Author Affiliations: Collaborative Innovation Center for Diagnosis and Treatment of Infectious Diseases; State Key Laboratory for Diagnosis and Treatment of Infectious Disease, First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou 310003, China (Lian JS, Hao SR, Chen DY, Wang YY, Yang JL, Jia HY and Huang JR); Yantai City Hospital for Infectious Diseases, Yantai 264000, China (Liu W)

Corresponding Author: Jian-Rong Huang, MD, Collaborative Innovation Center for Diagnosis and Treatment of Infectious Diseases; State Key Laboratory for Diagnosis and Treatment of Infectious Disease, First Affiliated Hospital, Zhejiang University School of Medicine, 79 Qingchun Road, Hangzhou 310003, China (Tel: +86-571-87236456; Email:hzhuangchina@sina.com)

 

© 2015, Hepatobiliary Pancreat Dis Int. All rights reserved.

doi: 10.1016/S1499-3872(15)60393-9

Published online July 2, 2015.

Contributors: HJR designed the study. LJS and LW performed the majority of experiments. HSR, CDY, WYY, YJL and JHY provided analytical tools and were also involved in editing the manuscript. LJS and LW wrote the manuscript, and contributed equally to this study. HJR is the guarantor.

Funding: This study was supported by grants from the National Grand Program on Key Infectious Diseases (2012ZX10002-004) and the National Natural Science Foundation of China (81100286).

Ethical approval: Not needed.

Competing interest: No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article.

BACKGROUND: Because of the diversity of the clinical and laboratory manifestations, the diagnosis of autoimmune liver disease (AILD) remains a challenge in clinical practice. The value of metabolomics has been studied in the diagnosis of many diseases. The present study aimed to determine whether the metabolic profiles, based on ultraperformance liquid chromatography-mass spectrometry (UPLC-MS), differed between autoimmune hepatitis (AIH) and primary biliary cirrhosis (PBC), to identify specific metabolomic markers, and to establish a model for the diagnosis of AIH and PBC.

 

METHODS: Serum samples were collected from 20 patients with PBC, 19 patients with AIH, and 25 healthy individuals. UPLC-MS data of the samples were analyzed using principal component analysis, partial least squares discrimination analysis and orthogonal partial least squares discrimination analysis.

 

RESULTS: The partial least squares discrimination analysis model (R²Y=0.991, Q²=0.943) was established between the AIH and PBC groups and exhibited both sensitivity and specificity of 100%. Five groups of biomarkers were identified, including bile acids, free fatty acids, phosphatidylcholines, lysolecithins and sphingomyelin. Bile acids significantly increased in the AIH and PBC groups compared with the healthy control group. The other biomarkers decreased in the AIH and PBC groups compared with those in the healthy control group. In addition, the biomarkers were downregulated in the AIH group compared with the PBC group.

 

CONCLUSIONS: The biomarkers identified revealed the pathophysiological changes in AILD and helped to discriminate between AIH and PBC. The predictability of this method suggests its potential application in the diagnosis of AILD.

 

(Hepatobiliary Pancreat Dis Int 2015;14:413-421)

 

KEY WORDS: autoimmune liver disease; biomarkers; metabolomics; autoimmune hepatitis; primary biliary cirrhosis; overlap syndrome

Autoimmune liver disease (AILD) is characterized by self-perpetuating inflammation of the liver with indeterminated causes.^[1] AILD includes autoimmune hepatitis (AIH), primary biliary cirrhosis (PBC), primary sclerosing cholangitis, and overlap syndrome (OS). AIH and PBC are two types of AILD with a higher incidence than PSC and OS.^[2] The diagnosis of AILD is challenging due to diverse clinical and laboratory manifestations.^{[3, 4]} The diagnosis of AILD requires a combination of clinical features, abnormal liver biochemical patterns, auto-antibodies and abnormal liver histology.^{[4, 5]} However, liver biochemical patterns and auto-antibody levels are not pathogenic or required for disease.^[6] Furthermore, atypical liver histology confounds the diagnosis.^{[7, 8]} Differentiating between AIH and PBC is difficult because of non-specific behaviors and the overlap of diseases, but the differentiation is important because the treatment regimens for AIH and PBC are different. Although PBC is currently treated with ursodeoxycholic acid,^{[9, 10]} AIH is mainly treated with prednisolone/prednisone with or without azathioprine.^[11]

 

In previous studies^{[12, 13]} on the diagnosis of AILD, proteomic analyses were performed to identify many auto-antibodies. However, auto-antibodies are insufficient to diagnose AILD.^{[6, 14, 15]} Antimitochondrial antibodies are present in >90% of PBC patients,^[16-18] and no antibody is highly specific in patients with AIH. Many scoring systems have been designed for the diagnosis of AILD. Three criteria for the diagnosis of AIH have been proposed by the International Autoimmune Hepatitis Group.^[19-21] Compared with the revised criteria in 1999, which were primarily designed for research purposes, the simplified scoring system has enhanced clinical applicability and practicability.^[22] The revised original scoring system performs well in patients with few or atypical features of AIH, whereas the simplified system is better at excluding the diagnosis in diseases with concurrent immune manifestations.^[23] However, the specificity of the revised original scoring system is not satisfactory, whereas the specificity of the simplified system is low. Furthermore, these scoring systems list liver biopsy results as a criterion. Although being a generally safe procedure, liver biopsy carries a risk of complications.^[24] Therefore, developing new methods to detect AILD accurately is important.

 

With the development of high-throughput methods, "omic" studies have become important to elucidate biological processes. As the products of cellular adjustment processes, metabolites levels are regarded as the ultimate readouts for genetic or environmental changes in biological systems.^{[25, 26]} Metabolomic methods have been used for clinical diagnosis and drug discovery. Using these methods, researchers have discovered many specific biomarkers and successfully diagnosed several diseases.^[27-29] Previous studies used metabolomic methods for the diagnosis of biliary tract cancer, renal cell carcinoma, and diabetes mellitus.^[30-32] These studies suggested that the accuracy and predictability of these methods can augment current diagnostic approaches. However, no study has characterized AILD using metabolomic methods.

 

Nuclear magnetic resonance spectroscopy and gas or liquid chromatography coupled to mass spectrometry (GC-MS and LC-MS, respectively) are common tools for metabolomic studies.^{[33, 34]} In contrast to nuclear magnetic resonance and GC-MS, LC-MS has higher resolution and sensitivity and is considered to be a powerful tool to obtain rapidly and effectively multiparametric metabolite profiles from biological fluids.^[35] In this study, we evaluated the feasibility of a metabolomic method to diagnose AILD. Our method was based on ultraperformance liquid chromatography (UPLC)-MS; we evaluated two common AILDs, AIH and PBC. This study aimed to determine whether the metabolic profiles were different between AIH and PBC, to identify specific metabolomic markers, and to establish a model for the diagnosis of AIH and PBC.

 

 

We selected 19 patients with AIH and 20 with PBC who were hospitalized at the First Affiliated Hospital of Zhejiang University School of Medicine (Hangzhou, China) from March 2009 to December 2010. The study protocol was approved by the Ethics Committee of the hospital and written informed consent was provided by all enrolled patients. The patients with AIH were diagnosed according to the revised criteria proposed by the International Autoimmune Hepatitis Group in 1999. Their scores confirmed definitely the diagnosis of AIH.^[20] Patients with PBC were diagnosed according to the entry criteria for clinical trials of PBC based on the recommendations of the American Association for the Study of Liver Diseases.^[36] Patients taking medication or supplements, or those with gallstones diseases or other factors that cause cholestatic abnormal liver profiles were excluded. In both groups, patients with primary sclerosing cholangitis, OS (such as PBC and AIH or AIH and primary sclerosing cholangitis), hepatitis virus infection, HIV co-infection, hepatocellular carcinoma, or diabetes were excluded. Twenty-five healthy persons who visited our hospital for physical check-ups were recruited as healthy controls. These individuals exhibited normal liver functions and had no evidence of disease. No statistically significant differences in age and gender were found among the PBC, AIH and control groups (Table 1, P>0.05).

 

At admission, before breakfast and initiation of drug treatment, blood was taken from 39 patients and 25 healthy individuals and centrifuged at 3600×g for 6 minutes at 4 ��. Blood serum was transferred into tubes and stored at -80 �� until analysis.

 

Acetonitrile, isopropyl alcohol, formic acid, methanol, leucine enkephalin (HPLC grade), bile acid, sphingomyelin, lysophosphatidylcholine and fat acid standards were purchased from Sigma-Aldrich (St. Louis, MO, USA). Distilled water was filtered through a Milli-Q system (Millipore, Bedford, MA, USA).

 

The serum samples were thawed at 4 ��. Eight quality control serum samples were prepared by mixing equal amounts of serum samples from 10 AIH patients, 10 PBC patients and 10 healthy controls. All serum samples were mixed with acetonitrile at a 1:3 ratio (v/v), shaken vigorously and centrifuged at 14 000 rpm for 20 minutes. The clear supernatants were placed into glass inserts in UPLC vials and randomized.

 

Chromatographic separation was performed using the 100×2.1 mm ACQUITY 1.7 µm column (Waters, Milford, MA, USA) in the ACQUITY UPLC system. In the positive ion mode, the mobile phase A was water/formic acid (99.9:0.1, v/v) and the mobile phase B was acetonitrile/isopropyl alcohol/formic acid (79.9:20:0.1, v/v). In the negative ion mode, the mobile phase A was water/formic acid (99.9:0.1, v/v) and the mobile phase B was methanol/formic acid (99.9:0.1, v/v). The column was maintained at 50 ��. The composition of the mobile phase B was changed from 3% to 97% in 16 minutes, maintained for 5 minutes, reached 100% in 1 minute, and was held for 5 minutes. The temperature of the sample manager was set to 4 ��, and the injection volume was 5 µL for each analysis. The column eluent was split to direct approximately 100 µL/min to the mass spectrometer.

 

MS was performed using a Waters Q-TOF Premier (Waters MS Technologies, Manchester, UK) with positive and negative ion modes. The nebulization gas was released at 450 L/h at a temperature of 350 ��. The cone gas was set to 50 L/h with a source temperature of 120 ��. For the positive ion mode, the capillary voltage was 3.0 kV and the sampling cone voltage was set to 40 V. For the negative ion mode, the capillary voltage was set to 2.5 kV and the sampling cone voltage was set to 40 V. MS was performed in the V optics mode with 9000 resolution. The acquisition rate of data was set to 0.3 second with a 0.02-second inter-scan delay. All analyses were acquired using the lock spray to ensure accuracy and reproducibility. The instrument was previously calibrated with sodium formate. The lock mass spray for precise mass determination was set using leucine enkephalin at m/z 556.2771 at a concentration of 0.5 ng/µL in the positive ion mode and m/z 554.2615 in the negative ion mode. The lock spray frequency was set to 5 seconds, and the lock mass was averaged over 10 scans. Mass spectra were acquired over the scanning range of m/z 50-1000. Metabolites were identified using LC-MS/MS. The product ions were detected over a scanning range of m/z 50-1000. The collision voltage was set between 5 and 60 V, and argon was used as the collision gas. Identification of the metabolites was confirmed by comparison with authentic standards.

 

We used Mass Lynx version 4.1 software to pre-process the data including retention time and m/z that were obtained from UPLC-MS analysis. The data were aligned and normalized during this process. To reflect the differences among the three groups, multivariate statistical analysis was performed using SIMCA-P+12.0. Pareto scaling was used in all of the models to avoid chemical noise. The data were exported and assessed using principal component analysis (PCA), partial least squares discrimination analysis (PLS-DA), and orthogonal partial least squares discrimination analysis (OPLS-DA). Potential biomarkers were selected according to the variable importance for project value, the loading plot, and the S-plot. SPSS version 16.0 was used for statistical analysis. For the identification of potential markers, the following databases were used: HMDB (www.hmdb.ca), METLIN (metlin.scripps.edu), Mass Bank (http://www.massbank.jp/), Pub Chem (http://pubchem.ncbi.nlm.nih.gov/) and KEGG (www.genome.jp/kegg/ligand.html). The PLS-DA model was established to discriminate between AIH and PBC cohorts. The receiver operating characteristic (ROC) analysis was used to cross-validate predicted Y values, while verifying the robustness of the metabolomic model.

 

To evaluate the predictive ability of the PLS-DA model for new samples, 12 independent samples were used, including samples from four patients with AIH, four patients with PBC and four patients with AIH-PBC OS. The diagnostic criteria for AIH and PBC were the same as mentioned above. AIH-PBC OS was diagnosed according to the Paris Diagnostic Criteria.^[37-39] Clinical characteristics of the samples used for external validation are summarized in Table 1. The sensitivity and specificity of the PLS-DA model for the prediction of new samples were calculated.

 

 

A total of 64 samples taken from 20 patients with PBC, 19 patients with AIH and 25 healthy controls were analyzed in the positive and negative ion modes. The plots of PCA scores showed that the quality control samples were tightly clustered, confirming that our method was reproducible. Variations in marker metabolites were further observed in quality control samples, which were used to evaluate the overall performance of our method.^[40] We calculated the coefficient of variance (CV) of all markers identified in quality control samples. The percentage CV ranged from 3.2% to 10.3% with a median CV of 6.7%, which showed satisfactory reproducibility (Table 2).

 

A total of 3019 peaks of positive ions and 3129 peaks of negative ions were detected using MassLynx by the same acquisition method. The data were imported into SIMCA-P+12 for PCA. Plots of the scores of PCA with positive and negative ion modes are shown in Fig. 1A and B. In the positive ion mode, the patients and controls were separated exclusively. However, the AIH and PBC groups were not clearly separated, resulting in poor discrimination. In the negative ion mode, distinct clustering was observed between the patients and controls. Furthermore, AIH and PBC patients were separated into two distinct regions.

 

Despite limitations, PCA can provide an overview of the pre-processed data.^[41] To better characterize the serum metabolite profile of AIH and PBC patients, supervised PLS-DA models were established based on the two disease groups with positive and negative ion modes. R²Y (cum) and Q² (cum) parameters, which indicated the fitness and predictability respectively, were used to evaluate the efficacy of the models. The results showed that the efficacy of the PLS-DA model using the positive ion mode was low (R²Y=0.492, Q²=0.292), indicating that the metabolomic profiling of human blood did not display significant differences between PBC and AIH patients. Conversely, the model using the negative ion mode displayed good efficacy (R²Y=0.991, Q²=0.943) (Fig. 1C). This finding prompted us to hypothesize that the two diseases are discernable using UPLC-MS with the negative ion mode. Therefore, the negative ion mode was used in our subsequent experiments.

 

Analysis of PLS-DA S-plot (Fig. 1D) and the variable importance for projects indicated several discriminatory metabolites that differentiated between AIH and PBC samples. An independent samples t test was performed, and variables without significant differences were eliminated between the AIH and PBC groups (P>0.05). The molecular ionization form (protonated or sodiated) and fragments corresponding to theoretical and standard fragments were used in marker identification.^[42]

One of the ions identified from the loading plot as being responsible for the variance in the data set was the [M-CH3]^- ion m/z 685.5319, eluting with a retention time of 16.41 minutes in the UPLC separation (Fig. 2A). According to its MS spectrum, the signal at m/z 685.5319 was derived from a [M+HCOOH-H]^- ion m/z 747.5549 as well as [M-CH3]^- ions generated by in-source fragmentation. MS/MS spectra of selected ion m/z 745.5549 were acquired to determine their fragmentation products and to confirm their structural assignments. MS/MS analysis of this [M+HCOOH-H]^- ion m/z 745.5549 gave rise to the major ions m/z 168.0465 (demethylated phosphoryl choline), m/z 685.5319 ([M-CH3]^-) and m/z 78.9639 (phosphoryl group) in the negative ion mode (Fig. 2B). The fragmentation pattern of the spectra obtained for sphingomyelin (SM) (d18:0/16:1) standard was similar to the MS/MS spectra (Fig. 2C). Thus, we identified this compound as SM (d18:0/16:1). Using this method, we identified five groups of markers, including four bile acids (BAs), three free fatty acids (FFAs), nine phosphatidylcholines (PCs), two lysolecithins (LPCs) and SM (d18:0/16:1). Fig. 3 shows the serum levels of glycochenodeoxycholic acid, FFA, LPC-16:0, PC-16:0/16:0, and SM during the PBC, AIH and healthy control groups. Table 2 shows the detailed list of the markers, and Fig. 1D indicates the identified markers using red triangles.

 

The goodness of fit (R²Y=0.991, Q²=0.943) of the original PLS-DA model and clustering of 100 Y-permutated models were visualized using a validation plot (Fig. 1E). The validation plot strongly indicated that the original model was valid and showed no signs of over-fitting. In addition, the area under the ROC curve of the PLS-DA model was calculated using cross-validated Y-predicted values, which were set to 1, and confirmed the validity of the model. A prediction value of Y≥0.5 was the optimal value for discrimination between the AIH and PBC groups with a sensitivity of 100% and a specificity of 100%.

 

The AIH and PBC groups were used to examine the predictive ability of the PLS-DA model for each disease, with a sensitivity of 100% and a specificity of 100% (Fig. 4).

 

 

The prognosis of patients with AILD is poor, but it can be improved by adequate therapy.^{[43, 44]} Early diagnosis of AILD is particularly important; however, current diagnostic methods such as autoantibody detection have limited sensitivity and specificity.^[6] In our study, a PLS-DA model was established using a highly reproducible UPLC-MS metabolomic method.

 

Low variations in the levels of peaks in quality control samples confirmed that our data were accurate and that our method was highly reproducible.^[40] In the S-plots of PCA with the negative ion mode, the group clusters were denser than those observed with the positive ion mode. These results indicated that some potential markers that might be used to differentiate the samples could not be ionized with the positive mode. Consequently, we used the negative ion mode to differentiate between the two disease groups. The blood metabolic phenotypes generated by AIH and PBC were significantly different from each other using UPLC-MS. In the PLS-DA model, the R²Y and Q² values were 0.991 and 0.943, respectively. The area under the ROC curve was 1 with a sensitivity of 100% and a specificity of 100%. This PLS-DA model accurately diagnosed AIH and PBC. The patients with AIH and PBC for external validation were correctly diagnosed by the PLS-DA model. These results confirmed the feasibility of the LC-MS-based serum metabolomic method for the diagnosis of PBC and AIH. Obviously, four samples of AIH-PBC OS for external validation did not integrate into any group in the T-predicted scatter plot of the PLS-DA model. This finding indicated that this disease has a unique metabolic profile. However, further investigation with more samples is required to validate this presumption.

 

A simplified OPLS-DA model was established using the identified markers and the top five variable importance for project value. These markers included glycochenodeoxycholic acid, FFA, LPC-16:0, PC-16:0/16:0, and SM. The R²Y and Q² values of the OPLS model were 0.872 and 0.810, respectively. The area under the ROC curve was 1 with a sensitivity of 100% and a specificity of 100% respectively. This result showed that the simplified model can also be used to diagnose AIH and PBC accurately.

 

In our study 5 groups of markers were identified, suggesting two pathways that mediate AILD. The first pathway mediates a metabolic disorder of BAs, whereas the second pathway mediates the derangement of lipid metabolism.

 

PBC is a chronic disease with unexplained liver interlobular bile duct damage. Cholestasis as a key feature of PBC is caused by ductal cholestasis in the early stages. Canalicular cholestasis is a secondary consequence of the outflow functional blockade.^[4] All BAs showed the same trend in our study and the intensity distribution of glycochenodeoxycholic acid in the three groups was typical among the BAs identified. As shown in Fig. 3A, glycochenodeoxycholic acid was significantly increased in the PBC group compared with that in the healthy control group. In contrast, serum glycochenodeoxycholic acid was mildly increased in the AIH group compared with that in the healthy control group. The mild increase of BA in AIH is caused by liver cell injury. When liver failure occurs, decreased hepatic clearance of BAs increases their serum level.^[45]

 

The derangement of lipid metabolism in our study was characterized by decreased serum levels of FFA, LPC, PC and SM. All markers of the four groups showed the same trend (Fig. 3B-E). We noted that this derangement was alleviated in the AIH group compared with the PBC group. These results were similar to those of a previous study showing lithocholic acid (LCA)-dependent decreases in lipid levels.^[46] The hepatic level of LCA was reported to elevate in patients with cholestatic liver disease.^{[47, 48]} Moreover, LCA decreased the levels of phospholipids, SM, FFAs, and triglycerides,^{[46, 49, 50]} and it was detoxified via hepatic biotransformation using two pathways. One is hydroxysteroid-sulfotransferase-mediated LCA sulfation, and the other is hydroxylation reaction that is catalyzed by the cytochrome P450 enzymes.^[51,52] When liver function was damaged, liver biotransformation function decreased, including hydroxysteroid-sulfotransferase-mediated LCA sulfation. Therefore, we assumed that higher levels of LCA might cause lipid metabolism disorders via the first pathway in AIH and PBC. Previous studies^{[53, 54]} indicated that the activation of cytochrome P450 correlates with farnesoid X receptor. Mammalian farnesoid X receptor, which is a transcription regulatory factor in bile salt synthesis, is activated by bile acids such as chenodeoxycholic acid or LCA.^{[50, 55]} The derangements of lipid metabolism are weakened in farnesoid X receptor-null mice compared with those in wild-type mice after LCA exposure.^{[6, 52]} As a cholestatic liver disease, the high level of BAs may induce farnesoid X receptor gene transcription in PBC. The level of BAs in AIH was lower than that of active farnesoid X receptor gene transcription (Fig. 3A). Therefore, we hypothesized that these pathways may lead to LCA poisoning in PBC patients, and that the first pathway is predominantly active in PBC patients. This hypothesis may explain why lipid metabolism disorder was milder in patients with AIH than those with PBC. This assumption needs to be validated in further investigations.

 

In conclusion, our study showed that metabolomic methods were accurate for the diagnosis of AIH and PBC. This study was first to use a metabolomic approach for the diagnosis of AILD; therefore follow-up investigations are required. We believed that these findings will help to develop a dependable diagnostic tool based on metabolomic methods for the diagnosis of AILD.

 

 

1 Czaja AJ. Autoimmune liver disease. Curr Opin Gastroenterol 2009;25:215-222. PMID: 19387256

2 Feld JJ, Heathcote EJ. Epidemiology of autoimmune liver disease. J Gastroenterol Hepatol 2003;18:1118-1128. PMID: 12974897

3 Mieli-Vergani G, Vergani D. Autoimmune paediatric liver disease. World J Gastroenterol 2008;14:3360-3367. PMID: 18528933

4 Poupon R. Primary biliary cirrhosis: a 2010 update. J Hepatol 2010;52:745-758. PMID: 20347176

5 Bayer EM, Schramm C, Kanzler S, Lohse AW. Autoimmune liver disease: diagnosis and therapy. Z Gastroenterol 2004;42: 19-30. PMID: 14997400

6 Bogdanos DP, Invernizzi P, Mackay IR, Vergani D. Autoimmune liver serology: current diagnostic and clinical challenges. World J Gastroenterol 2008;14:3374-3387. PMID: 18528935

7 Zen Y, Notsumata K, Tanaka N, Nakanuma Y. Hepatic centrilobular zonal necrosis with positive antinuclear antibody: a unique subtype or early disease of autoimmune hepatitis? Hum Pathol 2007;38:1669-1675. PMID: 17669466

8 Miyake Y, Iwasaki Y, Terada R, Onishi T, Okamoto R, Takaguchi K, et al. Clinical features of Japanese type 1 autoimmune hepatitis patients with zone III necrosis. Hepatol Res 2007;37:801-805. PMID: 17559422

9 Heathcote EJ. Management of primary biliary cirrhosis. The American Association for the Study of Liver Diseases practice guidelines. Hepatology 2000;31:1005-1013. PMID: 10733559

10 European Association for the Study of the Liver. EASL Clinical Practice Guidelines: management of cholestatic liver diseases. J Hepatol 2009;51:237-267. PMID: 19501929

11 Lamers MM, van Oijen MG, Pronk M, Drenth JP. Treatment options for autoimmune hepatitis: a systematic review of randomized controlled trials. J Hepatol 2010;53:191-198. PMID: 20400196

12 Song Q, Liu G, Hu S, Zhang Y, Tao Y, Han Y, et al. Novel autoimmune hepatitis-specific autoantigens identified using protein microarray technology. J Proteome Res 2010;9:30-39.

13 Tahiri F, Le Naour F, Huguet S, Lai-Kuen R, Samuel D, Johanet C, et al. Identification of plasma membrane autoantigens in autoimmune hepatitis type 1 using a proteomics tool. Hepatology 2008;47:937-948. PMID: 18306218

14 Czaja AJ. Autoimmune liver disease. Curr Opin Gastroenterol 2008;24:298-305. PMID: 18408457

15 Bogdanos DP, Mieli-Vergani G, Vergani D. Autoantibodies and their antigens in autoimmune hepatitis. Semin Liver Dis 2009;29:241-253.

16 Michieletti P, Wanless IR, Katz A, Scheuer PJ, Yeaman SJ, Bassendine MF, et al. Antimitochondrial antibody negative primary biliary cirrhosis: a distinct syndrome of autoimmune cholangitis. Gut 1994;35:260-265. PMID: 8307480

17 Liu B, Shi XH, Zhang FC, Zhang W, Gao LX. Antimitochondrial antibody-negative primary biliary cirrhosis: a subset of primary biliary cirrhosis. Liver Int 2008;28:233-239. PMID: 18251980

18 Mendes F, Lindor KD. Antimitochondrial antibody-negative primary biliary cirrhosis. Gastroenterol Clin North Am 2008; 37:479-484. PMID: 18499032

19 Johnson PJ MI. Meeting report: International Autoimmune Hepatitis Group. Hepatology 1993;18:998-1005.

20 Alvarez F, Berg PA, Bianchi FB, Bianchi L, Burroughs AK, Cancado EL, et al. International Autoimmune Hepatitis Group Report: review of criteria for diagnosis of autoimmune hepatitis. J Hepatol 1999;31:929-938. PMID: 10580593

21 Hennes EM, Zeniya M, Czaja AJ, Parés A, Dalekos GN, Krawitt EL, et al. Simplified criteria for the diagnosis of autoimmune hepatitis. Hepatology 2008;48:169-176.

22 Muratori P, Granito A, Pappas G, Muratori L. Validation of simplified diagnostic criteria for autoimmune hepatitis in Italian patients. Hepatology 2009;49:1782-1783.

23 Czaja AJ. Performance parameters of the diagnostic scoring systems for autoimmune hepatitis. Hepatology 2008;48:1540-1548. PMID: 18924244

24 Denzer U, Helmreich-Becker I, Galle PR, Lohse AW. Liver assessment and biopsy in patients with marked coagulopathy: value of mini-laparoscopy and control of bleeding. Am J Gastroenterol 2003;98:893-900. PMID: 12738474

25 Clayton TA, Lindon JC, Cloarec O, Antti H, Charuel C, Hanton G, et al. Pharmaco-metabonomic phenotyping and personalized drug treatment. Nature 2006;440:1073-1077. PMID: 16625200

26 Blow N. Metabolomics: Biochemistry's new look. Nature 2008;455:697-700. PMID: 18833281

27 Ellis DI, Dunn WB, Griffin JL, Allwood JW, Goodacre R. Metabolic fingerprinting as a diagnostic tool. Pharmacogenomics 2007;8:1243-1266. PMID: 17924839

28 Yin P, Wan D, Zhao C, Chen J, Zhao X, Wang W, et al. A metabonomic study of hepatitis B-induced liver cirrhosis and hepatocellular carcinoma by using RP-LC and HILIC coupled with mass spectrometry. Mol Biosyst 2009;5:868-876.

29 Sreekumar A, Poisson LM, Rajendiran TM, Khan AP, Cao Q, Yu J, et al. Metabolomic profiles delineate potential role for sarcosine in prostate cancer progression. Nature 2009;457:910- 914.

30 Wen H, Yoo SS, Kang J, Kim HG, Park JS, Jeong S, et al. A new NMR-based metabolomics approach for the diagnosis of biliary tract cancer. J Hepatol 2010;52:228-233. PMID: 20036026

31 Lin L, Huang Z, Gao Y, Yan X, Xing J, Hang W. LC-MS based serum metabonomic analysis for renal cell carcinoma diagnosis, staging, and biomarker discovery. J Proteome Res 2011;10:1396-1405. PMID: 21186845

32 Zhang J, Yan L, Chen W, Lin L, Song X, Yan X, et al. Metabonomics research of diabetic nephropathy and type 2 diabetes mellitus based on UPLC-oaTOF-MS system. Anal Chim Acta 2009;650:16-22. PMID: 19720167

33 Montoliu I, Martin FP, Collino S, Rezzi S, Kochhar S. Multivariate modeling strategy for intercompartmental analysis of tissue and plasma 1H NMR spectrotypes. J Proteome Res 2009;8:2397-2406. PMID: 19317465

34 Yu K, Sheng G, Sheng J, Chen Y, Xu W, Liu X, et al. A metabonomic investigation on the biochemical perturbation in liver failure patients caused by hepatitis B virus. J Proteome Res 2007;6:2413-2419. PMID: 17539670

35 Zelena E, Dunn WB, Broadhurst D, Francis-McIntyre S, Carroll KM, Begley P, et al. Development of a robust and repeatable UPLC-MS method for the long-term metabolomic study of human serum. Anal Chem 2009;81:1357-1364. PMID: 19170513

36 Silveira MG, Brunt EM, Heathcote J, Gores GJ, Lindor KD, Mayo MJ. American Association for the Study of Liver Diseases endpoints conference: design and endpoints for clinical trials in primary biliary cirrhosis. Hepatology 2010;52:349-359. PMID: 20578151

37 Kuiper EM, Zondervan PE, van Buuren HR. Paris criteria are effective in diagnosis of primary biliary cirrhosis and autoimmune hepatitis overlap syndrome. Clin Gastroenterol Hepatol 2010;8:530-534. PMID: 20304098

38 Chazouillères O, Wendum D, Serfaty L, Montembault S, Rosmorduc O, Poupon R. Primary biliary cirrhosis-autoimmune hepatitis overlap syndrome: clinical features and response to therapy. Hepatology 1998;28:296-301. PMID: 9695990

39 Boberg KM, Chapman RW, Hirschfield GM, Lohse AW, Manns MP, Schrumpf E; International Autoimmune Hepatitis Group. Overlap syndromes: the International Autoimmune Hepatitis Group (IAIHG) position statement on a controversial issue. J Hepatol 2011;54:374-385. PMID: 21067838

40 Want EJ, Wilson ID, Gika H, Theodoridis G, Plumb RS, Shockcor J, et al. Global metabolic profiling procedures for urine using UPLC-MS. Nat Protoc 2010;5:1005-1018.

41 Boccard J, Veuthey JL, Rudaz S. Knowledge discovery in metabolomics: an overview of MS data handling. J Sep Sci 2010;33:290-304. PMID: 20087872

42 Lin L, Yu Q, Yan X, Hang W, Zheng J, Xing J, et al. Direct infusion mass spectrometry or liquid chromatography mass spectrometry for human metabonomics? A serum metabonomic study of kidney cancer. Analyst 2010;135:2970-2978. PMID: 20856980

43 Hohenester S, Oude-Elferink RP, Beuers U. Primary biliary cirrhosis. Semin Immunopathol 2009;31:283-307. PMID: 19603170

44 Wiegard C, Schramm C, Lohse AW. Scoring systems for the diagnosis of autoimmune hepatitis: past, present, and future. Semin Liver Dis 2009;29:254-261. PMID: 19675998

45 Vlahcevic ZR, Pandak WM, Stravitz RT. Regulation of bile acid biosynthesis. Gastroenterol Clin North Am 1999;28:1-25. PMID: 10198776

46 Matsubara T, Tanaka N, Patterson AD, Cho JY, Krausz KW, Gonzalez FJ. Lithocholic acid disrupts phospholipid and sphingolipid homeostasis leading to cholestasis in mice. Hepatology 2011;53:1282-1293. PMID: 21480330

47 Fischer S, Beuers U, Spengler U, Zwiebel FM, Koebe HG. Hepatic levels of bile acids in end-stage chronic cholestatic liver disease. Clin Chim Acta 1996;251:173-186. PMID: 8862472

48 Setchell KD, Rodrigues CM, Clerici C, Solinas A, Morelli A, Gartung C, et al. Bile acid concentrations in human and rat liver tissue and in hepatocyte nuclei. Gastroenterology 1997;112:226-235. PMID: 8978363

49 Boyer JL, Trauner M, Mennone A, Soroka CJ, Cai SY, Moustafa T, et al. Upregulation of a basolateral FXR-dependent bile acid efflux transporter OSTalpha-OSTbeta in cholestasis in humans and rodents. Am J Physiol Gastrointest Liver Physiol 2006;290:G1124-1130.

50 Eloranta JJ, Kullak-Ublick GA. The role of FXR in disorders of bile acid homeostasis. Physiology (Bethesda) 2008;23:286-295. PMID: 18927204

51 Hofmann AF. Detoxification of lithocholic acid, a toxic bile acid: relevance to drug hepatotoxicity. Drug Metab Rev 2004;36:703-722. PMID: 15554243

52 Kitada H, Miyata M, Nakamura T, Tozawa A, Honma W, Shimada M, et al. Protective role of hydroxysteroid sulfotransferase in lithocholic acid-induced liver toxicity. J Biol Chem 2003;278:17838-17844. PMID: 12637555

53 Akiyama TE, Gonzalez FJ. Regulation of P450 genes by liver-enriched transcription factors and nuclear receptors. Biochim Biophys Acta 2003;1619:223-234. PMID: 12573481

54 Krasowski MD, Ni A, Hagey LR, Ekins S. Evolution of promiscuous nuclear hormone receptors: LXR, FXR, VDR, PXR, and CAR. Mol Cell Endocrinol 2011;334:39-48. PMID: 20615451

55 Jeffcoat R. Obesity-a perspective based on the biochemical interrelationship of lipids and carbohydrates. Med Hypotheses 2007;68:1159-1171. PMID: 17257774