Serum Metabolic Disturbances Associated with Acute-on-chronic Liver Failure in Patients with Underlying Alcoholic Liver Diseases: An Elaborative NMR-based Metabolomics Study.

OBJECTIVES: Acute-on-chronic liver failure (ACLF), which develops in patients with underlying alcoholic liver disease (ALD), is characterized by acute deterioration of liver function and organ failures are secondary to that. The clear understanding of metabolic pathways perturbed in ALD-ACLF patients can greatly decrease the mortality and morbidity of patients through predicting outcome, guiding treatment, and monitoring response to treatment. The purpose of this study was to investigate the metabolic disturbances associated with ACLF using nuclear magnetic resonance (NMR)-based serum metabolomics approach and further to assess if the serum metabolic alterations are affected by the severity of hepatic impairment.
MATERIALS AND METHODS: The serum-metabolic profiles of 40 ALD-ACLF patients were compared to those of 49 age and sex-matched normal-control (NC) subjects making composite use of both multivariate and univariate statistical tests.
RESULTS: Compared to NC, the sera of ACLF patients were characterized by significantly decreased serum levels of several amino acids (except methionine and tyrosine), lipid, and membrane metabolites suggesting a kind of nutritional deficiency and disturbed metabolic homeostasis in ACLF. Twelve serum metabolic entities (including BCAA, histidine, alanine, threonine, and glutamine) were found with AUROC (i.e., area under ROC curve) value >0.9 suggesting their potential in clinical diagnosis and surveillance.
CONCLUSION: Overall, the study revealed important metabolic changes underlying the pathophysiology of ACLF and those related to disease progression would add value to standard clinical scores of severity to predict outcome and may serve as surrogate endpoints for evaluating treatment response. Copyright:

INTRODUCTION

The spectrum of alcoholic liver diseases (ALD) includes steatosis (fatty liver), steatohepatitis (fatty liver with inflammation, also called alcoholic hepatitis), progressive liver fibrosis, and cirrhosis.[1] Acute-on-chronic liver-failure (ACLF) is a new clinical syndrome characterized by intense systemic inflammation and acute deterioration of liver function.[23] Currently, the biggest challenge in the clinical management of ACLF is the lack of reliable methods for rapid evaluation of disease severity, predicting therapeutic outcomes and survival. The Model for End-Stage Liver Disease (MELD)—which is commonly applied method for outcome prediction in patients with stable cirrhosis[45]—has several limitations with regard to outcome prediction in ACLF.[6] Therefore, recently a new organ failure based CLIF-SOFA (i.e., chronic liver failure sequential organ failure assessment) score have been developed.[7] Even though, the clinical decisions related to both selecting the appropriate treatment options as well as liver transplantation in ACLF patients are still driven by liver biopsy. Therefore, there is paucity and felt need of non-invasive surrogate markers to improve clinical diagnosis and prognosis of ACLF, monitoring treatment response and moreover to assess the severity of hepatic function. Metabolomics, because it allows rapid identification of metabolic perturbations in biological systems in response to a disease or therapeutic intervention, is increasingly being applied for identification of metabolic markers in body fluids (such as blood plasma/serum and urine) for improving the diagnosis and prognosis of human diseases.[8] Starting our efforts in this direction, we employed NMR based metabolomics approach to investigate altered metabolic profiles in the sera of ALD-ACLF patients and sought to identify perturbed metabolic pathways associated with severity of hepatic impairment.

MATERIALS AND METHODS

Recruitment of subjects

The study protocol was approved by the institutional research and ethical committee, Sanjay Gandhi Postgraduate Institute of Medical Sciences (SGPGIMS), Raibareli Road, Lucknow 226014, India (IEC Code: 2017-186-DM-99(B); File Number: PGI/BE/804/2017; Approval Date: October 30, 2017). The enrollment of subjects was carried out according to the norms of World Medical Association (WMA) declaration of Helsinki. An informed written consent was obtained from the guardians/kin of the patients after informing them the purpose of study. Relevant clinical and demographic details were collected for all the subjects in a custom-designed questionnaire. Serum samples were obtained from ALD patients (n = 40, 100% male with ascites) admitted in critical intensive care unit (ICU) for the management of ACLF (with Grade ≥ II based on CLIF-SOFA score[9]). Only patients with alcoholic acute hepatitis and alcoholic cirrhosis were considered and ACLF was diagnosed as per APASL[10] and ACLF grades as per CLIF C̈ SOFA criteria.[7] Exclusion criteria were age > 65 years, severe cardiopulmonary disease, chronic kidney disease (CKD) on dialysis, evidence of hepatocarcinoma or hepatic malignancy, infection with the human immunodeficiency virus, hepatitis B or C viruses, and a past history of acute decompensation during the previous 6 months. Blood samples were obtained within 1–3 days after the patient is stabilized in the ICU. Gender, age, the presence of ascites or hepatic encephalopathy, serum albumin bilirubin, the International Normalized Ratio, serum glutamic-oxaloacetic transaminase (SGOT) activity, serum glutamic pyruvic transaminase (SGPT) activity, and serum urea and creatinine levels were recorded at inclusion. For comparative analysis, the serum samples from 49 age matched normal control (NC) male subjects were collected after taking an informed consent. In each case, it was confirmed that the NC subjects are normotensive with no cardiovascular abnormalities and satisfying the above exclusion criteria. Each subject, either patient or healthy volunteer, provided a blood sample after overnight fasting for serum extraction. The serum was extracted as per the established protocol,[11] transferred into a sterile 1.5 mL microcentrifuge tube (MCT) immediately after the processing and stored at −80°C until the NMR experiments were performed.

Nuclear magnetic resonance measurements

The NMR experiments were performed and resulted NMR spectral features were analyzed as described previously.[131415]

RESULTS

Clinical and demographic details

The clinical and demographic characteristics of the subjects are summarized in Table 1. Based on the inclusion and exclusion criteria (as described in “Materials and Method” section), a total 89 subjects were involved in this study, 40 forming the disease group (ALD-ACLF) and 49 forming the NC group. The mean age of the ALD-ACLF and NC groups was 41.7 ± 7.54 and 45 ± 6.73 years, respectively. The main diagnoses for ACLF patients included in this study were sepsis (20%), HE, ascites (100%), and jaundice, whereas the main etiologic precipitant for ACLF was alcohol intake (within 4 weeks) with or without bacterial infection.

Table 1

Biochemical, clinical, and demographic characteristics of patients with ALD-ACLF and control cohorts recorded at inclusion

Variables/parameters	Case (n = 40)	Normal control (n = 49)
Age (in years)	41.7 ± 7.54	45 ± 6.73
Male gender	100%	100%
Alcoholic	40 (100%)	10 (20%)
Precipitant Active alcoholism	32 (80%)	–
Sepsis (SBP)	8 (20%)
Ascites	38 (95%)	–
Complete ascites mobilization	32 (80%)	–
Gastrointestinal bleeding (GIB)	2 (5%)
HE grading >2	27 (67.5%)	–
ACLF grade >2	28 (70%)	–
28 days nonsurvival	10	–
Blood urea (mg/dL)	39.5 ± 19.06	–
Hemoglobin g/dL	9.19 ± 1.68	–
TLC (cells/mm3)	11.25 ± 6.35	–
Platelet count (G/L)	87.69 ± 36.47	–
CRP/ESR	3.0 ± 2.9/36.4 ± 25.5	–
Total bilirubin (mg/dL)	17.77 ± 11.83	–
Albumin (g/dL)	2.6 ± 0.5	–
SGOT (SGPT) U/L	108.13 ± 66.927 (48.32 ± 26.31)	–
Serum sodium (serum potassium)	133.62 ± 5.54 (4.01 ± 0.44)	–
Creatinine (mg/dL) (INR)	1.53 ± 0.99 (2.79 ± 0.77)	–
U Na (U K)	34.66 ± 26.18 (48.2 ± 51.19)	–
CTP	12.2 ± 1.34	–
MELD	28.0 ± 7.0	–
Total OF	1.76 ± 0.72	–
CLIF SOFA	10.35 ± 1.66	–
CLIF C ACLF score	52.4 ± 6.4	–

mL = milliliter, mq/dL = milligram per deciliter, g/dL = gram per deciliter, U/L = Units per liter, G/L = Giga/Litre, TLC = total leucocyte count, SGOT = serum glutamic-oxalacetic transaminase (also known as enzyme aspartate aminotransferase, AST), SGPT = serum glutamic-pyruvic transaminase (also known as alanine aminotransferase, ALT), ALP = alkaline phosphate, INR = international normalized ratio, SBP = spontaneous bacterial peritonitis, CTP score = Child–Turcotte–Pugh score, MELD score = model for endstage liver disease score.[4512] SOFA = sequential organ failure assessment, CLIF = EASL chronic liver failure

Metabolic disturbances associated with acute-onchronic liver failure in patients with alcohol-related liver disease

The present study aims to compare the NMR based serum metabolic profiles of 40 ALD-ACLF patients and normal control (NC) subjects. The comparison of cumulative 1D 1H CPMG NMR spectra of serum samples obtained from ALD-ACLF patients (n = 40) and NC subjects (n = 49) provided an overview of metabolic variations (see Figure S1 in the electronic supplementary material, ESM). The majority of metabolites in the NMR spectra were identified and assigned by comparing the chemical shift and peak patterns with the 800 MHz database library of CHENOMX NMR suite.[15] These metabolite assignments were further validated using previously reported NMR assignments of serum/plasma metabolites in the literature in tandem with Biological Magnetic Resonance Data Bank (BMRB) database and The Human Metabolome Database (HMDB: http://www.hmdb.ca).[16171819] The exercise resulted into the unambiguous identification of 34 circulatory metabolites for concentration profiling in CHENOMX: 3-hydroxybutyrate (3HB), acetate, acetone, alanine, aspartate, choline, citrate, creatine, creatinine, fumarate, glucose, lactate, glutamate, glutamine, glycerol, glycine, isoleucine, leucine, lysine, mannose, methionine, phenylalanine, proline, propylene-glycol, pyruvate, succinate, threonine, trimethylamine, trimethylamine N-oxide (TMAO), tyrosine, urea, valine, sn-Glycero-3-phosphocholine (GPC), and histidine. Database chemical shifts and literature reports additionally provided identification of NMR signals of (a) N-acetyl-glycoproteins (NAG), (b) lipid and membrane metabolites, for example, choline, sn-glycero-3-phosphocholine (GPC), polyunsaturated fatty acids (PUFAs) and (c) lipoproteins, for example, high-density lipoprotein (HDL), low-density lipoprotein (LDL), and very low-density lipoprotein (VLDL) (see ESM, Figure S1).

Consistent with previous reports,[202122] the visual comparison of 1D 1H CPMG NMR spectra revealed altered serum levels for lipid and membrane metabolites in addition to evidently decreased serum levels of acute-phase proteins as inferred from low intensity NMR signal of N-acetyl glycoproteins at 2.01 ppm (See Electronic Supplementary Material [ESM], Figure S1A). However, the subtle metabolic differences in the sera of patients and NC subjects were not visually evident. Therefore, the NMR spectral features were subjected to multivariate data analysis to identify serum metabolic disturbances associated with ALD-ACLF. First, the CPMG data matrix containing normalized spectral features was analyzed using unsupervised PCA method, for evaluating initial grouping trends and class separation. Further, we employed supervised clustering method PLS-DA to reveal subtle metabolic differences among the study groups. The two dimensional score-plot derived PLS-DA model based analysis of normalized CPMG spectral features is shown in Figure 1A. Clearly evident that the serum samples of patients and control groups are well clustered and separated from each other indicating that the serum metabolic profiles of ACLD-ACLF patients are distinctively different from NC subjects. The PLS-DA model cross-validation parameters, R2 (explained variation) and Q2 (predictive capability) were significantly higher (R2 > 0.95, Q2 > 0.88, Figure 1B) suggesting that discriminatory model possesses a satisfactory fit with good predictive power. The metabolic features responsible for separating the study cohorts were identified using VIP score indexing. The VIP score plot highlighting top 25 metabolic changes with highest VIP score (VIP score ≥ 1.5) is shown in Figure 1C and the corresponding student t-test results are shown in electronic supplementary material (ESM, Figure S1B and S1C]). Compared to NC group, the serum levels of glucose, branched chain amino acids (valine, leucine), N-acetyl-glycoproteins, lipoproteins (HDL, LDL, and VLDL) and other lipid and membrane metabolites (such as choline, GPC, and PUFA) were decreased in ALD-ACLF patients, whereas those of glucose, lactate, acetate, TMAO, betaine, creatinine and methionine were increased in the patient cohort [Table 2].

Figure 1

(A,D) 2D score plots derived from PLS-DA model analysis involving normalized spectral features (A) and explicit concentration profiles (D) obtained for serum samples of ALD-ACLF patients (in red) and normal control subjects (in blue). The shaded or semi-transparent areas represent the 95% confidence regions of each group as depicted by their respective colors. (B,E) are barplots showing the three performance measures (prediction accuracy, multiple correlation coefficient R2 and the explained variance in prediction Q2) obtained after10 fold Cross Validation analysis. The validation parameters (R2 and Q2) obtained for PLS-DA model based on five and three components are displayed in the respective score plots in (A) and (D). (C,F) The VIP score plots derived from PLS-DA model based on five components in (C) and three components in (F). The symbol asterisk “*” represents the metabolic change is statistically significant as well

Table 2

Key serum metabolic profiles of discriminatory relevance evaluated for diagnostic potential in differentiating ALD-ACLF from NC using the receiver operating characteristic (ROC) curve analysis

Bin	Assignment	AUROC	P Value	Study A ACLF vs. CLF	Study B (severity)	Study C (LC vs. AAH)	Study D (severe CLF vs. mild CLF)
0.81	HDL	0.95	↓***	↓*			↓*
2.03	NAG+lipid	0.94	↓***
0.83	LDL	0.94	↓***
1.57	Lipid	0.93	↓***		↓*
3.21	Choline	0.92	↓***		↓*	↓*	↓*
3.19	GPC	0.92	↓***			↓*	↓*
2.01	NAG	0.91	↓***				↓*
3.27	Betaine	0.90	↑***
3.75	Glucose	0.88	↑***				↑*
5.31	PUFA	0.88	↓***
3.25	TMAO	0.85	↑***
3.77	Glucose	0.82	↑***
0.89	VLDL	0.82	↓***				–
3.03	Creatinine	0.81	↑***	↑*			↑*
0.87	LDLVLDL	0.79	↓***
0.95	Leucine	0.79	↓***				↑*
5.29	PUFA	0.78	↓***
2.05	NAG+lipid	0.78	↓***
1.99	Lipid	0.74	↓***
4.11	Lactate	0.62	–	↑*	↑*		↑*
3.23	Glucose+GPC	0.56	–
0.85	LDL	0.54	–
1.91	Acetate	–	–
2.13	Methionine	–	–		↑*	↑*	↑*
1.03	Valine	–	–

AUROC = area under ROC curve, CLF = chronic liver failure, ACLF = acute on chronic liver failure, LC = liver cirrhosis, AAH = acute alcoholic hepatitis. Study A: 500-MHz NMR-based serum metabolomics fingerprints of acute-on-chronic liver failure in patients with alcoholic cirrhosis.[20] Study B: NMR-based plasma metabolomics study,[23] which showed that NMR plasma levels accurately predict mortality in decompensated cirrhosis (DC). The NMR plasma profiles of nonsurvivors were attributed to reduced phosphatidylcholines and lipid resonances, with increased lactate, tyrosine, methionine, and phenylalanine signal intensities. Study C: Untargeted metabolomics study that compared the serum metabolic profiles of patients with acute alcoholic hepatitis and patients with liver cirrhotic.[24] Study D: NMR-based serum metabolomics study that compared the serum metabolic profiles of cirrhotic patients with mild chronic liver failure (CLF) condition and those with severe CLF condition.[25] *, **, and ***: Represent the metabolic change with p-value <0.05 <0.001 and <0.0001

Point to be mentioned here is that the metabolic differences derived from normalized spectral features may provide ambiguous information when the corresponding bin contains signals from multiple metabolites. For example, the spectral bin at 3.23 ppm, which mainly represents glucose, does have contribution of NMR signals from GPC and trimethylamine-N-oxide (TMAO). Other than this, the dominant signals of lipoproteins (LDL, VLDL), lipid/membrane metabolites, glucose and lactate C̈ which contribute to multiple spectral bins—does not render the subtle but significant metabolic changes to show off in the VIP score plot [Figure 1C]. Therefore, to extract further relevant information about metabolic changes, the concentration profiles of 34 circulatory metabolites were estimated from CPMG NMR spectra using CHENOMX NMR suite. The concentrations were further used to estimate circulatory ratios relevant in the present study context such as Glutamate-to-Glutamine ratio (EQR); Phenylalanine-to-Tyrosine ratio (PTR); Histidine-to-tyrosine ratio (HTR) and branched-chain amino-acid-to-tyrosine ratio (BTR; estimated as: (Leucine + Isoleucine + Valine)/Tyrosine; this is also referred as Fischer ratio[2627]). The resulted 38 serum metabolic entities were compared using supervised PLS-DA model based discriminatory analysis and the results are shown in Figure 1D–1F. The 2D score plot derived from PLS-DA model is shown in Figure 1D. Clearly evident that the samples are well clustered within their respective groups and the samples of two study groups are well separated suggesting that the serum metabolic profiles of ALD-ACLF patients are distinctively different from NC subjects [Figure 1D]. Further, the higher values of cross-validation parameters (R2 ~ 0.75, Q2 ~ 0.66) clearly established the goodness of separation between classes, and the statistical significance of the class-separating metabolic features [Figure 1E]. Figure 1F showing the VIP score plots highlights top 15 metabolic changes ranked according to their increasing discriminatory potential. Compared to NC, the sera of ACLF patients were characterized by decreased serum levels of various amino acids (valine, leucine, isoleucine, alanine, glycine, proline, threonine, glutamine, glutamate, aspartate, histidine and phenylalanine), and other circulatory metabolic entities such as glucose, acetate, acetone, citrate, fumarate, mannose, glycerol, trimethylamine, lactate, histidine-to-tyrosine ratio (HTR) and branched-chain-amino-acid-to-tyrosine ratio (BTR) whereas circulatory levels of methionine, trimethylamine-N-oxide and glutamate-to-glutamine ratio (EQR) were found significantly elevated in ALD-ACLF patients.

Following the identification of discriminatory serum metabolic entities (NMR variables), the ROC analysis was performed as a quantitative measure to evaluate their potential (i.e., specificity and sensitivity) for differentiating ACLF from NC cohort [Figure S2]. First, we computed ROC curves for all the top 25 normalized spectral features and the results are summarized in Table 2. Of various discriminatory metabolic entities, 14 metabolic entities (HDL, NAG+Lipid, LDL, lipid, choline, GPC, NAG, betaine, glucose, PUFA, TMAO, and glucose, VLDL, and creatinine) were found with AUROC value >0.8, suggesting these NMR-based serum signals show significant potential for differentiating ACLF patients from NC cohort. The representative ROC curve plots of 12 serum metabolic entities with the highest diagnostic potential are shown in Figure S2 in tandem with corresponding box plots to highlight the degree of metabolic alterations in ACLF compared to NC group.

Next, the concentration profiles of circulatory metabolites were evaluated for their diagnostic potential and the results of ROC curve analysis are summarized in ESM, Figure S3. Of 38, 12 circulatory metabolic entities (valine, isoleucine, leucine, HTR, BTR, histidine, alanine, trimethylamine, GPC, fumarate, threonine, and glutamine) were found with AUROC value >0.9 suggesting these could be potential biomarkers for clinical evaluation and surveillance of patients with ACLF in critical care. The representative ROC curve plots of 12 serum metabolic entities with the highest diagnostic potential are shown in ESM, Figure S3A in tandem with their respective box plots in ESM, Figure S3B to highlight the metabolic aberrations in ACLF compared to NC group. The implications of these metabolic changes (i.e., similarities or differences) in the pathogenesis of ACLF patients with underlying ALD have been discussed in detail below.

Concluding remarks

This study represents the first most elaborative high-field NMR-based metabolomics analysis of human sera collected from well‐characterized ACLF patients with pre-existing alcoholic liver diseases. The sera of ACLF patients were characterized by significantly decreased serum levels of various amino acids (except methionine and tyrosine) and those of lipid and membrane metabolites suggested severe nutritional deficiency in ACLF which may be attributed to several factors such as (a) poor oral intake and follow a low calorie diet, (b) fat malabsorption due to impaired mucosal barrier function,[28] and metabolic response to the stress of critical ill.[29] Malnutrition such as protein-energy malnutrition, muscle and adipose tissue depletion are common in patients with advanced liver disease[303132] and may progress further as liver function crumbles resulting into other critical complications like ascites, hepatic encephalopathy, infections, and death as evident in case of ACLF.[33] However, the clinical use of NMR based serum metabolomics approach as a diagnostic and prognostic tool for ACLF needs further validation studies on larger prospective cohorts of ACLF patients in a longitudinal manner to confirm these results and their association with clinical outcomes and severity of hepatic impairment. Nevertheless, the present study will form the basis for future clinical metabolomics studies aiming to identify metabolomics biomarkers for early diagnosis, predicting outcomes, differentiating clinical subtypes, determining the severity of the condition, monitoring treatment response, and guiding clinical trial testing.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.

(A) Stack plot of cumulative 1D 1H CPMG NMR spectra (left ranging from δ0.6– δ4.65 ppm and right ranging from δ5.0– δ9.0 ppm) recorded on sera of ALD-ACLF patients (in blue) and normal controls (NC, in red). The spectral peaks are labeled according to metabolic assignment. Key acronyms are: HDL: high-density lipoprotein; LDL: low density lipoproteins; VLDL: very low density lipoproteins; DMA: Dimethylamine; TMA: Trimethylamine; NAG: N-acetyl glycoproteins; 3HB: 3-hydroxybutyrate; GPC: glycerophosphocholine; TMAO: trimethyl-amine N-oxide. (B) The univariate student t-test performed with 38 estimated concentration profiles of circulatory metabolites to evaluate the statistically significant (pink dots) and insignificant (gray colored dots) metabolic differences between ACLF and NC groups (p-values <0.05 was used as the criterion of statistical signifcance). (C) Box-cum-whisker plots highlighting the statistically insignificant differences of metabolic profiles between ACLF (red) and NC (blue) groups. In each box plot, the box denote interquartile range, horizontal line inside the box denote the median, and bottom and top boundaries of boxes are 25th and 75th percentiles, respectively. Lower and upper whiskers are 5th and 95th percentiles, respectively

Top 12 key marker metabolites identified based on ROC curve analysis performed with all 25 normalized spectral features as tabulated in Table 2. The computed 95% confidence interval (CI) for individual marker metabolites is highlighted in the faint blue background over the ROC curve. The area under the receiver operating characteristic curve (AUROC) is shown in red to highlight the diagnostic potential of corresponding circulatory metabolite. The box-cum-whisker plots shown in the right side of each ROC curve plot clearly reveal metabolic change in patients with ALD-ACLF compared to NC. For each box plot showing quantitative variations of relative NMR signal integrals, the boxes denote interquartile ranges, the horizontal red line inside the box denotes the median, and the bottom and top boundaries of the boxes are the 25th and 75th percentiles, respectively. Lower and upper whiskers are the 5th and 95th percentiles, respectively

(A) Receiver operating characteristic (ROC) curve analysis performed for evaluating the diagnostic potential of various circulatory metabolites and their specific ratios for differentiating ALD-ACLF from NC group. The ROC plots of twelve circulatory metabolites identified with highest value of area under the ROC curve are shown here. (B) The box-cum-whisker plots showing quantitative variations for key circulatory metabolites of discriminatory potential (identified based on their top ranking in the VIP score plot and high AUROC values). In the box plots, the boxes denote interquartile ranges, horizontal line inside the box denote the median, and bottom and top boundaries of boxes are 25th and 75th percentiles, respectively. Lower and upper whiskers are 5th and 95th percentiles, respectively