate reaches roughly 1.5--2.5 mM. Almost every mitochondrial claim in the literature is a salicylate claim at the high end of that range. Almost every irreversible platelet claim is an aspirin claim at the low end.
That split is why 81 mg and 3 g are not the same drug at different strengths. They are different pharmacologies that happen to share a bottle.
------------------- --------------------- ------------------------------------ **Exposure** **Dominant species** **What actually happens**
81 mg daily Aspirin in platelets; Irreversible platelet COX-1 low-µM salicylate knockout; TXA2 collapse; partial PGE2 drop; 15-epi-lipoxin at COX-2; serotonin-dependent residual aggregation still visible
325 mg to Rising salicylate Endothelial COX-2 acetylation more anti-inflammatory complete; albumin-bound tryptophan grams displaced; TDO inhibition begins
Salsalate or Salicylate AMPK-β1 ligation; measurable proton high-dose aspirin leak; ACC phosphorylation; lower de (plasma salicylate novo lipogenesis; higher whole-body 1--3 mM) VO2
Overdose, \>3--5 mM Salicylate as toxin Frank uncoupling, KGDH/SDH restraint, heat, mixed respiratory alkalosis and anion-gap acidosis, PTP opening ------------------- --------------------- ------------------------------------
Baby aspirin lives in the first row. The metabolic-mitochondrial essay lives in the third and fourth. Conflating them is how the literature talks past itself.
2. The eicosanoid operating system aspirin was built to interrupt
Arachidonic acid in membrane phospholipid is liberated by cPLA2 and fed to cyclooxygenase. COX-1 is constitutive and is the only COX a platelet has. COX-2 is inducible in endothelium, monocytes, and inflamed tissue. Both make PGH2. Tissue-specific synthases then split PGH2 into TXA2, PGE2, PGI2, PGD2, and PGF2α.
The platelet's product is thromboxane A2. TXA2 binds TP receptors on neighboring platelets and on vascular smooth muscle. The platelet response is shape change, GPIIb/IIIa activation, and dense-granule release. The vessel response is constriction. Because the platelet has no nucleus, it cannot replace the COX-1 molecule aspirin acetylated. One pass through 81 mg therefore silences TXA2 production for the remaining life of that platelet, roughly seven to ten days. Nucleated endothelium can resynthesize cyclooxygenase; anucleate platelets cannot. Combined with presystemic (portal) acetylation of platelets on first pass, that is why 75--100 mg is relatively platelet-selective and why vascular PGI2 recovers while platelet TXA2 stays down. The TXA2/PGI2 ratio falls. That remains the cleanest oxygen-delivery effect of low-dose aspirin: fewer platelet plugs in capillaries, less TXA2-driven vasospasm. Endothelium is not exclusively a COX-2 organ --- some PGI2 is COX-1-derived --- so "COX-2 recovers in hours" is a useful shorthand, not a complete map of every vascular bed.
PGE2 is the more interesting eicosanoid for a metabolic hypothesis. It is not a clot molecule. It is a cAMP-raising, EP2/EP4-dependent instructor of innate immune cells. In experimental systems, PGE2 during monocyte-to-dendritic-cell maturation raises PKA activity and can suppress IL-12p70, tilting instruction away from a type-1 program. Aspirin does not have to stop platelets from docking to change the prostaglandin cargo those platelets can make. Separately, Liao and colleagues (NEJM 2012) found that regular aspirin after colorectal-cancer diagnosis was associated with longer survival only among patients whose tumors carried PIK3CA mutations --- a PI3K node that sits downstream of PGE2 receptors. That study is observational molecular-pathological epidemiology, not a randomized trial, and it has not been uniformly replicated. It is a clue, not a closed case. It is still difficult to explain a PIK3CA-restricted signal with "blood thinning" alone.
Aspirin does one other eicosanoid trick no other common NSAID copies. Acetylated COX-2 can no longer make PGH2, but it can still oxygenate arachidonate to 15R-HETE. Leukocyte 5-lipoxygenase converts 15R-HETE to 15-epi-lipoxin A4, the aspirin-triggered lipoxin (ATL). Chiang, Serhan and colleagues randomized 128 healthy adults to placebo or 81, 325, or 650 mg daily for eight weeks. Plasma ATL rose significantly in the 81 mg arm (mean change +0.25 ng/ml, P = 0.04), rose borderline in the 325 mg arm, and did not change significantly versus baseline in the 650 mg arm --- with large SDs in every arm. Across all three aspirin doses, ATL and TXB2 moved in opposite directions. ATL can stimulate endothelial NO and heme oxygenase-1 in experimental systems. In eNOS or iNOS knockout mice, aspirin loses much of its anti-inflammatory effect in IL-1β peritonitis (Paul-Clark et al.). So part of low-dose "anti-inflammation" is a resolution-and-NO story, not only prostaglandin subtraction. How much of that mouse NO requirement survives in human microcirculation is still an inference.
The first layer of the hypothesis is therefore not mitochondrial at all. It is a rewrite of the local chemical telegram that decides whether a capillary stays open.
3. Uncoupling, written as chemistry rather than metaphor
A mitochondrion makes ATP by pumping protons out through Complexes I, III, and IV, then letting them back in through ATP synthase. The proton-motive force is stored electrical and pH work:
An uncoupler is anything that lets protons return without passing through ATP synthase. The energy of substrate oxidation then leaves as heat instead of ATP. Electron transport accelerates to defend membrane potential. Oxygen consumption rises. ATP per O2 falls.
Classical protonophores --- DNP, CCCP, FCCP --- are lipophilic weak acids whose pKa sits near physiological pH and whose anion is charge-delocalized enough to cross the inner membrane. The cycle is:
1. The protonated acid HA crosses the membrane (high permeability).
3. The anion A− returns down the electrical gradient (the slow step).
4. It reprotonates in the intermembrane space and repeats.
Salicylate is a weaker version of that machine. Its pKa is about 2.8--3.0, so at cytosolic pH 7.2 almost all of it is A−. That would seem to disqualify it. Two structural facts save it. The ortho-hydroxyl forms an internal hydrogen bond that delocalizes the carboxylate charge and raises the anion's membrane permeability. And the protonated species HA, though scarce, is extremely membrane-permeant.
Gutknecht's bilayer work showed the dual behavior cleanly. At low pH salicylate behaves as an HA2−-type proton carrier: conductance scales with the square of concentration and peaks at pH = pKa. At neutral pH it behaves as a lipid-soluble anion. The mitochondrial cycle that follows is HA influx driven by the concentration gradient, plus A− efflux driven by voltage. Net result: protons appear in the matrix. Model calculations in that paper were already enough to explain the "loose coupling," uncoupling, and swelling seen in isolated mitochondria across the therapeutic-to-toxic range. Aspirin itself and benzoate are weaker because their anions permeate poorly. Acetaminophen does not do this at all --- one reason aspirin, not acetaminophen, is the Reye-associated drug.
Concentration windows
- Isolated mitochondria and model membranes: 1 mM salicylate can more than quadruple proton conductance in cited preparations. One review reports intracellular pH falling from 7.4 toward 7.2, 6.9, and 6.7 at 1, 2, and 4 mM in those systems. Complete uncoupling in isolated-mitochondria models typically sits at 2--5 mM. These numbers are experimental, not measured human-tissue pH after a tablet.
- Intact HEK-293 cells: above 1 mM, oxygen uptake rises. The rise is not additive with a saturating dose of DNP, which is the operational definition of a shared protonophore mechanism. Below 10 mM the AMP/ADP ratio does not have to change for AMPK to activate; above 10 mM it does, because ATP is now being wasted fast enough to be seen.
- Human high-dose aspirin or salsalate: plasma salicylate 1--3 mM. That is the overlap window --- mild leak plus AMPK ligation.
Mild uncoupling is not the same physiology as toxic uncoupling. At a small drop in membrane potential, reverse electron transport through Complex I falls, so superoxide production can fall even as VO2 rises. That is the argument made for low-dose DNP, for thyroid-driven loose coupling, and for UCP1 in brown fat. At a large drop, ATP collapses, the permeability transition pore opens in the presence of Ca2+, cytochrome c leaves, and the cell dies. Salicylate in isolated liver mitochondria is a proton/Ca2+ ionophore and potentiates Ca2+-stimulated PTP opening; cyclosporine A and ruthenium red only partly prevent it.
Overdose physiology --- hyperthermia, sweating, mixed acid-base disorder --- is this chemistry with the volume knob broken. The initial respiratory alkalosis of salicylate poisoning is central hyperventilation; the later anion-gap acidosis is lactate plus organic acids from a cell that can no longer keep ATP synthase coupled to the electron-transport chain, plus inhibition of Krebs dehydrogenases.
A second mitochondrial injury: KGDH
Salicylate does not only leak protons. In cardiac mitochondria it also inhibits ADP-stimulated NADH-linked (state 3) respiration by inhibiting α-ketoglutarate dehydrogenase. Salicylate is competitive at the α-ketoglutarate site. Acetylsalicylate is noncompetitive, consistent with on-enzyme acetylation after binding. So the same molecule that makes the membrane leakier also pinches the Krebs cycle at KGDH. State 4 rises; state 3 falls. That is not "more energy." That is a mitochondrion being asked to run faster through a narrower gate.
4. AMPK, substrate choice, and dual-fuel oxidation
AMPK is the cell's charge-state sensor. It is activated by AMP/ADP at the γ subunit, by upstream kinases at Thr172, and allosterically by ligands that bind the β1 subunit at the same site as the synthetic activator A-769662. Salicylate is such a ligand. It causes a small allosteric activation on its own, protects Thr172 from dephosphorylation, and antagonizes A-769662, which is how you know they share a site. Complexes containing β2, or a β1-S108A mutation, lose the effect.
The metabolic consequences of AMPK relevant here:
- Phosphorylation of acetyl-CoA carboxylase. ACC makes malonyl-CoA. Malonyl-CoA is the allosteric brake on CPT-1, the outer-membrane enzyme that lets long-chain fatty acyl-CoA into mitochondria. When ACC is phosphorylated, malonyl-CoA falls, CPT-1 opens, long-chain fat enters, and β-oxidation rises.
- Phosphorylation of hormone-sensitive lipase in a pattern that limits adipose lipolysis. The same kinase that opens the mitochondrial fat gate also reduces the flood of free fatty acids leaving fat cells. Circulating FFA fall. The tissue burns fat it already has rather than being force-fed more.
- Inhibition of mTORC1 and of some anabolic programs. Combined with EP300 competition, this is the caloric-restriction-mimetic claim.
- In 3T3-L1 preadipocytes, salicylate raises PGC-1α, NRF1, TFAM, mtDNA, citrate synthase, Complex I protein, and mitochondrial mass. Compound C blocks it. That is biogenesis, not just acute fuel switching.
Hawley's mouse work put the AMPK-β1 requirement under the lipid effects: salicylate that produced 1--2 mM plasma levels phosphorylated AMPK and ACC in liver and adipose, dropped circulating FFA, and dropped respiratory exchange ratio (more fat oxidation). Those lipid effects vanished in β1-null mice.
Steinberg's salsalate work then complicated the picture in a useful way. Glucose-homeostasis and liver-fat improvements survived AMPK-β1 knockout. What survived with them was an increase in oligomycin-insensitive respiration and a direct increase in mitochondrial proton conductance at clinical concentrations. Uncoupling, not AMPK, tracked the suppression of de novo lipogenesis. Brown adipose respiration rose even without UCP1. The only way whole-body carbohydrate oxidation and fat oxidation rise together is if the system's energy demand rose. A protonophore raises demand. AMPK then decides which substrate is used to meet it. Both can be true at salsalate concentrations. Neither is a description of 81 mg.
Aspirin as the intact ester has a separate fatty-acid-oxidation signature in cell culture. Five millimolar aspirin for 24 hours increased mitochondrial long-chain FAO, inhibited peroxisomal FAO, hyperacetylated mitochondrial proteins (aspirin is a stronger acetyl donor than acetyl-CoA in vitro), fragmented the mitochondrial network, and slowed the electron-transport chain. Medium-chain fat oxidation, which does not need CPT-1, did not change. SIRT3 knockout did not block the FAO increase. The authors' reading was compensatory: the cell drives long-chain fat through CPT-1 because the ETC and the network are impaired, not because the matrix FAO enzymes were activated. That paper is also the modern restatement of the Reye warning --- if the FAO machinery is already genetically weak, the same push becomes a bottleneck.
Substrate choice under salicylate is therefore three overlapping programs, not one: AMPK opens CPT-1; proton leak raises ATP demand so any available substrate is oxidized; acetylation plus ETC slowing forces compensatory long-chain FAO while peroxisomal FAO falls, with risk of bottleneck in FAO-deficient hosts.
5. Acetylation, EP300, autophagy, mitophagy
Aspirin donates acetyl groups. That is not only a COX fact. Mitochondrial proteins become more acetylated after aspirin exposure. Most of those marks do not appear to be the reason FAO rises, but they are a reminder that the drug is a chemical acetylating agent in a compartment that already runs an acetyl-CoA / SIRT3 economy.
The cleaner quality-control pathway is nuclear and cytosolic. Salicylate competes with acetyl-CoA at the catalytic domain of the acetyltransferase EP300. EP300 inhibition de-represses autophagy. The effect persists in AMPK-deficient cells, so it is a parallel program. In mice and in C. elegans, aspirin increases autophagic flux and, in heart, mitophagy as read by mito-Keima. That is the basis for calling aspirin a caloric-restriction mimetic. It is a high-exposure program. It is also the most plausible longevity biochemistry aspirin currently has, and it belongs to the millimolar window rather than to 81 mg.
Mild uncoupling and mitophagy are not in conflict. A slightly leakier, slightly depolarized mitochondrion is a better mitophagy substrate if the depolarization is uneven across the network. A quality-control reading of high-dose salicylate is: raise proton conductance a little, tag the worst organelles for removal, AMPK-biogenesis the replacements. That is a hypothesis, not a measured sequence at human 81 mg.
6. Serotonin as an operating system, not a neurotransmitter
About 95 percent of the body's serotonin is peripheral. Almost all of that is made by enterochromaffin cells in the gut by TPH1, released into the portal and systemic plasma, and immediately vacuumed up by platelets through SERT (SLC6A4). Platelets cannot synthesize serotonin. They store it in dense granules via VMAT at millimolar concentration, together with ADP, ATP, and Ca2+. What circulates free in plasma is a small residue. What people call "blood serotonin" is almost entirely platelet serotonin. Chronic SSRI use empties that store because SERT is the only refill pump.
This architecture is the normal operating procedure, and every piece of it has a job.
Hemostasis
When a platelet hits exposed collagen or a thrombin burst, dense granules exocytose. ADP hits P2Y12. Serotonin hits 5-HT2A on neighboring platelets. 5-HT2A is Gq-coupled. PLC-β hydrolyzes PIP2 to IP3 and DAG. IP3 releases Ca2+ from the dense tubular system. DAG activates PKC. The platelet that just received serotonin now releases more dense-granule contents. That is a local amplifier, not a systemic hormone. Serotonin also helps retain fibrinogen and thrombospondin on the platelet surface, which is a procoagulant surface effect distinct from aggregation itself.
TXA2, working through the same Gq family as 5-HT2A, is the parallel amplifier from the eicosanoid side. The two signals converge on PLC. This is why "aspirin resistance" can be a serotonin story: if the 5-HT2A limb is loud, killing the TXA2 limb does not silence the platelet. Pharmacometabolomics of 81 mg found exactly that --- higher pre- and post-aspirin serotonin tracked higher residual collagen-induced aggregation, and adding serotonin ex vivo restored reactivity after aspirin.
Vascular tone
The same 5-HT released onto the vessel wall is interpreted according to which receptor and which cell sees it.
- 5-HT2A on vascular smooth muscle: constriction. This is the dominant response in denuded or diseased arteries, including coronary arteries, and it is the receptor linked to coronary spasm and to pulmonary arterial constriction.
- 5-HT1B on endothelium: NO-dependent dilation, when the endothelium is intact.
- 5-HT7 on smooth muscle: a later depressor phase.
- 5-HT3 on vagal afferents: an early, brief depressor reflex.
Systemic 5-HT therefore produces a triphasic blood-pressure response: brief fall, pressor peak, slower fall. In a healthy capillary with intact endothelium the net can even be dilation. In a plaque-lined coronary or a platelet-plugged arteriole the net is constriction plus more aggregation. Immune-complex models make the point brutally: platelet-released serotonin, acting through 5-HT2 receptors, is required for the shock and the microvascular leak; TPH1 knockouts and SSRI-depleted platelets are resistant. In TRALI models the 5-HT2A limb on pulmonary arterial smooth muscle raises pulmonary pressure and edema. Peripheral serotonin is a context-dependent switch: repair signal in an intact vessel, occlusive and leaky signal in an injured or inflamed one.
Gut and portal metabolism
Enterochromaffin 5-HT drives motility through enteric 5-HT3/4 receptors. It also lands on hepatocytes, adipocytes, and pancreatic islets. Peripheral 5-HT can promote lipogenesis and suppress thermogenesis in some adipose beds, and it participates in insulin handling. The gut is therefore not only the factory for platelet serotonin. It is a metabolic organ whose 5-HT output changes hepatic and adipose fuel handling before a platelet ever fires.
Immune tone
Platelets docking onto monocytes and endothelium dump 5-HT onto cells that express 5-HT receptors. The same docking event dumps TXA2 and PGE2. Serotonin is part of the cargo that teaches the innate cell what kind of instructor to become. This is the seam between the dendritic-cell story and the serotonin story. They are not two essays. They are two readouts of one platelet synapse.
Brain, which is a separate pool
Central serotonin is made by TPH2 in raphe neurons. It does not freely equilibrate with platelet serotonin. Tryptophan availability does connect the pools, because TPH in both places competes with TDO and IDO for the same amino acid. That is the only honest bridge between "aspirin and mood" and "aspirin and platelets," and it runs through tryptophan, not through platelet 5-HT crossing the blood-brain barrier.
Mitochondria, directly
This is the least advertised part of serotonin physiology and the most relevant to an energy essay.
Tryptophan hydroxylase requires molecular oxygen. Serotonin synthesis is therefore itself an oxygen-consuming pathway. MAO-A, which degrades serotonin to 5-HIAA, sits on the outer mitochondrial membrane; the reaction produces hydrogen peroxide. So the life cycle of serotonin is physically bolted to mitochondria at synthesis (oxygen) and at disposal (MAO-A / H2O2).
In cortical neurons, 5-HT2A activation raises mitochondrial biogenesis and respiratory capacity through a SIRT1–PGC-1α axis. Both the hallucinogenic agonist DOI and the non-hallucinogenic 5-HT2A agonist lisuride increased mtDNA. SIRT1 knockout abolished the respiratory gains. That is central, not platelet, pharmacology, but it establishes that 5-HT2A is a mitochondrial-biogenesis receptor in at least one tissue. Reports that serotonin can raise inner-membrane potential, and that mitochondria contain the enzymes to convert serotonin to melatonin — itself an ETC-accelerating antioxidant — add a second, still-underdetermined intramitochondrial role. Serotonin also appears to tune the split between glycolytic and oxidative ATP from glucose. None of that needs to be oversold. It is enough to say that serotonin is not only a vessel and platelet signal. It is a fuel-partitioning signal with a documented mitochondrial-biogenesis limb.
Boundaries
Map ≠ terrain. Fuel · Circadian · Strength remain the board.
A working hypothesis, open to correction and revised as new evidence arrives. Last revised October 17, 2026.