This case-control study was conducted at the Clinic of Endocrinology and Metabolic Diseases, St George University Hospital in Plovdiv - a 1500-bed university-affiliated hospital serving urban and nonurban populations of approximately 1 million as a first-line and tertiary facility. The inclusion criteria for the case patients were as follows: 1) patients were recruited from the population of patients with newly diagnosed Hashimoto’s thyroiditis, who were referred to the Clinic of Endocrinology and Metabolic Diseases; 2) patients were aged 18 years or older at the time of diagnosis. The control group, with no history of HT, was also recruited in that clinic. Participants with a prior history of thyroid surgery, cognitive impairment, or inability to read or understand or who refused to sign the informed consent were excluded from the study. Patients who had already been diagnosed with thyroid disorders or who had already been treated for thyroid dysfunction were also excluded. To obtain more accurate data from initially selected patients, those with additional diseases of the cardiovascular, excretory, and digestive systems, with diabetes, or with other autoimmune diseases in which elevated TMAO values are expected were excluded. A total of 69 subjects (62 patients with HT and 7 healthy controls) agreed to participate. Body mass index (BMI) was calculated for each participant. Comorbidities, current medication consumption, and family medical history were assessed through medical records and interviews. The diagnosis of HT was based on medical history, symptoms, physical examination, laboratory tests, and ultrasound examination of the gland by a professional endocrinologist. Thyroid status (hyper-, hypo-, or euthyroid) was evaluated by thyrotropin (TSH) (normal 0.34 mIU/L–5.1 mIU/L), free thyroxine (fT4) (normal 11–23 pmol/L), and free triiodothyronine (fT3) (normal 3.8 pmol/L–6.3 pmol/L). The study was reviewed and approved by the Medical University of Plovdiv’s Scientific Ethics Committee (No. 4/04.05.2023) in accordance with the Helsinki Declaration. Informed consent was obtained from each patient.

TMAO (C₃H₉NO, according to the certificate of analysis with 100.0% purity) and its deuterated analogue (TMAO-d9 or C₃D₉NO, according to the certificate of analysis with 99.0% purity) were obtained from Cayman Chemicals (Ann Arbor, MI, USA). LC-MS-grade acetonitrile (ACN) and 98% formic acid were sourced from Honeywell (Charlotte, USA), while ammonium formate was supplied by Sigma-Aldrich (St. Louis, MI, USA). Deionized water in the lab was generated using the ELGA Veolia Chorus system (ELGA Lab Water, UK). The stock solutions of TMAO (1.0 g/L), TMAO-d9 (1.0 g/L), and the working internal standard solution of TMAO-d9 (3600 µg/L) were prepared in 80% ACN.

TMAO working solutions were prepared in 80% ACN at concentrations of 2.5, 4.0, 25.0, 75.0, 100.0, 125.0, and 150.0 mg/L for generating the calibration curve (CC) samples. Additionally, solutions with concentrations of 2.5, 7.5, 87.5, and 140 mg/L were used to prepare the lower limit of quantification (LLOQ) and quality control (QC) samples at three levels.

The CC and QC samples were prepared using plasma from healthy individuals. The plasma was separated and pooled immediately after centrifugation. Ten microliters of the corresponding CC and LLQC/QC working solutions were added to 490 microliters of plasma pool, mixed gently for 5 minutes, and frozen at −20°C. The resulting concentrations of CC and QC samples were 50.0, 80.0, 500.0, 1000.0, 1500, 2000, 2500, and 3000 µg/L; LLOQ/QC LOW, MID, and HIGH levels were 50 / 150, 1750, and 2800.0 µg/L.

In the sample preparation process, we utilized protein precipitation along with a sample dilution procedure. In a 2 mL microcentrifuge tube, 50 μL of human plasma (CC, QC, or patient sample) was combined with 10 μL of internal standard (IS) solution and gently vortexed for 5 minutes. Next, 440 μL of pure ACN was added, followed by vigorous vortex mixing for 6 minutes. After centrifugation at 12,500 g for 10 minutes, 50 μL of the upper organic layer was collected and transferred to a separate tube for vacuum evaporation at 45°C. In the final step, dry residue was re-dissolved in 400 µL of mobile phase (A) and injected for analysis.

The analysis was performed using a Thermo Ultimate 3000 chromatographic system coupled with a triple quadrupole mass spectrometer TSQ Quantum Access Max (Thermo Fisher Scientific, MA, USA). Chromatographic separation was performed using gradient elution on a core-shell Accucore™ HILIC analytical column (50×2.1 mm, 2.6 µm particles) from Thermo Fisher Scientific (MA, USA) with a flow rate of 0.35 mL/min. The mobile phase used was (A)—30 mM ammonium formate in 85% ACN and (C)—30 mM ammonium formate adjusted to pH 3 with formic acid. The gradient mode was as follows: up to 2.8 minutes (C) increased from 0 to 50%, held constant for 0.8 minutes, and then decreased to 0% in 0.4 minutes. The column was then reconditioned with the initial phase for 6 minutes. Detection was performed using heated electrospray ionization (HESI) in positive ionization mode, with a spray voltage of 3500 V, a vaporizer temperature of 300°C, a sheath gas setting of 40 arbitrary units, and a capillary temperature of 270°C. Ion sweep and aux gas flows were not used. The protonated molecules of the analyte and IS served as precursor ions for selected reaction monitoring (SRM), with transitions of m/z 76.08 → 58.2 for TMAO and 85.01 → 68.3 for TMAO-d9. The collision gas was argon; the collision energy was 15 V. Concentrations were determined using the background subtraction method.

The method’s selectivity was evaluated using 10 different individual human plasma matrices, employing the standard addition technique at two concentration levels (150 μg/L and 2800 μg/L) while ensuring a predefined normalized matrix effect within the range of 85%–115%. The Limit of Detection and Lower Limit of Quantification were set at 10 µg/L and 50 µg/L, respectively. Intra-day precision and accuracy were evaluated by analyzing five replicates at four concentration levels—LLOQ, LOW, MID, and HIGH. QC sample results were required to fall within the 85%–115% range for both within-run and between-run analyses, while LLOQ samples had to remain within ±20%. Between-day accuracy and precision were assessed using four sample replicates analyzed over four separate days at the same concentration levels, with acceptable variation set at ±20% for LLOQ samples and ±15% for all other samples. Post-preparative stability was assessed under various storage conditions, including 12 and 24 hours at room temperature in daylight and 12 and 24 hours at 4°C in the dark. Linearity was assessed in the defined CC range, with R²>0.996. The stability of the stock solution was assessed along with the short-term stability of working solutions at room temperature for 72 hours under both dark and light conditions, freeze-thaw stability over three cycles of 24 hours each, and long-term stability in plasma for over three months at −20°C, ensuring all results remained within 15% of theoretical values. Validation experiments were conducted in accordance with the current EMA/FDA industrial guidelines for LC-MS/MS bioanalysis, consisting of four consecutive analytical runs over four working days. Precision and accuracy were evaluated using a separate calibration curve (CC) for each run, with five replicates of the LLOQ and QC samples analyzed on the first day, followed by duplicate analyses of LLOQ and QC samples over the next three days. An additional set of experiments was conducted to confirm the method’s selectivity, matrix effect, and stability.

Numerical data are expressed as the median and range (minimum-maximum), while categorical variables are reported as counts and percentages. A nonparametric approach was applied, as most numerical variables did not follow a normal distribution, confirmed by the Kolmogorov-Smirnov test. To assess the existence of significant differences between HT subjects and healthy controls, the Mann-Whitney test (for numerical parameters) and the chi-square test or Fisher’s exact test were applied as appropriate (for categorical variables). Statistical analyses were conducted using SPSS 22.0 for Windows. A p-value below 0.05 was regarded as statistically significant.

Regardless of the shorter length of the chromatographic column used, our method lasts 10 minutes. We tried injecting a new sample after a shorter conditioning period but found instability and large analyte retention time variation. This is due to the slower kinetics of the HILIC mechanism compared to RPLC and the need for more time for column conditioning. The developed LC-MS/MS method for TMAO determination in blood plasma has linearity in the range of 50 μg/L–3000 μg/L with an R²=0.9959, and calculated accuracy (within-run min = 84%, max = 98.5%; between-run min = 84.0%, max = 112.7%) and precision (within-run = 2%–4.7%, and between-run = 3.3%–8.3%) were also within acceptable limits. The calculated normalized matrix effect, assessed using 10 individual plasma samples, ranged from 87.1% to 111.9%. Post-preparative stability at room temperature and at 4-8ºC was within the limits of −6.1% to 10%, while freeze/thaw stability ranged from 1.9% to 5.1%. Figs 1A and 1B show the mass spectrogram of the TMAO calibration solution with concentrations of 50 μg/L and 3000 μg/L. The mass spectrograms derived from the analysis of the remaining calibration and control levels are displayed in the Supplementary material.

The study involved 62 patients with autoimmune Hashimoto’s thyroiditis with a mean age of 36.9 years (min=18, max=82). Among these patients, 83.9% were women and 16.1% were men. Seven control subjects were also included. The mean fT3 and fT4 values in patients (4.88 pmol/L and 12.08 pmol/L) were lower than those in the control group (5.6 pmol/L and 14.9 pmol/L), and the differences were statistically significant (p=0.24 and p=0.07, respectively). The mean value of TMAO in patients with HT was higher than that in the control group (172.5 μg/L vs. 158.5 μg/L), but the difference was not statistically significant (p=0.800). The mean value of BMI of the patients with HT was 27.5. Based on this index, six groups were created. Notably, higher TMAO values were observed not only in overweight patients and those with varying degrees of obesity but also in underweight patients (Fig. 2). We found a positive correlation between BMI and TMAO concentration (r=0.380, p=0.002).

Participants were categorized based on the functional status of their thyroid gland, including those with hypofunction, those with a tendency to develop hypofunction, and euthyroid individuals. A comparative analysis was conducted on the mean values of TMAO across the three experimental groups and the control group.

The highest mean values (204.5 μg/L) were observed in hypothyroid patients, compared with those with a tendency to develop hypofunction (172.7 μg/L), euthyroid patients (161.9 μg/L), or healthy controls (119.8 μg/L) (Fig. 3). The patients were also examined for the presence of autoantibodies. Mildly elevated (<100 IU/mL) anti-TPO antibodies were found in 32.3% of the patients, of whom 3% had both anti-TPO and anti-TG autoantibodies. Nearly 39% of the patients had moderate elevations (100 IU/mL-500 IU/mL), with 16% having both antibodies. High levels (>500 IU/mL) were found in 29%, from which 13% had both autoantibodies elevated. None of the controls had elevated autoantibodies. Higher mean values of TMAO (206.7 µg/L vs. 150.5 µg/L) were found in the group with anti-TPO antibodies (Fig. 4).

We established a positive association between circulating TMAO and anti-TPO antibodies’ values (r=0.264, p=0.042).