Endpoint: mixed — mouse mechanism/atherosclerosis experiments (surrogate) + a human prospective cohort predicting MACE (hard).
Identifies choline, TMAO and betaine as gut-microbiota-dependent metabolites of dietary phosphatidylcholine. Dietary choline or TMAO supplementation promoted atherosclerosis in mice via upregulated macrophage scavenger receptors and foam-cell formation; germ-free mice confirmed the effect requires gut flora. Plasma TMAO levels also predicted cardiovascular risk in an independent human clinical cohort. This is the founding paper of the entire choline→TMAO→atherogenesis hypothesis.
relevance_note: the root primary for the whole TMAO-harm-pathway thread — establishes the mechanism and the first human predictive link, independent of the lipid pathway.
Methodology
Mixed design. HUMAN (rests on the Cleveland Clinic GeneBank/BioBank angiography cohort, D-4): untargeted plasma metabolomics in a prospective learning cohort (50 cases vs 50 matched controls, 3-year MI/stroke/death) and validation cohort (25 vs 25) identified choline, TMAO, and betaine; an independent 1,876-subject cohort tested dose-dependent association with prevalent CVD and angiographic burden. All human subjects were stable patients referred for elective coronary evaluation. MOUSE / mechanistic (own experiments, S-28): Apoe-/- mice fed control vs choline (0.5-1.0%) or TMAO (0.12%) chow for ~16 weeks; stable-isotope (d9-choline/d9-PC) tracer challenges in normal, antibiotic-treated, germ-free, and conventionalized mice; broad-spectrum-antibiotic suppression experiments; macrophage scavenger-receptor and foam-cell assays; and an F2 intercross mapping the FMO3 locus. COI: senior author Hazen holds patent/commercial interests in TMAO diagnostics (flagged across the Hazen-lab TMAO corpus).
Human observations (rest on D-4)
O-37 - Plasma choline, TMAO and betaine predict 3-year incident MI, stroke or death in a prospective angiography sub-cohort
Human observational finding, resting on the Cleveland Clinic GeneBank angiography cohort (D-4). Prospective incident-event design partly mitigates reverse causation, but n is small and the population is referral-selected.
Link to original
O-38 - In 1,876 angiography patients, plasma choline-TMAO-betaine show dose-dependent association with prevalent CVD and TMAO tracks coronary burden
Human observational finding on the GeneBank/BioBank angiography population (D-4). Cross-sectional/prevalent-CVD association in a sick, referral-selected cohort — confounding and reverse causation not excluded.
Link to original
Mouse / mechanistic observations (rest on S-28)
O-34 - Gut flora are obligately required to convert dietary phosphatidylcholine-choline to TMAO in mice (germ-free and antibiotic abolition, conventionalization restores)
Own isotope-tracer experiment establishing that intestinal microbiota are necessary for TMA/TMAO generation from dietary PC/choline.
Link to original
O-32 - Dietary choline or TMAO supplementation increases atherosclerosis in Apoe--- mice, correlating with plasma TMAO and not with plasma lipids
Own mouse experiment. The lipid-independence (no rise in cholesterol/TG/glucose) is central: it marks TMAO as a non-lipid atherogenic pathway.
Link to originalMethodology
Mixed design. HUMAN (rests on the Cleveland Clinic GeneBank/BioBank angiography cohort, D-4): untargeted plasma metabolomics in a prospective learning cohort (50 cases vs 50 matched controls, 3-year MI/stroke/death) and validation cohort (25 vs 25) identified choline, TMAO, and betaine; an independent 1,876-subject cohort tested dose-dependent association with prevalent CVD and angiographic burden. All human subjects were stable patients referred for elective coronary evaluation. MOUSE / mechanistic (own experiments, S-28): Apoe-/- mice fed control vs choline (0.5-1.0%) or TMAO (0.12%) chow for ~16 weeks; stable-isotope (d9-choline/d9-PC) tracer challenges in normal, antibiotic-treated, germ-free, and conventionalized mice; broad-spectrum-antibiotic suppression experiments; macrophage scavenger-receptor and foam-cell assays; and an F2 intercross mapping the FMO3 locus. COI: senior author Hazen holds patent/commercial interests in TMAO diagnostics (flagged across the Hazen-lab TMAO corpus).
Link to original
O-36 - Hepatic FMO3 expression and an FMO3-region SNP segregate with atherosclerosis, TMAO, and (inversely) HDL in a mouse F2 cross
Own mouse genetics: host genetic variation controlling TMAO production tracks atherosclerosis, an independent (non-dietary) line supporting a causal role for the TMAO axis.
Link to original
O-33 - Choline, TMAO and betaine upregulate macrophage scavenger receptors CD36-SR-A and promote foam-cell formation in mice
Own mouse/macrophage experiment establishing the proposed atherogenic mechanism (scavenger-receptor-driven foam-cell formation).
Link to original
O-35 - Antibiotic suppression of gut flora lowers plasma TMAO -100-fold and abolishes choline-induced ~3-fold atherosclerosis and macrophage-CD36 increase in mice
Own mouse experiment; the antibiotic-dependence of both TMAO and the atherosclerosis enhancement ties the atherogenic effect to the flora-generated metabolite.
Link to original
Interpretation
H-15 - Gut flora convert dietary phosphatidylcholine-choline to TMAO, which causally promotes atherosclerosis independent of lipids
The paper’s central mechanistic hypothesis, strongly supported in mice. Relevant to eggs because egg yolk is a major dietary source of phosphatidylcholine/choline.
Link to original
H-16 - Elevated plasma TMAO is a clinically meaningful (causal) cardiovascular risk factor in humans
Extrapolation from the mouse causality plus the human associations. Contested: human evidence here is observational in a selected high-risk cohort, and later chronic-feeding studies dispute whether dietary eggs even raise steady-state TMAO.
Link to original
Arguments
A-11 - Germ-free-antibiotic necessity plus TMAO-supplementation sufficiency plus host-genetic segregation jointly establish a causal choline--flora--TMAO--atherosclerosis chain in mice
Causality here is argued by convergence of manipulations that a confounded correlation could not survive. (1) Necessity: germ-free mice and antibiotic-pretreated mice make no d9-TMAO from labelled oral choline/PC, and conventionalization restores it — so gut flora are an obligate step. (2) Necessity for disease, not just for the metabolite: antibiotics that drop plasma TMAO >100-fold also abolish the ~3-fold choline-induced atherosclerosis and the macrophage/CD36 increase, tying the disease effect specifically to the flora-generated metabolite rather than to choline per se. (3) Sufficiency: feeding TMAO itself (bypassing the choline→TMA step) reproduces the atherosclerosis, and it does so with no rise in plasma cholesterol, triglycerides, or glucose — excluding the classical lipid pathway as the mediator. (4) A mechanistic handle: choline/TMAO/betaine upregulate macrophage scavenger receptors CD36 and SR-A and drive foam-cell formation, supplying a plausible cellular route from TMAO to plaque. (5) An independent, non-dietary line: in an F2 genetic cross, host hepatic FMO3 (which oxidizes TMA to TMAO) co-segregates with lesion area and plasma TMAO (and inversely with HDL), so genetic variation that changes TMAO changes atherosclerosis without any dietary manipulation — a mendelian-randomization-flavoured corroboration. No single line is decisive, but necessity + sufficiency + mechanism + genetics pointing the same way is the signature of a causal pathway rather than a shared-confounder correlation. Scope caveat: all of this is in Apoe-/- mice, an accelerated-atherosclerosis model, so it licenses the pathway’s existence and causal direction in mice, not automatically its magnitude in humans.
Validity (step 6)
status: approved, reason_if_not_false: checked.
The step is elementary experimental causal logic (necessity + sufficiency + a genetic instrument), traceable without specialist formalism, so checked and author-blind. Assuming the observations (O-32–O-36), the word “establish … in mice” is warranted because the design contains genuine do-operations at both ends of the chain, not just correlation:
- Necessity of the whole flora step: germ-free/antibiotic mice make no labelled TMAO from choline; conventionalization restores it. Necessity for disease: antibiotics that drop TMAO >100-fold also abolish the choline-induced atherosclerosis — tying the disease to the flora-made metabolite, not to choline per se.
- Sufficiency: feeding TMAO directly (bypassing the choline→TMA step) reproduces atherosclerosis, and does so with no lipid/glucose change — which closes the main alternative mediator (lipids) and pins the effect on TMAO itself.
- The obvious defeater to a pure-necessity argument — antibiotics abolishing disease via some non-TMAO microbiome effect — is killed by the sufficiency arm (TMAO alone suffices). Reverse causation is killed by the FMO3 F2-cross genetics (host genotype that raises TMAO raises plaque, with no dietary manipulation).
Necessity and sufficiency established by intervention jointly span the chain and survive the standard confounder/reverse-causation defeaters, so “establish a causal chain in mice” holds as stated. The claim is explicitly scoped to mice (existence/direction, not human magnitude), so it does not over-reach. Approved.
Link to original
A-12 - Human TMAO-CVD evidence is observational in a selected angiography cohort, so it supports but does not establish human causality
The human data have two parts with different strength. The prospective learning/validation sub-cohorts predict incident MI/stroke/death over 3 years, which does mitigate reverse causation (the events postdate the measurement). But the large-n confirmation (1,876) is a cross-sectional association with prevalent CVD burden, where reverse causation and shared-cause confounding (diet, renal function, gut-microbiome differences that co-vary with disease) are not excluded, and adjustment for traditional risk factors cannot remove unmeasured confounders such as kidney function (a strong determinant of plasma TMAO and of CVD). Crucially the entire human cohort is drawn from patients referred for elective coronary angiography — a sick, high-baseline-risk, non-general population — so both the effect size and its generality to healthy people are uncertain. What upgrades these associations toward a causal reading is not their internal design but their coherence with the mouse necessity/sufficiency chain: a metabolite shown to cause atherosclerosis in mice and associated with events in humans is more plausibly causal in humans than a bare human correlation would be. Hence the human observations support, but do not on their own establish, that plasma TMAO is a causal cardiovascular risk factor in people.
Validity (step 6)
status: approved, reason_if_not_false: checked.
Traced. The step is a deliberately modest one: (i) the large-n confirmation (O-38, n=1,876) is cross-sectional, so reverse causation and shared-cause confounding — kidney function being the salient unmeasured one, a strong determinant of both plasma TMAO and CVD — are not excluded; (ii) even the prospective sub-cohort (O-37), which does postdate measurement to events and so mitigates reverse causation, is drawn entirely from patients referred for elective coronary angiography, a selected sick high-risk population, so both effect size and generality to healthy people are uncertain. From these, “observational data in a selected cohort cannot by itself establish causality” follows straightforwardly. The second clause — that the associations gain causal weight “mainly by riding on the strong mouse mechanism” — is a valid consilience/coherence point: a metabolite already shown causal in a mouse intervention model plus associated with human events is more plausibly causal in humans than a bare correlation. Because the conclusion is explicitly “supports but does not establish,” there is no over-reach for a defeater to attack; the step is self-limiting and holds. Approved.
Link to original