Article pubs.acs.org/jpr
Metabonomic Profiles Discriminate Hepatocellular Carcinoma from Liver Cirrhosis by Ultraperformance Liquid Chromatography−Mass Spectrometry Baohong Wang,† Deying Chen,† Yu Chen,† Zhenhua Hu,‡ Min Cao,† Qing Xie,† Yanfei Chen,† Jiali Xu,† Shusen Zheng,‡ and Lanjuan Li*,† †
State Key Laboratory for Diagnosis and Treatment of Infectious Diseases, and ‡Key Lab of Multi-Organ Transplantation, The First Affiliated Hospital, College of Medicine, Zhejiang University, 79 Qing-chun Road, Hangzhou 310003, P.R. China S Supporting Information *
ABSTRACT: Hepatocellular carcinoma (HCC) is the fifth most common cancer worldwide and usually develops in patients with liver cirrhosis (LC). Biomarkers that discriminate HCC from LC are important but are limited. In the present study, an ultraperformance liquid chromatography−mass spectrometry (UPLC−MS)based metabonomics approach was used to characterize serum profiles from HCC (n = 82), LC (n = 48), and healthy subjects (n = 90), and the accuracy of UPLC−MS profiles and alpha-fetoprotein (AFP) levels were compared for their use in HCC diagnosis. By multivariate data and receiver operating characteristic curves analysis, metabolic profiles were capable of discriminating not only patients from the controls but also HCC from LC with 100% sensitivity and specificity. Thirteen potential biomarkers were identified and suggested that there were significant disturbances of key metabolic pathways, such as organic acids, phospholipids, fatty acids, bile acids, and gut flora metabolism, in HCC patients. Canavaninosuccinate was first identified as a metabolite that exhibited a significant decrease in LC and an increase in HCC. In addition, glycochenodeoxycholic acid was suggested to be an important indicator for HCC diagnosis and disease prognosis. UPLC−MS signatures, alone or in combination with AFP levels, could be an efficient and convenient tool for early diagnosis and screening of HCC in high-risk populations. KEYWORDS: serum, metabonomics, hepatocellular carcinoma (HCC), liver cirrhosis (LC), liquid chromatography−mass spectrometry (LC−MS)
■
INTRODUCTION Hepatocellular carcinoma (HCC) is the fifth most common cancer worldwide1 and often progresses from chronic hepatitis B-related liver cirrhosis (LC).2,3 The prognosis of HCC remains poor. The early and accurate detection of HCC from high-risk populations, such as those with LC, can improve the outcomes of HCC patients. Diagnostic imaging, such as ultrasound and computed tomography (CT), as well as serum alpha-fetoprotein (AFP) detection (at a cutoff of 20 ng/mL, AFP20) have been applied for the screening of high-risk patients for the early detection of HCC. However, these two approaches have limitations. For instance, the imaging diagnosis is operator- and equipmentdependent, and the use of this method makes it difficult to discriminate the cirrhosis node from small tumors in patients with a LC background. Additionally, about 1/3 of early stage HCC patients with small tumors (10 000 resolving power at m/z 556.2771). The instrument was previously calibrated with sodium formate; the lock mass spray for precise mass determination was set by leucine enkephalin at m/z 556.2771 with concentration 0.5 ng/μL in the positive ion mode. All analyses were acquired using the lock spray to ensure accuracy and reproducibility. Data were collected in centroid mode, the lockspray frequency was set at 5 s, and data were averaged over 10 scans.
MATERIALS AND METHODS
Chemicals
HPLC-grade acetonitrile and formic acid were purchased from Sigma-Aldrich (St Louis, MO). Distilled water was purified using a Milli-Q system (Millipore, Bedford, MA). Oleamide was obtained from J&K (Shanghai, China). Glycochenodeoxycholic acid (GCDCA), lysophosphatidylcholine (LPC), and phenylalanine standards were also purchased from SigmaAldrich. Enrolled Population and Sample Collection
This study protocol was approved by the Ethical Review Board of the first affiliated hospital of the medical school of Zhejiang University. Written informed consent and individual information were obtained from all of the subjects. A total of 220 subjects were selected that were enrolled in our hospital from January 2006 to December 2010. In the present study, two separate cohorts were included. The preliminary set was consisted of healthy controls (n = 70), patients with LC (n = 28), and patients with HCC (n = 23). Patients were diagnosed by clear clinical laboratory and imaging evidence such as blood chemistry, AFP assay, chest-X-ray, CT, hepatic angiography. These patients have been included in our previous fecal metabolomic and microbiomic study.11,16 Healthy controls, who visited for routine physical examination, were confirmed to have normal liver function and no viral hepatitis, alcohol or nonalcohol fatty liver, or other diseases. The validation set was comprised of healthy controls (n = 20), patients with LC (n = 20), and patients with HCC (n = 59). Besides the above clinical diagnosis method, the diagnosis of all patients was confirmed by histopathological examination by experienced pathologists in our hospital. Tumor stages were classified on the basis of the AJCC staging system. In total, 41 (70%) HCC patients were mainly at AJCC stage I and II, which means the early stage of HCC. LC was revealed in 50% (28/59) of HCC patients. Additionally, all patients were clinically diagnosed with HBVinduced hepatitis without other etiologies, such as autoimmune hepatitis, primary biliary cirrhosis, sclerosing cholangitis, hemochromatosis, a1-antitrypsin deficiency, Wilson’s disease, or alcoholic liver injury, and had not been on any medical treatment for 8 weeks prior to the sample collection. Each sample was numbered according to the collecting time and named and grouped according to the resident number in the hospital. Each subject had only one resident number in the hospital, by which the clinical information of subjects could be
Data Pretreatment and Statistical Data Analyses
All data generated from the UPLC−MS analyses were processed with the MarkerLynx within MassLynx software (Waters). These applications were used to detect, integrate, and normalize the intensities of the peaks to the sum of the peak intensities within that sample.6,17−19 The main parameters used 1218
dx.doi.org/10.1021/pr2009252 | J. Proteome Res. 2012, 11, 1217−1227
Journal of Proteome Research
■
are provided in the Supporting Information. The total peak area from a sample was assigned to a constant of 1000. The resulting multivariate data set, which consisted of a single matrix with RT−m/z pairs for each file, was analyzed with the SIMCA-P +12.0 (Umetrics AB, Sweden) using the multivariate data analysis to enable the easy visualization of any metabolic clustering of the different groups of samples. Prior to multivariate data statistics analysis, the normalized data were mean-centered and pareto-scaled.6,17−20 The unsupervised principal component analysis (PCA) was first utilized in all of the samples to visualize the general clustering and find the outlying data. The four outliers from HCC samples were deleted from the data set before further analysis and corresponded to patients who had symptoms of gastrointestinal tract hemorrhage prior to the sampling. Then, supervised partial least-squares-latent structure discriminate analysis (PLS-DA) was performed to identify biomarkers that contributed to the clustering observed in the PCA.10,18,19 To prevent model overfitting, the supervised models were validated with a permutation test that was repeated 200 times. Potential biomarkers of differentiating HCC from LC were selected according to the variable importance in the projection (VIP) values and the S-plot (Supporting Information, Figure S1). VIP values reflect the influence of each metabolite in the different groups, and the S-plot is a further loading plot and visual method that can be used for the selection of biomarkers. Variables farthest from the origin in S-plots contribute greatly to the difference between the groups. The statistical significance analysis was performed in SPSS 16.0 software. The Dunn’s post test was used to assess the statistical significance of the differences among the HCC, LC, and control samples. The Mann−Whitney test was used to evaluate the statistical significance of the differences between the control and HCC or LC samples (P < 0.01 or 0.05). The receiver operating characteristic (ROC) curve was performed to test the robustness of the PLS-DA model and evaluate the accuracy of identified potential biomarkers in distinguishing HCC from LC. The area under the ROC curve (AUC) was used to assess the candidate biomarkers of HCC, which is equal to the probability that this classifier will rank a randomly chosen positive instance higher than a randomly chosen negative one.21 The diagnostic efficiency of the biomarkers was also evaluated by sensitivity and specificity. Sensitivity was calculated as the ratio of the number of HCC samples that were correctly classified as HCC (true positives) to the total number of HCC samples, and specificity was calculated as the ratio of the number of LC samples that were correctly classified as LC (true negatives) to the total number of LC samples, as described by Chen et al.4 To further understand the pathogenesis of the disease and identify potential biomarkers, the HMDB (www.hmdb.ca) and KEGG (www.genome.jp/kegg/ligand.html) databases were queried with the exact mass of the potential biomarkers. Tandem mass spectroscopy analysis was then performed to identify potential biomarkers, and the results were compared with the available standard compounds in our lab or the MS/MS profiles available in the public database.
Article
RESULTS AND DISCUSSION
Distinguishing Metabolic Profiles between Biopsy-Proven HCC and LC
Diagnosis of HCC at an early stage is important for improved treatment outcomes, especially from patients with LC, from which 80% of HCC are inevitably developed. Meanwhile, determination of HCC from LC becomes much more difficult because regenerative nodules may mimic tumors in cirrhotic livers and elevated serum levels of AFP may also appear. Here, using the UPLC−MS-based serum global metabolite profiles detected all of the HCC patients with low serum AFP levels (AFP < 20 ng/mL) and provided a more exact analysis for cirrhotic patients than conventional AFP detection. Importantly, the metabonomic method used in this study established a noninvasive approach that revealed a global view of the metabolism of the whole body that was applicable to the surveillance of high-risk populations. Furthermore, the serum metabonomics strategy without pretreatment also enabled the process to be high throughput and cost-effective. To profile the “metabonome” during HCC progression, we investigated 220 serum metabolites from 82 patients with HCC, 48 patients with LC, and 90 healthy controls using UPLC−MS to identify a distinct metabolic profile as a fingerprint for predicting HCC. We started our preliminary research to examine whether there were potential biomarkers of HCC and clear classification tendency among metabolic profiles of patients with LC (n = 28), HCC (n = 23), and healthy controls (n = 70). The PCA scores plot based on the data of UPLC−MS positive ion mode analysis of serum samples showed a clear separation between the patients and healthy controls and even between patients with HCC and LC (Supporting Information, Figure S2). Although in the preliminary set the control subjects were composed of different ages ranging from 18 to 84, they clustered together, and this group showed significant differences in metabolites compared to HCC or LC patients (Supporting Information, Figure S2). To identify potential biomarkers contributing to the differences between the disease and control groups, supervised PLS-DA was subsequently performed, and the most important 20 variables were first selected according to their VIP value. Then, the univariate statistical methods were used to select significant candidate variables (data not shown). Subsequently, we designed the validated test in a separate cohort of 59 HCC, 20 LC, and 20 healthy controls to validate the efficiency of UPLC−MS profile in distinguishing HCC from LC and evaluate the efficiency of the candidate markers. A total of 3224 peaks were observed in the final reference list after merging the data from UPLC−MS ES+ mode. Each sample was profiled three times to ensure that each metabolite profile was stable and reproducible (Supporting Information, Figure S3). Evaluation of the unbiased serum metabonomics profiles indicated robust differences, not only between healthy controls and patients with HCC or LC, but also between biopsy-proven HCC and LC patients, highlighting the diagnostic potential of this noninvasive profiling approach. Highly confirmed patient populations are of central importance in the identification of disease-related biomarkers. Besides the clinical diagnosis methods used in the preliminary cohort, the diagnosis of patients in the validation set was all confirmed by histopathological examination by experienced pathologists in our hospital. In total, 41 (70%) HCC patients were mainly at AJCC stage I and II, which meant early stage 1219
dx.doi.org/10.1021/pr2009252 | J. Proteome Res. 2012, 11, 1217−1227
Journal of Proteome Research
Article
Table 1. Demographic and Clinical Information of the Enrolled Population in the Validation Studya patients with HCC parameters
healthy control
patients with LC
total
AFP < 200
AFP ≥ 200
number age (mean, range) sex (male/female) HBsAg ALT (U/l) TB (μmol/L) AFP (ng/mL)
20 42.5 (31−56) 18/2 negative 25.19 ± 4.01 16.19 ± 1.23 3.61 ± 0.25
20 46.5 (30−59) 18/2 positive 74.85 ± 15.21b 361.75 ± 53.02b 81.47 ± 46.45b
59 51 (21−70) 59/5 positive 74.09 ± 9.72b 82.33 ± 22.10b,c 6247 ± 2385b,c
28 51 (21−70) 34/4 positive 72.03 ± 13.08b 142.32 ± 36.55b,d 74.6 ± 15.66b
31 51 (28−63) 31/1 positive 77.22 ± 14.64b 44.48 ± 16.52b,c 15248.4 ± 417.33b
a
Abbreviations: HBsAg, hepatitis B virus surface antigen; ALT, alanine aminotransferase; TB, total bilirubin; AFP, alpha-fetoprotein. Normal values: ALT, 3−50 U/l; TB, 1.0−22.0 μmol/L; AFP, 0.05) (Figure 4B). The synthesis of CSA was demonstrated to occur from aspartate and ureidohomoserine
Figure 3. Internal group difference between HCC patients with or without high levels of serum AFP presented by PLS-DA plot based on the data from UPLC−MS profiling (positive ion mode) data. Legend: red, HCC-H patients (AFP ≥ 200 ng/mL); yellow, HCC-L patients (AFP< 200 ng/mL).
patients with different serum AFP levels might have the similar metabonomics profiles, and a different AFP level was not the factor for the metabonomic profiles that were used as the criteria for classifying HCC patients. Additionally, no significant correlation was found between the metabolic profiles and some of the pathophysiological data of HCC patients, such as age and sex (data not shown). Metabolite Biomarkers for Discrimination between HCC and LC and Their Biological Significance
HCC is considered to be a significantly advanced stage of LC that corresponds to a more aggressive condition that may progress from hepatitis. Therefore, the identification of specific biomarkers for the discrimination of HCC from LC would provide a significant advance toward improved therapy results. To this end, potential biomarkers were selected on the basis of the VIP value and the S-plots derived from the PLS-DA model, which were differentially affected by the different liver conditions and greatly contributed to the discrimination of 1222
dx.doi.org/10.1021/pr2009252 | J. Proteome Res. 2012, 11, 1217−1227
Journal of Proteome Research
Article
Table 3. Potential Biomarkers of HCC, Their Intensities in Metabolic Profiles, And Metabolic Pathwaysa no.
VIP
RT_m/z
1
14.8
0.68_293.0989
[M + H]+
adduct
canavaninosuccinate f
identified results
2
9.32
8.44_496.3411
[M + H]+
LPC −16:0e
3
9.25
0.67_315.0797
[M + Na]+
canavaninosuccinate f
4 5
7.15 6.88
0.52_125.9861 14.19_782.5639
[M + H]+ [M + Na]+
unkown 16:0/18:1-PCe
6
6.77
13.33_780.5517
[M + Na]+
16:0/18:2-PCe
7
6.39
0.67_160.0595
[M + H]+
8 9
5.82 5.74
1.61_120.0799 6.37_432.3088
10
5.04
13.34_758.5677
[M + H]+ [M − H2O + H]+ [M + H]+
canavaninosuccinate f fragment phenylalaninee GCDCAe 16:0/18:2-PCe
[M − 2H2O + H]+ [M + H]+
GCDCA
9.44_282.2785 12.91_828.5505
[M + H]+ [M + Na]+
oleamidee 16:0/22:6-PCe
4.19
13.12_804.5507
[M + Na]+
16:0/20:4-PCe
4.10
14.20_808.5801
[M + Na]+
18:0/18:2-PCe
11
5.02
6.37_414.2989
12
4.82
9.23_524.3709
13 14
4.64 4.45
15 16
e
LPC −18:0e
related pathway
HCC group
LC group
control group
organic acids metabolism phospholipid catabolism organic acids metabolism unkown phospholipid catabolism phospholipid catabolism organic acids metabolism gut flora metabolism bile acid metabolism
338.4 ± 21.2b,d
0.5 ± 0.2b
236.4 ± 18.7
435.5 ± 30.9b,d
221.4 ± 25.0b
956.8 ± 12.2
203.5 ± 15.4b,d
0.5 ± 0.1b
69.0 ± 5.6
105.8 ± 7.0b,d 80.8 ± 7.2b,d
4.5 ± 1.9b 44.3 ± 12.2
50.5 ± 3.4 89.2 ± 3.0
272.4 ± 16.2b,d
152.6 ± 24.3b
330.0 ± 6.9
0.1 ± 0.05b
34.0 ± 2.5
221.6 ± 53.7c 92.1 ± 12.9b
37.5 ± 0.8 0.8 ± 0.3
58.3 ± 12.6b
128.6 ± 3.7
50.7 ± 3.1b,d 85.4 ± 10.0b,d 52.5 ± 8.5b,d
phospholipid catabolism bile acid metabolism
123.6 ± 8.0b,d
phospholipid catabolism fatty acid metabolism phospholipid catabolism phospholipid catabolism phospholipid catabolism
129.1 ± 10.0b,d
55.3 ± 5.8b
306.5 ± 4.3
50.5 ± 10.8b,c 92.2 ± 8.8b,d
152.1 ± 36.3b 21.5 ± 4.4b
14.6 ± 3.9 190.1 ± 6.9
115.1 ± 7.6b,d
49.2 ± 9.1b
159.9 ± 5.1
79.1 ± 6.7b,d
42.4 ± 14.0b
109.4 ± 3.5
64.0 ± 9.7
b,d
160.7 ± 20.1
b
1.71 ± 0.2
a
Abbreviations: RT: retention time; m/z: mass-to-charge ratio; LPC: lysophosphatidylcholine; PC: phosphatidylcholine; GCDCA: glycochenodeoxycholic acid; VIP: variable importance in the project values. Notes: All data are presented as mean ± SD. bCompared with controls, P < 0.01. cP < 0.05. dCompared with LC group, P < 0.01. eMetabolites formally identified by standard samples or published data. f Metabolites putatively annotated by library searching.
by human liver extract.25 Earlier reports have suggested the presence of a metabolic pathway for the formation of guanidinosuccinate from CSA in mammalian liver and kidney.25−29 Guanidinosuccinate is also formed by transamidination from arginine to aspartate in perfused liver.30 This evidence demonstrated that CSA was synthesized and metabolized in the mammalian liver tissue. Taken together, we hypothesized that the markedly different trends of CSA in patients with or without biopsy-proven HCC might be due to the significantly different pathophysiological metabolism of liver cells in patients with LC and HCC. HCC progression is not a linearly developed disease but a much more complex process, which includes systemic deregulation in cell apoptosis, cell cycle, and other aspects of cell proliferation,14 and in turn is associated with significant metabolic alterations. The changed metabolites might be metabolized by tumor cells or by other organs in response to the presence of cancer or the hallmark of predisposing factors to HCC occurrence. In previous reports, identified metabolite markers associated with LC and HCC had the same direction of deviation, such as up- or down-expression.10,13,14,22,31 For example, 20 serum metabolite biomarkers of HCC were identified with the same alteration in LC and HCC utilizing the LC−MS and GC−MS analytical platforms.10 However, in the present study, we first identified the metabolite that was highly expressed in biopsyproven HCC patients but not in the biopsy-proven LC patients without HCC. The metabolites might be of significant practical use of discriminating HCC from LC, although before further extensive investigations, we could not know what type of tissue or cells contribute the relevant metabolic changes and interpret
the mechanism of alteration of the revealed metabolite markers in the progression from LC toward HCC. Therefore, we proposed the possible utility of CSA in monitoring the high-risk population for the early detection of HCC, especially for the discrimination of HCC from LC. The significantly lower levels of lysophosphatidylcholine (LPC) and phosphatidylcholine (PC), such as LPC (16:0), LPC (18:0), PC (16:0), and PC (18:0), which were observed in HCC and LC samples compared with healthy controls, are consistent with other reports7,10,12 (P < 0.01 or 0.05) (Table 3). However, the fold change in the HCC samples was significantly lower than that in LC patients (P < 0.05 or 0.01). The liver is the major site of LPC biogenesis,32 and the disturbance of hepatocellular metabolism of lipids is believed to contribute to the inflammatory status of patients with HBV infections, which, in turn, results in liver injury and tumorigenesis.10,14,33 Therefore, lipid metabolism might play a vital role in the development of HCC. PC and LPC are the two natural phospholipid classes that exist in a mixed form of molecular species and have important biochemical and biophysical functions.34 Moreover, LPC regulates a variety of biological processes, including cell proliferation, tumor cell invasiveness, and inflammation.7 Several mechanisms have been proposed to explain the PC/LPC disturbance, including the inhibition of phospholipase A2 or LCAT activity, the disturbance of lysolecithin acyltransferase activity, and the excess consumption of LPC in the inflammatory response.35,36 Low levels of LPCs imply an anti-inflammatory status in HCC patients, and markedly low levels of LPCs represent a severe immune suppression status in cirrhotic patients. Similar LPC 1223
dx.doi.org/10.1021/pr2009252 | J. Proteome Res. 2012, 11, 1217−1227
Journal of Proteome Research
Article
Figure 4. Intensities of candidate serum metabolite biomarker, canavaninosuccinate (CSA), in different groups and in discrimination of early HCC from LC. (A) Serum CSA were in significantly higher levels in patients with HCC and lower levels in patients with LC compared with those in healthy controls. (B) Serum CSA were in significantly higher levels in HCC with (HCC-H) or without higher AFP (HCC-L) compared with those in patients with LC (** compared with controls, p < 0.01; * p < 0.05; ## compared with LC group, p < 0.01; # p < 0.05). (C) Serum CSA as a candidate marker in discrimination of HCC from LC. Receiver operating characteristic (ROC) calculated using cross validated Y-predicted values of PLS-DA model, AFP20, AFP200, and CSA in discriminating HCC from LC and in discriminating (D) LC patients with HCC vs LC patients without HCC.
reabsorption process into the liver through hepato-enteric circulation. Therefore, the alteration of GCDCA is presumably due to both the liver injury and the different metabolic process involving gut microbiota in serum. A decrease in conjugated bile acids following an initial increase in bile acids has similar clinical significance as CT-derived liver volume/standardized liver volume,37 both of which represent the extent of hepatocyte loss. Additionally, bile acids markedly alter the expression of various genes involved in cholesterol and phospholipid homeostasis, which results in cell death and inflammation and leads to severe liver injury.32 In our previous study, the decreased levels of conjugated bile acids indicated the deterioration of bile acid conjugation, which suggested a poor prognosis for patients with liver failure.12 Bile acids, especially GCDCA, were also revealed to be specific biomarkers for LC patients, which was consistent with a previous report.14 Interestingly, in the present study, TB levels of HCC patients were significantly different between the HCC-L and HCC-H subgroups (Table 1). It has been speculated that the reactivation of AFP in HCC possibly indicates a functional loss of some tumor suppressors in the liver.38 The highly expressed TB could possibly influence the expression of some tumor suppressors in the liver of patients, which could potentially serve as a signature marker for the poor prognosis of HCC. The significantly elevated level of conjugated bile acids in patients with different stages (I to IV) were also revealed in a recent study, which was presumably associated with “acute” disruption of liver function at the early stage of tumorigenesis.10 Therefore, we suggested that the change in TB levels, such as
trends have also been found in other malignant diseases, such as leukemia, malignant lymphomas, renal cell carcinoma, gastrointestinal cancer, and other liver diseases.6 Therefore, we also suggest that serum LPC concentrations could be developed as a general indicator for malignant liver disease. Another potentially important biomarker, glycochenodeoxycholic acid (GCDCA), was significantly increased in both HCC and LC compared with the levels observed for healthy controls (P < 0.01) (Table 3). Interestingly, compared with LC, a significant decrease is observed in HCC patients. This was consistent with differences seen for TB (a traditional marker), the clinical examination data of the study cohort that have been explained above (Table 1). TB has long been used to assess possible liver dysfunction, and high levels of TB imply the decreased consumption of conjugated bile acids. GCDCA is one of the coagulated bile acids that are synthesized in hepatocytes from cholesterol through the activity of hepatic enzymes and excreted into the small intestine via the bile duct in a conjugated form.11 When hepatocyte levels are less than 30% of normal levels, the remaining hepatocytes compensate and perform the necessary daily functions. However, regeneration is depressed, which predicts a poor prognosis. Bile acid conjugation takes place in the hepatocytes, and a decrease in conjugated bile acids can represent massive necrosis of hepatocytes. Besides the direct impact of liver, the disturbance of gut flora in LC or HCC patients might also contribute to the higher level of GCDCA in serum because the deconjugation process takes place in intestinal tract with involvement of gut microbial enzymes followed by a 1224
dx.doi.org/10.1021/pr2009252 | J. Proteome Res. 2012, 11, 1217−1227
Journal of Proteome Research
Article
ion mode metabolic profiles showed good and clear separations between patients with HCC and LC (Figure 2C,D) and LC patients with or without HCC (Supporting Information, Figures S5 and S6). To test the robustness of the model and further validate whether this noninvasive UPLC−MS profiles or identified metabolites had the potential diagnostic potential, the ROC curve analysis was performed. ROC analysis for discrimination of HCC from LC, by the cross-validated predicted Y-values (CVPY) of the PLS-DA model based on data of the UPLC−MS profiles, revealed AUC of 100% (Figure 4C). An AUC of 100% was also obtained for the discrimination of LC patients suffering with or without HCC, which indicated that global UPLC−MS profiles resulted in 100% sensitivity and specificity for discriminating HCC from LC (Figure 4D). To increase the ease in which these methods can be integrated into a test kit, one or few indices would be more applicable. Therefore, CSA, which was lower in LC and higher in HCC (Figure 4A,B) and displayed the largest VIP value in the PLS-DA model of discriminating HCC from LC, was suitable to be an HCC candidate biomarker to differentiate HCC from LC. To test whether CSA could be used as the classifier in discriminating HCC from LC, ROC analysis was performed after the total peak area of the metabolic profiles from one sample was assigned to a constant of 1000. In comparison, using the traditional HCC biomarker AFP20 and AFP200 as the classifiers, AUCs for the patients with HCC or LC were 59.4 and 72.4%, respectively, which were much lower than that obtained by using serum CSA as the classifier (91.8%). Serum CSA apparently had a better performance than AFP in identifying HCC from LC (Figure 4C). In addition, the performance of serum CSA (AUC = 89.6%) as the classifier in detection of HCC was also better than AFP20 (AUC = 61.1%) and AFP200 (AUC = 74.7%) in LC patients suffering with or without HCC (Figure 4D). The optimized cutoff value of serum CSA intensity for the detection of HCC was selected from the above ROC analysis based on the Youden’s index, which was 62.8. Combined with the distribution of the data, 10 (the cutoff value) was selected and used to identify HCC in the 78 biopsy-proven LC and HCC patients (Table 4). In comparison, serum AFP was also
GCDCA, was of important clinical significance and could be an important indicator biomarker for HCC diagnosis and prognosis. A constitutive highly expressed TB that displays an abrupt decrease should be considered as an important clinical parameter that might imply the progression to advanced disease conditions, such as the deterioration of liver function or HCC. Oleamide, an amide of the fatty acid oleic acid (FAA), was also significantly increased in HCC patients (P < 0.05) and markedly increased in LC patients (P < 0.01) (Table 3). FAA regulates discrete processes in central and peripheral systems.12 An anti-inflammatory phenotype is observed in response to elevated peripheral FAA levels in mice via a cannabinoid receptor-independent mechanism.39 The metabolites with similar m/z values and retention times were found to be potential biomarkers for HBV and the induced acute deterioration of liver function.7 In our previous study, oleamide was also revealed as a risk factor for patients with liver failure,7 and its increase indicated the liver injury and anti-inflammatory status of LC and HCC patients. Oleamide is synthesized by cytochrome c in the rat kidney in a reaction that requires ammonium.40 The abnormal increase in oleic acid was also observed in hepatitis that proceeded cirrhosis and HCC and showed a tight correlation with bile acids.14 In this study, the trend in HCC and LC mirrored that of GCDCA. In addition, oleamide potentiates the GABA (A) (γ-aminobutyric acid) receptor that plays a role in regulating neuronal excitability throughout the nervous system.41 The markedly elevated oleamide might make significant contributions to the incidental occurrence of hepatic encephalopathy (HE) in LC patients, which vastly contributes to disease progression. We hypothesized that the oleamide was correlated with the TB, the high level of blood ammonium, and the HE. The markedly increased oleamide in LC disease could also be one discriminating factor against HCC. The obvious increase in phenylalanine in HCC (P < 0.05) and marked increase in LC patients were revealed (P < 0.01) (Table 3). A similar trend of phenylalanine levels in HCC patients was also observed in a recent study.10 Phenylalanine is a protein-derived aromatic amino acid that is catabolized by host gut microbiota to form phosphate-activated glutaminase (PAG).42 PAG is increased in cirrhotic patients and correlates with HE.43 Evidence suggested that gut microbiota was involved in host metabolism.44 Additionally, Sandler et al. found that the host response to translocated microbial products was associated with cirrhosis and predicted the progression to end-stage liver disease in patients with HBV or HCV infection.45 Moreover, we found that intestinal microbial communities were distinctly different in patients with LC compared with those in healthy controls.16,46 The different elevated levels of phenylalanine in LC and HCC patients might be the result of different microbial communities, which supports a potential role for gut microbiota in the pathogenesis and progression of liver diseases in patients with chronic HBV hepatitis. These data provide a new target for future therapies.
Table 4. Sensitivity and Specificity of Individual and/or Combined Use of UPLC−MS Profiles, CSA, and AFP for HCC Diagnosisa test
sensitivity (%)
specificity (%)
AFP20 AFP200 CSA CSA and/or AFP20 UPLS-MS
74 52 79.3 96.4 100
38 90 100 100 100
a
Abbreviations: CSA, canavaninosuccinate. Notes: The normalized intensities such as CSA in the UPLC−MS profiles were subjected to analysis after the total peak area from a sample was assigned to a constant of 1000; AFP20: AFP at a cutoff value of 20 ng/mL; AFP200: AFP at a cutoff value of 200 ng/mL.
Comparison of UPLC−MS Profiles and AFP Levels for the Detection of Early HCC Diagnosis
used as the classifier (the cutoff value was 20 ng/mL and 200 ng/mL, respectively). When AFP20 was used to diagnose HCC patients, a sensitivity of 74% and a specificity of 38% were determined for the detection of HCC. When AFP200 was used to diagnose HCC, a higher specificity (90%) and a lower sensitivity (52%) were determined, compared with those found
The early and accurate detection of HCC from high-risk populations, such as those with LC, is not only of crucial importance to improving the outcomes of HCC patients, but also the most challenging task in diagnosis of HCC. AFP has been long used as the serum marker for HCC. In the above analysis, the PLS-DA models derived from UPLC−MS positive 1225
dx.doi.org/10.1021/pr2009252 | J. Proteome Res. 2012, 11, 1217−1227
Journal of Proteome Research
Article
(2) Bosch, F. X.; Ribes, J.; Diaz, M.; Cleries, R. Primary liver cancer: Worldwide incidence and trends. Gastroenterology 2004, 127 (5 Suppl 1), S5−S16. (3) El-Serag, H. B.; Rudolph, K. L. Hepatocellular carcinoma: Epidemiology and molecular carcinogenesis. Gastroenterology 2007, 132 (7), 2557−76. (4) Chen, L.; Ho, D. W.; Lee, N. P.; Sun, S.; Lam, B.; Wong, K. F.; Yi, X.; Lau, G. K.; Ng, E. W.; Poon, T. C.; Lai, P. B.; Cai, Z.; Peng, J.; Leng, X.; Poon, R. T.; Luk, J. M. Enhanced detection of early hepatocellular carcinoma by serum SELDI-TOF proteomic signature combined with alpha-fetoprotein marker. Ann. Surg. Oncol. 2010, 17 (9), 2518−25. (5) Nicholson, J. K.; Lindon, J. C.; Holmes, E. ’Metabonomics’: Understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data. Xenobiotica 1999, 29 (11), 1181−9. (6) 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 (3), 1396−405. (7) Yang, J.; Zhao, X.; Liu, X.; Wang, C.; Gao, P.; Wang, J.; Li, L.; Gu, J.; Yang, S.; Xu, G. High performance liquid chromatography-mass spectrometry for metabonomics: Potential biomarkers for acute deterioration of liver function in chronic hepatitis B. J. Proteome Res. 2006, 5 (3), 554−61. (8) Yu, K.; Sheng, G.; Sheng, J.; Chen, Y.; Xu, W.; Liu, X.; Cao, H.; Qu, H.; Cheng, Y.; Li, L. A metabonomic investigation on the biochemical perturbation in liver failure patients caused by hepatitis B virus. J. Proteome Res. 2007, 6 (7), 2413−9. (9) Gao, H.; Lu, Q.; Liu, X.; Cong, H.; Zhao, L.; Wang, H.; Lin, D. Application of 1H NMR-based metabonomics in the study of metabolic profiling of human hepatocellular carcinoma and liver cirrhosis. Cancer Sci. 2009, 100 (4), 782−5. (10) Chen, T.; Xie, G.; Wang, X.; Fan, J.; Qiu, Y.; Zheng, X.; Qi, X.; Cao, Y.; Su, M.; Xu, L. X.; Yen, Y.; Liu, P.; Jia, W. Serum and urine metabolite profiling reveals potential biomarkers of human hepatocellular carcinoma. Mol. Cell. Proteomics 2011, 10 (7), No. DOI:10.1074/ mcp.M110.004945 . (11) Cao, H.; Huang, H.; Xu, W.; Chen, D.; Yu, J.; Li, J.; Li, L. Fecal metabolome profiling of liver cirrhosis and hepatocellular carcinoma patients by ultra performance liquid chromatography-mass spectrometry. Anal. Chim. Acta 2011, 691 (1−2), 68−75. (12) Hao, S.; Lian, J.; Xie, Q.; Chen, D.; Guo, Y.; Lu, Y.; Sheng, G.; Xu, W.; Huang, J.; Li, L. Establishing a metabolomic model for the prognosis of hepatitis B virus-induced acute-on-chronic liver failure treated with different live support system. Metabolomics 2011, 7 (3), No. DOI: 10.1007/s11306-010-0260-5. (13) Yang, J.; Xu, G.; Zheng, Y.; Kong, H.; Pang, T.; Lv, S.; Yang, Q. Diagnosis of liver cancer using HPLC-based metabonomics avoiding false-positive result from hepatitis and hepatocirrhosis diseases. J. Chromatogr., B: Anal. Technol. Biomed. Life Sci. 2004, 813 (1−2), 59−65. (14) Yin, P.; Wan, D.; Zhao, C.; Chen, J.; Zhao, X.; Wang, W.; Lu, X.; Yang, S.; Gu, J.; Xu, G. 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 (8), 868−76. (15) Soga, T.; Sugimoto, M.; Honma, M.; Mori, M.; Igarashi, K.; Kashikura, K.; Ikeda, S.; Hirayama, A.; Yamamoto, T.; Yoshida, H.; Otsuka, M.; Tsuji, S.; Yatomi, Y.; Sakuragawa, T.; Watanabe, H.; Nihei, K.; Saito, T.; Kawata, S.; Suzuki, H.; Tomita, M.; Suematsu, M. Serum metabolomics reveals gamma-glutamyl dipeptides as biomarkers for discrimination among different forms of liver disease. J. Hepatol. 2011, 55 (4), 896−905. (16) Chen, Y.; Yang, F.; Lu, H.; Wang, B.; Lei, D.; Wang, Y.; Zhu, B.; Li, L. Characterization of fecal microbial communities in patients with liver cirrhosis. Hepatology 2011, 54 (2), 562−72. (17) Barr, J.; Vazquez-Chantada, M.; Alonso, C.; Perez-Cormenzana, M.; Mayo, R.; Galan, A.; Caballeria, J.; Martin-Duce, A.; Tran, A.;
for AFP20. When CSA was used to predict HCC, a sensitivity of 79.3% and a specificity of 100% were determined. It is obvious that the use of CSA was more accurate than the AFP20 and AFP200 measurements. Intriguingly, the combined use of both CSA and AFP20 measurements increased the sensitivity of the diagnosis to 96.4% and the specificity to 100%.
■
CONCLUSIONS In the present study, the highly reproducible UPLC−MS, in conjunction with the multivariate data analysis and ROC curve analysis, was used for a global serum metabonomic profiling of advanced liver diseases and revealed significant changes in serum metabolites between biopsy-proven HCC and LC patients, discriminating HCC from LC patients with a sensitivity and specificity of 100%. Intriguingly, CSA was also found to perform better for HCC diagnosis than AFP20 (74%/38%) or even AFP 200 (52%/90%), displaying a sensitivity of 79.3% and a specificity of 100% for the biopsy-proven patient cohort. Importantly, the best results for discriminating HCC from LC patients were achieved when CSA was combined with serum AFP20 measurements (sensitivity, 96.4%; specificity, 100%). In addition, 13 potential biomarkers suggested disturbances in organic acids, phospholipids, fatty acid, bile acids, and gut flora metabolism in HCC patients. Although the present results established a metabolic linkage between biomarkers, such as CSA, LPCs, and GCDCA, and HCC, further studies will be carried out to validate those biomarkers, including larger cohorts of patients with different diseases, such as HBV-associated hepatitis, alcoholinduced liver diseases, and nonalcohol fatty liver diseases. In conclusion, this study confirmed the feasibility of using an UPLC−MS serum metabonomic method for the distinguishing HCC from LC and identified CSA as a new potential effective biomarker for the early detection of HCC from LC, alone or in combination with AFP.
■
ASSOCIATED CONTENT
* Supporting Information S
Supplemental quality control strategy, data preparation process, and supporting figures and tables. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*Phone: 86-571-87236759. Fax: 86-571-87236459. E-mail:
[email protected].
■
ACKNOWLEDGMENTS This study was supported by the Natural Science Foundation of China (30901190, 81121002), the Major National S&T Project for Infectious Disease (2012ZX10002-007), the National Program on Key Basic Research Project (2009CB522406 and 2007CB513003), and the Health Bureau of Zhejiang Province Foundation (2008QN010). We acknowledge Prof. Liang Li for the help in metabolites identification, and Prof. Jeremy K. Nicholson and the reviewers for the helpful and thoughtful comments for the manuscript.
■
REFERENCES
(1) Parkin, D. M. Global cancer statistics in the year 2000. Lancet Oncol. 2001, 2 (9), 533−43. 1226
dx.doi.org/10.1021/pr2009252 | J. Proteome Res. 2012, 11, 1217−1227
Journal of Proteome Research
Article
Wagner, C.; Luka, Z.; Lu, S. C.; Castro, A.; Le Marchand-Brustel, Y.; Martinez-Chantar, M. L.; Veyrie, N.; Clement, K.; Tordjman, J.; Gual, P.; Mato, J. M. Liquid chromatography-mass spectrometry-based parallel metabolic profiling of human and mouse model serum reveals putative biomarkers associated with the progression of nonalcoholic fatty liver disease. J. Proteome Res. 2010, 9 (9), 4501−12. (18) Coen, M.; Want, E. J.; Clayton, T. A.; Rhode, C. M.; Hong, Y. S.; Keun, H. C.; Cantor, G. H.; Metz, A. L.; Robertson, D. G.; Reily, M. D.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. Mechanistic aspects and novel biomarkers of responder and non-responder phenotypes in galactosamine-induced hepatitis. J. Proteome Res. 2009, 8 (11), 5175−87. (19) Qiu, Y.; Cai, G.; Su, M.; Chen, T.; Liu, Y.; Xu, Y.; Ni, Y.; Zhao, A.; Cai, S.; Xu, L. X.; Jia, W. Urinary metabonomic study on colorectal cancer. J. Proteome Res. 2010, 9 (3), 1627−34. (20) van den Berg, R. A.; Hoefsloot, H. C.; Westerhuis, J. A.; Smilde, A. K.; van der Werf, M. J. Centering, scaling, and transformations: Improving the biological information content of metabolomics data. BMC Genomics 2006, No. DOI: 10.1186/1471-2164-7-142. (21) Hsieh, S. Y.; He, J. R.; Yu, M. C.; Lee, W. C.; Chen, T. C.; Lo, S. J.; Bera, R.; Sung, C. M.; Chiu, C. T. Secreted ERBB3 isoforms are serum markers for early hepatoma in patients with chronic hepatitis and cirrhosis. J. Proteome Res. 2011, 10 (10), 4715−24. (22) Chen, J.; Wang, W.; Lv, S.; Yin, P.; Zhao, X.; Lu, X.; Zhang, F.; Xu, G. Metabonomics study of liver cancer based on ultra performance liquid chromatography coupled to mass spectrometry with HILIC and RPLC separations. Anal. Chim. Acta 2009, 650 (1), 3−9. (23) Mahadevan, S.; Shah, S. L.; Marrie, T. J.; Slupsky, C. M. Analysis of metabolomic data using support vector machines. Anal. Chem. 2008, 80 (19), 7562−70. (24) Tan, Y.; Yin, P.; Tang, L.; Xing, W.; Huang, Q.; Cao, D.; Zhao, X.; Wang, W.; Lu, X.; Xu, Z.; Wang, H.; Xu, G. Metabolomics study of stepwise hepatocarcinogenesis from the model rats to patients: potential biomarkers effective for small hepatocellular carcinoma diagnosis. Mol. Cell. Proteomics 2011, No. DOI: 10.1074/ mcp.M111.010694. (25) Koller, A.; Aldwin, L.; Natelson, S. Hepatic synthesis of canavaninosuccinate from ureidohomoserine and aspartate, and its conversion to guanidinosuccinate. Clin. Chem. 1975, 21 (12), 1777−82. (26) Koller, A.; Comess, J. D.; Natelson, S. Evidence supporting a proposed mechanism explaining the inverse relationship between guanidinoacetate and guanidinosuccinate in human urine. Clin. Chem. 1975, 21 (2), 235−42. (27) Takahara, K.; Nakanishi, S.; Natelson, S. Cleavage of canavaninosuccinic acid by human liver to form guanidinosuccinic acid, a substance found in the urine of uremic patients. Clin. Chem. 1969, 15 (5), 397−418. (28) Takahara, K.; Nakanishi, S.; Natelson, S. Studies on the reductive cleavage of canavanine and canavaninosuccinic acid. Arch. Biochem. Biophys. 1971, 145 (1), 85−95. (29) Sasaki, M.; Takahara, K.; Natelson, S. Urinary guanidinoacetateguanidinosuccinate ratio: An indicator of kidney dysfunction. Clin. Chem. 1973, 19 (3), 315−21. (30) Perez, G.; Rey, A.; Schiff, E. The biosynthesis of guanidinosuccinic acid by perfused rat liver. J. Clin. Invest. 1976, 57 (3), 807−9. (31) Yang, Y.; Li, C.; Nie, X.; Feng, X.; Chen, W.; Yue, Y.; Tang, H.; Deng, F. Metabonomic studies of human hepatocellular carcinoma using high-resolution magic-angle spinning 1H NMR spectroscopy in conjunction with multivariate data analysis. J. Proteome Res. 2007, 6 (7), 2605−14. (32) Matsubara, T.; Tanaka, N.; Patterson, A. D.; Cho, J. Y.; Krausz, K. W.; Gonzalez, F. J. Lithocholic acid disrupts phospholipid and sphingolipid homeostasis leading to cholestasis in mice. Hepatology 2011, 53 (4), 1282−93. (33) Ockner, R. K.; Kaikaus, R. M.; Bass, N. M. Fatty-acid metabolism and the pathogenesis of hepatocellular carcinoma: review and hypothesis. Hepatology 1993, 18 (3), 669−76.
(34) Chen, S.; Li, K. W. Mass spectrometric identification of molecular species of phosphatidylcholine and lysophosphatidylcholine extracted from shark liver. J. Agric. Food Chem. 2007, 55 (23), 9670−7. (35) Sullentrop, F.; Moka, D.; Neubauer, S.; Haupt, G.; Engelmann, U.; Hahn, J.; Schicha, H. 31P NMR spectroscopy of blood plasma: determination and quantification of phospholipid classes in patients with renal cell carcinoma. NMR Biomed. 2002, 15 (1), 60−8. (36) Lewen Jia, J. C.; Yin, P.; Lu, X.; Guowang, X. Serum metabonomics study of chronic renal failure by ultra performance liquid chromatography coupled with Q-TOF mass spectrometry. Metabolomics 2008, 4 (2), 183−9. (37) Yamagishi, Y.; Saito, H.; Ebinuma, H.; Kikuchi, M.; Ojiro, K.; Kanamori, H.; Tada, S.; Horie, Y.; Kato, S.; Hibi, T. A new prognostic formula for adult acute liver failure using computer tomographyderived hepatic volumetric analysis. J. Gastroenterol. 2009, 44 (6), 615−23. (38) Chen, F. Molecular signature of hepatocellular carcinoma, hope or hype in prognosis and therapy. Semin Cancer Biol. 2011, 21 (1), 1−13. (39) Cravatt, B. F.; Saghatelian, A.; Hawkins, E. G.; Clement, A. B.; Bracey, M. H.; Lichtman, A. H. Functional disassociation of the central and peripheral fatty acid amide signaling systems. Proc. Natl. Acad. Sci. U. S. A. 2004, 101 (29), 10821−6. (40) Driscoll, W. J.; Chaturvedi, S.; Mueller, G. P. Oleamide synthesizing activity from rat kidney: Identification as cytochrome c. J. Biol. Chem. 2007, 282 (31), 22353−63. (41) Yost, C. S.; Hampson, A. J.; Leonoudakis, D.; Koblin, D. D.; Bornheim, L. M.; Gray, A. T. Oleamide potentiates benzodiazepinesensitive gamma-aminobutyric acid receptor activity but does not alter minimum alveolar anesthetic concentration. Anesth. Analg. 1998, 86 (6), 1294−300. (42) Yap, I. K.; Angley, M.; Veselkov, K. A.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. Urinary metabolic phenotyping differentiates children with autism from their unaffected siblings and age-matched controls. J. Proteome Res. 2010, 9 (6), 2996−3004. (43) Romero-Gomez, M.; Ramos-Guerrero, R.; Grande, L.; de Teran, L. C.; Corpas, R.; Camacho, I.; Bautista, J. D. Intestinal glutaminase activity is increased in liver cirrhosis and correlates with minimal hepatic encephalopathy. J. Hepatol. 2004, 41 (1), 49−54. (44) Li, M.; Wang, B.; Zhang, M.; Rantalainen, M.; Wang, S.; Zhou, H.; Zhang, Y.; Shen, J.; Pang, X.; Wei, H.; Chen, Y.; Lu, H.; Zuo, J.; Su, M.; Qiu, Y.; Jia, W.; Xiao, C.; Smith, L. M.; Yang, S.; Holmes, E.; Tang, H.; Zhao, G.; Nicholson, J. K.; Li, L.; Zhao, L. Symbiotic gut microbes modulate human metabolic phenotypes. Proc. Natl. Acad. Sci. U. S. A. 2008, 105 (6), 2117−22. (45) Sandler, N. G.; Koh, C.; Roque, A.; Eccleston, J. L.; Siegel, R. B.; Demino, M.; Kleiner, D. E.; Deeks, S. G.; Liang, T. J.; Heller, T.; Douek, D. C. Host response to translocated microbial products predicts outcomes of patients with HBV or HCV infection. Gastroenterology 2011, 141 (4), 1220−30. (46) Xu, M.; Wang, B.; Fu, Y.; Chen, Y.; Yang, F.; Lu, H.; Xu, J.; Li, L. Changes of fecal Bifidobacterium species in adult patients with hepatitis B virus-induced chronic liver disease. Microb. Ecol. 2011, No. DOI: 10.1007/s00248-011-9925-5.
1227
dx.doi.org/10.1021/pr2009252 | J. Proteome Res. 2012, 11, 1217−1227