Korean patients with newly diagnosed or relapsed GD who visited the Department of Internal Medicine, Chungnam National University Hospital in Daejeon between January 2019 and July 2020 were recruited. All patients were treated with methimazole for 12 weeks to achieve euthyroidism. Plasma samples were prepared from all 18 patients at the initial visit and 12 weeks later at a follow-up visit. We selected treatment period for 12 weeks. Because we previously published a paper about thyrotoxic myopathy in patient with GD [14]. In that paper, we found that skeletal muscle function and thyroid hormone levels were recovered in most GD patients treated with methimazole for 12 weeks.

This study was approved by the Institutional Review Board of Chungnam National University Hospital (CNUH-2019-02-012). Each participant provided consent and documented by the Department of International Medicine of Chungnam National University Hospital in Korea.

Details have been described previously [14]. Briefly, for whole metabolite extraction, 10 µL of plasma was added to 240 µL of water and 250 µL of ice-cold methanol before being vortexed and centrifuged (14,000 ×g, 4°C, 15 minutes). The supernatant was collected in a 1.5 mL Eppendorf microtube, processed for the extraction of various types of compounds listed below, and used for LC-MS measurement. For water-soluble metabolites, including amino acids and nucleotides, 25 µL of the supernatant was diluted 3-fold with 0.1% formic acid. For acylcarnitines, 30 µL of the supernatant was added to 270 µL of ice-cold methanol, vortexed, sonicated, and centrifuged (14,000 ×g, 4°C, 15 minutes), and the supernatant was collected. For free fatty acids (FFAs), 25 µL of the supernatant was diluted 2-fold with ice-cold methanol. For bile acids, 50 µL of the supernatant was evaporated and dissolved with 25 µL of 20% methanol. For phospholipids, 5 µL of the supernatant was diluted 200-fold with 0.1% formic acid in 20% acetonitrile.

Plasma samples were analyzed using LC-MS on a LCMS-8060 instrument (Shimadzu, Kyoto, Japan), essentially as described previously [14]. To measure wide varieties of water-soluble metabolites, the prepared sample was separated on a Discovery HS-F5-3 column (150×2.1 mm, 3 μm particle size; Sigma-Aldrich, St. Louis, MO, USA) with mobile phases consisting of solvent A (0.1% formic acid) and solvent B (0.1% formic acid in acetonitrile). The column oven temperature was 40°C. The gradient elution program was as follows with a flow rate of 0.25 mL/min: 0-2 minutes, 0% B; 2-5 minutes, 0%-25% B; 5-11 minutes, 25%-35% B; 11-15 minutes, 35%-95% B; 15-25 minutes, 95% B; and 25.1-30 minutes, 0% B. The parameters for the heated electrospray ionization (ESI) source in negative/positive ion mode under multiple reaction monitoring (MRM) were as follows: drying gas flow rate, 10 L/min; nebulizer gas flow rate, 3 L/min; heating gas flow rate, 10 L/min; interface temperature, 300°C; desolvation line (DL) temperature, 250°C; heat block temperature, 400°C; and collision-induced dissociation (CID) gas, 270 kPa. For acylcarnitines, the sample was separated on a Luna HILIC column 200A (150×2.0 mm, 3 μm; Phenomenex, Torrance, CA, USA) with mobile phases consisting of solvent A (10 mM ammonium formate) and solvent B (acetonitrile/10 mM ammonium formate [9/1, v/v]). The column oven temperature was 40°C. The gradient elution program was as follows with a flow rate of 0.3 mL/min: 0-2.5 minutes, 100% B; 2.5-4 minutes, 100%-50% B; 4-7.5 minutes, 50%-5% B; 7.5-10 minutes, 5% B; and 10.1-12.5 minutes, 100% B. Acylcarnitine profiles were detected in positive ESI mode under precursor ion scan for m/z 85.5 by changing collision energy according to the lengths of fatty acids as follows: -20 for shortchain acylcarnitines (C0-C8); -35 for middle-chain acylcarnitines (C9-C12); and -45 for long-chain acylcarnitines (C12-C18). For FFAs, the sample was separated on a ACQUITY BEH Amide column (150×2.1 mm, 1.7 μm; Waters, Milford, MA, USA) with mobile phases consisting of solvent A (10 mM ammonium formate in 90% acetonitrile) and solvent B (0.1% formic acid in acetonitrile). The column oven temperature was 40°C. The gradient elution program was as follows with a flow rate of 0.4 mL/min: 0-5 minutes, 0% B; 5.1-7.5 minutes, 100% B; and 7.6-11 minutes, 0% B. FFA profiles were detected in negative ESI mode under selected ion monitoring. For bile acids, the sample was separated on a ACQUITY BEH Amide column (150×2.1 mm, 1.7 μm; Waters) with mobile phases consisting of solvent A (0.1% formic acid) and solvent B (acetonitrile). The column oven temperature was 50°C. The gradient elution program was as follows with a flow rate of 0.3 mL/min: 0-2 minutes, 20% B; 2-10 minutes, 80% B; 10-12 minutes, 80% B; and 12.1-15 minutes, 20% B. Bile acid profiles were detected in negative ESI mode under MRM. For phospholipids, the prepared sample was separated on a Kinetex C8 column (150×2.1 mm, 2.6 μm particle size; Phenomenex) with mobile phases consisting of solvent A (20 mM ammonium formate) and solvent B (acetonitrile/isopropanol [1:1, v/v]). The column oven temperature was 45°C. The gradient elution program was as follows with a flow rate of 0.3 mL/min: 0-1 minutes, 20% B; 1-2 minutes, 40% B; 2-25 minutes, 92.5% B; 25.1-35 minutes, 100% B; and 35.1-38 minutes, 20% B. Phospholipid profiles were detected in positive ESI mode under MRM.

Data processing was performed using the LabSolutions LC-MS software program (Shimadzu).

Peripheral blood was collected in heparin-coated tubes. Plasma levels of thyroid-stimulating hormone (TSH), T3, free thyroxine (T4), and TSH-binding inhibitor immunoglobulin (TBII) were measured by standard methods on an automated analyzer (Cobas 6000, Roche Diagnostics GmbH, Mannheim, Germany).

MetaboAnalyst was used for pathway analysis and performs metabolite set enrichment analysis, which includes a set of human and mammalian metabolites and a set of chemical grade metabolites. This module accepts a list of compound names, a list of compound names with concentrations, or a table of concentrations. The analysis is based on multiple libraries containing a set of approximately 9,000 biologically significant metabolites collected primarily from human studies, including more than 1,500 chemical classes. A portion of the data utilized for analysis could be analyzed because it was matched to the MetaboAnalyst database. The R program’s FactoMineR function was used to perform principal component analysis (PCA).

The analysis method was selected after proving the normalcy of the data in accordance with the distribution of each variable. Using the details of statistical tests included in the plot of ggstatsplot, an extension of ggplot2, violin plots could be created. The Student’s t test (two-sided) was used to compare variables between groups (before and after treatment) after assessing the assumption of normality. The Wilcoxon signed-rank test was used to compare paired variables between groups. Statistical analyses were performed using R software version 4.1.0 (R Foundation for Statistical Computing, Vienna, Austria). P<0.05 was considered significant.

The demographic and clinical characteristics of 18 patients with GD (male 33%, female 67%) at the initial visit are showed in Table 1. Twelve weeks after methimazole treatment, the thyroid hormone levels of all patients were within the normal range and were significantly different from those before treatment (Fig. 1A). The mean free T4, T3, TSH, and TBII levels of GD patients were similar in males and females (Fig. 1B). The age of the study participants was 22 to 64 years. Among all the measured metabolites, only nine plasma metabolites, including uracil, linolenic acid (C18:3), lysophosphatidylcholine (20:5), and cysteine, had positive correlations (P<0.05) with age in drug-naïve patients with GD. On the other hand, only palmitic acid (C16:0) had a negative correlation with age (Fig. 2).

Two hundred metabolites were measured in plasma samples of all participants with GD before and after methimazole treatment (Supplemental Table S1). PCA score plots showed separate clustering before and after 12 weeks of methimazole treatment in patients with GD (Fig. 3A). Creatine, stearic acid (C18:0), and palmitic acid (C16:0) were enriched in plasma of methimazole-naïve patients with GD. On the other hand, many LPC family members, 2-ketoglutaric acid, and nervonic acid (C24:1) were enriched after methimazole treatment (Fig. 3B).

The levels of only 76 plasma metabolites significantly differed before and after methimazole treatment. The levels of 61 metabolites were increased in drug-naïve patients with GD and the levels of 15 metabolites were increased after methimazole treatment. The levels of fatty acids including palmitic acid (C16:0), stearic acid (C18:0), carnitine, and oleic acid (C18:1), and amino acids including dimethylglycine, cysteine, and valine were elevated in patients with GD at the initial visit, but they were remarkably decreased upon methimazole treatment (Fig. 4). By contrast, the levels of nervonic acid (C24:1), LPC family members including LPC (16:0), (18:0), and (18:1), glycine, and creatinine were significantly increased upon methimazole treatment (Fig. 4 and Supplemental Fig. S1).

We observed changes of plasma metabolites before and after methimazole treatment in GD patients. To identify changes in metabolic pathways, we investigated which metabolic pathways differed between the initial visit and 12-week follow-up visit using pathway enrichment analysis. The aminoacyl-tRNA biosynthesis pathway was most significantly altered (Fig. 5). Moreover, various amino acid metabolic pathways, including valine, leucine, and isoleucine biosynthesis and glycine, serine, and threonine metabolism, differed before and after methimazole treatment in patients with GD (Fig. 5). In addition, phenylalanine, tyrosine, and tryptophan biosynthesis- and cysteine and methionine metabolism-related metabolite sets significantly differed between the initial visit and 12-week follow-up visit (Fig. 4B).