Altered urinary tryptophan metabolites in alcohol-associated liver disease

Abstract

Background:

Alcohol-associated liver disease (ALD) leads to millions of deaths worldwide annually. A few potential biomarkers have been discovered through metabolomics or proteomics. Tryptophan (Trp), one of the nine essential amino acids, has been extensively studied and has been shown to play significant roles in many mammalian physiological processes. However, Trp metabolism changes in ALD are not yet fully understood. Urine is an abundant and non-invasive source for disease biomarker discovery. The objective of the current study was to investigate whether the abundance of Trp metabolites in the urine of ALD patients is changed and if these changes in urine can serve as markers for differentiating between stages of ALD.

Method:

We quantified the concentration of Trp and its metabolites in the human urine samples of healthy controls (n = 18), patients with mild or moderate alcohol-related liver injury (n = 21), and patients with severe alcohol-associated hepatitis (AH) (n = 25) using both untargeted and targeted metabolomics.

Results:

Eighteen Trp metabolites were identified and quantified from the untargeted metabolomics data. We then developed a targeted metabolomics method to quantify the Trp and its metabolites. Our developed method can simultaneously quantify 22 Trp metabolites within 15 mins, and 17 metabolites were quantified from the human urine samples. The data acquired in these two platforms agree and show that the Trp concentration is not affected by the severity of ALD. However, the abundance of 10 Trp metabolites is correlated with the model for end-stage liver disease (MELD) score, and the abundance of 9 metabolites have significant abundance changes between healthy control and patient groups.

Conclusion:

Our data demonstrate that Trp metabolism is altered by excessive alcohol consumption even though the concentration of Trp is not affected. Two Trp metabolites, quinolinic acid and indoxyl sulfate, correlate highly with ALD stage.

INTRODUCTION

Metabolic rewiring is a major hallmark of alcohol-associated liver disease (ALD). During the past few years, studies on the metabolic signature of ALD have shown that the metabolism of lipids and polar metabolites is disturbed by chronic and heavy alcohol consumption (Yu et al., 2017, Qiu et al., 2020). Tryptophan (Trp), a polar metabolite, is an essential amino acid for protein synthesis. It is also a biochemical precursor of many microbial and host metabolites that play important roles in mammalian physiology, including gastrointestinal function, immunity, and nervous system function. Disruptions in Trp metabolism result in multiple human diseases, notably neurological, metabolic, infectious, psychiatric, intestinal disorders, and cancers (Fujigaki et al., 2017, Stone, 2020, Agus et al., 2018, Esteban-Zubero et al., 2017, Rothhammer et al., 2016, Trézéguet et al., 2021).

Trp metabolism occurs in the gastrointestinal tract through three major pathways. The first is the kynurenine pathway (KP) in both immune and epithelial cells, which metabolizes over 95% of free Trp to produce metabolites with distinct biological activities in the immune response and neurotransmission (Floc’h et al., 2011, Stone et al., 2013, Van der Goot and Nollen, 2013). The second pathway is the serotonin (5-hydroxytryptamine, 5-HT) production pathway in a specialized intestinal epithelial cell called enterochromaffin cells (ECCs). Trp is metabolized through these two pathways in the host cell. The end metabolites of the KP play important roles. For example, kynurenic acid and picolinic acid have neuroprotective effect, and xanthurenic acid and kynurenine are ligands of the aryl hydrocarbon receptor (AhR). AhR signaling is considered a key component of the immune response at barrier sites. It is thus crucial for intestinal homeostasis by acting on epithelial renewal, barrier integrity, and many immune cell types (Lamas et al., 2018). Importantly, the KP is also the pathway for NAD⁺ de novo synthesis. Quinolinic acid is converted into NAD⁺ via the enzyme, quinolinate phosphoribosyltransferase (QTRP), which is mainly expressed in the liver and kidney, the two major metabolism organs. NAD⁺ is a key coenzyme in energy metabolism, adaptive responses of cells to bioenergetics and oxidative stress, and genome stability. Its deficiency impairs energy generation and, ultimately, can play a role in multiple human diseases (Ralto et al., 2020, Zapata‐Pérez et al., 2021). The serotonin pathway produces the neurotransmitter, 5-HT. It is reported that more than 90% of 5-HT is produced in the gut (Reigstad et al., 2015, Yano et al., 2015, Agus et al., 2018). 5-HT produced in ECC is an important gastrointestinal signaling molecule that conveys signals from the gut to intrinsic or extrinsic neurons and influences intestinal peristalsis and motility, secretion, vasodilatation, and nutrient absorption (Agus et al., 2018). Therefore, 5-HT receptors are always the subjects of clinical investigations for treating gastrointestinal disorders, such as irritable bowel syndrome. In the third pathway, Trp is metabolized in gut microbiota and is transformed into indole and its derivatives. The metabolites produced in this pathway, such as indole-acetic acid, indole-propionic acid, and indole-acrylic acid, are the ligands of the AhR. Several AhR ligands are postulated to reduce ethanol-induced liver injury via induction of interleukin-22 (IL-22) (Hendrikx et al., 2019).

The diverse function of Trp metabolites in neurophysiology and immunology have been extensively studied. Scientific reports also show that Trp metabolites and enzymes involved in Trp metabolism play an important role in non-alcoholic fatty liver disease (Teunis et al., 2022). Indole derivatives induce AhR activation and improve alcohol-induced liver injury (Wrzosek et al., 2021). Even though some studies have evaluated aspects of Trp metabolism in ALD (Manna et al., 2011, Manna et al., 2015, Vidal et al., 2020), the global picture of the Trp metabolism change in ALD remains unclear. It therefore becomes important to reveal what metabolites can be detected as well as their quantities and how relevant these changes can be in the various stages of ALD in the urine of patients with ALD. We collected urine samples from patients exhibiting different stages of ALD and healthy volunteers. We first performed untargeted metabolomics to assess the metabolic profile differences between the representative sample groups. Thereafter, we evaluated targeted metabolomics to investigate the concentration changes of Trp and its metabolites in urine samples.

MATERIALS AND METHODS

Chemical and Reagents

A total of 290 metabolite standards were obtained from Sigma Aldrich (St. Louis, MO, U.S.A), Fisher Scientific (Waltham, MA, U.S.A) and Cayman Chemical (Ann Arbor, MI, U.S.A), including tryptophan and its metabolite standards, 5-hydroxy indole acetic acid (5-HIAA), indole-3-carboxaldehyde, indole-3-acetic acid, indoxyl sulfate, kynurenine, kynurenic acid, 3-hydroxy anthranilic acid (3-HAA), xanthurenic acid, 5-hydroxy tryptophan (5-HT), quinolinic acid, serotonin, melatonin, and tryptamine, indole, indole-3-acetamide, indole-3-propionic acid, indole-3-pyruvic acid, indole-3-acetaldehyde, indole-3-lactic acid, tryptophan, picolinic acid and indole-3-acrylic acid. Ammonium formate was purchased from Sigma Aldrich (St. Louis, MO, U.S.A). Acetonitrile and formic acid were purchased from Fisher Scientific (Waltham, MA, U.S.A). Both solvents were LC-MS grade. The analytical grade water was purified using Millipore Synergy UV system (Burlington, MA, U.S.A).

Stock Solution Preparation

Stock solutions of indole-3-lactic acid (10 mM), picolinic acid (20 mM), 5-HIAA (20 mM), melatonin (10 mM), indole (50 mM), indole-3-acetamide (20 mM), indole-3-acetic acid (10 mM), indole-3-aldehyde (25 mM), indole-3-propionic acid (20 mM), indole-3-acrylic acid (10 mM) and tryptamine (10 mM) were prepared using 100% acetonitrile; Stock solutions of indole-3-acetaldehyde (10 mM), 5-HT (10 mM), tryptophan (40 mM) and serotonin (5 mM) were prepared using acetonitrile/water (50:50, v/v); Stock solutions of indoxyl sulfate (20 mM), kynurenine (10 mM) and quinolinic acid (10 mM) were prepared using 100 % H₂O; Stock solution of xanthurenic acid (10 mM) was prepared using 0.1 mM NaOH; Stock solutions of kynurenic acid (25 mM) and indole-3-pyruvic acid (20 mM) were prepared using 100% dimethyl sulfoxide (DMSO); and Stock solution of 3-HAA (4 mM) was prepared using 100% ethanol. All standards and stock solutions were stored at −80 °C.

Patient Recruitment and Sample Collection

The clinical study was approved by the University of Louisville Institutional Review Board. Information on general patient definitions and recruitment is detailed in other publications (Xu et al., 2023, He et al., 2022). All study participants provided informed consent before the study, including appropriate authorization for data, sample collection and usage of urine samples for this investigation. Eighteen healthy controls and 46 patients with alcohol use disorder (AUD) and varying severity of ALD participated in this study, and their complete clinical history, physical examination, and clinical laboratory evaluation were collected after the study enrollment. The 46 individuals with AUD and ALD were categorized into two groups by the patients’ Model for End-stage Liver Disease (MELD) score. The first group was made up of subjects with non-severe ALD. These patients included those with elevated alanine aminotransferase (ALT), an elevated aspartate aminotransferase (AST), or an elevated total serum bilirubin who were hospitalized in an alcohol detox program. This group also included those with non-severe alcohol-associated hepatitis (AH), MELD < 20. Thus, this first group of patients (n = 21) with non-severe ALD spanned the spectrum from those with minimal liver enzyme abnormalities to moderate AH. The second group was made up of patients with severe AH (severe AH, MELD ≥ 20, n = 25). All patients with the diagnosis of AH met the clinical and laboratory guidelines as published by the NIAAA consortium on alcohol-associated hepatitis (Crabb et al., 2016). Detailed information about the patients and groups is listed in Table 1.

Table 1.

Demographic, drinking and liver indicators of study participants by group

Variables	HC (n=18)	Alcohol-Associated Liver Disease (ALD)				p-value a
		Non-severe ALD (n=21)		Severe AH (n=25)	Total Patients (n=46)
		AUD (n=8)	Moderate AH (n=13)
Age (years)	36 (24–60)	51 (39–67)	50 (34–65)	47 (27–66)	49 (27–67)	-
T. Bilirubin (mg/dL)	0.7 (0.4–1.3)	1.5 (0.8–2.6)	4.2 (1.2–18.2)	12.9 (3.7–34.2)	6.7 (0.8–34.2)	< 0.001
Male (Female)	12 (6)	4 (4)	7 (6)	20 (5)	31 (15)	-
International Normalized Ratio (INR)	N/A	1.2 (1.1–1.7)	1.5 (1.2–2.8)	2.0 (1.0–3.2)	1.7 (1.0–3.2)	-
Body Mass Index (BMI)	N/A	31.1 (27.8 – 33.4)	25.9 (20.6 – 41.1)	29.7 (22.4 – 50.5)	29.7 (20.6 – 50.5)	-
Aspartate Aminotransferase (AST, U/L)	27 (19–66)	59 (21–120)	119 (53–347)	88 (16–190)	90 (16–347)	< 0.001
Alanine Aminotransferase (ALT, U/L)	25 (16–109)	36.8 (14.0–60.0)	48 (18–194)	35 (16–66)	39 (14–194)	0.015
Alkaline phosphatase (IU/L)	52 (37–62)	124 (89–232)	173 (80–518)	144 (71–336)	148 (71–518)	< 0.001
Albumin (g/dL)	4.2 (3.8–4.3)	3.9 (2.6–4.9)	2.8 (1.9–4.5)	2.4 (1.4–4.3)	2.7 (1.4–4.9)	< 0.001
Creatinine (mg/dL)	0.88 (0.69–1.07)	0.69 (0.36–1.40)	0.68 (0.32–1.30)	0.89 (0.39–5.68)	0.79 (0.32–5.68)	0.082
MELD scores	N/A	9.2 (6.0–11.0)	16 (12–19)	24 (20–39)	18 (6–39)	-

Values are presented as mean (ranges). HC, healthy controls; ALD, alcohol-associated liver disease; AUD, alcohol use disorder; AH, alcohol associated hepatitis; MELD, model for end-stage liver disease; AUDIT, alcohol use disorders identification test; MDF, Maddrey’s discriminant function. Patients with non-severe ALD were made up of patients with AUD having some abnormality in liver testing, as well as AUD patients with liver disease advanced enough to meet criteria of moderate AH (12 < MELD < 20). Severe AH had a MELD ≥ 20.

^a:Mann–Whitney U test between healthy controls and patients with ALD/AH.

Each participant’s urine was collected at the research lab in the morning after the overnight fasting, along with the standard of care (SOC) clinical labs. Urine samples were stored at −80 °C until analyzed. All the de-identified data from study participants, who provided urine samples, were collected at baseline, and information on subsequent mortality was also obtained if available. Participant demographics (e.g., age, sex), drinking history, medical assessments at admission, and medical history were included in clinical data. Confirmatory tests for AH (laboratory and imaging) and markers of liver disease severity (MELD score) were collected and analyzed. A SOC laboratory panel specific for this study included, but was not limited to, the routine comprehensive metabolic panel (CMP, including liver injury panel), coagulation measures (PT+INR), and a complete blood count. The SOC labs were collected at the same time when research labs were collected. All samples and data mentioned above were analyzed at the University of Louisville, Louisville, KY, U.S.A.

Untargeted Metabolomics for Polar Metabolic Profiling

Extraction of polar metabolites.

Urine samples were thawed on ice. One hundred μL of urine from each sample was used for polar metabolite extraction. The detailed extraction method used in this study was reported in our previous work (Xu et al., 2023). Group-based pooled samples were also prepared by mixing 10 μL supernatant of each sample in the same group.

Parallel 2DLC-MS.

All samples were analyzed on a parallel two-dimensional liquid chromatography-mass spectrometry (2DLC-MS) system composed of a Thermo DIONEX Ultimate 3000^™ HPLC system coupled with a Thermo Q Exactive HF Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Fisher Scientific, Waltham, MA, U.S.A). The parallel 2DLC system was configured using a SeQuant ZIC^®-cHILIC column (2.1 × 150 mm, 3 μm) (Meck KGaA, Darmstadt, Germany) as the hydrophilic interaction liquid chromatography (HILIC) column and a Waters Acquity UPLC HSS T3 column (2.1 × 150 mm, 1.8 μm) (Milford, MA, U.S.A) as the reversed phase chromatography (RP) column. Both columns were connected with a 2-μL sample loop. The sample was simultaneously injected and separated on two columns. The eluate from the two columns was then mixed and delivered to the MS for measurement. Column compartment temperature was set to 40 °C. The mobile phase for HILIC separation was A: acetonitrile/H₂O (35:65, v/v) with 10 mM ammonium acetate (pH = 3.25, adjusted by acetic acid); B: 100% acetonitrile with 0.1% formic acid. The mobile phase for RP separation was A: 100% H₂O with 0.1% formic acid; and B: 100% acetonitrile with 0.1% formic acid. The flow rate was set to 0.4 mL/min. The parameters for mass spectrometry remained the same as we reported previously (He et al., 2019b). All samples were analyzed in random order under both positive mode ([M+H]⁺) and negative mode ([M-H]⁻). The full MS spectra were acquired for metabolite relative quantification. All pooled samples were analyzed under positive and negative modes to acquire MS/MS spectra using data dependent acquisition (DDA) at three collision energies (20, 40, and 60 eV), where the top six abundant ions in each full MS scan were selected for MS/MS spectrum acquisition. The specific parameters for MS/MS data acquisition were also reported in our previous work (He et al., 2019b).

Targeted Metabolomics for Tryptophan Metabolite Quantification

Tryptophan metabolites extraction.

Two hundred μL of urine was mixed with 800 μL methanol. After 2 min vortexing, the mixture was centrifugated at 14,000 rpm for 20 mins under 4 °C, and 400 μL of supernatant was transferred to a 1.5-mL Eppendorf tube. The supernatant was evaporated under the gentle nitrogen gas flow to remove the methanol and then lyophilized to dryness overnight. The dried extract was reconstituted into 35 μL 20% acetonitrile and transferred to an LC vial for LC-QqQ MS/MS analysis.

LC-QqQ MS/MS analysis.

Tryptophan metabolite quantification was performed using a Waters Acquity H-Class UHPLC system coupled with a Waters Xevo TQ-S micro triple quadrupole mass spectrometer (Waters, Milford, MA, U.S.A). The chromatographic separation was performed using a Waters Acquity UPLC HSS T3 column (2.1 × 150 mm, 1.8 μm). The mobile phase was A: H₂O with 0.1% formic acid (v/v) and B: acetonitrile with 0.1% formic acid (v/v). The flow rate was set to 0.4 mL/min. Column compartment temperature was set to 50 °C. The gradient started at 2% B for 0.3 min (0 – 0.3 min), increased to 10% B at 4 min (0.3 – 4.0 min), rapidly increased to 90% B at 5.5 min (4.0 – 5.5 min), then stayed at 90% B for another 2.5 mins (5.5 – 8.0 min), and finally dropped back to 2% B at 8.1 min to re-equilibrate the column for 7 mins (8.1 – 15.0 min). The total run time was 15 min per injection.

MS analysis was performed under both positive and negative electrospray ionization modes (ESI⁺ and ESI⁻ mode). Multiple reaction monitoring (MRM) was used to achieve the quantification for Trp and its metabolites. The specific MRM transition for each metabolite is listed in Table S1. In ESI⁺ mode, the capillary voltage was set to 3.18 kV. The cone voltages were compound-dependent, ranging from 2 V to 70 V. The collision energy was also compound-specific, ranging from 10 V to 32 V. The details of collision energy and cone voltages are also listed in Table S1. Nitrogen gas was applied on MS as de-solvation gas, and argon gas was used as the collision gas. The de-solvation temperature was 350 °C with de-solvation gas flow of 650 L/hr. In ESI⁻ mode, capillary voltage was changed to 3.37 kV, and the other parameters remained the same as in ESI⁺ mode.

MRM method development and evaluation.

The optimized MRM transitions for each metabolite was determined using three types of samples, including a blank; Trp and individual Trp metabolite; and a mixture of 22 metabolites (Trp and 21 Trp metabolites). Firstly, the top five abundant transitions were identified for each metabolite by direct infusion mass spectrometry using 10 μM authentic standard solution diluted from the stock solutions using 10% acetonitrile. The blank was 10% acetonitrile. The mixture contained 20 μM of each metabolite by diluting the stock solutions of the 22 metabolite standards using 10% acetonitrile. Each Trp metabolite in the mixture was determined by retention time, parent ion m/z, and MS/MS spectrum. The best transition of each metabolite for identification was determined by the transition uniqueness, abundance, and m/z value of the daughter ion. Specifically, the selected transition for each metabolite must be unique at its retention time; the most abundant transition was chosen when multiple unique transitions were provided; the transition with a larger m/z value of daughter ion was selected when multiple transitions had similar abundance.

External calibration.

The external calibration curve for each metabolite was generated using a set of mixtures that contained different concentrations of Trp and the standards of its metabolites with concentrations ranging from 0.02 μM to 100.00 μM. The limit of detection (LOD), lower limit of quantification (LLOQ), and upper limit of quantification (ULOQ) were determined to assess the sensitivity. The LOD of each metabolite was defined as its peak intensity in extracted ion chromatogram with a signal-to-noise ratio (S/N) ≥ 3.0 at its retention time. The peak area versus its concentration was plotted for each metabolite to generate a calibration curve, where the weighting factor was 1/X. The values of LLOQ and ULOQ were set at a concentration where the deviation of a calculated concentration and the true concentration is less than 20% (He et al., 2019b).

Stability.

Stability was assessed using pooled samples. After we extracted the metabolites from urine samples, a quality control (QC) sample was prepared by mixing 10 μL of each sample. The QC sample was injected into LC-QqQ MS between every 10 injections of urine samples. The relative standard deviation (RSD%) of all analytes’ concentrations determined by external standard calibration was used to assess the stability.

Data Analysis

2DLC-MS data for untargeted metabolomic profiling.

2DLC-MS data were first converted into mzXML format using the MSConvert (64-bit) software (Proteowizard, Palo Alto, CA, U.S.A). Spectrum deconvolution was performed using XCMS software (Tautenhahn et al., 2012). The MS/MS spectra of 290 purchased metabolite standards were generated as an in-house database and imported to mzVault software, a built-in component of Compound Discover software (version 3.2, Thermo Fisher Scientific, Inc., Germany). The identification of metabolites was performed using the Compound Discover software as described previously (He et al., 2019a). In brief, the 2DLC-MS/MS data were first matched to the in-house database by parent ion m/z, retention time, and MS/MS spectrum. Then, the spectra that did not match the in-house database were matched to the MS/MS spectra in the Compound Discover software. Identification of metabolites that do not have MS/MS spectra was achieved by matching precursor ion m/z and retention time to the corresponding information of the standards of Trp metabolites. The threshold of spectrum matching was set to m/z variation ≤ 5 ppm, retention time ≤ 9 s, and spectrum similarity score ≥ 0.5. MetSign software was used to process 2DLC-MS data for data alignment and normalization (Wei et al., 2011).

LC-QqQ MS/MS for targeted metabolomics.

MassLynx software (version 4.2, Waters, Milford, MA, U.S.A) was used to acquire and process LC-QqQ MS/MS data. The peak integration and external standard calibration curve were generated using the TargetLynx application, the built-in component of MassLynx software.

Statistical data analysis.

Statistical data analysis was performed using the SPSS software (version 25, IBM Corporation, Armonk, NY, U.S.A) and R (version 4.1.0, https://www.r-project.org/). Distributional assumptions of continuous outcomes were checked. Data transformation (e.g., log-transformation) was applied to meet the normality assumption, if needed. The univariate analysis of metabolite abundance among three study groups was performed using one-way ANOVA with Tuckey’s post-test. Group cross linear-by-linear association test (Mantel-Haenszel test for trend) was used for linear trend analysis. Receiver operating characteristic (ROC) analysis was used to classify patients based on the abundance of metabolites in the urine samples to estimate the true positivity of the effects and corresponding significances. Spearman’s rank correlation was used to measure the association of each metabolite with the clinical parameter MELD score. The thresholds of statistical significance tests were set as one-way ANOVA p ≤ 0.05, areas under the ROC curve (AUC) > 0.7 or < 0.3, Spearman’s rank correlation test coefficient |ρ| ≥ 0.246 (Ramsey, 1989). The error bars in histogram plots are shown as the standard error of the mean (SEM). Heatmap plot was created using OmicStudio (Lyu et al., 2023)

RESULTS

Trp is either metabolized through the host pathways (kynurenine pathway or serotonin pathway) in specific cells, e.g., immune and enterochromaffin cells, or direct transform through the microbial pathway in gut microbiota (Agus et al., 2018). More than 30 metabolites are involved in Trp metabolism. Figure 1 shows the Trp metabolic pathway containing the metabolites detected and quantified in this study.

Figure 1.

Tryptophan metabolic pathway and metabolite’s chemical structures. Metabolites in green represent the Trp metabolites in the host pathway. Metabolites in red represent the Trp metabolites in the microbial pathway.

Untargeted Metabolic Profiling

We categorized the 64 urine samples into three groups, healthy control (HC, n = 18), non-severe ALD (MELD < 20, n = 21), and severe AH (MELD ≥ 20, n = 25). The polar metabolites in the samples were extracted from all urine samples, and then analyzed by 2DLC-MS using untargeted metabolomics. After the spectra deconvolution, more than 8000 isotopic peaks were detected, and 380 were identified using either in-house or commercially available databases. Among the 380 identified metabolites, total 18 were the metabolites in the Trp pathway. Sixteen metabolites were identified by precursor ion m/z, retention time, and MS/MS spectrum matching. Due to low concentration of serotonin and indole-3-propionic acid, the MS/MS spectra of those two compounds were not acquired by DDA, therefore, they were identified by precursor ion m/z and retention time matching.

Figure 2 is the heatmap of the intensities of 18 Trp metabolites quantified by 2DLC-MS, where the relative intensity of each metabolite is normalized by z-score normalization. There is visually a significant trend of disease development, from HC to non-severe ALD to severe AH. For instance, indoxyl and indoxyl sulfate have stepwise decreasing trend of z-scores, indicating the pathway of indoxyl sulfate synthesis from Trp is significantly altered among the disease stages. Specific z-score for each metabolite is listed on supplementary material, Table S8.

Figure 2.

Heatmap of 18 Trp metabolites quantified by untargeted metabolomics. Row clustering uses median log2 intensity after z-score normalization across the sample categories from healthy control (HC), non-severe ALD (N-S) to severe AH (S). Z-score is calculated as z = (x-μ)/σ, where x is the peak area of a metabolite, μ is the mean value of the metabolite in all samples, and σ is the standard deviation.

Figure 3 shows the detailed results of one-way ANOVA, i.e., abundance changes of the 18 Trp metabolites between any two groups. Four Trp metabolites from non-microbial metabolism, kynurenine, anthranilic acid, quinolinic acid and xanthurenic acid, have significant abundance changes (p < 0.05) between HC and patients. The linear-by-linear association analysis indicates that kynurenine and quinolinic acid in the host pathway have a statistically significant increasing trend with increasing disease severity. In contrast, xanthurenic acid and anthranilic acid decrease with increasing ALD severity. In the microbial pathway, Four Trp metabolites have significant abundance changes (p < 0.05), including indole-3-aldehyde, indole-3-lactic acid, indole-3-acryloylglycine, and indoxyl sulfate. The linear-by-linear association analysis shows that indole-3-alydehyde and indole-3-lactic acid significantly increase with ALD severity. However, indole-3-acryloylglycine, and indoxyl sulfate have a significantly decreased trend with ALD severity. To evaluate the diagnostic performance of Trp metabolites, receiver operating characteristic (ROC) analysis shows that indoxyl sulfate and quinolinic acid are significantly changed in both HC vs. non-severe ALD and none-severe ALD vs. severe AH (Table S2). Areas under the ROC curve of indoxyl sulfate and quinolinic acid are 0.759 and 0.310 between HC and non-severe ALD; 0.726 and 0.244 between non-severe ALD and severe AH (Figures 4A and 4B).

Figure 3.

Inter-group abundance of Trp and its metabolites among healthy control (HC), non-severe ALD (N-S) and severe AH (S) in urine analyzed by untargeted and targeted metabolomics. Bar charts in green represent the relative intensity of Trp metabolites analyzed by untargeted metabolomics using 2DLC-MS. Bar charts in orange represent the absolute concentration of Trp metabolites determined by external standard calibration using LC-QqQ MS. The missing bars represent that a metabolite was not detected in the corresponding platform (* p<0.05; ** p <0.01; *** p <0.001 by one way ANOVA with Tuckey’s post test. # p <0.05; ## p <0.01; ### p <0.001 by linear-by-linear association test.)

Figure 4.

Receiver operating characteristic (ROC) analysis results of indoxyl sulfate and quinolinic acid via both untargeted profiling and targeted metabolomics platforms for differentiating healthy control (HC) from non-severe ALD (N S) and severe AH (S) from non-severe ALD. A1: indoxyl sulfate abundance between HC and N-S groups via untargeted profiling. A2: indoxyl sulfate abundance between S and N-S groups via untargeted profiling; B1: quinolinic acid abundance between HC and N-S groups via untargeted profiling; B2: quinolinic acid between S and N-S groups via untargeted profiling; C1: indoxyl sulfate abundance between HC and N-S groups via targeted metabolomics. C2: indoxyl sulfate abundance between S and N-S groups via targeted metabolomics; D1: quinolinic acid abundance between HC and N-S groups via targeted metabolomics; D2: quinolinic acid between S and N-S groups via targeted metabolomics.

Table S3 lists Spearman’s rank correlation coefficient ρ value between metabolite peak area and the patients’ MELD scores. Six metabolites, N’-formyl kynurenine, 5-HIAA, indole-3-aldehyde, indoxyl sulfate, quinolinic acid and kynurenic acid, are correlated with the MELD score with |ρ| > 0.246. Indoxyl sulfate is the only metabolite with a negative correlation ( = −0.52). None of the statistical methods (one-way ANOVA, Spearman’s rank correlation, liner-by-linear association analysis, and ROC analysis) showed that Trp has any significant differences between any two groups, suggesting that Trp is not affected by ALD severity.

Quantification of Trp Metabolites by Targeted Metabolomics

To confirm the results of Trp metabolites detected by untargeted metabolic profiling, we further developed a targeted metabolomics method, LC-triple quadrupole (QqQ) MS using multiple reaction monitoring (MRM). Our method can analyze 22 Trp metabolites in 15 min using the mixture of Trp metabolites prepared from the stock solutions. Figure S1 depicts the extracted ion chromatograms of the 22 Trp metabolites, of which 16 metabolites are detected under the ESI⁺ mode and the other 6 metabolites under ESI⁻ mode. Details about the developed LC-QqQ MS method are provided in the supplemental materials (Text S1).

When we used the developed method to analyze urine samples, only 17 metabolites were quantified from the urine samples. The other metabolites were not detected due to their low concentration in urine. Figure 3 depicts the concentration of the 17 metabolites among the three groups. One-way ANOVA analysis shows that 5-HIAA, kynurenine, picolinic acid, quinolinic acid, and xanthurenic acid have significant abundance changes (p < 0.05) between HC and patient groups (Figure 3). Notably, quinolinic acid can differentiate all three groups, i.e., HC, non-severe ALD, and severe AH. The linear-by-linear association analysis shows the concentration of these metabolites have a clear trend of change with ALD severity, i.e., 5-HIAA, kynurenine, and quinolinic acid have an increasing trend, while picolinic acid and xanthurenic acid have a decreasing trend.

In the microbial pathway, four metabolites, indole-3-aldehyde, indole-3-lactic acid, indoxyl sulfate, and indole-3-acetic acid, have significant concentration differences between HC and patients. However, no metabolites derived from gut microbiota could differentiate HC, non-severe ALD, and severe AH. Linear-by-linear association analysis shows that all four above-mentioned metabolites have a significant trend of abundance change. Indole-3-aldehyde and indole-3-lactic acid increase from HC to severe AH; indoxyl sulfate and indole-3-acetic acid decrease from HC to severe AH. The specific p-values of one-way ANOVA for the detected Trp metabolites are listed in Table S4.

Spearman’s rank correlation analysis shows indoxyl sulfate, indole-3-acetic acid, and picolinic acid, are negatively correlated with MELD score, and indoxyl sulfate has the highest correlation coefficient (ρ = −0.559) (Table S3). Quinolinic acid and serotonin are positively correlated with MELD score, and serotonin has the highest correlation (ρ = 0.433). Spearman’s rank correlation analysis of indoxyl sulfate and quinolinic acid agree with each other in untargeted and targeted metabolomics data.

The ROC analysis also indicates that indoxyl sulfate and quinolinic acid are significantly changed. The AUCs of these two metabolites are 0.714 and 0.108 between HC and non-severe ALD, and 0.828 and 0.259 between non-severe ALD and severe AH groups (Figures 4C and 4D). The concentration of indoxyl sulfate decreases with ALD severity, but quinolinic acid increases with increased ALD severity. The ROC results agree with the result of linear-by-linear association analysis. All results from the four statistical analysis methods show that Trp concentration does not significantly change with increased ALD severity, which agrees with the result of untargeted metabolomics. The AUC values from ROC analysis for all 17 Trp metabolites are listed in Table S5.

DISCUSSION

The purpose of untargeted metabolomics is to detect as many metabolites as possible. We applied a parallel 2DLC-MS to detect metabolites in full scan mode for untargeted metabolites. We then analyzed group-based pooled samples to acquire MS/MS spectra and identified 18 Trp metabolites from the untargeted metabolomics data. Targeted metabolomics analyzes a set of known metabolites, where the separation method and MS data acquisition method are optimized for these metabolites and therefore, the instrument’s sensitivity is increased for detecting the low abundance metabolites. We developed a LC-QqQ MS MRM method for analyzing the Trp and its metabolites, where each metabolite is identified based on its precursor ion m/z, retention time, and daughter ions. Our method can identify 22 Trp metabolites including Trp and we detected 17 in the urine samples. Each analysis method introduces a certain degree of variation to the data. Analyzing the same sample using multiple methods provides information of the accuracy and robustness of the analysis methods. We used two different analysis methods for untargeted metabolomics and targeted metabolomics, i.e., parallel 2DLC-MS vs LC-QqQ MRM MS. It is understandable that some metabolites detected in targeted metabolomics are not detected in the untargeted metabolomics. However, the metabolites detected in both analytical methods agree with each other (Figure 3), indicating good accuracy and stability of our two analysis methods.

Dietary and environmental factors influence the gut microbiome, gut barrier function and the gut:liver axis. Dysbiosis and disruption of the intestinal barrier increases gut barrier permeability. Consequently, this leads to the translocation of microbes and microbial products, such as lipopolysaccharide (LPS) and other metabolites, into the portal bloodstream, directly targeting the liver (Jiang et al., 2015, Knudsen et al., 2019, Ji et al., 2020). Changed tryptophan metabolism and the levels of Trp metabolites have been linked to increased inflammation and fibrosis (Chen et al., 2022). In ALD, the indole pathway metabolites, converted by the gut microbiome, have been shown to improve experimental liver injury through the AhR pathway (Wrzosek et al., 2021), and increased indole-3-lactic acid is considered to be a biomarker for diagnosis of early-stage liver injury in murine models of ALD (Manna et al., 2011, Manna et al., 2015). Similarly, indole metabolites appear to protect against pulmonary infection through AhR activation in alcohol-fed mice (Samuelson et al., 2021). The current study is the first effort to quantify all Trp metabolites from the urine of patients with ALD in order to determine the exact alterations of Trp metabolism in these patients.

Our results show that metabolites in all three Trp metabolism pathways are changed. The main products of the KP are kynurenic acid, picolinic acid, and NAD⁺. Kynurenic acid is synthesized from kynurenine by the enzyme, kynurenine aminotransferase (KAT). In contrast, the production of kynurenine is controlled by the two rate-limiting enzymes, indoleamine 2,3-dioxygenase (IDO) and tryptophan 2,3-dioxygenase (TDO), which are expressed in a tissue-specific manner. KAT also is the enzyme that converts 3-hydroxykynurenine to xanthurenic acid. In addition to kynurenic and xanthurenic acid, picolinic acid is another neuroprotective molecule in KP that is also decreased in the urine of ALD patients. Amino-ß-carboxymuconate-semialdehyde-decarboxylase (ACMSD) is the rate-limiting enzyme of picolinic acid production. Therefore, increased kynurenine and decreased kynurenic, xanthurenic, and picolinic acids detected in this study suggest the expression or the activity of KAT and ACMSD may be changed in patients with ALD. The KP is organized with at least four intersections, and the flux through various routes is unequal. That means that the production of different end-products of KP is not identical, and the balance between them can change (Modoux et al., 2021). The levels of kynurenic, xanthurenic, and picolinic acids are decreased, but the level of kynurenine is increased in ALD patients. The remaining kynurenine may be used for quinolinic acid synthesis. Interestingly, we indeed observed an increased level of quinolinic acid in patients with ALD in this study. Quinolinic acid is the precursor of NAD⁺ de novo synthesis; unfortunately, we did not detect NAD⁺ in the current study. Therefore, we do not know whether the level of NAD⁺ is increased. Quinolinic acid is a neurotoxic molecule, and the increase of quinolinic acid in patients with ALD suggests that the damaged liver function cannot remove it, or that the conversion of quinolinic acid is inhibited. The elevation of quinolinic acid may be induced by decreased expression or activity of QTRP. Our results also show that the variation of quinolinic acid could differentiate non-severe ALD from HC and severe AH from non-severe ALD.

5-Hydroxy-indoleacetic acid (5-HIAA) is uniquely the only metabolite that is significantly changed in the serotonin pathway, with a noticeable increase in patients with severe AH. As the end product of a neurotransmitter, 5-HIAA is increased in the brains of patients with infantile hydrocephalus (Gopal et al., 2008). Urinary 5-HIAA is used as a proxy for serotonin measurement (Lenchner and Santos, 2019). The increased 5-HIAA indicates that the metabolism of the neurotransmitter is increased in ALD patients.

In addition to the changes of metabolites in host pathways, the variations in Trp metabolism in microbiota are also significant. Indole-3-acetaldehyde, indoleacetic acid, indole 3-aldehyde, and tryptamine are ligands of AhR and are, therefore, involved in intestinal permeability, regulation of inflammation, and host immunity. Both indole 3-aldehyde and indoleacetic acid are ligands of AhR. Our data show that indole 3-aldehyde is significantly increased, but indoleacetic acid is notably decreased. The change in gut microbiota may be the reason for the different trends of these two metabolites. Indole-3-lactic acid is an anti-inflammatory molecule, and the increase of this metabolite in patients with ALD may demonstrate a compensatory attempt to attenuate gut inflammation. In addition, Trp is converted into indole in the gut, and indole is transferred to the liver, where it is metabolized into indoxyl sulfate. The stepwise reduction in indoxyl sulfate is closely correlated with ALD disease stage. ROC analysis shows that this metabolite could differentiate non-severe ALD from HC and severe AH from non-severe ALD. Indoxyl sulfate is an indole metabolite produced by bacteria expressing tryptophanase and has been suggested to help regulate stability of bacterial communities. In bone marrow transplant patients, low levels of urinary 3-indoxyl sulfate were associated with higher transplant-related mortality and worse overall survival (Weber et al., 2015). Low urinary levels were postulated to be a marker for gut dysbiosis. In patients in the ICU, low 3-indoxyl sulfate at 72 hours was associated with fewer non-ICU days, more deaths during one year of follow-up, and lower bacterial diversity, as assessed by the Shannon Index (Kuo et al., 2021). These studies are consistent with our findings of more severe disease with lower urinary indoxyl sulfate levels. While we did not measure bacterial diversity, patients with ALD frequently have gut bacterial dysbiosis and lower bacterial diversity (Li et al., 2019, Smirnova et al., 2020).

This is a small study and thus, the results should be carefully evaluated for reproducibility and large clinical studies should validate the results, while simultaneously identifying the role of metabolites that remained inconclusive in our study. As such, since the effect sizes are formidable, the results should adhere consistently in larger studies. Many neurotransmitters were observed in urine and it was an indicator that liver injury could be instrumental in the changes in their metabolism that are observed in urine. However, their role in neurological function also needs to be ascertained. We did not study the role of sex-dimorphism and this is not in the scope of this clinical investigation. Multimodel and multivariable analyses were also not used due to the small sample size.

In summary, our untargeted and targeted metabolomics data correspond well. They all show that alcohol consumption leads to changes in Trp metabolism, and notably, some Trp metabolites are closely correlated with ALD disease and its development. Most importantly, the increase of quinolinic acid and the decrease of indoxyl sulfate could differentiate the different stages of ALD, i.e., HC, non-severe ALD, and severe AH; thereby indicating that the abundance of these two metabolites could closely predict the course of disease development. This study justifies the use of urine (an abundant and non-invasive source) in diagnostic testing in ALD patients, especially those with more advanced disease.