Lipids were extracted by chloroform:methanol (2:1, v/v)40
. Cholesterol was quantified by GC/MS41
. Triglyceride was quantified by GPO reagent set (Pointe Scientific)42
. Cell DNA content was quantified by PicoGreen43
. RNA was isolated by TRIZOL® reagent (Invitrogen) and RNeasy Mini Kit (Qiagen). All reagents were purchased from Sigma unless otherwise specified.
Plasma samples and associated clinical data were collected as part of two studies involving stable non-symptomatic subjects undergoing elective cardiac evaluations at a tertiary care center. The first study, GeneBank, is a large well-characterized tissue repository with longitudinal data from subjects undergoing elective diagnostic left heart catheterization or elective coronary computed tomography angiography44
. The second study, BioBank, includes subjects undergoing cardiac risk factor evaluation/modification in a preventive cardiology clinic45
. CAD included adjudicated diagnoses of stable or unstable angina, MI or angiographic evidence of ≥ 50% stenosis of one or more epicardial vessels. PAD was defined as any evidence of extra-coronary atherosclerosis. Atherosclerotic CVD was defined as the presence of either CAD or PAD. All subjects gave written informed consent and the Institutional Review Board of the Cleveland Clinic approved all study protocols.
Discovery metabolomics analyses began with an unbiased search for plasma (fasting, EDTA purple top tube) analytes linked to CVD risk using a case:control design (Learning Cohort, N=50 cases and 50 controls). Cases were randomly selected from GeneBank subjects who experienced an MI, stroke or death over the ensuing three year period. An age- gender-matched control group was randomly selected from GeneBank subjects that did not experience a CVD event. An independent non-overlapping Validation Cohort (N=25 cases and 25 controls) was also from GeneBank. A third large (N=1876) independent study comprised of non-overlapping subjects then evaluated clinical associations of identified analytes. Approximately half (N=1020) of the subjects enrolled were from GeneBank and the remaining (N=856) were from BioBank. Similar patient characteristics within each cohort and the combined cohort are observed, as shown in Supplementary Table 4a.b
. The association of each PC metabolite and various cardiovascular phenotypes within each cohort (GeneBank and BioBank) are also similar (Supplementary Tables 4c–e
). All subjects in the large independent clinical study had similar inclusion and exclusion criterion, negative cardiac enzymes (troponin I < 0.03 ng/ml) and no recent history of MI or CABG. Estimate of glomerular filtration rate was calculated using the MDRD formula46
. Fasting blood glucose, C reactive protein, troponin I and lipid profiles were measured on the Abbott ARCHITECT platform (Abbott Diagnostics).
Plasma proteins were precipitated with 4 volumes of ice cold methanol and small molecule analytes within supernatants were analyzed following injection onto a phenyl column (4.6 × 250 mm, 5 μm Rexchrom Phenyl; Regis, Morton Grove, IL) at a flow rate of 0.8 ml/min using a Cohesive HPLC (Franklin, MA) interfaced with an PE Sciex API 365 triple quadrupole mass spectrometer (Applied Biosystems, Foster, CA) with Ionics (Ontario, Canada) HSID+, EP10+, XT+ redesigned source and collision cell as upgrades in positive MS1 mode. LC gradient (LC1) starting from 10 mM ammonium formate over 0.5 min, then to 5 mM ammonium formate, 25 % methanol and 0.1 % formic acid over 3 min, held for 8 min, followed by 100% methanol and water washing for 3 min at a flow rate of 0.8 ml/min was used to resolve analytes. Spectra were continuously acquired after the initial 4 minutes. Peaks within reconstructed ion chromatograms at 1 amu increments were integrated and both retention times and mass-to-charge ratio (m/z) of analytes were used for statistical analyses.
Selection criteria for determining analytes of interest were based upon the composite of MACE as the clinical phenotype, defined as incident myocardial infarction, stroke or death at 3 years, and included: (i) demonstration of a statistically significant difference between cases vs. controls using a Bonferoni adjusted two sided t-test (P<0.05); (ii) evidence of a significant (P<0.05) dose-response relationship between analyte level and clinical phenotype using Cochran-Armitage trend test; and (iii) a minimal signal-to-noise ratio of 5:1 for a given analyte.
Chemical characterization of unknown metabolites
To chemically define the structures of the plasma analytes selected for further investigation (i.e. analytes with m/z 76, 104 and 118 in positive MS1 mode), multiple approaches were used. Analytes of interest were isolated by HPLC, vacuum dried, re-dissolved in water and injected onto the same phenyl column with a distinct HPLC gradient (LC2, flow rate: 0.8 ml/min) starting from 0.2% formic acid over 2 min, then linearly to 18% acetonitrile containing 0.2 % formic acid over 18 min and further to 100% acetonitrile containing 0.2 % formic acid over 3 min. The targeted analytes were identified by their m/z and the appropriate fractions recovered. After removal of solvent, dry analytes were used for structural identification.
Samples analyzed by GC-MS were derivatized using Sylon HTP kit (HMDS + TMCS + Pyridine (3 : 1 : 9), Supelco). Derivitizations of TMAO and the plasma analyte at m/z 76 also included initial reduction by titanium (III) chloride47
and further derivatized by reaction with 2,2,2-trichloroethylchloroformate48
. Analyses were performed on the Agilent Technolgies 6890/5973 GC/MS in positive ion chemical ionization mode. The GC/MS analyses used a J&W Scientific (Folsom) DB-1 column (30 m, 0.25 mm inner diameter, 0.25-μm film thickness) for separations.
Quantitation of TMAO, choline and betaine
Stable isotope dilution LC/MS/MS was used for quantification of TMAO, choline and betaine. TMAO, choline and betaine were monitored in positive MRM MS mode using characteristic precursor - product ion transitions: m/z 76→58, m/z 104→60, and m/z 118→59, respectively. The internal standards TMAO-trimethyl-d9 (d9-TMAO) and choline-trimethyl-d9 (d9-choline), were added to plasma samples prior to protein precipitation, and were similarly monitored in MRM mode at m/z 85→68 and m/z 113→69, respectively. Various concentrations of TMAO, choline and betaine standards and a fixed amount of internal standards were spiked into control plasma to prepare the calibration curves for quantification of plasma analytes. TMA was similarly quantified from acidified plasma by LC/MS/MS using MRM mode.
Aortic root lesion quantification
Apolipoprotein E knockout mice (C57BL/6J.Apoe−/−) were weaned at 4 weeks of age and fed with either standard chow control diet (Teklad 2018) or a custom diet comprised of normal chow supplemented with 0.5% choline (Teklad TD.07863), 1.0% choline (Teklad TD.07864) or 0.12% TMAO (Teklad TD.07865) for 16 weeks. Mice were anesthetized with Ketamine/Xylazine prior to cardiac puncture to collect blood. Hearts were fixed and stored in 4% paraformaldehyde prior to frozen OCT sectioning and staining with Oil-Red-O and hematoxylin. Aortic root lesion area was quantified as the mean value of 6 sections39
. Aortic sections were immunostained with rat anti-mouse F4/80 antibody (ab6640, abcam) followed by goat anti-rat IgG-FITC antibody (sc-2011, Santa Cruz) and FITC conjugated CD36 mAb (Cayman Chemical) for F4/80 and CD36, respectively. Sections were mounted in Vectashield DAPI (H-1200, Vectashield) to take pictures under a Leica
DMR microscope (W. Nuhsbaum) equipped with a Q Imaging Retiga EX camera (Burnaby, BC, Canada). Use Image-Pro Plus Version 7.0 (MediaCybernetics) to integrate the positive staining area of F4/80 and CD36 in aorta.
Flow cytometry assays on scavenger receptors
Cell surface expression of scavenger receptors, SR-A1 and CD36, were quantified on peritoneal macrophages from female mice by flow cytometry after immunostaining with fluorochrome conjugated antibodies. Fluorescence intensity was quantified on a FACSCalibur flow cytometry instrument with FlowJo software (BD Biosciences). >10,000 total events were acquired to obtain adequate macrophages numbers. The following antibodies were used to stain macrophages: CD36 mAb FITC (Cayman Chemical), anti-mouse SR-AI/MSRA1 (R&D Systems), goat anti-rat IgG-FITC (Santa Cruz Biotechnology), Alexa Fluor® 647 anti-mouse F4/80 (eBioscience), Alexa Fluor® 647 anti-mouse CD11b (eBioscience) and the isotype controls, Alexa Fluor® 647 rat IgG2b (eBioscience), Alexa Fluor® 647 rat IgG2a (eBioscience), normal mouse IgA-FITC (Santa Cruz). Cells were incubated with antibodies for 30 min at 4°C and washed with 0.1% BSA in PBS. Cells with double positive staining for F4/80 and CD11b were gated as macrophage49–51
for the quantification of fluorescence intensity for CD36 and SR-A1 (Supplementary Figure 20
), with results normalized to F4/80.
C57BL/6J.Apoe−/− (B6.Apoe−/−) mice were purchased from the Jackson Laboratory and C3H/HeJ.Apoe
−/− (C3H.Apoe−/−) mice were bred by backcrossing B6.Apoe−/− to C3H/HeJ for 10 generations. The F2 mice were generated by crossing B6.Apoe−/− with C3H.Apoe−/− and subsequently intercrossing the F1 mice. Mice were fed Purina Chow containing 4% fat until 8 weeks of age, and then transferred to a Western diet (Teklad 88137) containing 42% fat and 0.15% cholesterol for 16 weeks until euthanasia at 24 weeks of age. Mouse atherosclerotic lesion area was quantified using standard methods39
. eQTL analyses were performed as previously described38
. Each individual sample was hybridized against the pool of F2 samples. Significantly differentially expressed genes were determined as previously described52
. Expression data in the form of mean log ratios (mlratios) were treated as a quantitative trait in eQTL analysis using Rqtl package for the R language and environment for statistical computing (http://cran.r-project.org/
Antibiotic knockdown of endogenous gut flora, germ-free mice, and conventionalization studies
An antibiotic cocktail (0.5 g/L vancomycin, 1 g/L neomycin sulfate, 1 g/L metronidazole, 1 g/L ampicillin) previously shown to be sufficient to deplete all detectable commensal bacteria37
was administered in drinking water ad lib. In additional studies, 8 week old female Swiss Webster Germ Free mice (Taconic SWGF) underwent an oral (gavage) choline challenge (see below) immediately following their removal from their germ-free microisolator shipper. After the choline or PC challenge, the germ-free mice were placed in conventional cages with non-sterile C57BL/6J female mice to facilitate transfer of commensal organisms. Four weeks later, the conventionalized mice underwent a second choline or PC challenge.
In vivo macrophage studies
C57BL/6J mice or B6.Apoe−/− mice were fed with either standard chow control diet (Teklad 2018) or a custom diet supplemented with 1.0% betaine (Teklad TD.08112), 1.0% choline (Teklad TD.07864), 0.12% TMAO (Teklad TD.07865) or 1.0% dimethylbutanol (DMB) supplemented in drinking water for at least 3 weeks. Elicited mouse peritoneal macrophages (MPMs) were harvested by peritoneal lavage with ice-cold PBS 3 days after intraperitoneal (i.p.) injection of 1 mL 4% thioglycollate. Some studies with mice were performed using a custom diet with low but sufficient choline content (0.07% total; Teklad TD.09040) vs. high choline diet (1.0% total; Teklad TD.09041) in the presence or absence of antibiotics. Choline content of all diets was confirmed by LC/MS/MS.
Foam cell staining
Foam cells were identified by microscopy cultured peritoneal macrophages on glass cover-slips following 6 hours in RPMI 1640 medium supplemented with 3% lipoprotein deficient serum. Cells were fixed with paraformaldehyde and stained with Oil Red O/hematoxylin53
. Cells containing >10 lipid droplets were scored as foam cells50
. At least 10 fields and 500 cells/condition were counted.
Real time PCR
Real time PCR of CD36, SR-A1 and flavin monooxygenases (FMOs) was performed using Brilliant II SYBR® Green QRT-PCR kit (Strategene). The forward and reverse primers, CD36, GAPDH, SR-A1, FMOs and F4/80, were synthesized by IDT based on sequences reported54–58
Synthesis of 1,2-dipalmitoyl-sn-glycero-3-phosphocholine-N,N,N-trimethyl-d9 (d9-DPPC) and preparation of lipid vesicles
d9-DPPC was synthesized by reacting 1,2 -dipalmitoyl-sn-glycero-3-phosphoethanolamine (Genzyme Pharmaceuticals) with per-deuteromethyliodide (CD3
I, Cambridge Isotope Laboratories)59–60
. The product was purified by preparative silica gel TLC and confirmed by both MS and NMR. Egg yolk lecithin (Avanti Polar Lipids) and d9-DPPC liposomes used for gavage feeding and i.p.
injection of mice were prepared by the method of extrusion through polycarbonate filters61
Metabolic challenges in mice
C57BL/6J mice were administered (gavage) unlabeled or the indicated stable isotope labeled choline or PC (egg yolk lecithin or d9-DPPC) using a 1.5 inch 20 gauge intubation needle. Choline challenge: gavage consisted of 150 μl of 150 mM choline-d9. PC challenge: gavage or i.p. injection of 300 μl 5 mg/ml of unlabeled PC or labeled d9-DPPC. Mice were fasted 12 hours prior to PC challenge. Plasma (50 μl) was collected via the saphenous vein from mice at baseline and post gavage or i.p. injection time points.
test and Wilcoxon rank sum test were employed to compare group means62–63
. Pearson correlation, Spearman rank correlation and Somers' Dxy correlation were used to investigate the correlation between two variables64–65
. Comparison of categorical measures between independent groups was done using χ2
. Odds ratios and 95% confidence intervals (CI) for cardiovascular phenotypes (history of MI, CAD, PAD, CVD and CAD+PAD) were calculated with R, Version 2.10.1 (www.r-project.org
), using logistic regression67
with case status as the dependent variable and plasma analyte as independent variable. Trend tests in frequencies across quartiles were done using Cochran-Armitage Trend tests68
. Levels of analytes were adjusted for traditional CAD risk factors in a multivariate logistic regression model including individual traditional cardiac risk factors (age, gender, diabetes, smoking, hypertension, lipids, CRP and estimated creatinine clearance) and medication usage (statin or other lipid lowering agents, antihypertensive agents including angiotensin-converting enzyme inhibitor, angiotensin receptor blocking agent, diuretic, calcium channel blocker or beta blocker, and aspirin or other platelet inhibitors).