Serum and Urinary Metabolomics Reflect the Early Stages of De Novo Metabolic Syndrome After Liver Transplant: A 2-Center Longitudinal Study

Abstract

INTRODUCTION:

Liver transplantation (LT) recipients are at high risk of developing de novo metabolic syndrome (MetS), which contributes to cardiovascular and cerebrovascular morbidity. This study investigated serum and urinary metabolic changes after LT to identify microbial and metabolic markers associated with MetS development.

METHODS:

We conducted a prospective, 2-center longitudinal study with biospecimen collection pre-LT and at 6 months, 1 year, and 2–9 years post-LT. Nuclear magnetic resonance spectroscopy was used to characterize serum and urine metabolomic profiles from 73 to 44 patients, respectively. MetS was defined as body mass index >30 kg/m² plus at least 1 additional metabolic abnormality.

RESULTS:

MetS prevalence increased from 11% pre-LT to 36% post-LT. Post-LT, serum metabolite profiles showed increased phosphocholines and lipid-CH₃ (low density lipoprotein), whereas urine profiles demonstrated higher levels of trimethylamine-N-oxide (TMAO) and phenylacetylglutamine. Patients who developed or had persistent MetS exhibited smaller increases in serum phosphocholines and lipid-CH₃ but greater elevations in urinary TMAO levels compared with patients who remained MetS-free.

DISCUSSION:

LT is followed by distinct metabolic shifts reflecting changes in both hepatic lipid metabolism and gut–liver microbial cometabolism. Elevated urinary TMAO, together with reduced serum phosphocholine and lipid-CH₃ responses, characterize patients who develop post-LT MetS and may serve as early biomarkers of cardiometabolic risk in LT recipients.

INTRODUCTION

Liver transplantation (LT) recipients are at risk of developing or experiencing worsening metabolic syndrome (MetS) within months after the procedure (1,2). Post-LT MetS is clinically important because it is associated with increased cardiovascular and cerebrovascular events (3), yet the underlying pathways that lead to its development remain incompletely understood. A clearer understanding of these pathways is needed to improve long-term metabolic outcomes in LT recipients (4).

Although immunosuppression, recovery-related reductions in physical activity, and lifestyle factors are frequently implicated, emerging work suggests that changes in host–microbiota interactions may also contribute to post-LT metabolic disturbances. Several microbially associated metabolites—including trimethylamine-N-oxide (TMAO) and phenylacetylglutamine (PAG)—have been linked to adverse cardiometabolic outcomes in nontransplant populations (5,6). These metabolites arise from gut microbial processing of dietary precursors followed by hepatic conversion, and their circulating levels therefore reflect combined microbial and hepatic metabolic activity. Whether similar alterations occur after LT, when both liver function and the gut–liver axis are undergoing substantial physiological adaptation, remains insufficiently characterized (7).

Previous single-center observations indicate that TMAO and PAG levels may increase after LT (8), but these findings require validation in larger and more diverse cohorts with long-term follow-up. Additional evidence from broader metabolomics studies supports the concept that metabolic signatures—including lipid-related and microbially derived metabolites—can reflect cardiometabolic risk and gut–liver interactions (9,10). Studies in cirrhosis and population-based cohorts further show that plasma metabolomic profiles are strongly linked to gut microbial composition (11), and that microbial, dietary, and genetic factors jointly shape inter-individual differences in circulating metabolites (12). Stool microbiota studies in chronic liver disease similarly demonstrate strong concordance between microbial taxa and plasma metabolite patterns across international cohorts (13).

Moreover, metabolomic approaches have revealed distinct metabolite patterns in cardiovascular disease, myocardial infarction, and unstable angina, offering mechanistic insights and potential biomarkers for risk stratification (14,15). Such observations underscore the broader relevance of metabolomics in detecting clinically meaningful metabolic perturbations. Urine metabolomics has likewise been shown to capture metabolic signatures associated with cardiometabolic phenotypes (5) and nuclear magnetic resonance (NMR)-based metabolomics has successfully predicted metabolic disorders in large community cohorts (16,17). Taken together, these findings provide a strong rationale for applying metabolomic profiling after LT.

This study aimed to characterize longitudinal changes in key microbial and metabolic biomarkers in serum and urine after LT and to evaluate their relationship with MetS development. We hypothesized that NMR-derived metabolic profiles would differ between pre-LT and post-LT states and that specific metabolites might be associated with the development or worsening of MetS.

METHODS

Study approval

All participants provided written informed consent. Study protocols were approved by the Institutional Review Boards of each participating institution and conducted in accordance with the Declaration of Helsinki.

Clinical cohort

Patients were consecutively enrolled from 2 transplant centers between June 9, 2011, and March 11, 2022, into a prospective biospecimen cohort. Samples were obtained before LT (mean 11 ± 6 months pre-LT) and at prespecified follow-up visits 6 months and annually after LT. Clinical data were collected longitudinally, and post-LT outcomes were assessed through chart review between January 11 and November 1, 2024. The date of chart review or death served as the study end point. Eligible participants were clinically stable, had not undergone retransplantation, and had no episodes of acute rejection or clinically significant infections in the preceding 6 months.

Definition of MetS

MetS was defined using an adapted version of the National Cholesterol Education Program—Adult Treatment Panel III criteria based on the presence of ≥3 of the following 4 components:

1. Central obesity: body mass index (BMI) >30 kg/m²

2. Hypertension: blood pressure ≥130/85 mm Hg or antihypertensive medication

3. Insulin resistance: fasting plasma glucose >100 mg/dL, hemoglobin A1c >5.7%, or antidiabetic treatment

4. Dyslipidemia: non-high density lipoprotein (HDL) cholesterol >130 mg/dL or lipid-lowering therapy

Because waist circumference, triglycerides, and HDL cholesterol were unavailable, we used BMI >30 kg/m² as a proxy for central obesity and non-HDL cholesterol >130 mg/dL as an indicator of atherogenic dyslipidemia. BMI >30 kg/m² is recognized in the International Diabetes Federation consensus definition as sufficient to assume central obesity when waist circumference is missing, and non-HDL cholesterol is recommended by major lipid guidelines as an alternative marker of cardiometabolic risk when fasting triglycerides or HDL-C are unavailable. These adapted criteria were prespecified and applied uniformly across all participants at baseline and at the end-of-study chart review.

For transition analyses, patients were categorized as persistent MetS (Y–Y), resolved MetS (Y–N), de novo MetS (N–Y), or no MetS (N–N). MetS status, death, and rejection events were determined during the 2024 chart review.

Sample collection

Fasted serum samples were collected at enrollment, 6 months post-LT, and annually thereafter. Urine samples (fasted, first morning void) were collected only at VCU at identical time points. All samples were aliquoted into 2 mL vials and stored at −80°C until batch preparation for NMR.

NMR spectroscopy studies

Aliquots of phosphate-buffered serum (300 μL) and urine (540 μL) were analyzed using standardized proton NMR spectroscopy protocols on a 600 MHz platform. This approach has been widely applied in cardiometabolic and hepatic research, including large-scale population metabolomics studies (6,16) and cirrhosis-focused microbial–metabolite analyses (11,13).

NMR spectra were processed initially using the NMR platform followed by the KnowItAll Informatics System, Metabolomics Edition v17.0 (Wiley Science Solutions, Hoboken, NJ). Regions corresponding to predefined metabolites were manually selected, whereas remaining spectral regions were automatically segmented using the IntelliBucket algorithm for consistent bin generation. Processed NMR spectra were normalized before multivariate analyses.

Univariate and multivariate statistical analysis

Metabolite levels were compared pre-LT and post-LT with and without MetS using t tests with Benjamini-Hochberg correction for multiple comparisons. Group comparisons were performed using partial least squares–discriminant analysis (PLS-DA), a supervised multivariate method commonly applied in metabolomics for discrimination of phenotype groups and biomarker discovery (18,19). Model quality was assessed by 5-fold cross-validation and expressed as Q². Metabolites contributing most to group separation were identified from model loadings. All analyses followed established chemometric workflows in MetaboAnalyst 6.0 (20).

RESULTS

Clinical cohort

A total of 73 patients were included in the serum NMR analyses, and 44 patients from 1 center contributed urine samples. Cohort characteristics are shown in Tables 1 and 2. All included patients received deceased-donor LT, with a mean waiting time of 11 ± 6 months from enrollment. Serum was available from all 73 participants at baseline, from 47 (64%) at 6 months, and from 33 (45%) at later than 1 year; urine samples were available from all 44 participants at baseline, from 37 (84%) at 6 months, and from 20 (46%) at 1 year or beyond (Supplementary Figure 1, Supplementary Digital Content 1, http://links.lww.com/CTG/B445).

Table 1.

Clinical details of LT subjects for serum NMR studies

Variable	All (N = 73)	Post-LT MetS (N = 27)a	Post-LT no-MetS (N = 46)	P value
Age, yr, mean (SD, range)	62.3 (9.68, 29–79)	63.9 (8.31, 48–79)	61.4 (10.33, 29–78)	0.285§
Sex, female/male, n (%)	54 (74)/19 (26)	21 (78)/6 (22)	33 (72)/13 (28)	0.382€
Ethnicity, Hispanic/non-Hispanic, n (%)	10 (14)/63 (86)	4 (15)/23 (85)	6 (13)/40 (87)	0.515€
Race, Caucasian/African–American/other, n (%)	61 (83)/8 (12)/4 (5)	22 (82)/2 (7)/3 (11)	39 (85)/6 (13)/1 (2)	0.152€
Cirrhosis etiology before LT, HCV/alcohol/HCV + Alc/MASLD/other, n (%)	36 (49)/10 (14)/7 (9)/9 (12)/11 (16)	14 (52)/3 (11)/3 (11)/3 (11)/4 (15)	22 (48)/7 (15)/4 (9)/6 (13)/7 (15)	0.276€
MELD at transplant, mean (SD, range)	23.2 (6.9, 12–45)	25.6 (7.2, 16–45)	22.0 (6.5, 12–40)	0.078§
HCC at LT	34 (47)	5 (19)	7 (15)	1.000€
BMI, kg/m2, mean (SD, range)	30.6 (5.2, 21–43)	32.9 (4.6, 25–43)	28.9 (5.1, 21–40)	0.0042§
Obese (BMI >30 kg/m2), yes, n (%)	38 (52)	22 (81)	17 (37)	<0.0001€
BP >85 diastolic or >130 mm Hg or pharmacological treatment of hypertensionb	35 (48)	22 (81)	13 (28)	<0.0001€
Fasting blood glucose >100 mg/dL or HbA1c >5.7% or on glucose lowering treatmentb	29 (39)	22 (81)	7 (15)	<0.0001€

Bold values represent statistically significant. P value: post-LT MetS vs post-LT no-MetS; § Student t test; € χ² test. Percentages are calculated within each column.

Alc, alcohol; BMI, body mass index; BP, blood pressure; HbA1c, hemoglobin A1c; HCC, hepatocellular cancer; HCV, hepatitis C virus; HDL, high density lipoprotein; LT, liver transplantation; MASLD, metabolic dysfunction-associated steatotic liver disease; MELD, model for end-stage liver disease; MetS, metabolic syndrome.

^aDefinition of post-LT MetS: presence of ≥3 metabolic abnormalities according to adapted ATP III–style criteria: BMI >30 kg/m² (proxy for central obesity), elevated blood pressure (≥130/85 mm Hg) or antihypertensive treatment, elevated fasting glucose (>100 mg/dL), HbA1c >5.7% or antidiabetic treatment, and dyslipidemia defined as non-HDL cholesterol >130 mg/dL or lipid-lowering treatment.

^bHypertension is defined as BP >85 diastolic or >130 mm Hg or pharmacological treatment of hypertension. Type 2 diabetes is defined as fasting blood glucose >100 mg/dL or HbA1c >5.7% or on glucose lowering treatment.

Table 2.

Clinical details of LT subjects for urinary NMR studies

Variable	All (N = 44)	Post-LT MetS (N = 16)a	Post-LT no-MetS (N = 28)	P value
Age, yr, mean (SD, range)	64.6 (9.3, 42–79)	68.4 (6.4, 59–79)	62.4 (10.0, 42–78)	0.037§
Sex, female/male, n (%)	35 (80)/9 (20)	15 (94)/1 (6)	20 (72)/8 (28)	0.186€
Ethnicity, Hispanic/non-Hispanic, n (%)	3 (7)/41 (93)	0/16 (100)	3 (11)/25 (89)	0.479€
Race, Caucasian/African American/other, n (%)	36 (81)/8 (19)/0	15 (94)/1 (6)/0	21 (75)/7 (25)/0	0.273€
Cirrhosis etiology before LT, HCV/alcohol/HCV + Alc/MASLD/other, n (%)	12 (27)/8 (18)/6 (14)/8 (18)/10 (23)	5 (31)/3 (19)/3 (19)/3 (19)/2 (12)	7 (24)/5 (17)/3 (10)/5 (17)/8 (31)	0.696€
MELD at transplant, mean (SD, range)	22.6 (6.2, 12–40)	24.4 (5.3, 16–37)	21.6 (6.5, 12–40)	0.146§
HCC at LT	12 (50)	5 (31)	7 (25)	0.855€
BMI, kg/m2, mean (SD, range)	31.2 (4.5, 21–40)	32.4 (3.7, 28–40)	30.4 (5.0, 21–37)	0.234§
Obese (BMI >30 kg/m2), yes, n (%)	21 (47)	11 (68)	10 (35)	<0.0001€
Hypertensionb	19 (43)	12 (75)	7 (25)	<0.0001€
Type 2 diabetesb	16 (36)	13 (81)	3 (10)	<0.0001€

Bold values represent statistically significant. P value: post-LT MetS vs post-LT no-MetS; § Student t test; € χ² test. Percentages are calculated within each column.

Alc, alcohol; BMI, body mass index; BP, blood pressure; HbA1c, hemoglobin A1c; HCC, hepatocellular cancer; HCV, hepatitis C virus; HDL, high density lipoprotein; LT, liver transplantation; MASLD, metabolic dysfunction-associated steatotic liver disease; MELD, model for end-stage liver disease; MetS, metabolic syndrome.

^aDefinition of post-LT MetS: presence of ≥3 metabolic abnormalities according to adapted ATP III–style criteria: BMI >30 kg/m² (proxy for central obesity), elevated blood pressure (≥130/85 mm Hg) or antihypertensive treatment, elevated fasting glucose (>100 mg/dL), HbA1c >5.7% or antidiabetic treatment, and dyslipidemia defined as non-HDL cholesterol >130 mg/dL or lipid-lowering treatment.

^bHypertension is defined as BP >85 diastolic or >130 mm Hg or pharmacological treatment of hypertension. Type 2 diabetes is defined as fasting blood glucose >100 mg/dL or HbA1c >5.7% or on glucose lowering treatment.

At baseline, 8 of 73 patients (11%) fulfilled criteria for MetS, increasing to 26 of 73 (36%) after LT. Based on transitions, 41 patients had no MetS either pre-LT or post-LT (N–N), 24 developed de novo MetS (N–Y), 6 improved from pre-LT MetS (Y–N), and 2 had persistent MetS (Y–Y). In the urine NMR subgroup, 16 had MetS post-LT and 28 did not.

Immunosuppressant use after LT

Patterns of immunosuppression were similar between groups. Six months post-LT, antimetabolite use was comparable in MetS and non-MetS patients (63.2% and 61.8%, respectively) and declined uniformly to 47.4% and 47.1% during follow-up, without group-specific differences (P > 0.5). Calcineurin inhibitor use remained nearly universal throughout the study and was therefore not evaluated statistically. mTOR inhibitor use was infrequent in both MetS (10.5%–5.3%) and non-MetS patients (11.8%–8.8%), without meaningful temporal changes (P = 1.00 for both). Corticosteroid exposure declined substantially after LT. Among non-MetS patients, usage fell from 58.8% at baseline to 20.6% at follow-up (P = 0.0026), whereas a similar but nonsignificant reduction occurred in the MetS group (from 42.1% to 10.5%, P = 0.0625).

Serum NMR results

Comparison of pre-LT and post-LT serum samples demonstrated marked metabolic shifts and differences associated with MetS status. Across the entire cohort, serum metabolic profiles differed significantly between pre-LT and post-LT (PLS-DA, 5-fold cross-validation, Q² = 0.61; Supplementary Figure 2A, Supplementary Digital Content 1, http://links.lww.com/CTG/B445). Post-LT profiles were characterized by increased signals corresponding to lipid-CH₂ (1.31–1.22 ppm), lipid-CH₃ associated with low density lipoprotein (LDL)/very low density lipoprotein (VLDL) (0.92–0.80 ppm), and phosphocholines (3.22–3.20 ppm), all reflecting alterations in circulating lipids and membrane-related metabolites. Concurrent decreases were observed in formate, acetate, tyrosine, phenylalanine, glycerol-phosphocholines, and lactate, representing changes in small organic acids, amino acids, and energy-related intermediates.

When stratified by MetS status, the principal differences between groups involved phosphocholines and lipid-CH₃ (LDL), metabolites central to phosphatidylcholine turnover. These signals increased after LT but did so to a lesser extent in patients who developed or had persistent MetS. Differences between MetS and non-MetS patients became most evident beyond 1 year post-LT, where post-LT phosphocholine and lipid-CH₃ levels were significantly lower in MetS patients (P < 0.01; Figure 1a, b). Multivariate modeling supported these findings: serum profiles from MetS patients pre-LT vs post-LT yielding discriminatory performance (Q² = 0.66); lipid-CH₂ and lipid-CH₃ (VLDL) were the major contributors to this separation. By contrast, patients without MetS exhibited slightly more overlap in multivariate space, with model performance (Q² = 0.57), although the same lipid-associated metabolites remained the leading discriminators.

Figure 1.

Relative serum levels of phosphocholine and lipid-CH₃ LDL at various LT time points according to post MetS (PostMetS Y: PreN-PostY [n = 15], PreY-PostY [n = 1], PostMetS N: PreN-PreN [n = 24], PreY-PostN [n = 4]). (a) Phosphocholine at pre-LT (baseline) (P = 0.12), 6 months post-LT (P = 0.047), 1-year post LT (P = 0.07), >1-year post-LT (P = 0.001). (b) Lipid-CH₃ LDL at pre-LT (baseline) (P = 0.03), 6 months post-LT (P = 0.09), 1-year post LT (P = 0.14), >1-year post-LT (P = 0.001). LDL, low density lipoprotein; LT, liver transplantation; MetS, metabolic syndrome; ns, not significant; *P < 0.05; **P < 0.01.

Urinary NMR results

Urinary metabolomics provided complementary insight into gut–liver metabolic interactions. Overall urinary profiles differed significantly before and after LT (PLS-DA, Q² = 0.58; Supplementary Figure 2B, Supplementary Digital Content 1, http://links.lww.com/CTG/B445). Post-LT, TMAO levels increased from 0.016 to 0.028 (P = 0.001) at 6 months, and PAG increased from 0.0038 to 0.0079 (P < 0.001), indicating substantial shifts in the handling of microbial-derived metabolites after transplantation.

The magnitude of these increases differed according to MetS status. In MetS patients, urinary TMAO levels increased by 0.019 at ≥6 months compared with 0.006 in non-MetS patients (P = 0.045), and by 0.029 vs 0.005 at ≥1 year (P = 0.036) (Figure 2). PAG levels demonstrated a similar, but not significant, pattern, increasing by 0.005 in MetS patients compared with 0.003 in those without MetS at ≥1 year.

Figure 2.

Relative urinary changes in TMAO at various LT time points according to post MetS (PostMetS Y: PreN-PostY [n = 15], PreY-PostY [n = 1], PostMetS N: PreN-PreN [n = 24], PreY-PostN [n = 4]). Significant differences were observed between PostMetS Y and PostMetS N at 6 months or later post-LT (P = 0.017), at 1 year or later post-LT (P = 0.021) and comparing signal levels with baseline at 6 months or later (P = 0.045) and at 1 year or later (P = 0.036). LT, liver transplantation; MetS, metabolic syndrome.

Multivariate modeling further distinguished MetS-associated urinary signatures. In MetS patients, TMAO was the principal discriminator (Q² = 0.58), whereas in non-MetS patients the separation was slightly weaker (Q² = 0.44) and driven primarily by an unassigned signal with a chemical shift of 1.82 ppm.

DISCUSSION

This study evaluated serum and urinary metabolic profiles in liver transplant recipients and examined how these profiles relate to the development or worsening of MetS in a 2-center longitudinal cohort. The findings demonstrate that several metabolites reflecting lipid metabolism and gut–liver axis activity change after LT and differ between patients who develop MetS and those who do not. These results suggest that NMR-based metabolomics may offer additional insight into post-transplant metabolic function and may help identify individuals at higher risk of adverse metabolic trajectories.

Approximately 1 in 3 patients developed de novo MetS after LT. Identifying these patients earlier in their post-transplant course remains challenging with current clinical tools. In this context, noninvasive serum and urine metabolomics may complement existing assessments by capturing broader host and microbial metabolic processes. In our cohort, serum phosphocholine and lipid-CH₃ signals increased after LT but rose to a lesser extent in patients with MetS. Lipid-CH₃ (LDL/VLDL) reflects circulating lipoprotein-associated lipid moieties, and phosphocholines participate in membrane and lipoprotein turnover. The attenuated rise of these signals in MetS patients suggests differences in post-LT lipid metabolism, although the specific biological pathways underlying these patterns cannot be determined by metabolomics alone.

Post-LT increases in urinary TMAO, and to a lesser but nonsignificant degree in PAG, were more pronounced in MetS patients. These metabolites arise from gut microbial processing of dietary nutrients followed by hepatic conversion and conjugation, and their circulating levels therefore reflect integrated gut–liver metabolic activity (7). Previous studies in nontransplant populations have associated elevated TMAO levels with cardiometabolic risk (21), but the relevance after LT is less clear. The larger post-LT increases observed in MetS patients in our study indicate that metabolic pathways involving these metabolites differ between subgroups. Whether these differences are driven by diet, host–microbiota interactions, hepatic metabolic recovery, or medication exposures cannot be conclusively determined from our data (11). Considering serum and urine findings together, LT induces broad metabolic shifts encompassing both hepatic lipid metabolism and microbial cometabolism, with more pronounced microbial-derived metabolite elevations in patients who develop MetS. Pathway analyses also implicate phenylalanine, tyrosine and tryptophan biosynthesis, and glycerophospholipid metabolism.

These findings should be interpreted in the context of several important considerations. Certainly, this study, and any further study, would benefit from increased sample sizes at more similar time points. This would enable mixed-effect models to be used, in addition to PLS-DA multivariate analyses combined with univariate analysis, to account for confounding factors such as age, differences in medication including antimetabolite usage, and dietary factors. Serum and urine metabolomics reflect systemic metabolite pools and do not distinguish the proportional contributions of gut microbial activity, hepatic metabolism, diet, or immunosuppressive therapy, all of which are known to shape the plasma metabolome. Although all samples were collected in a fasting state from patients who were clinically stable, metabolites such as TMAO and PAG arise from complex, multistep processes that involve both microbial and hepatic pathways. The study lacked parallel microbiome sequencing data, which limits direct inference about microbial composition or function. Furthermore, although metabolomics can capture differences between patient groups, it cannot establish causality, and the observed associations should be viewed as hypothesis-generating. Changes in immunosuppressive regimens may also influence metabolic outcomes. Calcineurin inhibitors and antimetabolites were used consistently across groups, whereas corticosteroid exposure declined substantially over time as is routine clinical practice. This was more in patients without MetS. Given the known metabolic effects of corticosteroids, differences in tapering patterns could have contributed to some of the observed metabolic variability, although at 6 months post-LT the impact is unclear. However, the direction and magnitude of such effects cannot be determined from this analysis.

Overall, this study shows that de novo MetS is common after LT and is accompanied by distinct serum and urinary metabolite patterns. Patients who developed MetS displayed a blunted increase in lipid-related serum signals and a greater rise in gut-derived urinary metabolites such as TMAO. These findings suggest that post-LT metabolic dysfunction may involve alterations in both hepatic lipid pathways and gut–liver metabolic interactions. Monitoring selected metabolites could potentially enhance risk stratification for MetS and cardiometabolic complications in LT recipients. Future studies integrating longitudinal diet assessment, immunosuppressive drug exposure, microbiome profiling, and targeted metabolic analyses will be essential to clarify the mechanisms underlying these metabolic changes and to determine whether they can be modified to improve long-term outcomes.

CONFLICTS OF INTEREST

Guarantor of the article: Jasmohan S. Bajaj, MD, MS, FACG.

Specific author contributions: J.S.B., and E.V.: conceptualized the study, obtained funding, supervised and validated the study. I.J.C., and A.G.: performed metabolomic and bioinformatics analysis. All other authors were involved in study conduct. I.J.C., and M.M.: wrote the first draft and all authors revised and approved the final version.

Financial support: This work was partly supported by grants from NCATS R21TR002024 to J.S.B. and E.C.V., and VA Merit Review grants I0CX001076 and I01CX002472 to J.S.B. We thank the Roger Williams Institute of Liver Studies, Foundation for Liver Research for core funding of the NMR aspects of this project. We thank the Centre for Biomolecular Spectroscopy, King's College London for NMR data acquisition, which is funded by the Wellcome Trust and British Heart Foundation (ref. 202767/Z/16/Z and IG/16/2/32273).

Potential competing interests: None to report.

Data availability statement: Due to restrictions in place by our IRB, individual level data beyond which is already in the figures are not available to be released.

Study Highlights.

WHAT IS KNOWN

• ✓ Metabolic syndrome (MetS) is a major cause of morbidity postliver transplant.

• ✓ Predicting this outcome is difficult using clinical criteria alone.

WHAT IS NEW HERE

• ✓ In a 2-center prospective study, serum and urine metabolomics at baseline and over time were associated with MetS development or progression after liver transplantation.

• ✓ Specific gut-derived metabolites (trimethylamine-N-oxide) were associated with MetS along with serum phosphocholine and lipid-CH₃ (low density lipoprotein).

Supplementary Material

ct9-17-e00968-s001.docx ^{(1MB, docx)}

ABBREVIATIONS:

Alc

alcohol

BMI

body mass index

BP

blood pressure

HCC

hepatocellular cancer

HCV

hepatitis C virus

HDL-C

high density lipoprotein

LT

liver transplant

MASLD

metabolic dysfunction-associated steatotic liver disease

MELD

model for end-stage liver disease

MetS

metabolic syndrome

NMR

nuclear magnetic resonance

N-N

No-No

N-Y

No-Yes

PAG

phenylacetylglutamine

PLS-DA

partial least squares discriminant analysis

TMAO

trimethylamine-N-oxide

VCU

Virginia Commonwealth University

Y-N

Yes-No

Y-Y

Yes-Yes