Gut bacteria-related metabolite TMAO linked to heart disease risk in patients with possible artery blockages

This analysis is part of a large ongoing prospective diagnostic study (NCT01838148↗, clinicaltrials.gov) designed to advance the early detection of fCAD [26, 27]. The local ethics committee approved the study, which was carried out according to the principles of the Declaration of Helsinki. All patients provided written informed consent. The authors designed the study, gathered, analyzed and vouched for the data and analysis, wrote the paper, and made the decision to submit it for publication. Reported data follow STARD guidelines for studies of diagnostic accuracy [28].

Patients were recruited from 2010 to 2016 at the University Hospital of Basel, Switzerland. Enrolled patients were suspected to have fCAD and were referred for rest/stress myocardial perfusion single-photon emission tomography/computer tomography (MPI-SPECT). MPI-SPECT/CT was the preferred cardiac imaging technique in patients with a wide range of pre-test probabilities for fCAD during that time. Patients requiring chronic dialysis were excluded.

The likelihood for the presence of fCAD was quantified by the integrated clinical judgment of the treating cardiologist using a visual analogue scale (VAS) ranging from 0 to 100% twice: once before stress testing integrating all medical information available at that time, such as age, sex, cardiovascular risk factors, previous cardiac history, symptoms and baseline ECG data; second, after stress testing, integrating symptoms the patient experienced during exercise/stress, the workload achieved, and ECG changes recorded during exercise/stress. The cardiologist was blinded to both biomarker measurements and MPI-SPECT images at the time of assessment.

Venous non-fasting blood samples for determination of TMAO, choline, betaine, carnitine, and high-sensitivity cardiac troponin (hs-cTn) T [23, 26, 29], an established cardiovascular biomarker associated with fCAD, were obtained at rest, before stress testing. After centrifugation, samples were frozen at − 80 °C until assayed in a blinded fashion in a dedicated core laboratory. Serum TMAO, choline, betaine and carnitine were quantified using stable isotope dilution LC/MS/MS analyses as previously described using a Shimadzu Nexera Ultra High Performance Liquid Chromatograph (UHPLC) system interfaced with Shimadzu 8050 Triple Quadrupole Mass Spectrometer [1, 3, 18]. Hs-cTnT was measured with the Elecsys System on the Modular Analytics E170 or the Cobas e 602 (Roche Diagnostics, Rotkreuz, Switzerland). Limit of blank and limit of detection of this assay are 3.0 ng/L and 5.0 ng/L, respectively. The upper reference limit (URL) of a healthy reference population was 14 ng/L with an imprecision corresponding to 10% coefficient of variation (CV) at 13 ng/L [29]. Creatinine was measured by either the University Hospital of Basel central laboratory, Risch laboratories and Rothen laboratories. eGFR (estimated glomerular filtration rate) was calculated using the CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) equation provided within the transplantr-package in R. Cystatin-C was measured by SomaLogic using the SomaScan^® assay, which uses single-stranded DNA-based protein affinity reagents that are modified to mimic amino acid chains, enhancing protein-nucleic acid interaction [30]. These so-called SOMAmer reagents are selected against proteins in their native folded state, which after binding and other processing steps can then be quantified by DNA quantification techniques, providing a readout in relative fluorescent units (RFU) directly proportional to the amount of the protein [30].

Expert interpretation of MPI-SPECT/CT images combined with information obtained from invasive coronary angiography and whenever available fractional flow reserve measurements were used in the adjudication of fCAD.

All patients underwent a routine rest/stress dual isotope (201Tl for rest, 99mTc sestamibi for stress) or single isotope (99mTc sestamibi for stress and rest) MPI-SPECT protocol as described previously [27, 31, 32]. MPI-SPECT images were scored semi-quantitatively using a 17-segment model with a 5-point scale (0 = normal, 1 = mildly reduced tracer uptake, 2 = moderately reduced tracer uptake, 3 = severely reduced tracer uptake and 4 = no uptake). The summed rest score (SRS) and summed stress score (SSS) were calculated based on the 17 segments in the rest and stress images. The difference of SRS and SSS yielded the summed difference score (SDS), whereby an SDS of at least two or a positive transient ischemic dilation ratio (TID) was considered as inducible myocardial ischemia. Two readers derived SSS and SRS by visual assessment and compared with the software (QGS) result. Differences in the visual assessment by the two readers were resolved by consensus. In case of equivocal findings from MPI-SPECT and coronary angiography, two independent cardiologists (one interventional cardiologist, one general cardiologist) that were blinded to biomarker results reviewed the case. A positive perfusion scan was overruled when coronary angiography showed normal coronary arteries and a negative perfusion scan was overruled if coronary angiography (within three months) revealed a high-grade coronary lesion (> 75% or fractional flow reserve (FFR) lower than 0.80) [23, 26, 29].

The prognostic endpoints were all-cause death, cardiovascular death and MI during long-term follow-up. Patients were contacted by telephone or standardized follow-up letter 1 year, 2 years and 5 years after enrolment with ongoing follow-up. In case of an event, further information was obtained from hospital records, general practitioner/cardiologists records, or the national death registry blinded to biomarker concentrations.

Normality testing was done using visual assessments (histograms and QQ-plots) and as Shapiro–Wilk test. Continuous variables are presented as median and respective interquartile range (IQR) and categorical variables are presented as frequencies and respective percentages. The Agresti-Coull method was used to calculate confidence intervals of proportions. Measures of central tendency of the biomarkers were compared by Mann–Whitney U test. Baseline characteristics were compared by Kruskal test for continuous data and Fisher test for categorical data. Diagnostic accuracy of the clinical assessment, TMAO, betaine, choline, carnitine, and their combination for fCAD was quantified by the area under the receiver operating curve (AUC) and compared with the method described by DeLong et al. [33]. Based on previous findings showing an association between TMAO concentration and prevalent CAD [6, 11, 34], subgroup analysis was performed stratified according to the presence or absence of previously known CAD. We hypothesized that TMAO concentration would have highest diagnostic accuracy in patients without known CAD. Using logistic regression with fCAD as outcome, log2-transformed concentrations of TMAO or one of its precursors (betaine, choline, carnitine) were adjusted for patient characteristics, risk-factors and treatment, and combined with the quantified clinical assessment of the treating physician before and after stress testing. Multicollinearity was checked by the variance inflation factor (VIF) using the vif-function of the car package in R-Studio.

Kaplan–Meier curves for concentrations of TMAO, betaine, choline and carnitine below and above the median as well as stratified by the upper reference limit of the established biomarker hs-cTnT (URL 14 ng/L), were constructed and compared by log rank testing. Quantification of the predictive accuracy of TMAO and precursors was done by time dependent AUCs whilst accounting for censoring (timeROC package) [35]. Cox regression analysis was used to evaluate whether the analyzed biomarkers were independent predictors of patient characteristics and risk factors. Assumptions for Cox regression were tested using Schoenfeld residuals. Statistical analyses were performed with R version 4.0.2. All hypothesis testing was two-tailed, and a p value < 0.05 was considered statistically significant.

Overall, 1726 consecutive patients with suspected functionally relevant coronary artery disease (fCAD) were included in this analysis (Supplemental Fig. 1), 478 patients (28%) were adjudicated to have fCAD. A total of 421 (24%) patients underwent coronary angiography with 284 patients within the 3 months after enrolment. Anamnesis revealed 764 (44%) to have known history of CAD. During follow-up, 88 patients experienced an incident MI, 223 patients died overall, 115 of which died due to cardiovascular reasons (CV death). Characteristics of patients stratified by fCAD are shown in Table 1. Stratification of patients by experience of either an AMI, all-cause or CV death, is shown in Supplemental Table 1. Patients with fCAD tended to be older and a higher proportion were male. A significant proportion of patients had diabetes, hypertension, and a history of cardiovascular disease. To portrait renal state of the patients, baseline eGFR and cystatin-C were added to the baseline tables. As expected, eGFR was significantly lower in patients with fCAD and cystatin-C significantly higher in patients with fCAD.

Median TMAO concentrations were significantly higher in patients with fCAD than in those without fCAD (5.33 IQR [3.55, 8.80] µM vs. 4.66 IQR [3.08, 7.24] µM, p < 0.001, Fig. 1A). Stratification by a history CAD revealed that this phenomenon was exclusively present in patients without known CAD (5.36 IQR [3.71, 8.57] µM vs 4.38 IQR [2.94, 6.83] µM, p < 0.001, Fig. 1B). Similar findings emerged for the three precursors betaine, choline and carnitine (Fig. 1C–H).

In patients with normal eGFR (n = 689), TMAO and the three precursors were significantly higher in patients adjudicated to have fCAD (Supplemental Fig. 2A–D). The baseline characteristics of this subgroup are presented in Supplemental Table 2.

TMAO concentrations were significantly, albeit weakly, correlated with hs-cTnT (Spearman’s rho 0.32, p < 0.001), NT-proBNP (Spearman’s rho 0.22, p < 0.001) and age (Spearman’s rho 0.25, p < 0.001). These correlations were weaker for betaine, choline and carnitine (Supplemental Fig. 3).

In the overall cohort, diagnostic accuracy for detection of fCAD was quantified and showed modest but statistically significant value for TMAO with a ROC AUC of (0.56, 95% CI 0.53–0.59, p < 0.001) and its precursors (betaine 0.55, 95% CI 0.52–0.58, p = 0.002; choline 0.54, 95% CI 0.51–0.57, p = 0.007; carnitine: 0.55, 95% CI 0.51–0.58; p = 0.001; Supplemental Fig. 4). Similar findings emerged in the subgroup of patients without known CAD (TMAO 0.59, 95% CI [0.54–0.63], p < 0.001; betaine 0.54, 95%-CI [0.50–0.58], p = 0.039; choline: 0.56, 95% CI [0.51–0.60], p = 0.006; carnitine 0.57, 95% CI [0.52–0.61], p = 0.002; Supplemental Fig. 4B). TMAO and its precursors did not significantly increase the AUC provided by the quantitative clinical assessment by the treating physician before (AUC 0.61, 95% CI [0.58–0.64]) and after (AUC 0.65, 95% CI [0.62–0.68]) stress testing (p > 0.05 for all comparisons of VAS + TMAO (or precursors) versus VAS alone, Supplemental Fig. 4C + 4D). TMAO (OR 1.19, 95% CI [1.08–1.31], p < 0.001) and its precursors were significant predictors of fCAD in the univariable model, with TMAO remaining a significant predictor even after adjusting for age, sex and CAD history (OR 1.12, 95% CI [1.01–1.24], p = 0.036, Supplemental Table 3). When adjusting for renal function (i.e., taking Cystatin-C into the model) or when adjusting for further for pre-defined patient characteristics, cardiovascular risk factors and medical history (model 2), neither TMAO nor its precursors remained significant predictors. In patients with normal eGFR, TMAO and the three precursors were only significant predictors in the univariable model but were no longer significant after adjusting further (Supplemental Table 3).

The median follow-up time was 1827.5 days (IQR [756, 1908]). The cumulative event incidence for 5-years death was 12.9% (223 events), for 5-years CV death 6.7% (115 events) and for 5-years MI 5.1% (88 events), and the composite 5-years MACE (MI and cardiovascular death) was 11.8% (203 events).

In Kaplan–Meier analyses, TMAO concentrations above the median had substantially higher all-cause, CV mortality and mace events versus patients below the median (P < 0.001, each; Supplemental Fig. 5 and 6). Similar results emerged for choline and carnitine.

Kaplan–Meier analyses revealed that stratification according to hs-TnT and TMAO concentrations below/above the URL provided incremental prognostic value with the highest all-cause and CV mortality observed in patients with both hs-cTnT and TMAO above the URL (Fig. 2A).

Subgroup analysis (n = 919) of marker-combination with low and normal eGFR levels, Kaplan–Meier survival analyses revealed significantly highest cumulative all-cause death events in patients with low eGFR combined with a TMAO above the median (Fig. 2B).

Time-dependent ROC curve analysis showed a consistent and moderate-to-good discriminative performance of TMAO and choline for 5-years all-cause death (e.g. TMAO at 2-years AUC: 0.67, 95% CI [0.61–0.73]; Choline at 2-years AUC: 0.64, 95%-CI [0.57–0.70]) and 5-years CV death (TMAO at 2-year, AUC: 0.73, 95% CI [0.65–0.81]; Choline at 2-year AUC: 0.67, 95% CI [0.59, 0.76]), and modest accuracy for the other markers and endpoints (Fig. 3). Furthermore, Cox-regression analyses on continuous log-transformed markers are shown in Table 2. TMAO, Carnitine and Choline remained independent predictors in the full model without adjustment of renal function (model 2). When adjusting further for renal function (model 2 + cystatin-c) both TMAO and carnitine remained independent predictors for 5-years CV death. In the cox-regression analysis performed on median stratified marker (Supplemental Table 4), only TMAO remained a significant predictor of all-cause death and CV death, even when adjusted fully and taking renal function into account (model 2 + cystatin-c).

Below is the link to the electronic supplementary material.

Dr. Walter reports a research grant from the Swiss Academy of Medical Sciences and the Bangerter Foundation (YTCR 23/17). Dr. Twerenbold reports receiving research support from the Swiss National Science Foundation (P300PB_167803), the Swiss Heart Foundation, the Swiss Society of Cardiology, the University Hospital of Basel, as well as speaker honoraria/consulting honoraria from Roche Diagnostics, Abbott Diagnostics, Siemens, Singulex and Brahms. Dr. Nestelberger received speaker honoraria from Beckman-Coulter. Dr. Koechlin has received a research grant from the University of Basel, the Swiss Academy of Medical Sciences and the Gottfried and Julia Bangerter-Rhyner Foundation, as well as the “Freiwillige Akademische Gesellschaft Basel”, outside the submitted work. Professor Mueller reports receiving research support from the Swiss National Science Foundation, the Swiss Heart Foundation, the KTI, the European Union, the Stiftung für kardiovaskuläre Forschung Basel, the University of Basel, the University Hospital Basel, Abbott, Beckman Coulter, Biomerieux, Brahms, Ortho Diagnostics, Roche, Siemens, Singulex, Sphingotec, as well as speaker honoraria/consulting honoraria from Abbott, Amgen, Astra Zeneca, Biomerieux, Boehringer Ingelheim, BMS, Brahms, Cardiorentis, Novartis, Roche, Sanofi, Siemens, and Singulex. Professor Hazen and Professor Wang were supported in part by grants from the National Institutes of Health and the Office of Dietary Supplements (P01HL147823, HL103866↗, HL126827↗, HL130819↗ and the Leducq Foundation). Mass spectrometry studies were performed on instruments housed in a facility supported in part by a Center of Excellence Award by Shimadzu Scientific Instruments. Professor Hazen and Professor Wang report being named as co-inventor on pending and issued patents held by the Cleveland Clinic relating to cardiovascular diagnostics and therapeutics and being eligible to receive royalty payments for inventions or discoveries related to cardiovascular diagnostics or therapeutics from Cleveland HeartLab, Quest Diagnostics, and Procter & Gamble. SL Hazen also reports being a paid consultant for Procter & Gamble and having received research funds from Procter & Gamble and Roche Diagnostics.

The Basel VIII study has been approved by the local ethics committee of Basel, Switzerland that is the Ethikkommission Nordwest- und Zentralschweiz (EKNZ) with the number (PB_2019-00001/EKBB 100/04).

Each patient enrolled to the Basel VIII study has agreed to participate in the study via a written consent form.

All authors agreed for publication of this manuscript.

Please contact the corresponding author regarding data availability.

Please contact the corresponding author to request codes.