1 NON TARGETED METABOLOMICS AND TARGETED PROTEOMICS TOWARD THE DISCOVERY OF NOVEL HEPATOCELLULAR CARCINOMA BIOMARKERS By ASEM I. FITIAN A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UN IVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2013
2 2013 Asem I. Fitian
3 To my parents Naela and Issa
4 ACKNOWLEDGMENTS I would need to draft a second thesis to fully express my gratitude to the individuals who made this thesis a reality. I first want to thank my graduate advisor R oniel Cabrera for his patience and encouragement, for continuously challenging me with intellectually stimu lating projects and for d emonstrating what good human character is about on a daily basis I next need to thank David R. Nelson for opening his doors and granting me VIP access to his laboratory and for providing the framework and the funding for this m as I am so fortunate to have been amidst and have visionary leadership during my time in the lab. I also want to extend my gratitude to my committee member Robert L. Cook for his guidance encouragement and for always asking gre at questions questions which cultivated my critical thinking during this project. His perspective enabled me to discover nuances in our projects where I may have overlooked them and he served as yet another catalyst of my personal and professional growth d uring this effort. The completion of this work would also not have been possible without the guidance of my program director Richard Snyder who always put my colleag first. In general, I learned as much about how to be a good hum an being as I learned how to be a good scientist from my interaction with these individuals. I also want to thank Yiling Xu for teaching me the proteomic te chniques presented in this work and I extend my gratitude to Cecilia Lopez from the Henry Baker Lab for helping me with the hierarchical cluster analysis of our metabolomics data. The National Science Foundation fellowship that made this experience a possibility also merits recognition.
5 I conclude by extending my love and gratitude to my mother Naela and my father Issa Getting to this point may not have been possible if it were not for their moral leadership, guidance, patience, and unflinching love and support I also want to thank my brother Amer my sisters Bayan and Feda, and my uncle Osama for their friendship, support, and for always putting my best interests first The bonds we share make my efforts a million times easier. Finally, I thank God for giving me this tremendous opportunity.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST O F TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 SYSTEMATIC RE VIEW OF HCC METABOLOMICS STUDIES ............................. 14 Challenges in Liver Cancer Diagnosis ................................ ................................ .... 14 Metabolomics: A Powerful Biomarker Discovery Tool ................................ ............ 15 HCC Metabolomics: A Systematic Review ................................ ............................. 16 HCC Metabolomics: Summary of Trends ................................ ................................ 26 Limitations in HCC Metabolomics Studies ................................ .............................. 27 Conclusions and Future Outlook ................................ ................................ ............. 29 2 NON TARGETED METABOLOMICS OF LIV ER CANCER AND CIRRHOSIS ....... 38 Introduction ................................ ................................ ................................ ............. 38 Materials and Methods ................................ ................................ ............................ 40 Patients ................................ ................................ ................................ ............ 40 Metabolomics Analysis ................................ ................................ ..................... 40 Statistical Analysis ................................ ................................ ............................ 42 In Vitro Validation ................................ ................................ ............................. 43 Medium preparation ................................ ................................ ................... 44 Cell culturing ................................ ................................ .............................. 44 ELISA ................................ ................................ ................................ ......... 45 Results ................................ ................................ ................................ .................... 46 Patients ................................ ................................ ................................ ............ 4 6 Metabolomic Profiles ................................ ................................ ........................ 46 Pathways Involved in Stepwise Hepatocarcinogenesis ................................ .... 47 Lipid metabolism ................................ ................................ ........................ 47 Bile acids ................................ ................................ ................................ .... 48 Acylcarnitines ................................ ................................ ............................. 48 Lipid signalling molecules ................................ ................................ .......... 48 Oxidative stress homeostasis ................................ ................................ .... 49 Fibrinogen cleavage peptides ................................ ................................ .... 50 Amino acids and protein metabolites ................................ ......................... 50 Other metabolites ................................ ................................ ....................... 51 Receiver Operator Characteristic Analysis ................................ ....................... 52 In Vitro Validation ................................ ................................ ............................. 54
7 Discussion ................................ ................................ ................................ .............. 54 Biomarkers of HCC ................................ ................................ .......................... 54 Biomarkers of Cirrhosis ................................ ................................ .................... 61 Conclusion and Future Directions ................................ ................................ ........... 63 3 TARGETED PROTEOMICS: SOLUBLE CD25 AS POTENTIAL HCC BIOMARKER ................................ ................................ ................................ .......... 74 Introduction ................................ ................................ ................................ ............. 74 Materials and Methods ................................ ................................ ............................ 75 Study Population ................................ ................................ .............................. 75 Serum Preparation and sIL 2R ELISA for sCD25 Quantification ...................... 76 Statistical Analysis ................................ ................................ ............................ 77 Results ................................ ................................ ................................ .................... 77 Clinical Characteristics of HCC Patients and DCs ................................ ............ 77 Relationship Between sCD25 Level and Extent of Liver Disease ..................... 78 Utility of sCD25 in Predicting HCC Presence ................................ ................... 78 sCD25 as a Marker for Early Stage HCC ................................ ......................... 79 Correlation between Tumor Burden and sCD25 Level ................................ ..... 79 Discussion ................................ ................................ ................................ .............. 80 LIST OF REFERENCES ................................ ................................ ............................... 88 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 101
8 LIST OF TABLES Table page 1 1 Systematic review of current HCC metabolomics studies. ................................ 32
9 LIST OF FIGURES Figure page 2 1 Clinical characteristics of HCC patients and disease controls ........................... 66 2 2 Random forest supervised classification analysis ................................ ............. 67 2 3 Unsup ervised hierarchical cluster of HCC and DC metabolomes.. ..................... 69 2 4 Major aberrant pathways and putative biomarkers of HCC and cirrhosis .......... 70 2 5 Sensitivity and specificity of putative biomarkers of HCC and cirrhosis.. ............ 72 3 1 Clinical characteristics of DCs (n=61) and HCC patients (n=143). ..................... 84 3 2 Levels of sCD25 and AFP in NHC, DC, and HCC.. ................................ ............ 85 3 3 sCD25 ROC curve for presence of HCC ................................ ........................... 86 3 4 sCD25 ROC curve for early HCC ................................ ................................ ...... 86 3 5 Correlation of sCD25 and AFP with tumor burden ................................ ............. 87 3 6 sCD25 capacity to detect early HCC lesions in the setting of cirrhosis ............. 87
10 LIST OF ABBREVIATIONS A ASLD American Association for the Study of Liver Diseases AC Acid ceramidase AFP Alpha fetoprotein AFP L3 Lectin bound AFP AUC Area under the curve BCLC Barcelona Clinic for Liver Cancer BMI Body mass index CP Child Pugh CPT Carnitine palmitoyltrans ferase CSA Canavaninosuccinate DC Disease control DCA Dicarboxylic acid DCP Des gammacarboxyprothrombin ESI Electrospray ionization FFA Free fatty acid GC Gas chromatography GGT Gamma glutamyltranspeptidase GRO Growth related oncogene alpha H 2 O 2 Hydrogen peroxide HBV Hepatitis B HCC Hepatocellular carcinoma HCV Hepatitis C
11 HETE H ydroxyeicosatetraenoic acid IL 2 Interleukin 2 IVD In vitro diagnostic kDa Kilo Dalton LC Liquid chromatography LOX Lipoxygen ase LPC Lysophosphatidylcholine LPE Lysophosphatidylethanolamine MELD Model for End Stage Liver Disease MMP Matrix metalloproteinase MRM Multiple reaction monitoring MS Mass spectrometry MS MS Tandem mass spectrometry M/Z Mass to charge ratio NAFLD Non alcoholic fatty liver disease NASH Nonalcoholic steatohepatitis NHC Normal healthy control NMR Nuclear magnetic resonance PHGDH Phosphoglycerate dehydrogenase PPAR Peroxisome proliferator activated receptor alpha QTOF Quadrupole time of flight RF Random forest ROC Receiver operator characteristic
12 ROS Reactive oxygen species RSD Relative standard deviation RT Retention time S1P Sphingosine 1 phosphate sCD25 S oluble CD25 SD Standard deviation SELDI Surface enhanced laser desorption/ionization SHMT Serine hydroxymethyltransferase SPH Sphingosine SPHK Sphingosine kinase SRM Selected reaction monitoring TCA Tricarboxylic acid cycle TLC Thrombin light chain TOF Time of flight UPLC Ultrahigh performance liquid chromatography XO Xanthine oxidase
13 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Maste r of Science NON TARGETED METABOLOMICS AND TARGETED PROTEOMICS TOWARD THE DISCOVERY OF NOVEL HEPATOCELLULAR CARCINOMA BIOMARKERS By A sem I. F itian August 2013 Chair: Roniel Cabrera Major: Medical Sciences Translational Biotechnology Hepatocellular c arcinoma has a five year survival rate of 10%. Early detection of HCC improves patient outcome but is exacerbated by the poor diagnostic accuracy of the principal HCC biomarker alpha fetoprotein (AFP) Reso urce poor areas of sub Saharan Africa and East Asia suffer the and c cause of cancer related death in the United States, there is a global urgent need for an inexpensive, repro ducible, and accurate HCC rapid diagnostic test that streamline s earl y detection and improve s the curative outlook for HCC patients Using liver cancer and cirrhosis patient serum, non targeted liquid chromatography/mass spectrometry metabolomics was emplo yed toward the discovery of HCC diagnostic biomarkers. Serendipitously during the metabolomics endeavor, potential biomarkers of cirrhosis were also identified. In a t argeted proteomics investigation, ELISA and multivariate statistics were used to characte rize the utility of immune factor soluble CD25 in diagnosing the presence and early stages of HCC.
14 CHAPTER 1 SYSTEMATIC REVIEW OF HCC METABOLOMICS STUDIES Challenges in Liver Cancer Diagnosis l cancer, possessing a five year survival rate of 10 % that results in between 250,000 to 1,000,000 deaths per year (1 2). HCC culminates from a preexisting long term condition of cirrhosis in 90% of cases (3) and cirrhosis patients are among the best chara cterized individuals at high risk for developing cancer. Notwithstanding the opportunity for clinical surveillance of these patients, the dismal survival rate persists and HCC has now emerged as the fastest rising cause of cancer related death in the Unite d States (4 5). HCC patients in sub Saharan Africa face a particularly grim outlook, with >90% of patients in rural areas of these regions progressing within the first year of HCC onset (6). A major hindrance to successful early diagnosis of HCC stems from the substandard accuracy of the principal HCC diagnostic modalities. Alpha fetoprotein (AFP) is the main biomarker for HCC and despite the inexpensive and reproducible nature of the AFP blood test, its sensitivity of 25 65% (7 8) for HCC has ex acerbated e arly detection of this cancer. The low sensitivity of AFP can be explained by the fact that up to 40% of HCC and cirrhosis patients have normal AFP levels, and that AFP is often elevated in patients without HCC (9 11). Moreover, only 10 20% of patients wit h early stage HCC have elevated AFP levels (12). Exclusion of AFP as an HCC unreliability to accurately screen for early HCC (13). Efforts to overcome this substandard perfo rmance have resulted in the identification and commercialization of novel HCC biomarkers des gamma carboxyprothrombin (DCP) and lectin bound AFP
15 (AFP L3). These markers are ineffective when used alone as HCC biomarkers, however, and even when combined with AFP still demonstrate poor sensitivity for HCC, particularly in the detection of lesions <3cm (14) While magnetic resonance imaging and computed tomography offer better accuracy in HCC diagnosis, these sophisticated diagnostic modalities are both economi cally and logistically incompatible with the resource ality rate (1 5 17 ). Improving e arly HCC detect ion and patient outcome globally requires fulfilling of the urgent need for a reproducible, inexpensive, a nd accurate HCC diagnostic test. Metabolomics: A Powerful Biomarker Discovery Tool Numerous genomic and proteomic screens have been employed to identify potential biomarkers for HCC ( 18 24 ) but to date the markers identified in these studies have not been cli nically fruitful. Because the liver is the hub of carbohydrate, amino acid, and lipid metabolism ( 25 26 ), chronic liver diseases undoubtedly disrupt normal metabolic function. A metabolomics analysis of HCC and cirrhotic tissue can therefore elucidate the metabolic pathways most relevant to the hepatocarcinogenic process and identify metabolites showing promise as HCC biomarkers. Metabolomics is the comprehensive identification of all small metabolites <2kDa in a tissue sample. Through combined gas or liqui d chromatography/mass spectrometry or nuclear magnetic resonance (NMR) instrumentation, metabolomics platforms enable investigators to rapidly screen hundreds of metabolites in a large series of biofluid or solid tissue samples and are capable of simultane ously detecting both acidic (polar) compounds like amino acids and organic (non polar) species such as lipids. Metabolomics platforms are useful translational research tools because they reveal intergroup metabolite expression pattern differences in an aut omated, rapid, high throughput, quality controlled, and
16 reproducible manner. Metabolomics facilitates rapid identification of diagnostic markers, prognostic markers, and lead drug target pathways and their implementation in in drug discovery divisions of p harmaceutical giants underscores their important role in the lead target generation realm (27) HCC Metabolomics : A Systematic Review At least fifteen human HCC metabolomics studies and one rodent HCC metabolomics study have been reported ( 28 42 ). These st udies illustrate the key pathways involved in stepwise hepatocarcinogenesis and reveal metabolites that may have utility as biomarkers of HCC and cirrhosis The findings of these studies are described below and are summarized in Table 1 1 These works reve al deregulation of bile acid metabolism, fatty acid oxidation by way of the carnitine palmitoyltransferase (CPT) shuttle system, amino acid metabolism, and glycerophospholipid metabolism in HCC vs. cirrhosis. Among the first HCC metabolomics studies was performed in 2007 when Yang and colleagues employed high resolution magic angle spinning 1 H nuclear magnetic resonance to delineate the metabolomic profile differences of low grade HCC, high grade HCC, and non involved adjacent cirrhosis using patient biopsy samples (28 ). The grou p reported higher levels o f glycerophosphocholine, choline, creatine, the glycerophospholipid phosphorylethanolamine, glutamine, and glutamate, and decreased gl ucose in HCC vs. cirrhosis. They then compared the metabolomic expression patterns of low grade HCC vs. cirrhosis to eluci date the pathways most prominently involved in the formatio n of early HCC. There were i ncreased levels of lactate, alanine, leucine, glutamate, glutamine, phosphatidylcholine (PC) and PE but lower levels of glucose, and
17 glycogen in HCC vs. uninvolved contr ol tissue These trends were subsequently observed in the comparison between high grade HCC and cirrhosis. The elevated metabolites were directly related to tumor burden, while the diminished compounds exhibited an inverse correlati on. Bile acid elevation s in HCC have been reported previously ( 43 44 ) and may be explained by HCC invasion and obstruction of the bile duct. Bile duct blockage can impede adequate transfer of bile acids to the small intestine thereby impairing sufficient absorption and digestion of fats and leading to a buildup of both bile acids and lipids in the hepatic tumor microenvironment. The finding of increased amino acids in HCC vs. control tissue is consistent with numerous studies implicating elevated amino acids and the enzymes resp onsible for their production in cancer initiation and progression ( 45 48 ). In one large scale metabolomics analysis of 60 cancer cell lines, Jain and colleagues identifi ed glycine as the most significantly and consistently upregulated metabolite in ca ncer cells vs. healthy lines (48 ). Enhanced amino acid production is consistent with the metabolic remodeling hallmark of cancer known as the Warburg effect, which involves a shift from TCA cycle and oxidation to a heightened reliance on glycolysis for energy production (49) Amino acids are important glycolytic enzyme activators, and one recent study demonstrated that serine was an activator of pyruvate kinase M2 ( 50 ), the cancer isoform of glycolyt ic enzyme pyruvate kinase responsible for conversion of phosphoenolpyruvic acid ( PEP ) to pyruvate. The conc omitant observation of elevated lactate in HCC vs. cirrhosis is co nsistent with the Warburg hypothesis. Upregulated amino acids in HCC may al so be explained by the greater protein turnover that transpires in rapidly dividing tumors versus the surrounding non cancerous tissue. The upregulation of choline, the
18 head group of many phospholipids that comprise the plasma membrane, not only reflects i ncreased plasma membrane synthesis demand by the growing tumor but is also consistent with the observed elevations in bile (cholic) acid concentration in HCC vs. control tissue since choline is an important constituent of these fat emulsifying compounds. T he destruction of cholinergic receptors in amyloid plaques and neurofibrillary tangles of the brain has also been linked to encephalopathy, a common neurological disorder witnessed in HCC patients (51) The link between hepatic encephalopathy (HE) and HCC remains unclear, but the parallel elevations of choline and glutamine in HCC patients vs. cirrhosis controls may partially be associated with a cataclysmic loss in cholinergic receptor concentration and the initial stages of HE onset, respectively ( 52 ). M arked alterations in lipid expression between HCC patients and cirrhosis controls were also hallmarks of a recent metabolom ics study by Patterson et al .et al. ( 29 ). HCC patients and cirrhosis controls were gender and sex matched and had Child Pugh A, emble matic of well compensated cirrhosis. Patient plasma samples were processed on an ultrahigh performance liquid chromatography/electrospray ionization quadrupole time of flight mass spectrometry (UPLC/(ESI)QTOF MS) platform. The group reported significantly downregulated levels of lysophosphatidylcholines (LPC) and free fatty acids (FFA) in HCC vs. cirrhosis. LPC(14:0), LPC(20:3), LPC(22:6) and very long chain fatty acids FFA(24:0) (lignoceric acid) and FFA(24:1) (nervonic acid) all trended lower in HCC. The decreases of lignoceric and nervonic acid in HCC vs. cirrhosis were especially patent and may reflect peroxisome proliferator activated receptor alpha (PPAR ) induced enhancement of peroxisomal oxidation. Previous
19 report s implicating heightened PPAR activity in HC C support this hypothesis (53 55 ). Decreases in these FFAs may also be related to increased activity of lignoceryl CoA ligase, the enzyme responsible for very long chain fatty acid catabolism and one acted upon by PPAR (56 ) As mentioned pr eviously, LPCs are the glycerophospholipid building blocks of cell membranes and elevations in these metabolites may reflect the heightened metabolic needs of growing HCCs. LPCs are also major l ipids bound to human albumin (57 ). Decreased serum albumin is a signature of l iver cirrhosis and liver cancer and elevations of systemic LPCs may be attributed to the shortage of app ropriate albumin binding sites that results in increased circulating levels of these metabolites. This study also showed significant ele vations of bilirubin and biliverdin in HCC vs. cirrhosis, a phenomenon consistently reported in HCC metabolomics studies and witnessed in the clinical management of chronic liver disease patients. An insignificant elevation of bile (cholic) acids in HCC wa s also observed, a trend consistent with the aforementioned 1 H NMR study. Moreover, an elevation of bile acids coincides with the findings of a study by Hayashi and colleagues that showed that peroxisomal oxidation may serve an anabolic role in the production of bile acids (58 ). Aberrant lipid and bile acid metabolism were also hallmarks of HCC vs. cirrhosis in a recent w ork by Ressom and colleagues (36 ). Using UPLC/QTOF MS, serum metabolomic profiles we re compared in 78 HCC patients representing various etiologies and 184 cirrhosis disease controls. Patients were matched by age, gender, race, etiology, Model for End Stage Liver Disease (MELD) scores, and AFP levels. Mass spectrometry peak areas showed th at in contrast to the findings of the above mentioned studies, the bile acids glycodeoxycholic acid (GDCA), glycocholic acid (GCA) and
20 taurochenodeoxycholic acid (TCDCA) all trended significantly lower in the HCC group vs. cirrhosis controls. Among the oth er trends observed was overexpression of LPCs and lys ophosphatidylethanolamines (LPE ) in HCC vs. cirrhosis, and upregulation of sphingosine 1 phosphate (S1P) in HCC. The overexpression of LPCs is in agreement with the Patterson study while the upregulation of S1P, a signalling lipid, was a novel finding in HCC metabolomics. S1P has been heavily implicated in promoting the progression of seve ral cancers including HCC (59 60 produced via acid ceramidase (AC) activity on cerami de. Ceramides are shown to possess apoptotic effects, while sphingosine 1 phosphate is demonstrated as an anti apopt otic and angiogenic molecule (61 ). This cell turnover control mechanism is known ator of cell death homeostasis. Higher S1P in HCC may also reflect an independent enhancement of SPHK activity. One study demonstrated the antitumor property of a selective SPHK2 inhibitor in HCC xenografts (62 ), implicating SPHK as a promoter of HCC progr ession. Heightened AC activity that results in increased sphingosine may lead to a larger reservoir of S1P via sphingosine kinase (SPHK) and may promote the establishment of a microenvironment conducive to HCC initiation. Further metabolomics work by Xiao, Ressom and colleagues (29) revealed metabolomic differences between 40 HCV associated HCC patients and 49 cirrhosis controls using UPLC/QTOF MS. The team also studied the link between metabolite levels and tumor burden. In contrast to their above mentione d study, this analysis found significant downregulation of glycerophospholipids (LPCs and LPEs) but again observed decreases in bile acids. Among the novel findings of this investigation
21 included decreased levels of acylcarnitines in HCC vs. cirrhosis. Res ults also showed that as tumor burden worsened, the expression of acylcarnitines and bile acids trended significantly downward. Stage II and III HCC exhibited lower levels of these metabolites in comparison to stage 1 (staging based on the American Joint C ommittee on Cancer Tumor Lymph Node Metastatic Disease (TNM) system). The downregulation of fatty acids, acylcarnitines, and bile acids in HCC vs. cirrhosis supports the cancer Warburg effect involving a metabolic shift from TCA cycle and mitochondrial o xidation to a heightened reliance on glycolysis for energy production. To undergo oxidation in the mitochondrial matrix, free fatty acids (fatty acyl CoA) must link with cytosolic carnitine via CPT shuttle system enzymes to form acetylcarnitines. Acetylc arnitines are capable of penetrating the inner mitochondrial membrane and once inside the matrix, CPT enzymes liberate fatty acyl CoA allowing oxidation to ensue. D ecreased concentration of acylcarnitine s in HCC vs. cirrhosis suggests impair ment of CPT1 mediated formation of these compounds from FFA and carnitine. Bile acids are synthesized in the liver and aid in fatty acid absorption and digestion and their downregulation in HCC may reflect this metabolic shift away from oxidation or reduced de novo bile acid production caused by the obliteration of healthy hepatocytes during chronic liver disease. Di minished bile acids may also reflect constitutive activation of farnesyl X receptor (FXR), a bile acid activated nuclear receptor that is also activated by a variety of other lipids including eicosanoids (63) FXR silences Cyp7A1 catalyzed production of bile acids and is implicated in promoting progression of HCC by multiple studies (64 65 ). In the Fujino et al study of FXR induced promotion of HCC
22 progr ession, siR mediated supplementation of FXR enhanced HepG2, Huh7, and HLE HCC cell line progression while FXR knockdown halted this progression This trend of diminished bile acids in HCC vs. cirrhosis in the Xiao work was also observed in a recent LC/QTO F MS study using serum from 82 HBV associated HCC patients, cirrhosis pa tients, and healthy controls (38 ). Wang et al reported strong analysis also showed a striking 677 fold elevation in canavaninosuccinate (CSA) level in HCC patients vs. their cirrhosis counterparts ( P < .01). Subsequent receiver operator characteristic analysis revealed that the sensitivity and specificity of AFP and CSA for distinguishing the HCC patie nts from cirrhosis controls were: AFP 20ng/mL 74% and 38%; AFP 200ng/mL 52% and 90%; CSA 79.3% and 100%; combined CSA and AFP 20 ng /mL 96.4% and 100%. CSA is the precursor to fumarate, a key metabolite of the TCA cycle, and elevation of CSA may reflect impairm ent in the formation of TCA intermediates and promotes the above referenced Warburg theory involving a metabolic shift from oxidative to anaerobic energy production in cancer microenvironments, which are often more hypoxic than healthy tissue (66 ). Becaus e HCC is linked to a variety of diverse etiologies including viral hepatitis, alcoholic cirrhosis, non alcoholic fatty liver diseas e (NAFLD), steatohepatitis, and aflatoxin B1 metabolomics is useful for revealing the variation in metabolism these etiologi es cause. Given the limited knowledge of the metabolomic differences between hepatitis B (HBV) and hepatitis C associated HCC, Zhou and colleagues compared the serological metabolomic expression patterns of HCV cirrhosis associated HCC and HBV cirrhosis HC C (39 ). The group also looked at whether metabolomic alterations in
23 HCV HCC and HBV HCC were significantly different from cirrhosis and healthy control metabolomes. The study revealed massive downregulation of LPCs in HBV HCC and HCV HCC versus cirrhosis c ontrols, with a greater magnitude decrease of LPC expression in HBV infected HCC patients vs. cirrhosis controls than their HCV HCC counterparts. The analysis also showed significant elevation of bile acids, heme pigmentation compounds bilirubin and bilive rdin, upregulation of acylcarnitines and downregulati on of glycerophospholipids in cirrhosis patients vs. healthy controls, suggesting that these metabolites were signatures of the onset of cirrhosis. Metabolomic profile comparison between HBV cirrhosis pa tients vs. patients with HBV only revealed a similar global downregulation of LPCs in the cirrhosis cohort vs. viral hepatitis controls. The findings of this study suggest a progressive downregulation of LPCs during the course of progression from viral hep atitis to cirrhosis and a bottomed out expression occurring with HCC. This trend may reflect first the substantial cell death that occurs during cirrhosis resulting in diminished LPC levels, followed by a subsequent massive turnover of residual LPCs by the growing HCC that further exhausts the LPC reservoir. The magnitude of LPC decrease in HBV HCC vs. cirrhosis was greater than HCV HCC vs. cirrhosis, suggesting that HBV exerts a more prominent influence on LP C metabolism than HCV Other classes of metaboli tes routinely implicated in metabolomics studies include amino acids and their derivatives, a notable example of which are the glutamyl peptides. A recent capillary electrophoresis time of flight mass spectrometry analysis involving sera obtained from HCV associated HCC patients, cirrhosis patients, HBV and chronic hepatitis C patients, and healthy volunteers showed markedly sig nificant
24 variations in glutamyl e xpression among these groups (35 ). No differences in glutamyl peptide expression were observed between HCC and cirrhosis controls, but several significant alterations were witnessed in the HCC vs. viral hepatitis, HCC v s. NHC, cirrhosis vs. viral hepatitis, and cirrhosis vs. NHC comparisons. In general, HCC glutamyl expression was increased in compari son to healthy controls, while glutamyl peptides were decreased in HCC vs. viral hepatitis. glutamylglycine, gluta mylalanine, glutam ylvaline and glutamylserine, glutamyltaurine, glutamylleucine, and glut amyllysine were all strongly ( .0001 < P <.001) downregulated in HCC vs. viral hepatitis B and C infection. glutamyl peptides are precursors to glutathione, the chief antioxidant compound primarily synthesized in the liver. The respective elevation of these intermediates in HCC and cirrhosis patients vs. NHC suggests that increased oxidative stress contributing to liver dysfunction calls for heighted producti on of these precursors to combat this deteriorative process. This is consistent with reports implicating oxidative damage as a key pathway in HCC progression and one that increases patient vulner ability for HCC recurrence (67 6 8). glutamyl peptides are also liberated in free form by gamma glutamyl transpeptidase (GGT) mediated breakdown of glutathione. GGT is an enzymatic signature of liver disease and a marker routinely used in the clinic to assess the severity of liver dysfuncti on. Excess breakdown of glutathione may explain the relative elevation of glutamyl peptides in cirrhosis vs. NHC and the impaired oxidative stress neutralization commonly witnessed in HCC. Protein metabolism perturbations can also be detected through s ur face enhanced laser desorption/ionization time of flight mass spectrometry (SELDI TOF MS)
25 metabolomics. Wu et al. tapped this technology to investigate the metabolomic expression profiles of HBV associated HCC patients, cirrhosis patients, and healthy cont rols (31 ). Their results reveal the upregulation of two proteins, growth related oncogene alpha (GRO ) and thrombin light chain (TLC), in HCC patients vs. cirrhosis controls. To validate these putative markers, serum from an alternative series of HCC, cirrhosis, healthy, and cancer control patients was subjected to SephdexTM Peptide 10/300GL, HitrapTM CM and Mono Q 5/50 GL liquid chromatography for protein separation. The separated products were purified using SDS PAGE and their identities were confirmed using electrospray ionization mass spectrometry (ESI MS). In this alternative series of patients, GRO was upregulated four fold in HCC patients vs. cirrhosis controls and its level directly correlated with HCC tumor burden. GRO also trended higher in gastric, nasopharyngeal and lung cancers relative to cirrhosis patients, but the magnitude of GRO ele vation in cancer controls vs. cirrhosis was smaller than the elevation witnessed in HCC vs. cirrhosis, suggesting that GRO may play a more prominent role in HCC progression. Interestingly, TLC was 1.4 times more elevated in HCC patients vs. cirrhosis con trols but downregulated in the cancer control patients vs. cirrhosis, suggesting that TLC is a unique signature of HCC. T he combined sensitivity and specificity of GRO +TLC+AFP for discriminating HCC patients from cirrhosis and healthy controls was 91.7% and 92.7% respectively. At the 400ng/mL cutoff for HCC diagnosis AFP had a sensitivity of 69% and a specificity of 83%. GRO is a chemokine involved in invoking leukocyte cell migration and is associated with pro inflammatory processes angiogenesis, a nd cancer (69 71 ). Its elev ation may be a signature of viral hepatitis associated HCC immune deregulation
26 involving heightened monocyte migration and increased inflammation, and a subsequent increased likelihood for successful host evasion of tumor prophyl actic mechanisms (62). TLC is a protein cleavage fragment that is generated from matrix metalloproteinas e associated (MMP) peptide cleavage (31 ). Its upregulation may be explained by simultaneous E cadherin loss and MMP activation by Twist1 which has been s hown to promote HCC expansion (7 3). Its utility as a cancer biomarker was also shown in a SELDI TOF MS study of gastric cancer by Ebert and colleagues, where TLC accurately distinguished gastric cancer patients from patients without cancer with a sensitiv ity of 8 9.9% and a specificity of 90% (7 4). Taken together, the aberrations in protein and amino acid expression in HCC revealed by metabolomics may be tapped toward the development of clinically fruitful HCC diagnostics. HCC Metabolomics: Summary of Trend s Several metabolomic expression themes emerged from these investigations: LPCs were significantly diminished in HCC vs. cirrhosis in three out of five studies reporting significant alteration of LPC expression among these groups Bile acids GCA, GCDCA an d GDCA were decreased in all three studies reporting significant differences among HCC vs. cirrhosis patients Acylcarnitines trended lower in HCC vs. cirrhosis Amino acids trended higher in HCC in all studies reporting significant expression differences between HCC and cirrhosis Heme pigmentation molecules bilirubin and biliverdin were significantly upregulated in HCC. Elevation of LPCs in HCC vs. cirrhosis is consistent with increased demand of glycerophospholipids by the growing tumor. Further, LPCs a re major lipids bound to albumin, and the loss of albumin commonly witnessed in chronic liver disease and HCC may mean a shortage of docking sites for LPCs, resulting in increased systemic levels
27 of these glycerophospholipids. Diminished bile acids in HCC vs. cirrhosis reinforces the Warburg effect commonly witnessed in cancer studies involving a shift in energy production away from oxidative processes like oxidation and the TCA cycle toward anaerobic glycolysis, which is more suitable to the hypoxic or a noxic tumor microenvironment. Moreover, the loss of bile acids in HCC significantly impairs fatty acid absorption and digestion and is consistent with the general trend of increased lipids in HCC vs. cirrhosis. Limitations in HCC Metabolomics Studies Nota ble contradictions of reported pathway trends in HCC vs. contro ls included upregulation of LPC, fatty acids, acylcarnitines and bile acids in some studies and downregulation of these respective metabolite classes in others. This heterogeneity is likely due to differences in study design. Some studies matched patients by age and gender, but the majority of referenced HCC metabolomics studies did not extensively characterize HCC and cirrhosis patients by demographic and clinical characteristic parameters. Onl y one study reported Model for End Stage Liver Disease (MELD) scores for HCC and cirrhosis patients, and one study indicated the Child Pugh status of their HCC or cirrhosis cohorts, which exacerbates efforts to pinpoint expression patterns to. Body mass in expression profile, also went unreported for HCC, cirrhosis, and healthy study subjects in all studies cited in this review. BMI can have a significant influence on the relative metabo lite expression differences between patients, particularly with regard to adiposity. Failure to control for BMI may explain the noticeably different trends of LPC, FFA, and
28 bile acid expression in HCC vs. controls among these studies. It is likely that the patients recruited for these metabolomics studies had wide ranging BMIs. Differences in HCC etiology also went largely unreported in the studies discussed and may have strongly contributed to the discord of metabolic expression profiles in these studies. To limit the influence of potential cofounders such as comorbidity, BMI, age or gender, and etiology on HCC metabolomes, patient clinical characteristics should be controlled for more conscientiously in future HCC metabolomics studies. The critical metab olomic comparison between HCC patients and cirrhosis controls was reported in 8/15 studies referenced in this review. Given that the majority of primary liver cancer cases occur in patients with a preexisting condition of cirrhosis, the metabolomic compari son between HCC and cirrhosis is more clinically relevant than the comparison of HCC vs. NHC HCC is a complex heterogeneous disease and HCC patients often present with multiple comorbidities. It is therefore likely that marked metabolomic differences will be observed between HCC patients and healthy subjects. Differences between HCC patients and cirrhosis controls are more subtle than HCC vs. NHC and it would be expected that serious metabolomic differences exist between HCC/NHC and cirrhosis/NHC Given th at HCC arose in the background of cirrhosis in all reported HCC metabolomics studies it is impossible to determine whether the alterations witnessed in HCC vs. NHC are related to HCC or cirrhosis. To verify the putative biomarkers discovered in these stud ies, the groups primarily employed principal component analysis (PCA), orthogonal projection to latent structures (OPLS), supervised projection to latent structures discriminant analysis (PLS
29 DA), or the random forest machine learning algorithm. In general the groups which performed HCC metabolomics did so with the focus on the potential standardization of metabolomics as a high throughput diagnostic platform to be used in the clinical setting. While t hese approaches classified appropriate groups with exce llent accuracy and high area under the curve values the urgent need for a biomarker blood test was not conscientiously addressed in these studies. A lmost none of the studies verified the metabolite expression patterns in vitro through enzyme linked immuno sorbent assay (ELISA) or western blotting techniques, and no study validated the expression patterns using in vivo models. This may reflect the fact that conventional in vitro assays lack the sensitivity needed to accurately quantify small metabolites weig hing <1kD in a tissue sample. While metabolomics based diagnostics may be of benefit to industrialized novel HCC in vitro diagnostic can improve disease monitoring globally and can seamlessly be employed in resource poor a reas A potential avenue for fulfilling this global need while employing sensitive metabolites is, therefore, the development of targeted in vitro assays possessing sensitivity for small metabolites that rival s the sensitivity of a mass spectrometer. Conclusions and Future Outlook The results of these studies show inconsistencies in the trends of bile acids, LPCs, heme pigmentation molecules, acylcarnitines and fatty acids in HCC vs. cirrhosis. There was never theless considera ble overlap in the pathway s identified by these studies which best distinguished HCC from cirrhosis Future metabolomics studies with larger and better characterized patient cohorts should help resolve these differences.
30 Other metabolites that showed promise as biomarkers of HCC included canavaninosuccinate, which showed a striking 680 fold elevation in HCC patients vs. cirrhosis controls and outperformed AFP in sensitivity and specificity. Sphingosine 1 phosphate (up), GRO (up), and thr ombin light chain (up) were other novel putative HCC bi omarkers that had superior predictive utility for HCC than AFP Although the comparison of HCC vs. cirrhosis was not reported in 7/15 studies, the majority of these studies reported significant altera tions between cirrhosis patients and healthy volunteers and the se results reveal a panel of putative markers for cirrhosis. Currently, the gold standard of cirrhosis and HCC diagnosis is the biopsy, an invasive procedure, and an accurate cirrhosis biomarke r may eliminate the patient discomfort and expenses associated with this procedure (76) Among the putative markers of cirrhosis include the purine metabolites hypoxanthine and inosine ( 41 ), both significantly downregulated five and six fold respectively i n cirrhosis vs. NHC. glutamylalanine, glutamylvaline, glutamylglutamine, glutamylphenylalanine and glutamylcitrulline were also significantly elevated in cirrhosis vs. NHCs. Increases in the glutamyl peptides in cirrhosis vs. NHC and HCC vs. NHC indicates heighted production of glutathione to combat the oxidative damage process commonly implicated as a promoter of tumor initiation and progression. glutamyl peptides are also liberated in free form by gammaglutamyl transpeptidase (GGT) mediated breakdown of glutathi one. GGT is an enzymatic signature of liver disease and a marker routinely used in the clinic to assess the severity of liver dysfunction. The elevation of glutamyl peptides in cirrhosis vs. NHC may therefore conversely be explained by excessive
31 breakdow n of glutathione resulting in impaired oxidative stress neutralization that is commonly witnessed in HCC. The findings of these analyses show a potential role for LPCs, bile acids, acylcarnitines, amino acids, GRO lignoceric acid, nervonic acid, thromb in light chain, and S1P in distinguishing HCC patients from cirrhosis controls. Diagnostic panels measuring for fasted amino acid levels in HCC patients may also be of benefit in early HCC detection, but validation studies are needed. Diagnostic panels mea suring for acylcarnitines, LPCs, or bile acids may be of utility but consensus regarding their expression pattern in HCC vs. cirrhosis was not achieved from the results of these metabolomics studies and must first be established. glutamyl peptides, and nucleic acid degradation products hypoxanthine and/or inosine may also be of utility as markers of cirrhosis. Future HCC metabolomics study designs must compare the expression patterns of HCC patients vs. cirrhosis controls. Given that HCC culminates from progressively worsening chronic liver disease, metabolomics will also prove useful for unveiling the pathways involved in th e stepwise hepatocarcinogenesis. Moreover, pathways involved in the progression from chronic viral hepatit is infection/alcohol injury/NAFLD to cirrhosis are of particular value for identification of potential cirrhosis biomarkers that can be used in a non invasive manner In vitro or in vivo validation and metabolomic expression response to treatment are avenu es that remain largely unexplored in HCC metabolomics and are needed to bolster the findings of these studies. Opportunities for metabolomic profiling of well matched HCC vs. cirrhosis samples are fertile and should be endeavored.
32 Table 1 1. Systematic r eview of current HCC metabolomics studies. P<.05 indicates significant alteration. Author Tissue (organism) HCC Etiology Significantly Altered Metabolites in HCC Patients vs. Cirrhosis Controls Main Pathways distinguishing HCC from Cirrhosis Significantly Altered Metabolites in Cirrhosis Patients vs. Healthy Volunteers Main Pathways distinguishing Cirrhosis from Healthy Volunteers Zhou L, Wang Q, Yin P, et al Serum (human) Cirrhosis (of unknown origin) Not reported N/A Tryptophan GCA GCDCA LPC 0:0/ 14:0 FFA (C16:1) FFA (C16:0) FFA (C18:2) FFA (C18:1) FFA (C18:0) FFA (C20:5) FFA (C20:4) FFA (C20:2) FFA (C22:6) FFA (C22:5) C16:1 CN C10 CN C10:1 CN C8 CN C6 CN Tryptophan metabolism Carnitine palmitoyl transferase shuttle system Bile acid Gao H, Lu Q, Liu X, et al. Serum (human) Not reported Not reported N/A Isoleucine Leucine Valine Acetate N acetylglycoproteins Acetoacetate Pyruvate Glutamine a ketoglutarate choline taurine glycerol unsa turated lipid tyrosine 1 methylhistidine phenylalanine Amino acid metabolism TCA cycle
33 Table 1 1. Continued Author Tissue (organism) HCC Etiology Significantly Altered Metabolites in HCC Patients vs. Cirrhosis Controls Main Pathways distinguishi ng HCC from Cirrhosis Significantly Altered Metabolites in Cirrhosis Patients vs. Healthy Volunteers Main Pathways distinguishing Cirrhosis from Healthy Volunteers Patterson AD, Maurhofer O, Beyoglu D, et al. Plasma (human) HCC n=20: EtOH n=6 HBV n=3 HCV n=5 NASH n=3 NASH/alcoholic steatohepatitis n=1 HH n=2 LPC(14:0) LPC (18:1) LPC (20:4) LPC(20:3) LPC(22:6) FFA(24:0) lignoceric acid FFA(24:1) nervonic acid bilirubin biliverdin Lysophosphatid ylcholine (glycerophosph olipid metabolism) Very long chain fatty acid metabolism Heme pigment metabolites Not reported N/A Li S, Liu H, Jin Y, Lin S, Cai Z, Jiang Y Serum (mouse) EtOH induced cirrhosis, HCC xenografts (HepG2 hindquarter injection) Not reported N/A Leucine Phenylpyruvic acid Phenylalanine Tryptophan LPE(16:0) LPE(18:0) LPC (16:0) LPC(20:1) LPC(22:6) PC(16:0/18:3) PC(12:1/24:3) PC(16:0/20:4) PC(16:0/22:6) PC(18:0/20:4) SM(d18:0/16:1) Amino acid metabolism LPE metabolism LPC metabolism PC metabolism Soga T, Sugimoto M, Honma M, et al. Serum (human) H CV Investigated, but none found N/A Glutamylalanine Glutamylvaline Glutamylglutamine Glutamylphenylalanine Glutamylcitrulline Methionine sulfoxide Oxidative stress homeostasis
34 Table 1 1. Continued Author Tissue (organism) HCC Eti ology Significantly Altered Metabolites in HCC Patients vs. Cirrhosis Controls Main Pathways distinguishing HCC from Cirrhosis Significantly Altered Metabolites in Cirrhosis Patients vs. Healthy Volunteers Main Pathways distinguishing Cirrhosis from Health y Volunteers Xiao JF, Varghese RF, Zhou B, Nezami Ranjbar MR, et al. Serum (human) HCV dihydroxy cholan 24 oic acid dihydroxy cholest 24 en 26 oic acid GCA GDCA GCDCA TCDCA Linoelaidyl carnitine Oleoylcarnitine Palmitoyl carnitine O octanoyl R carnitine LPC(20:1) LPC(20:4) PE(20:4/18:1) 4E;15Z Bilirubin IXa 15,16 dihydrobiliverdin 3 ganidinopropionicacid Tetracosahexaenoic acid 3 hydroxy eicosanoic acid Oleamide Phe Phe Bil e acid biosynthesis Long chain carnitine metabolism Small peptide metabolism Not investigated N/A Wan g B, Chen D, Chen Y, et al. Serum (human) HCC n=82: HBV cirrhosis n=41 HBV only n=41 LPC 16:0 LPC 18:0 16:0/18:1 PC 16:0/18:2 PC 16:0/20:4 PC 16:0/22:6 PC 18:0/18:2 PC Phenylalanine GCDCA Canavaninosuccinate Phospholipid catabolism Gut flora metabolism Bile acid metabolism Organic acid metabolism LPC 16:0 LPC 18:0 16:0/18:1 PC 16:0/18:2 PC 16:0/20:4 PC 16:0/22:6 PC 18:0/18:2 PC Oleamide Phenylalanine GCDCA Canavaninosuccinate Phospholipid catabolism Fatty aci d metabolism Gut flora metabolism Organic acid metabolism
35 Table 1 1. Continued Author Tissue (organism) HCC Etiology Significantly Altered Metabolites in HCC Patients vs. Cirrhosis Controls Main Pathways distinguishing HCC from Cirrhosis Significantly Altered Metabolites in Cirrhosis Patients vs. Healthy Volunteers Main Pathways distinguishing Cirrhosis from Healthy Volunteers Zhou L, Ding L, Yin P, et al. Serum (human) HCC n=69: HBV HCC n=38 HCV HCC n=31 LPC(14:0) HCC C LPC (16:1) HCC C LPC(18:3) HCC B HCC C LPC(18:2) HCC B HCC C LPC(18:1) HCC B HCC C LPC(18:0) HCC C LPC(20:5) HCC C LPC(20:4) HCC B HCC C LPC(20:3) HCC B HCC C LPC(20:2) HCC C LPC(20:0) HCC B HCC C LPC(22:6) HCC C LPC(22:5) HCC B HCC C LPC(22:4) HCC B HCC C Phospholipid catabolism Phenylalanine GCA GDCA Bilirubin LPE( 18:2) LPC(22:6) LPC(18:2) LPC(20:4) LPC(16:0) LPC(18:0) C18:1 CN Amino acid metabolism Bile acid metabolism Heme pigment Phospholipid catabolism Carnitine palmitoyl transferase shuttle system Chen T, Xie G, Wang X Serum (human); Urin e (human) HCC n=82: Cirrhosis n=66 No cirrhosis n=16 Not reported N/A Inositol 2,2_ Bipyridine Methionine Arginine Stearic acid Palmitic acid Citric acid 2 piperidine carboxylic acid 5 Hydroxy tryptophan Tyrosine TCA cycle Amino a cid metabolism Fatty acid metabolism
36 Table 1 1. Continued Author Tissue (organism) HCC Etiology Significantly Altered Metabolites in HCC Patients vs. Cirrhosis Controls Main Pathways distinguishing HCC from Cirrhosis Significantly Altered Metabolites i n Cirrhosis Patients vs. Healthy Volunteers Main Pathways distinguishing Cirrhosis from Healthy Volunteers Cao H, Huang H, Xu W, et al Fecal (human) HCC n=23; Cirrhosis n=23 Not reported N/A Chenodeoxycholic acid dimeride Urobilin Urobilinogen 7 ketolithocholic acid LPC C18:0 LPC C16:0 Bile acid metabolism Heme pigmentation molecules Microbiome metabolism Glycerophospholipid metabolism Yin P, Wan D, Zhao C, et al. Serum (human) Not reported Not reported N/A Hypoxanthine Inosine Bilirubin Glycocholic acid GCDCA Taurine Dihydrosphingosine LPC C18:2 LPC C18:3 LPC C16:1 LPC C18:0 LPC C16:1 L Acetylcarnitine 6 Methylnicotinic acid Purine synthesis Heme pigmentation molecules Bile aci d metabolism Glycerophospholipid metabolism Carnitine palmitoyl transferase shuttle system Wu FX, Wang Q, Zhang ZM, et al. Serum (human) HBV GRO Thrombin light chain Protease cleavage Not reported N/A
37 Table 1 1. Continued Author Tissue (organism) HCC Etiology Significantly Altered Metabolites in HCC Patients vs. Cirrhosis Controls Main Pathways distinguishing HCC from Cirrhosis Significantly Altered Metabolites in Cirrhosis Patients vs. Healthy Volunteers Main Pathways distinguishing Ci rrhosis from Healthy Volunteers Ressom HW, Xiao JF, Tuli L, et al. Serum (human) HCC n =78: HCV 67% HBV 15% Alcoholism 29% NASH 13% Cryptogenic (8%) Autoimmune (3%) GDCA TCA TCDCA Sphingosine 1 phosphate LPC (16:0) LPC (17:0) LPC (18:0) LPC (15:0) LPC (22:6) LPE (22:6) LPE (20:4) LPE (20:3) PS Bile acid metabolism Spingolipid metabolism Glycerophospho lipid metabolism Not investigated N/A Yang Y, Li C, Nie X, et al. Biopsy (human) HCC n=17: Cirrhosis n=9 No cirrhosis n=8 Glucose Creatine PE Glutamine Glutamate PC+GPC Glycolysis Glycerophospho lipid metabolism Amino acid metabolism Bile acid metabolism Not investigated N/A
38 C HAPTER 2 NON TARGETE D METABOLOMICS OF LIVER CANCER AND CIRRHOSIS Introduction Recent advances in analytical chemistry and biostatistical methods have placed the emerging field of metabolomics at the frontier of lead marker generation. Because the liver is the principal organ of amino acid, lipid, and carbohydrate metabolism and given the stepwise hepatocarcinogenic process involving transition from chronic viral hepatitis to cirrhosis and ultimately culminating in establishment of HCC, HCC is an ideal ca ndi date for metabolomics profile comparison studies Metabolomics is the identification and quantification of all small molecules < 1 kD in a tissue sample ( 75 ) Metabolomics integrates liquid or gas chromatography/mass spectrometry (LC/MS, GC/MS) or nuclear m agnetic resonance (NMR) with mass spectral and biostatistical analysis software to provide a seamless, validated and high throughput method for obtaining and comparing the full complement of metabolites in tissue samples derived from multiple groups. Whil e still in its relative infancy, HCC metabolomics has gained significant traction in recent years as demonstrated by the above referenced studies A large number of these studies however, looked solely at metabolomic expressi on pattern differences between HCC patients and healthy volunteers. Because >90% of HCC cases are diagnosed in patients with cirrhosis, an inquiry into metabolome differences between HCC and cirrhosis is a more clinically relevant comparison and can better elucidate the mechanisms invo lved in the progression from cirrhosis to HCC. Furthermore, these studies focused on the utility for metabolomics itself to function as a novel clinical diagnostic modality While this technology shows promise in predicting
39 patient disease status based on his/her metabolomic expression profile, the sophisticated nature of this technology is economically and logistically incompatible with (6) Hence, the focus of these studies should als o be on the utility for metabolites or the enzymes that produce them to function as biomarkers of HCC. In this study, we compared the global serum met a bolomes of 30 hepatitis C associated HCC patients (HCC), 27 hepatitis C cirrhosis disease control patient s (DC) and 30 normal healthy controls (NHC) using non targeted ultrahigh performance liquid chromatography electrospray ionization tandem mass spectrometry (UPLC/(ESI)MS MS) metabolomics to identify biomarkers of HCC Multivariate statistics and the random forest supervised classification analysis (RF) were employed to narrow down the list of potential markers to a subset of metabolites that showed the most promise as HCC biomarkers. We then teste d the utility of these putative markers to accurately diagnos e HCC using receiver operator characteristic (ROC) curve analysis and ELISA We report significant aberrations in a number of pathways potentially involved in the progression from HCV cirrhosis to HCC including amino acid metabolism, purine and pyrimidine nucleotide processing, reactive oxygen species homeostatic pathways, lipid signaling cascades, and lipid metabolism. We further identified several metabolites that show promise as biomarkers for cirrhosis. While the liver biopsy is a highly accurate approa ch for diagnosing advanced l iver fibrosis and cirrhosis, it is not, by virtue of its invasive nature a gold standard approach (76) T he identification of accurate cirrhosis biomarkers may facilitate development of equally accurate and noninvasive cirrhosi s in vitro diagnostics.
40 Materials and Methods Patients This study was approved by the University of Florida Institutional Review Board and written informed consent was obtained from all study participants prior to their enrollment. HCC patients were diagno sed and staged according to the Barcelona Clinic Liver Cancer Staging System (BCLC). The study cohort included 30 baseline HCC patients whose cancer arose exclusively from HCV cirrhosis, 27 HCV cirrhosis disease control patients (DC), and 30 healthy non di abetic controls (NHC). We recruited only HCC and DC patients with histologically proven HCV related cirrhosis given that HCV cirrhosis is the leading cause of HCC in Western countries. Patient clinical characteristics were obtained from the medical record and included the following parameters: Demographic information, body mass index (BMI), AFP level, Child Pugh score (CP), and Model for End Stage Liver Disease score (MELD). HCC patient samples were collected from patients with HCC at stage A (n = 13), B (n = 10) and C (n = 7) and all were nave to cancer treatment at the time of blood withdrawal. All DC patients in the study were enrolled into our surveillance program, received serial cross sectional imaging every six months, and had no liver masses on enro llment. NHCs were recruited from the LifeSouth Community Blood Center in Gainesville, FL. Blood was drawn from patients presenting to clinic after overnight fast, whole blood samples were immediately processed for serum and serum samples were stored in our biorepository at 80 o C until retrieval. Metabolomics Analysis For metabolomic profiling of patient serum, we employed an integrated, non targeted UPLC/(ESI) MS platform described in a previous publication (77 ) Upon retrieval, serum
41 samples were thawed on ice and 100 L of serum was aliquotted per subject. Large proteins were precipitated for exclusion with methanol containing four standards that enabled monitoring of extraction efficiency. Samples were then evacuated into a collection chamber in vacuo using an automat ed liquid handler (Hamilton MicroLab STAR). The metabolomics platform utilized three operation modes to ensure fu ll coverage of the metabolome: G as chromatography mass spectrometry (GC MS) to detect lipids and organic phase biomolecules, UPLC/(ESI)MS MS in the positive mode optimized for detection of basic species, and U PLC/ (ESI)MS MS in the negative mode optimized for detection of acidic species. Serum samples were aliquotted into three equal parts, randomly sorted, and processed on their respective inst rument on three consecutive days such that each run consisted of exactly one samples. GC/MS samples were separated and quantified on a Thermo Finnigan Trace DSQ fast scanning single quadrupole mass spectrometer using electron impact i onization. The samples destined for GC/MS analysis were re dried under vacuum desiccation for a minimum of 24 hours prior to being derivatized under dried nitrogen using bistrimethyl silyl triflouroacetamide (BSTFA). The GC column was 5% phenyl and a 40 to 300C temperature ramp was applied in a 16 minute period The instrument was tuned and calibrated for mass resolution and mass accuracy on a daily basis. The UPLC/(ESI)MS MS portion of the platform was based on a Waters Acquity UPLC and a Thermo Finnig an linear trap quadrupole fourier transform (LTQ FT) mass spectrometer, which had a linear ion trap front end and a Fourier transform ion cyclotron resonance mass spectrometer backend. Process coefficients of variation including instrument performance, chr omatography, and mass calibration were checked using reference
42 standards to ensure quality on a daily basis. The median relative standard deviation (RSD) of instrument variability was 5% and total process variability RSD was 12%. Samples were reconstituted in 50 L of 6.5 ammonium bicarbonate in water (pH 8) and 50 L of .1% formic acid in water for the ESI positive and negative analyses, respectively. All reconstitution solvents contained instrument internal standards used to monitor instrument performance and fu nctioned as retention index (RI) markers. A pooled quality control matrix sample was also created to ensure low process variability during the analysis. For ions with counts greater than 2 million, an accurate mass measurement was performed. Accurate mass measurements could be made on the parent ion as well as fragments. The typical mass error was less than 5 ppm. Chromatographic peaks, retention time, mass/charge (m/z) data, and MS spectra were obtained for all compounds in the serum samples. To verify th e identity of detected metabolites, proprietary mass spectra interpretation software cross referenced the generated MS spectra with ~1500 authenticated standard spectra housed in an internal library and matches were confirmed by a trained operator. Statist ical A nalysis project.org). Unsupervised hierarchical clustering analysis of HCC and DC metabolomes was performed using dChip (Harvard, Cambridge, MA). Random forest analysis, a superv ised machine learning classification technique reporting on the consensus of a large number of decision trees, was employed to (1) predict HCC, DC, and NHC classes on the basis of global metabolomic expression profiles and (2) identify metabolites most imp ortant to the group classification for elucidation of potential HCC
43 biomarkers. Metabolites with the largest mean decrease accuracy were deemed as putative biomarkers. The mean decrease accuracy describes the magnitude loss of class prediction accuracy fol lowing removal of a metabolite from the random forest analysis. The greater the decrease in class prediction accuracy, the more valuable the marker is deemed to the class prediction and hence, potentially, to the disease. Following log transformation and v ariance normalization of the metabolite levels, intergroup fold differences for HCC vs. DC and DC vs. NHC were generated, and a two sided t test was then used to identify metabolites whose expression differed significantly ( P < .05) between compared groups To qualify for designation as potential biomarkers, metabolites must have met following criteria: (1) > 80% expression in HCC group and > 50% expression in DC/NHC (2) intergroup fold difference significance of P < .05 (3) false discovery rate q < .10. ROC analysis was performed using MedCalc (Mariakerke, Belgium) on potential HCC biomarkers and on biomolecules showing the greatest promise as biomarkers of cirrhosis. In Vitro Validation To validate the in vitro expression pattern of rate limiting enzymes inv olved in the production of putative biomarkers of HCC, we tested for the expression patterns of xanthine oxidase, the enzyme that forms xanthine from hypoxanthine; phosphoglycerate dehydrogenase, the rate limiting enzyme of serine biosynthesis; sphingosine the lipid signalling molecule generated from acid ceramidase activity on ceramide; and sphingosine 1 phosphate, the phosphorylated version of sphingosine implicated as an anti apoptotic molecule. These levels were determined in serum from a randomized gr oup of 24 HCC and 12 DC patients from our metabolomics cohort, t hree human HCC cell lines, and one human renal cancer cell line. The HCC lines consisted of Huh7,
44 HepG2, and Hep3B. AHCN, a renal cell carcinoma cell line, was used as a cancer control. Mediu m preparation Fetal bovine serum (FBS), penicillin and L glutamine were removed from 20 o C and (Cellgro) and non essential amino acids (NEAA) taken from 4 o C, were thawed in a water bath at 37 o C for 5 min 200 mL of DMEM was then poured into a 150 mL bent neck cell culture flask (Corning) under a sterile hood. 20 mL of FBS, 2mL penicillin, 2mL L glutamine and 2mL NEAA were then added to the flask Contents were mixed and filtered using a filter bottle system (Corning; Cat# 430767). Cell cult uring Cell lines contained in 2mL cryogenic vials (Corning) were removed from our internal liquid nitrogen repository and thawed in a 37 o C water bath. 2 3mL of medium was added to a 15mL tube corresponding to each cell lin e cells were transferred into their respective tube, and tubes were capped loosely and incubated at 37 o C for 30min Tubes were then shaken gently and centrifuged at 1600 RPM for 6 min at 23 o C. Supernatant was suctioned away using a vacuum and each cell li ne pellet resuspended with 1mL of medium. Cells were counted using a 10:1 trypan blue:suspension ratio. Cells were then transferred to a Petri dish sized according to the cell count (small Petri dish for low cell count, larger dish to acco mmodate greater c ell quantity) and incubated 18 30h at 5% CO 2 and 37 o C. To alleviate cell crowding, cells were split by washing the Petri dish with PBS at least 2x and subjected to 1mL trypsin EDTA. After a 2min incubation at 37 o C, cells were viewed on under a microscope t o observe successful detachment from the plate. Cell suspensions were then placed in 15mL tubes
45 corresponding to each cell line and pelleted at 1600 RPM for 6min. Supernatant was subsequently removed, pellets were resuspended with 3mL medium, and cells wer e counted as described above. Cells were then split equally onto two Petri dishes, labeled and dated, and placed back in the incubator for 18 30h. Growth was monitored over a period of 3 me that corresponding to concentrations of 0.5x10 6 cells/mL, 1.0 x10 6 cells/mL, and 1.5 x10 6 cells/mL. Cells at these appropriate concentrations were then transferred to a 12 well plate and incubated. 12 well plates were then centrifuged at 1600 RPM for 6m in on a plate holder, and 500 L of supernatant of each cell line concentration was transferred to 2mL cryogenic vials and stored at 20 o C until ELISA ELISA Xanthine oxidase (XO), phosphoglycerate dehydrogenase (PHGDH), sphingosine, and sphingosine 1 phosp hate sandwich ELISA kits (Cusabio) were used to measure the expression pattern of these enzymes/metabolites in human serum (HCC n=24, DC n=12) and cell line supernatants (n=5) ELISA was performed according to Briefly, microtit er plates coated with anti human antibody were inculca ted with human serum and cell line supernatant and subjected to horseradish peroxide, a substrate solution that upon addition induced a color change. The intensity of the colored product was directly pr oportional to the level of enzyme for XO and PHGDH or metabolite for sphingosine and sphingosine 1 phosphate present in each well. Plates were read at 450nm using a Spectramax 190 plate reader (Molecular Devices, Sunnyvale, CA) and analyzed using SoftMax P ro 6 (Molecular Devices).
46 Results Patients Most of our patients were Caucasian males and the mean ages of HCC patients and cirrhosis controls were 60 and 55 years respectively ( Figure 2 1 ). All patients were matched by age, gender, and body mass index (BMI ) and had Child Pugh A status and MELD scores < 12, emblematic of well compensated cirrhosis. To minimize the influence of comorbidity on metabolite expression patterns, no patient had other cancers and all were non diabetic and did not have ascites at the time of blood withdrawal. 28/30 HCC patients had a BMI <30, the United States obesity cutoff, which minimized the influence of excess adiposity on the generated biochemical profiles. All patients were abstinent from alcohol and illicit drug use at the time of enrollment. Metabolomic P rofiles The metabolomics analysis detected a total of 485 biomolecules. 107 metabolites were significantly altered in the HCC vs. DC ( P < .05) but none of these metabolites exhibited >3 fold elevation or downregulation vs. DC. In contrast, 245 metabolites were significantly altered in DCs vs. NHC, with eight among them exhibiting >10 fold overexpression. In the random forest analysis comparing all three groups, RF accurately classified 100% of the NCs, 70% of DCs, and 70% of HCC patients (Figure 2 2) For the RF comparison of HCC vs. cirrhosis, RF gave a mean prediction accuracy of 82%, a satisfactory improvement over the 50% accuracy that would be expected from random ass ignment to either group The most differentially expressed metabolites in HCC patients vs. DCs belonged to pathways of amino acid metabolism, lipid signaling, lipid metabolism, acylcarnitine metabolism, nucleic acid processing, and oxidati ve stress homeostasis and by virtue of their strong association with HCC, t hese metabolites
47 assume putative status as HCC biomarkers (Figure 2 3) Pathways of importance for cirrhosis were bile acid metabolism, acylcarnitine metabolism, fatty acid metabolism, fibrinogen cleavage peptides, hemoglobin catabolism, and dipeptide meta bolism Hierarchical cluster analysis of HCC and DC metabolomes is shown in Figur e 2 4 The heat map shows a major node of separation which partial ly stratifies HCC patients on the left and places DCs on the right. Nevertheless, the hierarchical cluster di d not reveal markedly distinct metabolomic expression patterns between HCC patients and DCs. Pathways Involved in Stepwise Hepatocarcinogenesis Lipid metabolism The liver numerous HCC me tabolomics analyses have implicated aberrations in lipid metabolism as sign atures of HCC development (29, 36, 39, 78 ). In our study we witnessed a strongly significant upregulation of multiple lipid classes in HCC vs. DC and DC vs. NHC. Monohydroxy fatty a cids, dicarboxylic fatty acids (DCA), glycerophospholipids, sphingolipids, and acylcarnitines all trended higher in HCC vs. DC. This is in contrast to what would be expected from BMI data for the respective groups showing greater adiposity in DC vs. HC C. A n especially remarkable upregulation of DCA was seen in the comparison of cirrhosis controls vs. NHC, where nonanedioate (azelate), undecanedioate, and decanedioate (sebacate) were overexpressed in cirrhosis 54, 36, and 26 fold respectively Multivariate s tatistics indicated that the P value for each of these metabolite fold differences was 0. We also witnessed global downregulation of glycerophospholipids in DCs vs. NHC. Medium chain fatty acid levels were not
48 statistically different in HCC vs. DC, but thi s class of lipid metabolites was strongly upregulated in DC vs. NHC. A massive downregulation of lysophospholipids (LPC) was also witnessed in DC vs. NHC and HCC vs. NHC a trend observed in most HCC metabolomics studies (29, 38 39, 41 42) Of the 33 LPCs detected 28 were significantly downregulated in DC vs. NHC and 23 were significantly diminished in HCC vs. NHC. There was an increase of LPCs in HCC vs. DC but this was not significant (P=.21). Bile acids Bile acids, which are routinely implicated as mark ers of liver disease in HCC metabolomics studies (36 39), exhibited vast overexpression in cirrhotics vs. NHC. Taurochenodeoxycholate (TCDCA), taurocholate and glycocholate were 17.8, 15, and 4 fold upregulated in DC vs. NHC and are joined by other cholic acids that were elevated in cirrhosis These metabolites were also markedly elevated in HCC vs. NHC with similar fold differences for each aforementioned bile acid. Acylcarnitines Alone, fatty acyl CoAs (fatty acids) cannot enter the mitochondria for b oxi dation and must first be converted to acylcarnitines by enzymes of the carbamoyl phosphate transport shuttle system. In our study, acetylcarnitine, glutarylcarnitine, and succinylcarnitine were significantly upregulated in HCC patients vs. DCs. In DC vs. N HC, medium chain acylcarnitines averaged two fold overexpression while long chain acylcarnitines were downregulated. Lipid signalling molecules Eicosanoids are twenty carbon autocrine and paracrine lipid signalling molecules that perform a myriad of signal transduction roles, notably in the cycloxygenase (COX)
49 and lipoxygenase (LOX) pathways, and these metabolites are best known for regulating pain and inflammation. Eicosanoids consist of leukotrienes, prostaglandins and thromboxanes and are derived from ar achidonic acid or linoleate. In our study, eicosanoids were significantly overexpressed in HCC patients vs. DCs and downregulated in DCs vs. NHC. These differences included the two prostaglandin parent molecules arachidonate (20:4n6) which was significant ly elevated in HCC vs. DC but downregulated in DC vs. NHC, and dihomo linoleate (20:2n6) which showed significant elevation in HCC vs. DC Linoleate/arachidonate derived lipid signaling molecules 13 HODE+9 HODE (2.54 fold increase in HCC:DC), 15 HETE (2.8 ) and 12 HETE (1.7) were significantly upregulated in HCC vs. DC. Further, 12 HETE and 15 HETE were among metabolites with the highest mean decrease accuracy in random forest analysis comparing HCC vs. DC. Oxidative stress homeostasis The liver is the ma jor site of glutathione (GSH) synthesis and principal organ of oxidative stress homeostasis. Key HCC vs. DC trends among oxidative stress associated metabolites included 1.62 fold upregulation of xanthine ( P =9.35x10 5 ), a purine degradation intermediate wh ose production is accompanied by generation of reactive oxygen species (ROS) hydrogen peroxide (H 2 O 2 ) and 1.56 fold elevated 2 hydroxybutyrate (2HB) ( P <.0003) a n important neutralizing agent of reactive oxygen species which complements glutathione during periods of marked oxidative stress Gamma glutamyl peptides, which are liberated via gammaglutamyl transpeptidase mediated breakdown of glutathione, were also significantly upregulated in HCC vs. DC. Xanthine and 2HB were among the best class predictors in the random forest analysis of HCC vs. DC.
50 Fibrinogen cleavage peptides Our study identified two unknown fibrinogen cleavage fragments, denoted by amino acid sequence, that were strongly upre gulated in DC vs. NHC indicating potential importance of these metabolites in distinguishing cirrhosis patients from healthy subjects ADSGEGDFXAEGGGVR (ADS) differed from DSGEGDFXAEGGGVR (DSG) by the presence of a single alanine on the N terminus. DSG and ADS exhibited 11.8 and 2.9 fold elevation in DC vs. NHC, respe ctively. Moreover, DSG levels were 35% downregulated in HCC vs. DC ( P =.046). Amino acids and protein metabolites Numerous studies have linked amino acid metabolism aberrations to cancer development (17 21) including several recent HCC metabolomics studies (38, 40, 42 ). Consistent with these reports, we found significant upregulation of serine, aspartate, glycine, phenylalanine, and glutamate in HCC vs. HCV cirrhosis, and these metabolites had the strongest P values and lowest false discovery rates (q value) in the intergroup metabolite expression level comparison between HCC patien ts and DCs Serine was the most significantly altered amino acid between HCC patients and DCs ( P = 7.0 x 10 6 ) and was the best class predictor in the random forest analysis of HCC patients and cirrhosis controls. A more reactive form of serine, homoserine, also trended higher with 1.62 fold overexpression (P = .0002) in HCC vs. DC and was the fourth most important metabolite for accurate class prediction in the HCC vs. DC random fo rest analysis. Glycine was also strongly associated with HCC, exhibiting a 1.36 fold increase in HCC patients versus DCs ( P <2.58 x10 5 ) and likewise ranked among the top in the supervised random forest classification of HCC vs. DCs. The trend of overexpres sed
51 amino acids in HCC vs. DC was further seen among amino acids valine (P = .0409), tyrosine (P = .0211), asparagine (P = .0108) and tryptophan (P =.0036). The reverse trend in amino acid expression was true for the comparison between DCs and NHC s with HCV cirrhotics exhibiting strongly significant downregulation of serine, glycine, aspartate, glutam ate and phenylalanine We also observed a consistent downregulation of dipeptides pyroglutamylglycine, glycylvaline, alanylalanine and asparty leucine in DC v s. NHC In HCC vs. DC, pyroglutamylglycine and glycylvaline trended 1.95 and 1.48 times higher, and of the only four metabolites significantly downregulated in HCC vs. DC, three were dipeptides: phenylalanyltryptophan, phenylalanylserine, and leucylalanine Other metabolites Hemoglobin catabolites bilirubin and biliverdin are routinely implicated as important class distinguishers in HCC metabolomics studies (29, 32, 39, 41) and elevated bilirubin is a clinically valuable hallmark of liver decompensation (7 9) Consistent with this fact, we observed significantly elevated levels of heme catabolites bilirubin, biliverdin, and urobilinogen in DC vs. NHC and HCC vs. NHC. There was no significant difference between the levels of heme catabolites in HCC patients v s. DCs, however. 2 pyrrolidinone, a gamma aminobutyric acid (GABA) metabolite with an unknown role in cirrhosis or cancer exhibited a strongly significant, progressive elevation from NHC to cirrhosi s and cirrhosis to HCC 1,2 propanediol, a ketogenesis in termediate, was 5.8 times upregulated in DC vs. NHC ( P <1.0x10 15 ) We also witnessed significant elevation of purine and pyrimidine metabolites in HCC vs. DCs.
52 Receiver Operator Characteristic Analysis To gauge the diagnostic utility of the putative HCC an d cirrhosis biomarkers, we generated receiver operator characteristic (ROC) curves using MS peak areas corresponding to metabolite concentr ation in each patient The HCC vs. DC comparison was used during ROC analysis of HCC biomarkers, and DC vs. NHC was u sed for cirrhosis ROC curves. The ROC analyses showed that the diagnostic utility of cirrhosis markers was better than the utility for putative HCC biomarkers to accurately discriminate between HCC patients and DCs. Given the striking upregulation dicarbox ylates and bile acids in HCV cirrhosis patients vs. healthy volunteers and similar trends among other potential markers of cirrhosis, we performed ROC analysis on the capacity for these metabolite levels to discriminate cirrhotics from NHCs. The results of this ROC analysis reveal that azelate, which was 54 times upregulated in cirrhosis patients vs. NHC, possessed 100% sensitivity, 100% specificity and AUC=1.00 ( P =0.00) when used to distinguish DC from NHC. Sebacate was elevated 26 fold in DC vs. NHC and R OC analysis showed that this DCA also possessed a sensitivity of 100% and a specificity of 100% with an AUC=1.00 and P=0.00 when used to discriminate between these groups. A third DCA, undecanedioate, was elevated 36 fold in DC vs. NHC and had a sensitivit y of 78% and a specificity of 90% (AUC=.860, P<.0001) when distinguishing HCV cirrhotics from NHC. Two other DCAs, 2 hydroxyglutarate and hexadecanedioate, were strong predictors of cirrhosis in the DC vs. NHC ROC analysis. 2 pyrrolidinone also possessed 1 00% sensitivity and 100% specificity (AUC=1.00, P=0.00) for distinguishing DCs from NHCs. Aspartate was the most sensitive HCC biomarker, possessing a sensitivity of 100 and a specificity of 52 for HCC (area under the curve, AUC=.789, P<.0001). In
53 comparis on, AFP had a sensitivity of 63.3 and a specificity of 83.6 for detecting the presence of HCC in the same study cohort (AUC=.760, P <.0001). Glycine also had a high sensitivity of 83.3 and a specificity of 63.0 (AUC=.801, P <.0001) for HCC. Serine had a se nsitivity of 73.3 and a specificity of 85.2 when used to distinguish HCC patients from DCs (AUC=.831, P used to distinguish early HCC patients (stage A, n=13) from the disease controls, possessing a sensitivity and specificity of 76.9 and 88.9, respectively (AUC: .875, P <.0001). Xanthine, a purine degradation metabolite accompanied by production of reactive oxygen species (ROS) hydrogen peroxide (H 2 O 2 ), had a sensitivity of 63.3 and a specificity of 88.9 for detecting HCC presence (AUC=.790, P <.0001). The 15 lipooxygenase product 15 HETE, a lipid signalling eicosanoid involved in inflammation and endothelial cell adhesion, was also identified as a potentially important biomarker for HCC in the ROC an alysis, possessing a sensitivity of 83.3 and a specificity of 59.3 for HCC presence (AUC=., P =.0011). 15 HETE (sensitivity/specificity=73Our study also revealed a litany of potentially important lipid signatures of HCC, and multiple HCC metabolomics studi es have implicated aberrations in lipid metabolism as hallmarks of the neoplastic HCC process. 2 hydroxypalmitate (2HP) possessed the highest mean decrease accuracy among lipids in the HCC vs. DC RF analysis and ROC analysis showed that 2HP possessed a sen sitivity of 80 and a specificity of 63 when used to distinguish HCC patients from cirrhosis controls (AUC=.736, P <.0004). Given the previously implicated role of sphingosine 1 phosphate in HCC progression and that sphingosine had the second highest mean de crease accuracy among lipids in the HCC vs. DC RF analysis, we tested its utility as an HCC biomarker using ROC curve analysis
54 and found that it possessed a sensitivity of 58.3 and a specificity of 86.7 for distinguishing HCC patients from DCs (AUC=.731, P <.0058). In V itro Validation The levels of sphingosine trended higher with increasing tumor burden s R=.225 ) but these increases were not significant (P=.29) At a cutoff of 29ng/mL, sphingosine had a sensitivity of 50% and a specificity of 75% for distinguishing HCC patients from disease controls, but these values were again insignificant (AUC=.531, P=.765). There was no difference between sphingosine 1 phosphate expression in HCC patients vs. DCs and the level of S1P did not exhibit any differe ntial expression with regard to tumor burden. ROC analysis showed that S1P was a poor predictor of HCC in the comparison with DCs (Sensitivity/specificity: 63%/67%, AUC=.538, P=.377), but its ability to distinguish early HCC from cirrhosis controls showed improvement (75%/67%, AUC=.636, P=.24). The comparison of xanthine oxidase expression in HCC vs. DC failed to demonstrate a significant difference in the expression patterns between thes e groups. ROC analysis further showed that in this group of patients, XO was not good distinguisher of HCC vs. DC (AUC=.542, P=.711). Interestingly, the median XO level in HCC cell lines was 4.7ng/mL, while the level of XO in AHCN was 28.9ng/mL. Discussion Biomarkers of HCC Our study identified significant aberrations in lip id, amino acid, and nucleotide metabolism in HCC vs. cirrhosis and cirrhosis vs. NHC. The following metabolites were significantly elevated in HCC vs. DC: Arachidonic acid derived eicosanoids 12 HETE and 15 HETE produced through 12 lipoxygenase, 15 LOX, an d cytochrome P450
55 pathways; the Gamma glutamyl peptides which act as substrates and catabolic end products of GGT catalyzed reversible production of ROS antioxidant glutathione; purine and pyrimidine catabolites xanthine and hypoxanthine, adenine, uridine, and N methyladenosine; amino acids serine, glycine, aspartate, phenylalanine; N acetyl amino acids involved in epigenetic regulation; glycerophospholipids; acylcarnitines, lipid intermediates of the mitochondrial carbamoyl phosphate transferase shuttle sy stem; the sphingolipid sphingosine. 12 HETE and 15 HETE are products of lipoxygenase and cytochrome P450 enzymes and exert both pro and anti inflam matory effects (80) 12 HETE is reported to augment tumor metastatic potential through activation of protein kinase C ( 81 ) It also promotes vasoconstriction and is involved in stimulating the aggregation of neutrophils and platelets ( 82 ). Ki 67 and GST n immunohistochemical staining by Xu and colleagues showed that 12 LOX was significantly overexpressed in rat H CC tumors, HepG2 and L02 cell lines. The group subsequently demonstrated a dose dependent attenuation of HCC development and induction of apoptosis by the 12 LOX inhibitor baecalein (83) A nearly identical study performed by Wong and colleagues confirmed the findings of the Xu work (84) The endogenous production of 12 HETE by tumors of the stomach, lung and colon and by melanoma cells has also been observed, and 12 HETE was further shown to enhance tumor cell metastatic potential ( 85 ). 15 HETE induces va soconstriction and is implicated in edema (86) and is a central anti apoptosis molecule in pulmonary arterial smooth muscle cells through a variety of mechanisms including inducible nitric oxide synthase and K+ channel regulation ( 87 88 ) 15 HETE was rece ntly shown to promote HCC growth and
56 metastasis through the phosphoinositide 3 kinase/protein kinase B/heat shock protein 90 (PI3K/Akt/HSP90) pathway (89) Our metabolomic data shows that 12 HETE and 15 HETE were significantly downregulated in DC vs. NHC b ut experienced a rebound in expression in the HCC group with both compounds exhibiting significant elevation in HCC vs. DC ROC analysis of these markers showed good sensitivity but their potential substandard specificity would diminish their utility as b iomarkers of HCC. Notwithstanding, their reemergence in HCC suggests a potential role for these roles of 12 HETE and 15 HETE and their enzymes in HCC remain largely uncharacter ized and are worthy of further investigation. Oxidative stress through accumulation of ROS species superoxide, hydrogen peroxide (H 2 O 2 ), and hydroxyl radical is known to promote the development of HCC primarily by enhancing DNA damage (90) Our data revea ls significant elevation of oxidative stress related metabolites xanthine and the gamma glutamylpeptides in HCC vs. DC. Xanthine is produced from hypoxanthine by xanthine oxidase, and the production of xanthine is accompanied by generation of H 2 O 2 One stu dy reported a significant elevation of xanthine oxidase activity in HCC ( 91) while others showed a decrease in XO activity in HCC (92). The c oncomitant elevation of hypoxanthine in our study suggests overactivity of xanthine oxidase in HCC. Interestingly, one study showed that the elevation of ROS in hepatoma cells was a prerequisite for heightened glycolysis (93), consistent with the Warburg theory of cancer. While oxidative stress is routinely implicated in cancer and in HCC, the role the xanthine oxidase pathway in HCC is unexplored and needs better characterization. Its elevation in HCC vs. cirrhosis may
57 reflect increased oxidative stress that is a hallmark of the n eoplastic HCC microenvironment, and oxidative stress may synergistically promote the enhan ced glycolysis exhibited by growing tumors. Gamma glutamyl antioxidant that is produced primarily in liver and acts as a reducing agent to neutralize ROS. Glutathione is catalytically produ ced by gamma glutamylpeptidase (GGT), a clinical signature of chronic liver disease. GGT breakdown of glutathione re sults in the liberation of free glutamyl peptides. Glutathione scavenging of ROS is catalyzed by glutathione s transferase P1 (GSTP1) and studies show that impairment of GSTP1 promotes hepatocellular carcinoma development (9 4 95 ) Failure of GSTP1 in HCC would impair ity to fulfill its role in oxidative stress homeostasis, exacerbating oxidative stress in liver and increasing the likelihood for DNA damage to transpire. Decreased activity of liver glutathione peroxidase, another important free radical scavenger, has als o been observed in HCC (9 6 ) Elevations in glutathione in HCC vs. cirrhosis are often reported and the increase of gamma glutamylpeptides witnessed in our study may reflect bolstered GSH manufacture aimed at overcoming the deficiency of functional GSTP1 or glutathione peroxidase. Another oxidative stress metabolite, 2 hydroxybutyrate, was also strongly elevated in HCC vs. cirrhosis and was among the most important metabolites for class prediction by random forest. When overwhelming oxidative stress eclipses glutathione antioxidant capacity, the liver (9 7 ) The upregulation of 2HB may be a signature of elevations in ROS that summon emergency LDH production of 2HB to complement the antiox idative effects of GSH. ROC analysis
58 of Xanthine, 2HB, and the gamma glutamyl peptides demonstrates that these oxidative stress metabolites may have utility as ROS related biomarkers of HCC. Multivariate analysis and random forest showed that serine, aspar tate and importance to HCC development may also be linked to its role as an allosteric activator of pyruvate kinase M2 (PKM2), the principal cancer isoform of PK which conver ts phosphoenolpyruvate to pyruvate. Serine not only activates this enzyme upon binding, but depletion of serine results in an attenuation of PKM2 activity. Among the hallmarks of the cancer microenvironment is hypoxia and numerous studies have demonstrated oxygen, a phenomenon known as the Warburg effect. PKM2 has been implicated as a critical arbitrator of cancer growth and proliferation. The elevation of serine in HCC vs glycolytic process. A more reactive form of serine, homoserine was also among the most significantly overexpressed markers of HCC. The role of this amino acid in human physio logy is unknown, but a study by Gazarian and colleagues showed that (98) The study proposed that homoserine is produced through an emergency pathway aiding in overwhelmed methionine catabolism and proposed that this metabolite is liberated through S adenosylmethionine (SAM) cleavage into homoserine and methylthioadenosine. The increased pool of amino acids in HCC vs. DC is consistent with the anabolic needs of growing HCCs.
59 Glycine is also a vital anabolic metabolite of growing cancers. Recently, Jain et al analyzed the metabolic profile of 60 cancer cell lines using LC/MS MS (48) and found that glycine biosynthesis was among the most important pathways for the development of a broad spectrum of ca ncers. Further, higher expression of glycine biosynthesis pathway enzymes was associated with increased mortality in breast cancer patients. Glycine is synthesized in the mitochondria by serine hydroxymethyl transferase 2 (SHMT2). Jain and colleagues stopp ed cancer cell proliferation by both antagonizing glycine uptake by the cancer cells and by impairing mitochondrial biosynthesis of glycine via RNAi mediated knockout of SHMT2. When glycine was added into the medium, proliferation was rescued. To track the destination of the SHMT hyperproduced glycine, the team conducted a metabolic tracing experiment using ( 13 C)glycine and saw that the glycine was being incorporated into purine nucleotides of the rapidly proliferating cancer cells. This suggests that the o verproduction of glycine in the mitochondria of cancer cells may be a pivotal anabolic process that through enhanced nucleotide synthesis necessary for active mitosis galvanizes the pre neoplastic lesion for progression to HCC. Consistent with this theory, we observed significant upregulation of several purine and pyrimidine metabolites in HCC vs. DC These metabolites were downregulated in DC vs. NHC but experienced a rebound in concentration in HCC patients. This ma y reflect the metabolically enhanced sta te of the HCC microenvironment Further aberrations of lipid metabolism were seen in HCC vs. DC and DC vs. NHC, revealing a litany of metabolites with putative utility in the diagnosis of HCC and cirrhosis. Sphingosine is the precursor of sphingosine 1 ph osphate, a metabolite
60 increasingly becoming recognized as an important facilitator of cancer development. In agreement with a recent HCC metabolomics study identifying aberrations in sphingosine metabolism in HCC ( 36 ) we observed a significant elevation of sphingosine in HCC patients vs. DCs (P=.0017) Ceramide is routinely shown to be a tumor suppressing lipid and a facilitator of tumor cell apoptosis, while S1P is implicated in HCC proliferation and has been shown to silence apoptosis signalling. The cera mide/sphingosine/sphingosine 1 (61) and this mechanism plays a crucial role in cell turnover homeostasis. Sphingosine is ultimately produced from ceramide in a reversible m anner by acid ceramidase ( 99 ). The observed elevation in sphingosine in the HCC patients vs. cirrhosis controls may reflect the establishment of bolstered cell survival and proliferative capacity in cirrhotic tissue mediated through enhanced acid ceramidase pathway activity and pro moting the formation of neoplastic HCC tissue. Interestingly, the initiation of the sphingosine production pathway involves the combination of palmitoyl CoA and serine via serine palmitoyltransferase to produce the biosynthetic intermediate 3 ketosphingani ne. The observed elevation in serine levels in our HCC patient group enhancement of HCC proliferation through PKM2, but also the heightened activity of pathways such as AC. The syn ergy of these phenomena may be chief components of the pre neoplastic microenvironment that increases the cell survival capacity and hence the likelihood for HCC initiation. One potential HCC therapeutic strategy is directed at AC suppression, which allow s for ceramide accumulation and inhibition of sphingosine production
61 simultaneously resulting in increased apoptosis and reduced cell survival The inhibition of AC resulting in accumulation of ceramide / t runcation of sphingosine levels has been shown to bo th sensitize HCC cells to chemotherapy and shrink HCC tumors in vivo ( 100 ) In a similar study involving melanoma cell response to treatment decarbazene, AC suppression resulted in better tumor response while the inducible overexpression of AC bolstered tu mor resistance to the treatment (101 ). G ene expression profiling of hepatitis C induced HCC by Mas and colleagues showed that AC was among the top 10% of the most overexpressed genes in HCC samples vs. normal uninvolved liver tissues ( 102 ) Other studies a imed at sphingosine kinase inhibition have shown a synergistic abrogation of HCC development when sorafenib is combined with selective inhibition of SPHK2 ( 60 103 ). Our metabolomics data is in agreement with a growing body of evidence suggesting that ther apies targeting the ceramide/sphingosine axis may be beneficial in abrogating HCC tumor growth. Biomarkers of Cirrhosis The DC vs. NHC comparison revealed 65 significantly altered pathways, which is less than the number that significantly differed in HCC v s. DC. Notwithstanding, the metabolite fold differences between these groups were more remarkable than those in the HCC vs. DC comparison. We observed a tremendous overexpression of dicarboxylic acids in DC vs. NHC and ROC curves of azelate and sebacate sh ow that each of these metabolites could accurately distinguish DCs from NHCs with 100% sensitivity and 100% specificity (AUC=1.00, P =0.00). DCAs are toxic very long chain fatty acids that inhibit mitochondrial oxidation and serve as substrates for PPAR oxidation (104) The overex pression of DCAs in DCs vs. NHC and HCC vs. NHC may
62 be related to the impairment of CPT shuttle system, as shown by the buildup of acylcarnitines in HCC and DC vs. NHC. In lieu of a deregulation in CPT function, PPAR activity would be enhanced to compensate for the loss of mitochondrial oxidation, and this enhancement is shown to promote HCC development (53 55). Given that DCAs are substrates of PPAR oxidation, their elevati on may be a signature of a metabolic shift from mitochondrial to peroxisomal fatty acid oxidation. The injurious effects of this shift involve the fact that H 2 O 2 a ROS, is liberated with the oxidation of every DCA. Hence, overreliance on this pathway may increase the oxidative stress, contributing to greater likelihood of injurious effects. Further, DCAs can be activated by microsomal dicarboxylyl CoA synthetases, and the products of this reaction proceed through H 2 O 2 generating reactions ( 104 ) The high u pregulation of DCA s in HCC vs. NHC and DC vs. NHC hence potentially reinforces the link between ROS and hepatocarcinogenesis. Interest ingly, azelate is also produced by naturally occurring skin yeast Malasezzia furfur (105) Previous studies have shown a r elationship between altered gastrointestinal function and skin disorders (106) an interplay termed the gut brain skin axis S tudies have also shown that the onset of can cer can disrupt the skin flora (107) More importantly, however, is the potential production of DCAs by gut microflora. Although it is unknown whether gut bacteria produce DCAs, aberrant gut microflora expression is becoming increasingly implicated as a signature of a variety of diseases and the elevation of DCAs may in fact be related to increased volume of gastrointestinal bacteria reflecting more favorable microenvironment of cirrhosis and HCC.
63 Further aberrations in lipid metabolism were witnessed through strong elevations in bile acids TCDCA, TCA, GHCA, GCA and GHCA were seen in DC vs. NHC and ROC analysis confirmed that these bile acids possessed strong diagnostic utility for cirrhosis. Several studies have shown that total bile acids have high diagnostic utility for cirr hosis and hepatobiliary disease (108 110). Other significantly elevated markers included two unkno wn fibrinogen cleavage byproducts. The accelerated breakdown of fibrinogen is strongly implicated in cirrhosis ( 111 ) and is in agreement with the increased pr othrombin time (INR) that is a signature of impaired extrinsic pathway of coagulation and is among the major clinical indications of liver disease Highly overexpressed acylcarnitines in both HCC vs. DC and DC vs. NHC was also observed and is consistent wi th recent HCC metabolomics studies showing an elevation of these metabolites i n chronic liver disease patients Aberrant acylcarnitine metabolism has been implicated as an important mechanism of cirrhosis onset (112 113). Increases in acylcarnitines in HCC patients vs. NHC and cirrhotics vs. NHCs may reflect failure of the enzyme CPTI to catalyze the unification of free fatty acid with carnitine Conclusion and Future Directions At least fifteen HCC metabolomi cs studies have been conducted but among these studies only seven showed metabolomic differences between HCC and DC (28, 29, 31, 36 39). Given that HCC is a complex heterogeneous disease that arises from a preexisting long term condition of cirrhosis and is often accompanied by significant comorbidity, the comparison of HCC vs. NHC will invariably show substantial metabolomic differences. To demonstrate this fact, of the 485 metabolites detected in our analysis, 260 metabolites were significantly up or downregulated in HCC vs. NHC
64 but in contrast, 107 m etabolites were significantly altered in HCC vs. DC. The HCC vs. DC contains less patent differences but is more clinically relevant because HCC arises in the setting of cirrhosis in >90% of cases and because the HCC surveillance paradigms target individua ls with cirrhosis Furthermore, just half of the previous HCC metabolomics studies reported the etiology of cirrhosis. HCC emerges from viral hepatitis B or C associated cirrhosis, alcoholic cirrhosis, nonalcoholic steatohepatitis, primary biliary cirrhosi s, and non alcoholic fatty liver disease. These distinct etiologies each likely leave a unique metabolomic fingerprint in the afflicted patient which may explain the inconsistent expression trends reported in contemporary HCC metabolomics investigations In addition, previous studies did not match HCC patients and cirrhosis controls by appropriate etiology, further confounding interpretation of the phenomena witnessed. MELD and Child Pugh scores were also absent from these studies and it is unclear whether the metabolomic expression profiles are reflective of well compensated or decompensated cirrhosis. Further, no HCC metabolomic studies to date have metabolomic expression data, wi th more obese individuals exhibiting higher levels of lipids and amino acids. Our study addresses these limitations by matching HCC and DC patients according to histologically proven HCV associated cirrhosis, age, gender, race, MELD, Child Pugh score, and BMI. Our patients were also abstinent from alcohol and illicit drug use at the time of enrollment, further controlling for potential confounders. While this study reported on the differences between HCC vs. cirrhosis and cirrhosis vs. NHC, we did not eval uate the metabolomes of chronic HCV patients without cirrhosis. HCV only vs. HCV cirrhosis would likely better illustrate the
65 metabolomic expression differences than the comparison of DC vs. NHC. And, while ROC curves provided validity to the markers with promising results, these curves were generated o mass spectrometry peak area. Absolute quantitation methods are required to better establish the diagnostic utility of these putative markers. Finally, our validation efforts s howed that PHGDH levels trended significantly lower from DCs through the different HCC stages (R= .335, P=.035). Our validation did not identify any significant alterations in sphingosine, S1P, xanth ine oxidase, and none of the ELISA data demonstrated good sensitivity and specificity for distinguishing HCC patients vs. DCs Notwithstanding, we used a low number of patients in the validation (HCC n=24, DC n=12) and that these values may not be reflective of the true trend s between HCC and DC F uture validati on of metabolomics data will require larger cohorts of patients and should also measure enzyme activity. Our study identified a panel of metabolites in that may be of utility as biomarkers of HCC including amino acids, oxidative stress markers, and purine and pyrimidine nucleotides. We f urther showed putative utility for dicarboxylic acids and bile acids to function as biomarkers of cirrhosis. V alidation of these markers are planned.
66 Figure 2 1. Clinical characteristics of HCC patients and HCV cirrhosis disease control subjects.
67 Figure 2 2 Random forest supervised classification model A) Random forest supervised class prediction analysis of HCC patients, HCV cirrhosis DCs, and NHC. Random forest accurately placed 70% of HCC patients in their appropria te groups on the basis of their global metabolomic profile expression profile. 70% of DCs were also accurately classified and 100% of NHCs were correctly predicted as NHC. The mean predictive accuracy of this model for all comers was 80%. B) Random forest analysis of HCC vs. DC only Class prediction accuracy for HCC patients was 73% and 70% for DC. Serine was the most important metabolite for distinguishing HCC patients from cirrhosis controls and more broadly, amino acids played important roles in the ran dom forest classification.
68 A B
69 Fig ure 2 3 Unsupervised hierarchical clustering analysis of HCC and DC metabolomes. A major node of separation separated disease control patients (green) toward the right of the cluster and HCC patients (violet) at the left of the cluster. HCC patients (n=30) denoted by violet H. HCV cirrhosis disease controls (n=27) indicated by green D The UNFL data corresponds to patient identifiers. Red signifies overexpression, blue represents downregulation and white corresp onds to no change
70 Figure 2 4 Most significantly altered metabolites in HCC and cirrhosis. Dashes in 2 4B denote non significant p values. LC/MS pos represents positive mode electrospray ionization (ESI) and neg corresponds to negative mode ESI. P<. 05 used as the significance cutoff. q value represents the false discovery rate, which describes the confidence in the p value. m/z, mass to charge ratio. RT, retention time. A) Markers most strongly associated with HCC (P<.05, q<.10). B) Putative biomarke rs of cirrhosis.
71 Figure 2 4 Continued
72 Figure 2 5 Sensitivity and specificity of putative biomarkers of HCC and cirrhosis. Mass spectromet ry peak areas used to construct the receiver operator characteristic curves. P value < .05 used as a cutof f for significance. A) Biomarkers of HCC. B) Biomarkers of cirrhosis. A
73 B Figure 2 5 Continued.
74 CHAPTER 3 TARGETED PROTEOMICS: SOLUBLE CD25 AS POTENTIAL HCC BIOMARKER Introduction The immune system employs various defense mechanisms to inhibit ca ncer proliferation. However, a hallmark of cancer is the ability to exploit these defenses and ultimately eclipse tumor immunity (115). Soluble CD25 (sCD25) is an immune factor that is part of the immune suppressive network of HCC and holds potential promi se as a biomarker for HCC. sCD25 is produced after proteolytic release from the membrane subunit (CD25) of the interleukin (IL) 2 receptor. When CD25 is present on the T affinity IL 2 receptor that allows optimal IL 2 signaling for T cell activation and proliferation (116). We have previously shown that the level of sCD25 in the serum of patients with HCC is directly correlated with the degree of tumor burden (117). In addition to the correlation with HCC burden, sCD25 also has novel functional properties with an ability to inhibit, in a dose related manner, antitumor T cell responses. The release of sCD25 with its immune inhibitory properties is another level of immune regulation invo lved in HCC development and progression (118 120). In the present study, we evaluated the correlation between the serum level of sCD25 and liver pathology using a well defined cohort of patients with HCC and patients with advanced liver fibrosis. We hypoth esized that sCD25 has the potential to be an effective biomarker for the presence and early detection of HCC. We determined the level of sCD25 in healthy subjects (NCs), disease controls (DCs) with advanced fibrosis and patients with HCC. We then determine d the sensitivity and specificity of sCD25 in order to distinguish patients with HCC from controls with
75 advanced fibrosis and cirrhosis, evaluated the efficacy of sCD25 in detecting early HCC, of AFP. We concluded our study by revisiting the connection between sCD25 level and tumor burden observed in our previous study using this larger cohort of patients with HCC. Materials and Methods Study Population The study protocol was approved by the Uni versity of Florida Institutional Review Board and we obtained written informed consent from all participants in the study. The study included 143 patients with HCC in the setting of cirrhosis, 61 liver DCs and 30 NCs. Cirrhosis in the HCC patients develope d from v arious primary etiologies (Figure 3 1 ). The DC etiologies included patients with chronic HCV or HBV infection, chronic HCV or HBV related cirrhosis, HCV and HBV co infection, alcohol abuse, NAFLD a nd cryptogenic cirrhosis All DCs were evaluated for stage of fibrosis with a liver biopsy and serum collection was performed on the same day as the biopsy. Of the 61 DCs, 54 had ci rrhosis and these patients had a fi (121). All DCs with cirrhosis and patients with HCC enrolled in this study had well compensated cirrhosis compatible with Child Pugh A classification. No patient had a performance status >1 and th e majority of HCC patients had a performance status score of 0. Blood samples from gender and age matched NCs (n=30; 15 males, 15 females) were obtained from the local blood bank (Life South, Gainesville, FL, USA). HCC was diagnosed according to the noninv asive radiological criteria per the AASLD guidelines (13). The staging of HCC was performed using the Barcelona Clinic Liver Cancer (BCLC) staging system. Patients diagnosed with stage A HCC had a single lesion <5 cm or 2 3 lesions <3 cm in
76 size. Multinodu lar lesions >5 cm were characteristic tumor features in patients with stage B HCC. Macroscopic vascular invasion or metastatic disease was established in patients with Stage C HCC (122). Within the group of DCs, 44 patients had HCV related cirrhosis confirmed by histopathology. The patients with confirmed HCV related cirrhosis were enrolled into our surveillance program, received serial cross sectional imaging every six months and had no liver masses on enrollment and 12 months afte r enrollment. The following clinical data were obtained for each HCC patient: age, gender, ethnicity, etiology of HCC, BCLC stage, AFP level and Model for End Stage Liver Disease (MELD) score. For the DCs, we obtained age, gender, ethnicity, etiology of li ver disease, MELD score and AFP data for 46 of the 61 DC patients. AFP was measured in 46 of the 61 DC patients. Laboratory hepatic function data needed for MELD score calculation was not obtained for 10 of the 61 DCs at the time of their enrollment Serum Preparation and sIL 2R ELISA for sCD25 Quantification Whole blood samples were collected on the clinic date when patients were diagnosed with HCC and processed for serum isolation. Then, using sIL 2R ELISA (Bender MedSystems, Vienna, Austria), we processed fresh samples for sCD25 in duplicate using our previous approach (117). Briefly, microtiter plates coated with anti human sIL 2R antibody were inculcated with serum containing sCD25 and subjected to horseradish peroxide, a substrate solution that upon addition induced a color change. The intensity of the colored product was directly proportional to the level of sCD25 present in each well. Plates were read at 450nm using the SpectraMax190 reader (M olecular Devices, Sunnyvale, CA, USA). Measurements of sCD25 above the upper
77 limit of the calibration range (20,000 pg/ml) were diluted by half using buffer from the manufacturer. Statistical Analysis Data for sCD25 and AFP levels are expressed as box plot s with medians SD. Receiver operator characteristic (ROC) curves with respective points of maximal accuracy for sensitivity and specificity were generated to determine biomarker performance. The Parikh model for combined sensitivity and specificity was u sed to determine combined diagnostic utility of sCD25+AFP (123). The multiple regression test was used to evaluate the correlation between clinical parameters and sCD25 level. We used the Mann Whitney U test to assess the significance of group differences in the level of sCD25. Spearman's rank correlation coefficient was used to examine the correlation between HCC stage and the level of sCD25. P<0.05 was considered to indicate a statistically s ignificant result. Statistical data were analyzed using MedCalc version 126.96.36.199 (MedCalc Software, Mariakerke, Belgium). Results Clinical Characteristics of HCC Patients and DCs In our study, the majority of HCC patients were male (79%) and Caucasian (82% ). African Americans, Hispanics and Asians were less common (Figure 3 1 ). A near equal distribution of patients was present across all stages of HCC, with 48 patients having stage A disease (early HCC), 45 patients being in the stage B subset (intermediate HCC) and 50 patients having stage C cancer (advanced HCC). The average age of the HCC patients was 63.6 years. Of the HCC patients, 60% had chronic HCV infection, making HCV the predominant etiology of their liver disease.
78 Other primary causes of HCC included alcoh ol related cirrhosis (n=13), chronic HBV infection (n=7) and NAFLD/non alcoholic steatohepatitis (NASH; n=13). The majority of HCC patients (97%) had a MELD score <15. Most HCC patients (67.13%) had an AFP level <400ng/ml. For the DCs, the main etiology of their liver disease was chronic HCV infection and nearly all patients ha d an AFP level <400ng/ml sCD25 levels in HCC patients. Relationship Between sCD25 Level and Extent of Liver Disease The serum levels of sCD25 were detected at a significantly higher level in HCC patients than NCs and DCs (Figure 3 2) The median value of sCD25 in the HCC patients (6,955 pg/ml) was significantly higher than that of the DCs and NCs (P<0.0001). The median level of sCD25 in the DC group (4,310 pg/ml) was higher than that in the NCs (2,098 pg/ml), but the difference was not significant (P=0.0676). The levels of sCD25 of HCC patients were not correlated with age, gender or MELD score. AFP levels in HCC patients. AFP levels were obtained for DC and HCC patients but not the N Cs. The difference between the median AFP level in HCC patients (49.4 ng/ml) and that in DCs (5.3 ng/ml) w as significant (P<0.0001 ). For the DC and HCC patients, there was no correlation between the AFP levels and age, gender or MELD score. Utility of sCD2 5 in Predicting HCC Presence After showing that sCD25 was able to differentiate HCC from controls with cirrhosis (P<0.0001), we then determined its capacity to detect the presence of HCC. This analysis showed that at a cutoff value of 2,180 pg/ml, sCD25 ha d a sensitivity of 92.3% and a specificity of 37.7% for detecting HCC (Figure 3 3 ). By comparison, AFP had a sensitivity of 53.8% and a specificity of 86.8% at a cutoff value of 32.3 ng/ml. The area under the curve (AUC) values for sCD25 and AFP were 0.685 and 0.755,
79 respectively. At 20 ng/ml, the recommended clinical cutoff value for AFP used in clinical practice, the sensitivity of AFP was 60.1% and the specificity was 81.8% (AUC=0.733). sCD25 as a Marker for Early Stage HCC We evaluated the performance of sCD25 in detecting early HCC by comparing the level of sCD25 in patients with BCLC stage A HCC with the sCD2 5 responses of DC patients (Figure 3 4 ). In this ROC analysis, an optimal cut off value of 2,859 pg/ml for sCD25 had a sensitivity of 89.6% and a specificity of 39.3% with an AUC of 0.630 (P<0.0001). By comparison, at a cutoff value of 20 ng/ml, AFP had a sensitivity of 41.7% and a specificity of 8 2.6% (AUC=0.630, P=0.0257 ). Correlation between Tumor Burden and sCD25 Level We evaluated the correl ation between levels of sCD25 and tumor burden (Figure 3 5) We also examined the correlation between AFP and tumor burden. We observed a progressive elevation in the level of sCD25 with increasing HCC stage (R=0.213, P<0.0160). The patients with early sta ge HCC (stage A) had the lowest median level of sCD25 (6,339 pg/ml), while patients with multinodular HCC (stage B) had an intermediate median level of sCD25 (7,365 pg/ml). The patients with advanced HCC (stage C) had the highest median l ev el of sCD25 (8,8 89 pg/ml ). We also found a strong positive correlation between A FP l evel and stage of HCC ( R=0.513, P<0.0001).Using the cutoff value of 2,859 pg/ml for sCD25 and 20 ng/ml for AFP, we evaluated the correlation betweensCD25 and size of HCC in patients with e arly HCC (Figure 3 6) In this analysis, we divided the subset of HCC patients with early HCC into patients with a single small tumor <2cm and patients with single tumors <5cm in size. In this analysis, sCD25 retained a high sensitivity in those patients w ith the smallest lesions (<2cm) and
80 in pa tients with lesions <5cm We also found that the sensitivity for AFP was low and increased with size of HCC. Discussion Given that the majority of HCC patients present with advanced disease, there is a pressing need for an effective biomarker that detects the presence and early stages of HCC at a better capacity than AFP. In the present study, we found that sCD25 effectively distinguished HCC patients from healthy and DC subjects. The levels of sCD25 in HCC patients were significantly higher than those in NCs and DCs with advanced liver fibrosis. This analysis found that sCD25 possesses a sensitivity of 92% at a cutoff value of 2,180 pg/ml for the presence of HCC and warrants additional investigation as a potential sc reening test. Furthermore, sCD25 also retained this high sensitivity (90% at a cutoff value of 2,899 pg/ml) for detecting HCC in a subset of patients with early stage HCC, highlighting its potential use as a screening tool in those at high risk for HCC. Wh en comparing sCD25 with AFP, we found that sCD25 has a higher sensitivity than AFP in detecting the presence of HCC, particularly in patients with early HCC. We also observed a positive correlation between the level of sCD25 and the degree of tumor burden of HCC, with levels of sCD25 progressively increasing from early (stage A) to advanced stage (stage C) HCC. This correlation was consistent with our previous study (117), which revealed a positive correlation between serum levels of sCD25 and tumor burden. When combined with AFP, the sensitivity and specificity of sCD25+AFP were 96.4 and 93.2, respectively, underlying the potential complementary value sCD25 offers with AFP testing. These findings suggest that the measurement of serum levels of the immune ma rker sCD25 may improve earlier detection of HCC and could potentially be a useful novel prognostic
81 marker. A number of serum markers have been evaluated to detect HCC, including AFP, lectin bound AFP (AFP L3 ) and des co mmonly used serological assay to detect HCC is the blood test for AFP. However, AFP at the clinically recommended cutoff value of 20 ng/ml suffers from poor sensitivity (124). Studies have suggested DCP and AFP L3 as potential biomarkers for HCC, particula rly when used in a complementary fashion (125 127). However, a multicenter phase II biomarker study, using a total of 836 patients with 50% of the patients being controls with cirrhosis and 50% having HCC, showed the sensitivity of 60% for AFP to be better than those of DCP and AFP L3% (128). For those patients with early stage HCC, the sensitivity of 65% for AFP was more sensitive than DCP and AFP L3%. Studies that combined the three markers AFP L3%, DCP and AFP have failed to show a substantial improvement in sensitivity, even wh en using a cohort of HCC patients with large, unresectable tumors (129). When patients with early HCC were analyzed from a nested case control study of 39 HCC cases developing during the randomized Hepatitis C Antiviral Long term Treatment (HALT C) trial, the sensitivity dropped to 47% for DCP and 61% for AFP (130). Numerous studies have shown that the sensitivity of these markers drops as a function of decreasing tumor size, highlighting an insufficient sensitivity for these markers at detecting the onset of cancer at its earliest stage (131 133). The need to be able to detect HCC at its earliest stage was further demonstrated in the HALT C trial, which analyzed the serum levels of AFP and DCP for 12 months prior to the diagnosis of HCC. DCP and AFP had good specificities (94 and 75%, respectively), but possessed markedly low sensitivities, of 47 and 43%, respectively.
82 In the present stud y, sCD25 was more sensitive than AFP in distinguishing patients with early HCC from cirrhosis control patients. Moreover, our study showed that AFP level was not correlated with early stage HCC lesions smaller than 3cm (P=0.1148). This finding is consisten t with the conclusions of other studies (27) demonstrating a lack of sensitivity of the AFP serological test when used in the screening for early tumors. While AFP had poor sensitivity in our study, it did show a high specificity for both HCC presence and early HCC in comparison to the low specificity of sCD25 in these analyses. The inadequate specificity of sCD25 is a limitation that requires further evaluation through the recruitment of a larger HCC cohort, since an ideal biomarker should possess high sen sitivity and specificity. Combining sCD25 with AFP resulted in significant improvement of sensitivity and specificity and this combination may be of benefit. AFP continues to be widely used but concern over its poor performance as a marker has led to its e xclusion as a recommended test for screening patients at high risk for HCC in the current practice guidelines from the AASLD (13,134, 135). Currently, the main screening strategy recommended is serial liver ultrasonography. However, this screening modality is not being effectively used since the majority of patients at the highest risk for HCC development are not undergoing surveillance (136). Ultrasonography also poses significant challenges related to availability and operator experience in interpreting i mages from cirrhotic livers and obese patients (137). The identification of a novel biomarker for HCC that detects early cancer and is capable of overcoming these limitations may improve surveillance efforts and clinical outcomes. While this study was not designed to evaluate the influence of the etiology of underlying liver disease or other clinical factors on sCD25,
83 we did not identify a correlation between clinical parameters and levels of sCD25 level. Most importantly, our study controlled for underlyin g liver function by enrolling only HCC patients with Child Pugh A cirrhosis. Biases were further eliminated through the blind implementation of bioassay procedures. Our study demonstrated a correlation between sCD25 level and tumor burden, suggesting its p otential use as predictor of prognosis at baseline. Future studies analyzing sCD25 responses in samples obtained during a surveillance program may provide further insight on the utility of sCD25 in surveillance for early HCC detection. We currently lack a reliable serum marker for the early detection of HCC. In accordance with the phase specific biomarker standardization model delineated by Pepe et al we highlight the progress of our initial study using the immune marker of sCD25 (138). Previously, we as sessed the performance of sCD25 in a small group (n=60) of HCC patients. In the present study, we expanded our analysis to a larger cohort of HCC patients (n=143) and again observed the previously shown marked elevation of sCD25 in HCC patients in comparis on to the levels manifested in healthy controls and controls with cirrhosis (11). Our findings show that sCD25 distinguished HCC from appropriate controls and that this marker identified the presence of HCC more effectively than AFP, particularly in patien ts with early tumors. The high sensitivity of sCD25 suggests it holds promise as a marker for early HCC which is an area of unmet need. To further characterize the utility of sCD25 in detecting early stages of HCC tumor development, larger longitudinal and validation studies are planned.
84 Figure 3 1. Clinical characteristics of DCs (n=61) and HCC patients (n=143).
85 Figure 3 2 Levels of sCD25 and AFP in NHC, DC, and HCC. A) Log transformed expression of sCD25 (pg/ml). Levels of sCD2 5 were significantly different between HCC patients and DCs but did not differ significantly between DCs and NHC. Respective sCD25 medians of 2,098, 4,310, and 6,955 pg/ml for NHC, DCs, and HCC patients B) AFP level (ng/ml) in HCC patients and disease con trols. AFP expression was strongly elevated in HCC patients vs. cirrhosis controls. M edian level of 5.3 ng/ml for DC subjects and 49.4 ng/ml for HCC patients sCD25 AFP
86 Figure 3 3 Receiver operating characteristic curve analysis reveals i ndependent and combined sensitivity and specificity of sCD25 and AFP Red dot denotes point of maximal accuracy. A) sC D25 ROC curve. B) AFP ROC curve Figure 3 4 ROC curves for sCD25 Soluble CD25 outperformed AFP in accurately resolving HCC patients from cirrhosis contro ls, possessing a sensitivity of 89.6. The specificity of AFP was significantly higher than that of sCD25.
87 Figure 3 5 Correlation of sCD25 and AFP with tumor burden. Tumor burden based on the BCLC staging system. A) sCD25 exhibited a direct and significa nt relationship to tumor burden. B) AFP correlated more strongly with tumo r burden than sC D25 Figure 3 6. Capacity of sCD25 to detect small early HCC lesions in the setting of cirrhosis.
88 LIST OF REFERENCES 1. Altekruse SF, McGlynn KA, Reichman ME. Hepato cellular carcinoma incidence, mortality, and survival trends in the United States from 1975 to 2005. J Clin Oncol. 2009 Mar 20;27(9):1485 91. 2. El Serag HB. Hepatocellular carcinoma. N Engl J Med. 2011 Sep 22;365(12):1118 27. 3. Hussain SM, Semelka RC, Mitche ll DG. MR imaging of hepatocellular carcinoma. Magn Reson Imaging Clin N Am 2002;10:31 52. 4. El Serag HB, Rudolph KL. Hepatocellular carcinoma: epidemiology and molecular carcinogenesis. Gastroenterology 2007;132:2557 76. 5. El Serag HB, Mason AC. Rising inci dence of hepatocellular carcinoma in the United States N Engl J Med. 1999;340:745 50. 6. Kew MC. Epidemiology of hepatocellular carcinoma in sub Saharan Africa. Ann Hepatol. 2013 Mar;12(2):173 82. 7. Cedrone A, Covino M, Caturelli E, et al Utility of alpha f etoprotein (AFP) in the screening of patients with virus related chronic liver disease: does different viral etiology influence AFP levels in HCC? A study in 350 western patients. Hepatogastroenterology. 2000 Nov Dec;47(36):1654 8. 8. Gupta S, Bent S, Kohlwe s J. Test characteristics of alpha fetoprotein for detecting hepatocellular carcinoma in patients with hepatitis C. A systematic review and critical analysis. Ann Intern Med. 2003 Jul 1;139(1):46 50. 9. Ryosuke Tateishi, Yoshida H, Matsuyama Y, Mine N, Kondo Y, Omata M. Diagnostic accuracy of tumor markers for hepatocellular carcinoma: a systematic review. Hepatol Int. 2008 March; 2(1): 17 30. 10. Bialecki ES, Di Bisceglie AM. Diagnosis of hepatocellular carcinoma. HPB (Oxford). 2005; 7(1): 26 34. 11. Chen CH, Lin ST, Kuo CL, Nien CK. Clinical significance of elevated alpha fetoprotein (AFP) in chronic hepatitis C without hepatocellular carcinoma. Hepatogastroenterology. 2008 Jul Aug;55(85):1423 7. 12. Lai Q, Melandro F, Pinheiro RS, et al Alpha fetoprotein and novel tumor biomarkers as predictors of hepatocellular carcinoma recurrence after surgery: a brilliant star raises again. Int J Hepatol. 2012;2012:893103.
89 13. Bruix J, Sherman M; Practice Guidelines Committee, American Association for the Study of Liver Diseases. M anagement of hepatocellular carcinoma. Hepatology. 2005 Nov;42(5):1208 36. 14. Weitz IC, Liebman HA. Des gamma carboxy (abnormal) prothrombin and hepatocellular carcinoma: a critical review. Hepatology. 1993 Oct;18(4):990 7. 15. Wild CP, Hall AJ. Primary prevent ion of hepatocellular carcinoma in developing countries. Mutat Res. 2000 Apr;462(2 3):381 93. 16. El Serag HB, Davila J. Surveillance for hepatocellular carcinoma: in whom and how? Therap Adv Gastroenterol. 2011 January; 4(1): 5 10. 17. Venook AP, Papandreou C, Furuse J, de Guevara LL. The incidence and epidemiology of hepatocellular carcinoma: a global and regional perspective. Oncologist. 2010;15 Suppl 4:5 13. 18. Qiao B, Wang J, Xie J, et al Detection and identification of peroxiredoxin 3 as a biomarker in hep atocellular carcinoma by a proteomic approach. Int J Mol Med. 2012 May;29(5):832 40. 19. Marshall A, Lukk M, Kutter C, Davies S, Alexander G, Odom DT. Global Gene Expression Profiling Reveals SPINK1 as a Potential Hepatocellular Carcinoma Marker. PLoS One. 20 13;8(3):e59459. 20. Cabrera R, Fitian AI, Ararat M, et al Serum levels of soluble CD25 as a marker for hepatocellular carcinoma. Oncol Lett. 2012 Oct;4(4):840 846. 21. Woo HG, Park ES, Thorgeirsson SS, Kim YJ. Exploring genomic profiles of hepatocellular carcin oma. Mol Carcinog. 2011 Apr;50(4):235 43. 22. Shang S, Plymoth A, Ge S, et al Identification of osteopontin as a novel marker for early hepatocellular carcinoma. Hepatology. 2012 Feb;55(2): 483 90. doi: 10.1002/hep.24703. 23. Rodrigues RR, Barry CT. Gene pathway analysis of hepatocellular carcinoma genomic expression datasets. J Surg Res. 2011 Sep;170(1):e85 92. 24. Pleguezuelo M, Lopez Sanchez LM, Rodriguez Ariza A, et al Proteomic analysis for developing new biomarkers of hepatocellular carcinoma. World J Hepatol 2010 Mar 27;2(3):127 35. 25. Nguyen P, Leray V, Diez M, et al Liver lipid metabolism. J Anim Physiol Anim Nutr (Berl). 2008 Jun;92(3):272 83.
90 26. Mitra V, Metcalf J. Metabolic functions of the liver. Anaesthesia and Intensive Care Medicine. 10(7): 334 335. 27. Xu EY, Schaefer WH, Xu Q. Metabolomics in pharmaceutical research and development: metabolites, mechanisms and pathways. Curr Opin Drug Discov Devel. 2009 Jan;12(1):40 52. 28. Yang Y, Li C, Nie X, et al 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 Jul;6(7):2605 14. 29. Patterson AD, Maurhofer O, Beyoglu D, et al Aberrant lipid metabolism in hepatocellular carcinoma reveal ed by plasma metabolomics and lipid profiling. Cancer Res. 2011 Nov 1;71(21):6590 600. 30. Chen T, Xie G, Wang X, et al Serum and urine metabolite profiling reveals potential biomarkers of human hepatocellular carcinoma. Mol Cell Proteomics. 2011 Jul;10(7):M110.004945. 31. Wu FX, Wang Q, Zhang ZM, et al Identifying serological biomarkers of hepatocellular carcinoma using surface enhanced laser desorption/ionization time of flight mass spectroscopy. Cancer Lett. 2009 Jul 8;279(2 ):163 70. 32. Cao H, Huang H, Xu W, et al Fecal metabolome profiling of liver cirrhosis and hepatocellular carcinoma patients by ultra performance liquid chromatography mass spectrometry. Anal Chim Acta. 2011 Apr 8;691(1 2):68 75. 33. Gao H, Lu Q, Liu X, et al Application of 1H NMR based metabonomics in the study of metabolic profiling of human hepatocellular carcinoma and liver cirrhosis. Cancer Sci. 2009 Apr;100(4):782 5. 34. Wu H, Xue R, Dong L, et al Metabolomic profiling of human urine in hepatocellular carc inoma patients using gas chromatography/mass spectrometry. Anal Chim Acta. 2009 Aug 19;648(1):98 104. 35. Soga T, Sugimoto M, Honma M, et al glutamyl dipeptides as biomarkers for discrimination among different forms of liver disease. J Hepatol. 2011 Oct;55(4):896 905. 36. Ressom HW, Xiao JF, Tuli L, et al Utilization of metabolomics to identify serum biomarkers for hepatocellular carcinoma in patients with liver cirrhosis. Anal Chim Acta. 2012 Sep 19;743:90 100.
91 37. Xiao JF, Varghese RS, Zhou B, et al LC MS based serum metabolomics for identification of hepatocellular carcinoma biomarkers in Egyptian cohort. J Proteome Res. 2012 Dec 7;11(12):5914 23. 38. Wang B, Chen D, Chen Y, et al Metabonomic profiles discriminate hepatocellular carcinoma from liver cirrhosis by ultraperformance liquid chromatography mass spectrometry. J Proteome Res. 2012 Feb 3;11(2):1217 27. 39. Zhou L, Ding L, Yin P, et al Serum metabolic profiling study of hepatocellular carcinoma infected with hepatitis B or hepatitis C virus by using liquid chromatography mass spectrometry. J Proteome Res. 2012 Nov 2;11(11):5433 42. 40. O, et al Tissue metabolomics of hepatocellular carcinoma: Tumor energy metabolism and the role of transcriptomic classification. Hepatology. 2013 Mar 5. doi: 10.1002/hep.26350. [Epub ahead of print] 41. Yin P, Wan D, Zhao C, et al A metabonomic study of hep atitis B induced liver cirrhosis and hepatocellular carcinoma by using RP LC and HILIC coupled with mass spectrometry. Mol Biosyst. 2009 Aug;5(8):868 76. 42. Li S, Liu H, Jin Y, et al Metabolomics study of alcohol induced liver injury and hepatocellular carc inoma xenografts in mice. J Chromatogr B Analyt Technol Biomed Life Sci. 2011 Aug 15;879(24):2369 75. 43. El Mir MY, Badia MD, Luengo N, Monte MJ, Marin JJ. Increased levels of typically fetal bile acid species in patients with hepatocellular carcinoma. Clin Sci (Lond). 2001 May;100(5):499 508. 44. El Houseini ME, Amer MA, El Din AS, El Sherbiny M, Hussein TD, Mansour O. Evaluation of Serum Total Bile Acids in the Diagnosis of Hepatocellular Carcinoma. Journal of the Egyptian Nat. Cancer Inst., 2000 Dec;12(4): 30 7 313. 45. Luo J. Cancer's sweet tooth for serine. Breast Cancer Res. 2011;13(6):317. 46. Locasale JW, Grassian AR, Melman T, et al. Phosphoglycerate dehydrogenase diverts glycolytic flux and contributes to oncogenesis. Nat Genet 2011;43:869 74. 47. Possemato R, Ma rks KM, Shaul YD, et al. Functional genomics reveal that the serine synthesis pathway is essential in breast cancer. Nature 2011;476:346 50. 48. Jain M, Nilsson R, Sharma S, et al. Metabolite profiling identifies a key role for glycine in rapid cancer cell pr oliferation. Science 2012;336:1040 4.
92 49. Kim JW, Dang CV. Cancer's molecular sweet tooth and the Warburg effect. Cancer Res. 2006 Sep 15;66(18):8927 30. 50. Chaneton B, Hillmann P, Zheng L, et al. Serine is a natural ligand and allosteric activator of pyruvate kinase M2. Nature. 2012 Nov 15;491(7424):458 62. 51. Szutowicz A, Tomaszewicz M, Bielarczyk H. Disturbances of acetyl CoA, energy and acetylcholine metabolism in some encephalopathies. Acta Neurobiol Exp (Wars). 1996;56(1):323 39. 52. Kreis R, Farrow N, Ross BD. Localized 1H NMR spectroscopy in patients with chronic hepatic encephalopathy. Analysis of changes in cerebral glutamine, choline and inositols. NMR Biomed. 1991 Apr;4(2):109 16. 53. Meyer K, Jia Y, Cao WQ, Kashireddy P, Rao MS. Expression of peroxisome prol iferator activated receptor alpha, and PPARalpha regulated genes in spontaneously developed hepatocellular carcinomas in fatty acyl CoA oxidase null mice. Int J Oncol. 2002 Dec;21(6):1175 80. 54. Lalwani ND, Reddy MK, Qureshi SA, Reddy JK. Development of hepa tocellular carcinomas and increased peroxisomal fatty acid beta oxidation in rats fed [4 chloro 6 (2,3 xylidino) 2 pyrimidinylthio] acetic acid (Wy 14,643) in the semipurified diet. Carcinogenesis. 1981;2(7):645 50. 55. Reddy JK, Hashimoto T. Peroxisomal beta oxidation and peroxisome proliferator activated receptor alpha: an adaptive metabolic system. Annu Rev Nutr. 2001;21:193 230. 56. Christensen E, Woldseth B, Hagve TA, et al Peroxisomal beta oxidation of polyunsaturated long chain fatty acids in human fibrob lasts. The polyunsaturated and the saturated long chain fatty acids are retroconverted by the same acyl CoA oxidase. Scand J Clin Lab Invest Suppl. 1993;215:61 74. 57. Ojala PJ, Hirvonen TE, Hermansson M, Somerharju P, Parkkinen J. Acyl chain dependent effect of lysophosphatidylcholine on human neutrophils. J Leukoc Biol. 2007 Dec;82(6):1501 9. 58. Hayashi H, Miwa A. The role of peroxisomal fatty acyl CoA beta oxidation in bile acid biosynthesis. Arch Biochem Biophys. 1989 Nov 1;274(2):582 9. 59. Furuya H, Shimizu Y Kawamori T. Sphingolipids in cancer. Cancer Metastasis Rev. 2011 Dec;30(3 4):567 76. doi: 10.1007/s10555 011 9304 1.
93 60. Bao M, Chen Z, Xu Y, et al Sphingosine kinase 1 promotes tumour cell migration and invasion via the S1P/EDG1 axis in hepatocellular car cinoma. Liver Int. 2012 Feb;32(2):331 8. 61. Yang L, Yue S, Yang L, et al Sphingosine Kinase/Sphingosine 1 Phosphate (S1P)/S1P Receptor Axis is Involved in Liver Fibrosis Associated Angiogenesis. J Hepatol. 2013 Mar 2. pii: S0168 8278(13)00143 8. doi: 10.10 16/j.jhep.2013.02.021. [Epub ahead of print] 62. Beljanski V, Lewis CS, Smith CD. Antitumor activity of sphingosine kinase 2 inhibitor ABC294640 and sorafenib in hepatocellular carcinoma xenografts. Cancer Biol Ther. 2011 Mar 1;11(5):524 34. 63. Thompson J, Zha ng J, Vanden Heuvel J. Activation of farnesoid X receptor (FXR) by bioactive lipids. 64. Fujino T, Takeuchi A, Maruko Ohtake A, et al Critical role of farnesoid X receptor for hepatocellular carcinoma cell proliferation. J B iochem. 2012 Dec;152(6):577 86. 65. K umagai A, Fukushima J, Takikawa H, Fukuda T, Fukusato T. Enhanced expression of farnesoid X receptor in human hepatocellular carcinoma. Hepatol Res. 2013 Jan 10. doi: 10.1111/hepr.12047. [Epub ahead of print] 66. Kim Y, Lin Q, Glazer PM, Yun Z. Hypoxic tumor microenvironment and cancer cell differentiation. Curr Mol Med. 2009 May;9(4):425 34. 67. Marra M, Sordelli IM, Lombardi A et al Molecular targets and oxidative stress biomarkers in hepatocellular carcinoma: an overview. J Transl Med. 2011 Oct 10;9:171. 68. Su zuki Y, Imai K, Takai K, et al Hepatocellular carcinoma patients with increased oxidative stress levels are prone to recurrence after curative treatment: a prospective case series study using the d ROM test. J Cancer Res Clin Oncol. 2013 Feb 15. 69. Hannelie n V, Karel G, Jo VD, Sofie S. The role of CXC chemokines in the transition of chronic inflammation to esophageal and gastric cancer Biochim Biophys Acta. 2012 Jan;1825(1):117 29. doi: 10.1016/j.bbcan.2011.10.008. 70. Kong J, Kong L, Kong J, et al After insu fficient radiofrequency ablation, tumor associated endothelial cells exhibit enhanced angiogenesis and promote
94 invasiveness of residual hepatocellular carcinoma. J Transl Med. 2012 Nov 21;10:230. 71. Bechara C, Chai H, Lin PH, Yao Q, Chen C. Growth related on cogene alpha (GRO alpha): roles in atherosclerosis, angiogenesis and other inflammatory conditions. Med Sci Monit. 2007 Jun;13(6):RA87 90. 72. Smith DF, Galkina E, Ley K, Huo Y. GRO family chemokines are specialized for monocyte arrest from flow. Am J Physiol Heart Circ Physiol. 2005 Nov;289(5):H1976 84. 73. Zhao XL, Sun T, Che N, et al Promotion of hepatocellular carcinoma metastasis through matrix metalloproteinase activation by epithelial mesenchymal transition regulator Twist1. J Cell Mol Med. 2011 Mar;15(3 ):691 700. 74. Ebert MP, Lamer S, Meuer J, et al Identification of the thrombin light chain a as the single best mass for differentiation of gastric cancer patients from individuals with dyspepsia by proteome analysis. J Proteome Res. 2005 Mar Apr;4(2):586 9 0. 75. Dettmer K, Aronov PA, Hammock BD. Mass spectrometry based metabolomics. Mass Spectrom Rev. 2007 Jan Feb;26(1):51 78. 76. Bedosa P, Carrat F. Liver biopsy: The best, not the gold standard. J Hepatol 2009;50: 1 3. 77. Evans AM, DeHaven CD, Barrett T, Mitchell M, Milgram E. Integrated, nontargeted ultrahigh performance liquid chromatography/electrospray ionization tandem mass spectrometry platform for the identification and relative quantification of the small molecule complement of biological systems. Anal Chem 2009 Aug 15;81(16):6656 67. 78. Wu JM, Skill NJ, Maluccio MA. Evidence of aberrant lipid metabolism in hepatitis C and hepatocellular carcinoma. HPB (Oxford). 2010 Nov;12(9):625 36. 79. Stuart KE, Anand AJ, Jenkins RL. Hepatocellular carcinoma in the United Sta tes. Prognostic features, treatment outcome, and survival. Cancer. 1996 Jun 1;77(11):2217 22. 80. Uderhardt S, Krnke G. 12/15 lipoxygenase during the regulation of inflammation, immunity, and self tolerance. J Mol Med (Berl). 2012 Nov;90(11):1247 56. 81. Liu B, Timar J, Howlett J, Diglio CA, Honn KV. Lipoxygenase metabolites of arachidonic and linoleic acids modulate the adhesion of tumor cells to
95 endothelium via regulation of protein kinase C. Cell Regul. 1991 Dec;2(12):1045 55. 82. O'Flaherty JT, Showell HJ, Beck er EL, Ward PA. Neutrophil aggregation and degranulation: Effect of arachidonic acid. Am J Pathol. 1979 May;95(2):433 44. 83. Xu XM, Yuan GJ, Deng JJ, et al Inhibition of 12 lipoxygenase reduces proliferation and induces apoptosis of hepatocellular carcinoma cells in vitro and in vivo. Hepatobiliary Pancreat Dis Int. 2012 Apr;11(2):193 202. 84. Wong BCY, Wang PW, Cho HC, et al.12 Lipoxygenase inhibition induced apoptosis in human gastric cancer cells Carcinogenesis (2001) 22(9): 1349 1354. 85. Chen YQ, Duniec ZM, Liu B, et al. Endogenous 12(S) HETE production by tumor cells and its role in metastasis. Cancer Res. 1994 Mar 15;54(6):1574 9. 86. Burhop KE, Selig WM, Malik AB. Monohydroxyeicosatetraenoic acids (5 HETE and 15 HETE) induce pulmonary vasoconstriction and ede ma. Circ Res. 1988 Apr;62(4):687 98. 87. Nie X, Song S, Zhang L, Qiu Z, Shi S, Liu Y, et al 15 Hydroxyeicosatetraenoic acid (15 HETE) protects pulmonary artery smooth muscle cells from apoptosis via inducible nitric oxide synthase (iNOS) pathway. Prostagland ins Other Lipid Mediat. 2012 Jan;97(1 2):50 9. 88. Li Y, Li Q, Wang Z, Liang D, et al 15 HETE suppresses K(+) channel activity and inhibits apoptosis in pulmonary artery smooth muscle cells. Apoptosis. 2009 Jan;14(1):42 51. 89. Ma J, Zhang L, Zhang J, et al 15 Lipoxygenase 1/15 hydroxyeicosatetraenoic acid promotes hepatocellular cancer cells growth through protein kinase B and heat shock protein 90 complex activation. Int J Biochem Cell Biol. 2013 Mar 6;45(6):1031 1041. 90. Marra M, Sordelli IM, Lombardi A, et al Molecular targets and oxidative stress biomarkers in hepatocellular carcinoma: an overview. J Transl Med. 2011 Oct 10;9:171. doi: 10.1186/1479 5876 9 171. 91. Lin CC, Yin MC. B vitamins deficiency and decreased anti oxidative state in patients with liver ca ncer. Eur J Nutr. 2007 Aug;46(5):293 9. 92. Stirpe F, Ravaioli M, Battelli MG, Musiani S, Grazi GL. Xanthine oxidoreductase activity in human liver disease. Am J Gastroenterol. 2002 Aug;97(8):2079 85.
96 93. Shi DY, Xie FZ, Zhai C, Stern JS, Liu Y, Liu SL. The role of cellular oxidative stress in regulating glycolysis energy metabolism in hepatoma cells. Mol Cancer. 2009 Jun 5;8:32. 94. Li T, Zhao XP, Wang LY, et al Glutathione s transferase p1 correlated with oxidative stress in hepatocellular carcinoma. Int J Med Sc i. 2013;10(6):683 90. doi: 10.7150/ijms.5947. Epub 2013 Apr 3. 95. Sherman M, Campbell JA, Titmuss SA, Kew MC, Kirsch RE. Glutathione S transferase in human hepatocellular carcinoma. Hepatology. 1983 Mar Apr;3(2):170 6. 96. Casaril M, Gabrielli GB, Dusi S, Nicol i N, Bellisola G, Corrocher R. Decreased activity of liver glutathione peroxidase in human hepatocellular carcinoma. Eur J Cancer Clin Oncol. 1985 Aug;21(8):941 4. 97. Shimazu T, Hirschey MD, Newman J, et al. Suppression of Oxidative Stress by Hydroxybutyrate, an Endogenous Histone Deacetylase Inhibitor. Science 2013: 339 (6116), 211 214. 98. Gazarian KG, Gening LV, Gazarian TG.L Homoserine: a novel excreted metabolic mar ker of hepatitis B abnormally produced in liver from methionine. Med Hypotheses. 2002 Apr;58(4):279 83. 99. Huang W, Chen C, Lin Y, and Lin C. Apoptotic Sphingolipid Ceramide in Cancer Therapy. Journal of Lipids (2011);Article ID:565316. doi:10.1155/2011/5653 16 100. Morales A, Pars R, Villanueva A, Llacuna L, Garca Ruiz C, Fernndez Checa JC. Pharmacological inhibition or small interfering RNA targeting acid ceramidase sensitizes hepatoma cells to chemotherapy and reduces tumor growth in vivo. Oncogene. 2007 Feb 8;26(6):905 16. 101. Bedia C, Casas J, Andrieu Abadie N, Fabris G, Levade T. Acid ceramidase expression modulates the sensitivity of A375 melanoma cells to dacarbazine. J Biol Chem. 2011 Aug 12;286(32):28200 9. 102. Mas VR, Maluf DG, Archer KJ, Yanek K et al. Ge nes involved in viral carcinogenesis and tumor initiation in hepatitis C virus induced hepatocellular carcinoma. Mol Med 2009 Mar Apr;15(3 4):85 94.
97 103. Beljanski V, Knaak C, Zhuang Y, Smith CD. Combined anticancer effects of sphingosine kinase inhibitors and sorafenib. Invest New Drugs. 2011 Dec;29(6):1132 42. 104. Vamecq J, Draye JP, Brison J. Rat liver metabolism of dicarboxylic acids. Am J Physiol. 1989 Apr;256(4 Pt 1):G680 8. 105. Falco OB. Dermatology. Springer Publishing (2000); p 319. 106. Bowe WP, Logan AC. Acne vulgaris, probiotics and the gut brain skin axis back to the future? Gut Pathog. 2011 Jan 31;3(1):1. 107. McBride ME, Duncan WC, Bodey GP, McBride CM. Microbial skin flora of selected cancer patients and hospital personnel. J Clin Microbiol. 1976 January; 3( 1): 14 20. 108. Barnes S, Gallo GA, Trash DB, Morris JS. Diagnositic value of serum bile acid estimations in liver disease. J Clin Pathol. 1975 Jun;28(6):506 9. 109. Neale G, Lewis B, Weaver V, Panveliwalla D. Serum bile acids in liver disease. Gut. 1971 Feb;12(2) :145 52. 110. Festi D, Morselli Labate AM, Roda A, Bazzoli F, Frabboni R, Rucci P, Taroni F, Aldini R, Roda E, Barbara L.Diagnostic effectiveness of serum bile acids in liver diseases as evaluated by multivariate statistical methods. Hepatology. 1983 Sep Oct;3 (5):707 13. 111. Tytgat GN, Collen D, Verstraete M. Metabolism of fibrinogen in cirrhosis of the liver. J Clin Invest. 1971 Aug;50(8):1690 701. 112. Amodio P, Angeli P, Merkel C, Menon F, Gatta A. Plasma carnitine levels in liver cirrhosis: relationship with nutri tional status and liver damage. J Clin Chem Clin Biochem. 1990 Sep;28(9):619 26. 113. Alonso de la Pea C, Rozas I, Alvarez Prechous A, Pardias MC, Paz JM, Rodriguez Segade S. Free carnitine and acylcarnitine levels in sera of alcoholics. Biochem Med Metab Bi ol. 1990 Aug;44(1):77 83.
98 114. Morris C, O'Grada C, Ryan M, Roche HM, Gibney MJ, Gibney ER, Brennan L. The relationship between BMI and metabolomic profiles: a focus on amino acids. Proc Nutr Soc. 2012 Nov;71(4):634 8. 115. Nakamoto Y, Guidotti LG, Kuhlen CV, Fowle r P and Chisari FV: Immune pathogenesis of hepatocellular carcinoma. J Exp Med. 1998;188: 341350. 116. Cacalano NA and Johnston JA: Interleukin 2 signaling and inherited immunodeficiency. Am J Hum Genet. 1999;65: 287 293. 117. Ca brera R, Ararat M, Cao M, et al. Hepatocellular carcinoma immunopathogenesis: clinical evidence for global T cell defects and an immunomodulatory role for soluble CD25 (sCD25) Dig Dis Sci. 2010;55: 484 495. 118. Hoechst B, Ormandy LA, Ballmaier M, et al. A new population of myeloid derived suppressor cells in hepatocellular carcinoma patients induces CD4(+)CD25(+)Foxp3(+) T cells. Gastroenterology. 2008;135: 234 243. 119. 13. Arun B, Cu rti BD, Longo DL, et al. Elevations in serum soluble interleukin 2 receptor levels predict relapse in patients with hairy cell leukemia. Cancer J Sci Am. 2000;6: 21 24. 120. Ma nshouri T, Do KA, Wang X, et al. Circulating CD20 is detectable in the plasma of pati ents with chronic lymphocytic leukemia and is of prognostic significance. Blood. 2003;101: 2507 2513. 121. Knodel l RG, Ishak KG, Black WC, et al. Formulation and application of a numerical scoring system for assessing histological activity in asymptomatic chron ic active hepatitis. Hepatology. 1981;1: 431 435. 122. Llovet JM, Br C and Bruix J. Prognosis of hepatocellular carcinoma: the BCLC staging classification. Semin Liver Dis. 1999;19: 329 338. 123. Parikh R, Mathai A, Parikh S, Chandra Sekhar G, Thomas R. Understand ing and using sensitivity, specificity and predictive values. Indian J Ophthalmol. 2008 Jan Feb;56(1):45 50. 124. Trevisani F, D'Intino PE, Morselli Labate AM, et al. Serum alpha fetoprotein for diagnosis of hepatocellular carcinoma in patients with chronic liv er disease: influence of HBsAg and anti HCV status. J Hepatol. 2001;34: 570 575. 125. Sassa T, K umada T, Nakano S and Uematsu T. Clinical utility of simultaneous measurement of serum high sensitivity des gamma carboxy prothrombin and Lens culinaris agglutinin A reactive alpha fetoprotein in patients with small hepatocellular carcinoma. Eur J Gastroenterol Hepatol. 1999;11: 1387 1392. 126. Shimauchi Y, Tanaka M, Kuromatsu R, et al. A simultaneous monitoring of Lens culinaris agglutinin A reactive alpha fetoprotein an d des gamma carboxy
99 prothrombin as an early diagnosis of hepatocellular carcinoma in the follow up of cirrhotic patients. Oncol Rep. 2000;7: 249 256. 127. Marrero JA, Su GL, Wei W, et al: Des gamma carboxypro thrombin can differentiate hepatocellular carcinoma from nonmalignant chronic liver disease in American patients. Hepatology. 2003;37: 1114 1121. 128. Marrero JA, Feng Z, Wang Y, et al: Alpha fetoprotein, des gamma carboxyprothrombin, and lectin bound alpha fetoprotein in early hepatocellular carcinoma. Gastroe nterology. 2009;137: 110 118. 129. Carr BI, Kanke F, Wise M and Satomura S: Clinical evaluation of lens culinaris agglutinin reactive alpha fetoprotein and des gamma carboxy prothrombin in histologically proven hepatocellular carcinoma in the United States. Dig Dis Sci. 2007;52: 776 782. 130. Di Bisceglie AM, Sterling RK, Chung RT, et al; HALT C Trial Group: Serum alpha fetoprotein levels in patients with advanced hepatitis C: results from the HALT C trial. J Hepatol. 2005;43: 434 441. 131. Nakamura S, Nouso K, Sakaguchi K, et al: Sensitivity and specificity of des gamma carboxy prothrombin for diagnosis of patients with hepatocellular carcinomas varies according to tumor size. Am J Gastroenterol. 2006;101: 2038 2043. 132. Sterling RK, Wright EC, Morgan TR, et al: Frequency of elevated hepatocellular carcinoma (HCC) biomarkers in patients with advanced hepatitis C. Am J Gastroenterol. 2012;107: 64 74. 133. Lok AS, Sterling RK, Everhart JE, et al; HALT C Trial Group: Des gamma carboxy prothrombin and alpha fetoprotein as biomarkers f or the early detection of hepatocellular carcinoma. Gastroenterology. 2010;138: 493 502. 134. el Houseini ME, Mohammed MS, Elshemey WM, Hussein TD, Desouky OS and Elsayed AA: Enhanced detection of hepato cellular carcinoma. Cancer Control. 2005;12: 248 253. 135. Yu en MF and Lai CL: Screening for hepatocellular carcinoma: survival benefit and cost effectiveness. Ann Oncol. 2003;14:1463 1467. 136. Davila JA, Henderson L, Kramer JR, et al: Utilization of surveillance for hepatocellular carcinoma among hepatitis C virus infe cted veterans in the United States. Ann Intern Med. 2011;154: 85 93. 137. Daniele B, Bencivenga A, Megna AS and Tinessa V: Alpha fetoprotein and ultrasonography screening for hepatocellular carcinoma. Gastroenterology. 2004;127(Suppl 1): S108 S112.
100 138. Pepe MS, Et zioni R, Feng Z, et al: Phases of biomarker devel opment for early detection of cancer. J Natl Cancer Inst. 2001;93: 1054 1061.
101 BIOGRAPHICAL SKETCH Asem Fitian was born in 1987 to Naela and Issa After hopping around five states in America and graduatin g high school in Pittsburgh, Pennsylvania he settled in Gainesville where he earned a Bachelor of Science in b iology at the University of Florida. After finishing his undergraduate study he got his first taste of biomedical Liver Research Unit laboratory. He instantly fell in love with it and, eager to continue cultivating the fertile opportunities within the Liver Unit, he entered the Tr anslational Biotechnology M.S. P rogram at the UF College of Medicine where he focused hi s research efforts on the identification of novel liver cancer biomarkers He also pursued a g raduate m inor in b usiness a dministration to gain broader understanding regulatory and industrial framework. He graduated with his M.S. in August 2013 Asem improve the quality of life of those around him and he i ntends to couple his translational research background with compassionate care as a physician