The present study was conducted as a prospective observational study and patients in whom CABG was planned using the LITA through median sternotomy were screened for study eligibility. With an estimated number of patient enrollment as 10 per month based on our institutional clinical volume, the study was designed to enroll patients either until 20 study patients were included or until 3 months of enrollment period was completed to minimize any bias from a prolonged storage period of blood samples. Patient enrollment was started on August 31, 2017 and enrollment of 20 patients was completed on November 15, 2017. During the study period, 24 patients were screened and 4 of these patients were excluded because there was a substantial risk of manipulating the heavily calcified ascending aorta. Mean age of the study patients was 63.8 ± 10.8 years and 14 patients were men (Table 1). Nineteen patients underwent isolated CABG and the other patient underwent combined CABG and aortic valve replacement.

Under general anesthesia and hemodynamic monitoring, the LITA was harvested after median sternotomy and intravenous heparin (300 IU/kg) was injected before cutting the distal end of the LITA to maintain an activated clotting time greater than 300 seconds. After pericardiotomy, 2 mL of whole blood was simultaneously drawn from the ascending aorta and the LITA. Aorta sample was drawn by direct puncture using 2 mL syringe with 21 gauge needle, and LITA sample was collected by shedding to opened 2 mL syringe. Both samples were injected to ethylenediaminetetraacetic acid-coated vacutainers (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) through 21 gauge syringe needles. All sample procedures were carefully performed to avoid any hemolysis. Immediately after sampling, plasma was separated by centrifugation at 3,000 rpm for 10 minutes at 4°C and stored in a −72°C liquefied nitrogen (N2) freezer until sample preparation for MS.

Extraction solvents, including high-performance liquid chromatography grade of isopropanol, acetonitrile, and water were purchased from J.T. Baker Chemical Co. (Phillipsburg, NJ, USA). The chemicals used for derivatization including fatty acid methyl ester (FAME) mixtures, pyridine, methoxamine (MeOX) hydrochloride, and N-Methyl-N-(trimethylsilyl)trifluoroacetamide (MSTFA) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Reference standards and internal standards used for identification and semi-quantification of selected markers were also obtained from Sigma-Aldrich.

Untargeted metabolomic analysis was performed using a high-resolution mass analyzer GC-TOF-MS (LECO Corporation, St. Joseph, MI, USA). The samples were prepared as described previously.11 Briefly, 50 µL of plasma samples was extracted using 1 mL of degassed extraction solution (acetonitrile:isopropanol:H2O = 3:3:2). For the quality control (QC) samples, 100 µL of each sample was pooled after the first extraction. The extracted samples were evaporated under a N2 evaporator and subjected to a second extraction with 400 µL of extraction solution (acetonitrile:H2O = 1:1). Completely dried samples were then derivatized with 10 µL of methoximation solution (20 mg/mL MeOX hydrochloride in pyridine) at 30°C for 90 minutes in a shaking incubator and cooled at room temperature. The samples were further derivatized with 90 µL of mixture solution (5% FAME in MSTFA) at 70°C for 45 minutes and cooled again. The prepared samples were then transferred into GC injection vials.

For the GC analysis, a 1 µL aliquot of each prepared sample was injected with a front inlet split ratio of 20. The QC samples were injected after every ten samples to ensure the quality of the analysis. The metabolites from the samples were separated through Rtx-5MS columns (Restek Corporation, Bellefonte, PA, USA). The GC oven temperature was increased from 50°C at a rate of 20°C per minutes until the oven temperature reached 350°C to separate the metabolites. The mass spectrometer was set to detect metabolites ranging from 50 to 800 m/z (mass per charge ratio) with an acquisition voltage and a rate of 1,750 and 20 spectra/second, respectively. The transfer line and ion source temperatures were set to 280°C and 250°C, respectively.

Metabolomics data analysis was performed using Metaboanalyst 4.0, a web-based software.13 Metabolites with over 50% missing values were excluded in the following analyses. Interquartile range was applied for the data filtering and Pareto scaling was applied as a data pretreatment method for normalization.1415 Statistical analyses, such as principal component analysis (PCA) and the volcano plot,161718 were performed to discover metabolic markers showing significant differences between the ITA and ascending aorta. Briefly, the PCA is an unsupervised multivariate analytical method using an unpaired data set of samples method to visualize subtle similarities or differences among complex datasets.16 The volcano plot is a combination analysis of fold change (FC) and t-test which visualizes the difference between datasets by the Y-axis with log10 (P value) and the X-axis with number of significant pairs which meet the FC cutoff criteria defined below. For the volcano plot analysis, false discovery rate (FDR)-adjusted P values from paired t-tests were used to control FDR in multiple tests.19 For selection of marker candidates the mean initial cutoff values of FC were set to be higher than 1.1 or lower than 0.9 with a significant threshold count percent set to be greater than 60%. The final markers were selected based on cutoff values of the volcano plot with a mean FC > 1.2 or < 0.8 and a significant threshold count percent > 75%.2021

ChromaTOF 4.6 (LECO Corporation) was used to assign the name of markers. Three commercially available libraries, including the NIST/EPA/NIH Mass Spectral Library (Version 2.2), LECO-Fiehn Rtx5 and Wiley (9th edition), were used22 with a cut-off value of 70% to match the minimum similarity of the detected spectrum.

Identified metabolites were reviewed through the literature and their main metabolic pathways were determined using the Kyoto Encyclopedia of Genes and Genomes pathway database (http://www.genome.jp/kegg/pathway.html) to select other potential metabolic markers.16

Prior to the marker quantification, confirmation on assigned name of the markers was conducted by comparing mass spectrum with commercially available reference standards. For the quantification, GC-TOF-MS based semi-quantification was performed using the commercially available reference standards of identified metabolites. Dodecanoic acid methyl ester, eicosanoic acid methyl ester, and octadecanoic acid methyl ester from the FAME mixtures were used as the internal standards. The peak area was used for the quantification. For quantification of the hypothetical markers, the concentration was determined using an enzyme-linked immunosorbent assay (ELISA) kit (NOVATEINBIO, Woburn, MA, USA) analyzed through a VersaMaxPLUS microplate reader (Molecular Device, San Jose, CA, USA). The quality of the ELISA was treated as accurate if the quantification of the metabolite was duplicated, including a standard curve, and the coefficient of variation was less than 15% in all samples.

All statistical analyses were conducted using the GraphPad Prism 7.0 software package (GraphPad Software, San Diego, CA, USA). A paired t-test was used for statistical comparison between the concentrations from the ITA and ascending aorta. The P values of less than 0.05 were considered statistically significant in the paired t-test.

The study protocol was reviewed and approved by the Institutional Review Board (IRB) at Seoul National University Hospital (approval No. 1708-058-877). Individual consent under IRB-approved protocols was obtained from all study patients.

A total of 851 peaks were detected by GC-TOF-MS. Tightly clustered QC samples in the PCA plot showed that GC-TOF-MS analysis was conducted properly. No obvious separation was observed between the ITA and ascending aorta groups in the paired PCA plot (Fig. 1). In volcano plot analysis, a total of nine peaks met the initial criteria (Fig. 2). Among the nine selected peaks, the peak shape and mass spectrum comparison with the commercial library excluded four markers that were considered as derivatization reagents or artifacts. L-cysteine, propanoic acid, cholesterol, picolinic acid and L-glutamine were selected as possible markers (Table 2). In the subsequent volcano plot analysis with the final cut-off criteria, the L-cysteine showed a statistically higher peak intensity in the ITA than the ascending aorta (mean FC = 1.86, P = 0.02) (Fig. 3).

Based on the MSEA, three intermediate metabolites in the ‘cysteine and methionine metabolism pathway' linked with L-cysteine, such as L-methionine, L-homocysteine and L-cystine, were selected for the subsequent analyses. Hydrogen sulfide (H₂S), one of the final products, was also selected as one of the hypothetical markers. The GC-TOF-MS based semi-quantification demonstrated that plasma concentrations of L-cysteine, L-methionine, and L-cystine were 2.0-, 1.4- and 1.2-fold higher, respectively, in the ITA than in the ascending aorta (P < 0.001 in each, Table 3 and Fig. 4). L-homocysteine was not detected in the samples. ELISA showed that the plasma concentrations of H₂S were significantly higher in the ITA than in the ascending aorta (P = 0.001) (Table 3).