ARTICLE pubs.acs.org/jpr
Urinary Metabolic Biomarkers of Hepatocellular Carcinoma in an Egyptian Population: A Validation Study Mohamed I. F. Shariff,*,† Asmaa I. Gomaa,‡ I. Jane Cox,§ Madhvi Patel,† Horace R. T. Williams,† Mary M. E. Crossey,† Andrew V. Thillainayagam,† Howard C. Thomas,† Imam Waked,‡ Shahid A. Khan,† and Simon D. Taylor-Robinson† †
Liver Unit, Division of Diabetes Endocrinology and Metabolism, Department of Medicine, Imperial College London, 10th Floor QEQM Building, St Mary’s Hospital Campus, South Wharf Road, London, W2 1NY, United Kingdom ‡ National Liver Institute, Menoufiya University, Egypt § Institute of Hepatology, Foundation for Liver Research, 69-75 Chenies Mews, London WC1E 6HX, United Kingdom
bS Supporting Information ABSTRACT:
The advent of metabonomics has seen a proliferation of biofluid profiling studies of patients with hepatocellular carcinoma. The majority of these studies have been conducted in single indigenous populations making the widespread applicability of candidate metabolite biomarkers difficult. Presented here is a urinary proton nuclear magnetic resonance spectroscopy study of mainly hepatitis C virus infected Egyptian patients with hepatocellular carcinoma, which corroborates findings of a previous study from our group of mainly hepatitis B-infected Nigerian patients with hepatocellular carcinoma. Using multivariate statistical analysis, in the form of orthogonal signal-corrected partial least squared discriminant analysis, the sensitivity and specificity of the technique for distinguishing patients with tumors from healthy controls and patients with cirrhosis was 100%/94% and 81%/71%, respectively. Discriminatory metabolites included glycine, trimethylamine-N-oxide, hippurate, citrate, creatinine, creatine, and carnitine. This metabolic profile bears similarity to profiles identified in the Nigerian cohort of subjects indicative of tumor effects on physiology, energy production, and aberrant chromosomal methylation. This is the first study to identify similarly altered urine metabolic profiles of hepatocellular carcinoma in two etiologically and ethnically distinct populations, suggesting that altered metabolism as a result of tumorogenesis is independent of these two factors. KEYWORDS: hepatocellular carcinoma, liver cancer, urinary biomarkers, magnetic resonance spectroscopy
’ INTRODUCTION Hepatocellular carcinoma (HCC) is a solid tumor of the liver and the most prevalent primary liver cancer, the fifth commonest, and the third most common cause of cancer-related death worldwide.1-5 The highest incidence rates are in western and central Africa and eastern and southeastern Asia. In the developed world, the incidence of HCC, apart from in Japan, is comparatively low, but there has been a steady overall increase across most western nations over the last two decades.6-9 Hepatitis B and C viruses (HBV and HCV) in the presence of cirrhosis are the most potent risk factors for HCC.1 Egypt has the highest prevalence of hepatitis C virus (HCV) worldwide, with population estimates ranging between 6 and 28%, largely owing to the use of unsterilised needles in a widespread schistosomiasis treatment program in the 1950s1970s using intravenous tartar emetic. HCV has been estimated r 2011 American Chemical Society
to be the main etiological factor of HCC in 40-50% of cases10,11 and is the second most frequent cause of cancer incidence and mortality among men. The only established tumor marker of HCC is serum R-fetoprotein (AFP), a fetal glycoprotein, normally undetectable soon after birth. Unfortunately, AFP has a sensitivity and specificity of less than 70%,12-15 and international guidelines have therefore concluded that AFP is an inadequate surveillance test for HCC.16 Urine and serum have been shown to contain a wealth of metabolic information that may be altered due to underlying disease.17-19 Our group has shown that in a Nigerian population, where HBV is endemic, urinary metabolic profiles can distinguish Received: November 1, 2010 Published: January 31, 2011 1828
dx.doi.org/10.1021/pr101096f | J. Proteome Res. 2011, 10, 1828–1836
Journal of Proteome Research
ARTICLE
Table 1. Subject Demographics characteristic
a
healthy controls
cirrhosis
p-value
HCC
n
17
14
16
Median age (range) (years)
41.0
54.0
51.5
0.01a and 0.37b (Mann-Whitney test for nonparametric data)
Male n (%)
9/17 (53%)
11/14 (79%)
15/16 (94%)
0.02a and 0.32b (Fisher’s test for categorical data)
Ethnicity
Egyptian
Egyptian
Egyptian
HCV Abþ
0
7/14 (50%)
11/16 (69%)
HCC versus healthy controls. b HCC versus cirrhosis.
between patients with HCC, cirrhosis and healthy controls, based on a set of key metabolites (creatinine, carnitine, creatine and acetone).20 Studies in Chinese populations by Chen and colleagues, using ultra high performance liquid chromatography-mass spectrometry (UPLC-MS) and by Wu and colleagues, using gas chromatography-MS (GC-MS) have also identified urinary metabolic changes associated with HCC. The metabolites identified, however, are not concordant across these studies although multivariate analytical models using these data have successfully discriminated disease groups,21,22 suggesting that although the reported metabolites may not be concordant, differentiating patients with tumors is consistent based on broad metabolic profiles. Metabolic profiling studies on HCC tissue23,24 and serum25,26 have also highlighted global discriminatory metabolic alterations in patients with HCC. These studies draw on the hypothesis derived by Otto Warburg in the 1920s that tumor cells have altered metabolic profiles to that of healthy cells, primarily due to altered mitochondrial respiration and glycolytic pathways.27 It is also probably the case that tumors have secondary effects on host physiology that are detectable through biofluid metabolic profiling. A urine diagnostic test for HCC would allow widespread noninvasive screening and surveillance programmes to be implemented in the developing world, where current diagnostic tests (AFP and liver ultrasound) are prohibitively expensive. All of the studies described have been undertaken in single indigenous populations where there is usually a single etiological pathogen of liver disease and HCC. As a result, the results may not hold true for other populations or etiologies. In this study, we present a urinary metabolic profiling study of Egyptian subjects with HCC and liver disease of diverse etiology that has identified corroboratory metabolic changes to our previous study in a Nigerian population. This is the first corroboratory HCC metabolic profiling study in two diverse populations, suggesting that urinary testing for HCC is a tangible possibility across populations.
’ EXPERIMENTAL METHODS Patient Selection
Egyptian urine and serum samples were collected from patients attending the National liver Institute, Menoufiya University, Shbeen El Kom, Egypt. Ethical approval was granted by the research ethics committees at the National Liver Institute, Menoufiya University and the Hammersmith Hospital Campus, Imperial College London . A total of 58 patients were recruited for study between 2005 and 2006: 18 patients with HCC (diagnosed by two imaging techniques showing early arterial enhancement and rapid washout in portal phase or one imaging with serum AFP >400 ng mL-1); 20 patients with clinically or histologically confirmed cirrhotic liver disease; and 20 healthy Egyptian control subjects. Eleven samples
Table 2. Hepatocellular Carcinoma and Cirrhosis Aetiological Factor etiology
HCC
cirrhosis
HCV
11
7
HBV
0
1
Schistosomiasis
0
4
Budd Chiari syndrome Idiopathic
0 5
2 0
16
14
Total
from all three cohorts were identified as “outliers” according to principal component analysis (PCA), a multivariate analytical technique. This was indicative of aberrant spectral findings which unfairly influenced the multivariate data and these samples were excluded from further analysis as detailed in the Results section. This left 16 samples in the HCC group, 14 in the cirrhosis group and 17 in the healthy control group. Subject demographics are shown in Table 1. The median age of the HCC group was not significantly older than the cirrhosis group (p = 0.37), but was significantly older than the healthy controls (p = 0.01). There were significantly more males in the HCC group (15/16) than the healthy control group (9/17) (p = 0.02), but not compared to the cirrhosis group (11/14) (p = 0.32). Staging of HCC stage was according to the Okuda staging system, based on tumor volume and hepatic decompensation, stage I representing early disease and stage III, advanced disease.28 One of the HCC subjects had stage I disease, 10 stage II and 5 stage III. The majority of patients with HCC and 50% of patients with cirrhosis were HCV antibody positive: 11/16 (69%) and 7/14 (50%), respectively. The remaining patients with HCC or cirrhosis had diverse etiologies (Table 2). All healthy controls were of Egyptian origin and had no history of liver disease. Urine Sample Collection
Random 5 mL urine samples were collected into plain glass tubes and stored at -80 °C, 2 to 4 h after collection in Egypt, until air transportation on dry ice to Imperial by A.I.G, M.M.E.C and S.D.T-R. Samples were thawed in London and prepared according to standard methodology:29 400 μL of urine were mixed with 200 μL of buffer solution (0.2 M Na2HPO4/0.2 M NaH2PO4, pH 7.4), and 60 μL of 3-trimethylsilyl-(2,2,3,3-2H4)-1-propionate (TSP)/D2O solution (final concentration of TSP = 1 mM) were added. The TSP served as an internal chemical shift reference (δ 0.00 ppm) and the D2O provided a field lock. The buffered urine sample was left to stand for 10 min and then centrifuged at 13 000 g for 10 min. Five-hundred fifty microliters of supernatant were transferred into a 5 mm diameter glass NMR tube (Wilmad LabGlass, Vineland, NJ, USA) for proton nuclear magnetic resonance (1H NMR) spectroscopy. Samples remained in a sample queue on the NMR autoanalyzer for up to 4 h until data acquisition. 1829
dx.doi.org/10.1021/pr101096f |J. Proteome Res. 2011, 10, 1828–1836
Journal of Proteome Research
ARTICLE
Table 3. Serum Biochemistry Profiles of Subject Cohorts HCC (range)
p-values (Mann-Whitney)
16
14
AFP (IU mL-1)
745 (174-5000)
77 (20-174)
106 (44-318)
150* (97-229)
62 (44-80)