Background
Metabolic syndrome is now considered a major public health issue worldwide and the rising incidence of metabolic syndrome correlates with the high prevalence of nonalcoholic fatty liver disease (NAFLD) [
1,
2]. NAFLD encompasses a clinicopathological spectrum of diseases ranging from simple to progressive steatosis, nonalcoholic steatohepatitis (NASH), and ultimately cirrhosis and hepatocellular carcinoma (HCC) [
3]. The surge in the number of cases with NAFLD is projected to lead to an increase in the number of patients with NAFLD-related HCC [
4]. The most worrisome issue is the onset of HCC in NAFLD patients who do not yet have cirrhosis [
5]. At present, a liver biopsy is essential for the diagnosis of NASH and for prognostic risk stratification of NAFLD patients. Thus, noninvasive biomarkers indicating NAFLD activity over time that can be used as a prognostic factor for subsequent development of HCC are of great interest.
As is often the case with human biomarker studies, patients exhibit clinical heterogeneity based on habits, nutrition, comorbidities, and therapeutic interventions. In addition, the fact that HCC develops from NAFLD in humans over several decades presents difficulties in discovering serum biomarkers that reflect time-dependent processes of the disease. Many rodent models of NAFLD induced by genetic manipulations or diets have been developed [
6]. However, these rodent models do not completely replicate the histological and clinical features of NAFLD in humans. By contrast, pigs are non-primate mammals that closely resemble humans in terms of anatomy, genetics, and physiology as well as lipid metabolism. As model organisms, they allow serial examinations to be conducted using the same experimental set [
7,
8]. To generate a swine model of NAFLD associated HCC model, we adopted Microminipig (MMP) registered with the Japanese Ministry of Agriculture, Forestry and Fisheries as an experimental strain of swine. Recently, MMP was utilized as a model of atherosclerosis induced by feeding a high fat and high cholesterol diet, which means this strain must be as a useful animal model for understanding the pathophysiology of a variety of metabolic syndrome [
9].
In recent years, mass spectrometry-based proteomics has been utilized for biomarker research in various diseases. Many proteins in the blood are synthesized in the liver, and the abundance and structure of many of these proteins change in response to the longitudinal course of liver diseases, including NAFLD [
10,
11]. Thus, understanding the blood proteome corresponding to the liver pathology of different stages of the disease has the potential to provide unique information about disease-associated biomarker proteins. Although numerous biomarkers for NAFLD have been identified by proteomic techniques using animal models, there are few biomarkers that have validated the clinical utility in human cohorts [
12]. On the other hand, established clinical scoring system for NAFLD progression could predict advanced fibrosis but not development of HCC [
13].
Using a successfully established swine model of NAFLD associated HCC on MMPs, we have found a significant alteration of serum proteome in terms of multiple protein complex (MPC) formation as disease progress. Among those proteins, inter-alpha-trypsin inhibitor heavy chain 4 (ITIH4) plays an important role in predicting the course of NAFLD over time, from simple steatosis to NASH and HCC. Most importantly, the current swine study was validated with human studies to show that serum ITIH4 can be a robust and an expecting biomarker of the disease.
Methods
Animals and diets
Three male MMPs aged 3–4 months were purchased from Fuji Micra, Inc. (Shizuoka, Japan). In NAFLD group, two pigs were fed a high-fat diet (HFD) in which 47% of the total calories were from fat, 21.5% calories from fructose, and 16% calories from protein (Research Diets, New Brunswick, NJ, USA). In the HFD, methionine and choline were provided at 3500 and 822 ppm concentrations, respectively. The diet also consisted of 2% cholesterol and 0.7% sodium cholate by weight The third pig was fed a normal diet at 3% of body weight per day as a control animal. The normal diet was composed of less than 10.0% carbohydrate, greater than 13.0% protein, and 2.0% fat by weight (Marubeni Nisshin Feed Inc., Tokyo, Japan). The three pigs were maintained at 24 ± 3 °C under a 12-h light/dark cycle with free access to water.
Experimental protocol
Twelve weeks after starting dietary intervention, two pigs that were fed the HFD were intraperitoneally administered diethylnitrosamine (DEN; 60 mg/kg body weight) every 2 weeks until 36 weeks. On the contrary, a pig that was fed a normal diet and was not administered DEN was used as the control. Body weight, abdominal girth, and blood pressure were measured every 4 weeks. Blood pressures were measured at the foreleg using the Manschette method. Blood analysis and liver biopsies were performed every 12 weeks after 12 h of fasting from the last feeding. Blood samples were collected from the sinus venarum cavarum. At liver biopsies, pigs underwent laparotomy with mid-line incision under general anesthesia. Gross observation of the whole liver was performed in order to check for the presence or absence of tumor development. Liver tissues were incised after the addition of transfixation ligatures with a 3–0 braid absorbable suture on the line just proximal to a scheduled cutting line. A portion of each sample was fixed with 10% formaldehyde and embedded in paraffin wax. Another portion was snap-frozen in liquid nitrogen immediately after sampling. All experiments and measurements were performed under general anesthesia. As premedication for general anesthesia, animals were sedated with intramuscular medetomidine hydrochloride (0.08 mg/kg body weight), midazolam (0.08 mg/kg body weight), and atropine sulfate (0.03 mg/kg body weight). General anesthesia was induced and maintained by inhalation of 1–2% isoflurane. Pulse oximetry and noninvasive blood pressure were monitored during experiments. The physical conditions of the animals were checked every day. If abnormal symptoms such as general fatigue, decreased activity, frequent vomiting, or respiratory distress were observed, pigs were killed before the predetermined day. After 60 weeks, the pigs were killed by cutting the inferior vena cava under deep anesthesia using isoflurane. All protocols were approved by the Ethics Committee of Animal Care and Experimentation at Kyoto University (MedKyo16619).
Serum biochemistry
10 ml of serum was separated by centrifugation at 3500 rpm for 10 min and stored at − 80 °C immediately. Serum triglyceride (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and blood sugar levels were measured using standard methods. Serum liver biochemical parameters were also measured by a local clinical laboratory.
Histological assessment of NAFLD activity and tumor characteristics
Paraffin-embedded tissues were sectioned at a thickness of 4 μm and stained with hematoxylin and eosin or Masson trichrome stain. Pathologists reviewed stained glass slides to provide NAFLD activity scores (NAS). The NAS was created as an unweighted score for steatosis (0–3), lobular inflammation (0–3), and ballooning (0–2). In addition, fibrosis was described as stage 1 to 4 [
14]. Based on these results, the samples at each time point from the three pigs were divided into four groups according to total NAS and presence or absence of HCC; NAS = 0, NAS = 1–3, NAS ≥ 4 without HCC, and NAS ≥ 4 with HCC group.
Immunohistochemistry
After deparaffinization and rehydration, antigen retrieval was applied by microwaving for 10 min with 0.01 M citric acid buffer. The endogenous peroxidase was destroyed by hydrogen peroxide. Primary antibodies were reacted at 4 °C overnight. The secondary antibody (EnVision System-HRP; Dako, Glostrup, Denmark) was incubated for 1 h at room temperature. Subsequently, staining with 3,3′-diaminobenzidine (Dako) for 3 min was performed. For immunohistochemistry of tumor lesions, rabbit polyclonal antibodies for GS (11037–2-AP; Proteintech, Chicago, IL, USA; 1:200 dilution), GPC3 (ab66596; Abcam, Cambridge, UK; 1:200 dilution), HSP70 (10995–1-AP; Proteintech; 1:200 dilution), and ARG1 (AV45673; Sigma-Aldrich, Steinheim, Germany; 1:200 dilution) were used as primary antibodies.
Two-dimensional blue native/SDS gel electrophoresis (2D BN/SDS-PAGE)
The 2D BN/SDS-PAGE permits high-resolution separation of the serum MPC. First, 2 μl of serum was placed on a centrifugal filter column (Amicon Ultracel 3 K; Millipore, Billerica, MA, USA) with 500 μl of blue native buffer. The buffer consisted of 6-aminohexanoic acid, 200 mM Bistris, 500 mM EDTA, 5 M NaCl, and 20% glycerol. After centrifugation for 90 min at 15,000 rpm at 4 °C, the column was inverted in a microcentrifuge tube and centrifuged to collect the sample. After the preparation of serum, samples were mixed with BN loading buffer (2% (w/v) Coomassie 250 G and 750 mM aminocaproic acid: 10% of the sample volume). In the first-dimensional BN-PAGE, 12 μl of sample was applied per lane to a 4 to 15% gradient gel (Bio-Rad, Hercules, CA, USA), and electrophoresis was performed at 50 to 80 V at 4 °C with BN running buffer (25 mM Tris and 192 mM glycine). Next, the first-dimension gel was incubated for 30 min in reducing solution (20% (v/v) glycerol, 25% (v/v) 1 M Tris, and 1% (w/v) dithiothreitol). For further separation in the second dimension with sodium dodecyl sulfate (SDS)-PAGE, the lanes from the first-dimension gel were cut out and placed onto 12% gels (Bio-Rad) at the position of the teeth of a normal gel comb. The SDS-PAGE was performed at 100 V for 100 min at room temperature with SDS running buffer (25 mM Tris, 192 mM glycine, and 0.1% (w/v) SDS). After completion of gel electrophoresis, analytical gels were stained with colloidal Coomassie blue to visualize protein spots.
Matrix-assisted laser desorption ionization-time of flight tandem mass spectrometry (MALDI-TOF MS/MS)
Protein spots excised from 2D gels were dehydrated with acetonitrile and dried by centrifuge for 20 min. The gel pieces were rehydrated in 10 μl trypsin (Roche, Mannheim, Germany) solution (5 μg/ml in 50 mM NH4HCO3 and 5 mM CaCl2) on ice for 45 min and then incubated at 37 °C overnight. Peptides were extracted twice using 5% HCOOH in 50% acetonitrile and dried. Digested peptides were dissolved in 0.1% tri-fluoroacetic acid and desalted using a C18 ZipTip (Millipore). The eluted peptide was applied to a MALDI target plate (MTP AnchorChip; Bruker Daltonik, Bremen, Germany) with 5 mg/ml α-cyano-4-hydroxycinammic acid (Bruker Daltonik) dissolved in 0.1% tri-fluoroacetic acid and 50% acetonitrile and was dried at room temperature. The target plate was mounted onto an AutofleX II device (Bruker Daltonik), and MALDI-TOF MS was performed in the positive ion reflector 1–3 kDa mode. Several precursor peaks with signal-to-noise ratios > 5 from each MS spectrum were selected manually and subjected to tandem mass spectrometry (LIFT-TOF/TOF) analysis. The generated mass spectra were identified by searching the database with the MASCOT server 2.3 (Matrix-Science, Boston, MA, USA). Mascot search conditions were as follows; Database: SwissProt, Taxonomy: Mammalia, Enzyme: trypsin, Global Modification: Carbamidomethyl (C), Variable Modification: Oxidation (M), tolerance: 300 ppm in MS, 0.8 Da in MS/MS.
Western blotting
To validate the MS/MS results, we performed western blotting analyses using sera from both groups from 0, 12, 24, 36, 48, and 60 weeks. 2 μl of each sample applied per lane of a 4 to 15% gradient gel (Bio-Rad), and electrophoresis was performed at 50 to 80 V at room temperature with SDS loading buffer. The gels were transferred onto polyvinylidenedifluoride membranes in transfer buffer (25 mM Tris, 192 mM glycine, 0.01% (w/v) SDS, and 20% (v/v) methanol) at 12 V for 1 h. After blocking with 5% skim milk in phosphate-buffered saline for 1 h, the membranes were probed with a primary antibody overnight at 4 °C followed by horseradish peroxidase-conjugated secondary antibody (Cell Signaling Technology, Danvers, MA, USA) for 1 h at room temperature. Signals were detected using a Chemiluminescence Kit (Chemi-Lumi One; Nacalai Tesque, Kyoto, Japan) and visualized by the digital imager. Rabbit anti-ITIH4 antibody (24069–1-AP; Proteintech, Chicago, IL, USA; 1:1000 dilution), anti-ceruloplasmin (Cp) antibody (ab110449; Abcam, Cambridge, UK; 1:1000 dilution), and anti-haptoglobin (Hpt) antibody (NBT-MFG-102; Cosmo Bio Co., Ltd., Tokyo, Japan; 1:1000 dilution) were used as primary antibodies.
Enzyme-linked immunosorbent assay (ELISA)
ELISA tests were used to quantify the protein concentration of ITIH4 in serum samples at all time points in both groups. The assay was carried out using an ITIH4 ELISA kit based on the sandwich ELISA principle (LifeSpan BioSciences, Seattle, WA, USA). 100 μl of serum sample was applied per well and incubated at 37 °C for 2 h. Biotinylated detection antibody was reacted at 37 °C for 1 h, and avidin-horseradish peroxidase conjugate was then added. Color development was induced by the addition of 3,3′,5,5′-tetramethyl benzidine substrate. The optical density of the well was measured at a wavelength of 450 nm.
Immunohistochemistry of the liver
Expression levels of ITIH4 protein in the liver were analyzed by immunohistochemistry. Immunohistochemistry was performed using a rabbit anti-ITIH4 antibody (Proteintech; 1:200 dilution) in the liver tissues of NAFLD and control pigs at 0, 12, 36, and 60 weeks.
Human cohort study
We obtained serum samples from four clinically classified groups. The HCC with NAFLD group included 55 HCC patients associated with NAFLD who underwent hepatectomy for HCC at Kyoto University Hospital between January 2007 and December 2016. These patients were characterized as negative for hepatitis C virus (HCV) antibody, hepatitis B surface antigen, and chronic alcohol consumption. For the simple steatosis (SS) group, we used serum samples from 40 liver transplantation donors who were diagnosed as SS by histological examination. They did not have a history of chronic alcohol consumption. For the NASH group, serum samples were obtained from 40 NASH patients without HCC who underwent liver biopsy and were diagnosed as NASH by pathologists at Hyogo College of Medicine Hospital. In addition, 21 HCV and 14 hepatitis B virus (HBV)-related HCC patients who underwent hepatectomy between January 2015 and December 2016 were included in the virus-related HCC group. All samples were collected before surgery and stored at − 80 °C until use. Using 2 μl of each serum sample, western blotting was performed with anti-ITIH4 antibodies (sc-515,353; Santa Cruz Biotechnology, Santa Cruz, CA, USA; 1:1000 dilution). The expression level of ITIH4 was quantified using average pixel intensity by densitometry. The diagnostic performance of ITIH4 intensity regarding NAFLD progression and HCC development was evaluated by multivariate receiver operating characteristic (ROC). In order to assess the relationship between serum ITIH4 levels and patients’ prognosis, 55 patients in the HCC with NAFLD group were divided into the low ITIH4 and high ITIH4 groups and their overall survival (OS) was compared. This study was approved by the Ethics Committee of Kyoto University Graduate School and Faculty of Medicine (approval code: R0261–1).
Statistical analyses
Categorical variables were compared by Pearson’s chi-squared test. Continuous variables were compared using Student’s t-tests. OS curves for the low ITIH4 and high ITIH4 groups were estimated using the Kaplan-Meier method, and the results were examined using the log-rank and Wilcoxon tests. With regard to OS, univariate and multivariate hazard ratios (HR) were estimated using the Cox’s hazards regression model. All P-values were two-sided, and differences with a P < 0.05 were considered statistically significant. All statistical analyses were performed using JMP 8.0 software (SAS Institute, Cary, NC, USA).
Discussion
To the best of our knowledge, our study is the first report of a pig model replicating HCC associated with NAFLD. Serum proteomics on swine HCC with NAFLD model implicated ITIH4 as a non-invasive biomarker reflecting NAFLD progression as well as subsequent HCC development. Most importantly, the results in the swine study have been validated in human cohort studies; human ITIH4 in the sera were significantly elevated in patients with HCC with NAFLD.
An HCC pig model associated with NAFLD has not been reported previously. Our pigs were fed an originally formulated HFD containing a higher percentage of fructose that induces hepatic lobular inflammation [
17], 2% cholesterol, and 0.7% sodium cholate, which are traditionally present in an atherogenic diet. As a result, animals exhibited obesity, hypertension, hyperglycemia, dyslipidemia, and atherosclerosis. Based on these systemic phenotypes, their livers showed histological features mimicking those of human NAFLD, including hepatocyte ballooning and lobular inflammation, which are mandatory features for a NASH diagnosis [
18‐
20]. Liver tumors exhibited positive staining of GS, GPC3, HSP70, and ARG1, which are markers for human well-differentiated HCC [
21,
22]. However, microvesicular steatosis was observed in the pigs; this is a less common feature (10%) in human NAFLD than macrovesicular steatosis [
23]. Previous NASH pig models showed resistance to the development of macrovesicular steatosis at 24 weeks of HFD feeding [
24]. While the reason for this remains unclear, the period of lipid accumulation in the liver might have an impact on the histological confirmation of steatosis. In addition, serum TG levels were not elevated in our pigs. This is consistent with the results of previous studies that used HFD-fed pig models [
24‐
26]. In another pig model, more than 18 months of HFD feeding achieved serum triglyceride elevation [
27]. Thus, longer duration studies may reveal hypertriglyceridemia.
In order to detect serum proteomic biomarkers that predict the pathological processes of HCC associated with NAFLD, we adopted a BN/SDS 2D-PAGE technique that is useful for identifying altered MPC profiles in serum as well as individual proteins involved in disease pathogenesis. In this study, ITIH4, Cp, and Hpt, which are characterized as acute phase proteins mainly secreted from the liver, were assumed to coexist in the same MPC. These proteins are mostly characterized as high density lipoprotein (HDL) associated proteome and previous studies utilizing shotgun proteomic analysis identified acute-phase response proteins in HDL as strongly implicating the lipoprotein in inflammation and the innate immune system [
28,
29]. In the current MMP model, ITIH4 was found as one of the most altered proteins in that fraction especially in the late stage of disease progression. Another line of evidence supported this finding that ITIH4 is known as one of newly found BMI-associated loci, which is especially found in Asian not found in European ancestry populations [
30]. This evidence is highly suggestive that ITIH4 found in MMP pig model can be utilized as a useful biomarker for Japanese population.
ITIH4 is a 120-kDa serum glycoprotein secreted primarily by the liver [
31]. It is a member of the inter-alpha-trypsin chain family of proteins, a family involved in stabilization of the extracellular matrix [
32,
33]. In the current pig model, serum ITIH4 levels were significantly associated with high NAS and were further elevated throughout HCC development. The significance of ITIH4 was further confirmed by immunohistochemistry of pig liver specimens, showing that it is synthesized in both cancer lesions and the parenchyma in advanced NAFLD. We also confirmed that serum ITIH4 levels were elevated according to NAFLD progression in another pig fed an HFD and administered a lower dose of DEN (20 mg/kg body weight). Our results implicate upregulated serum ITIH4 as a biomarker reflecting NAFLD progression and subsequent HCC development. In contrast, serum ITIH4 levels were not associated with the severity of fibrosis in the pigs. These results were consistent with the findings of the human serum validation experiment. Thus, serum ITIH4 levels may have clinical utility for assessing the risk of HCC development in non-cirrhotic NAFLD patients.
In the validation experiments using human samples, we evaluated the highest levels of serum ITIH4 in HCC with NAFLD patients. We also verified the diagnostic efficacy of serum ITIH4 for HCC with NAFLD. Although the optimal cut-off level of serum ITIH4 could identify the HCC with NAFLD patients from NASH patients with 86% sensitivity, several NASH patients had high levels of serum ITIH4 in spite of the absence of HCC development. In the prognostic significance of ITIH4, HCC with NAFLD patients who had preoperatively higher serum ITIH4 levels exhibited poorer prognoses after hepatectomy. Previous report showed that serum ITIH4 level was not independently associated with OS in hepatitis virus-related HCC patients [
34]. Combined with our results that the ITIH4 expression in hepatocytes increased in the parenchyma of swine as NAFLD progressed, upregulation of serum 120 kDa ITIH4 may have the potential to predict HCC development from NASH, especially in the pre-cancerous state of NASH. Furthermore, we demonstrated 35 kDa fragment of serum ITIH4 was elevated in NASH patients. 120 kDa ITIH4 is cleaved by plasma kallikrein and cleaved 35 kDa fragment is assumed to remain intact [
35]. Because upregulated 120 kDa ITIH4 caused by NAFLD progression was cleaved by kallikrein in serum, 35 kDa fragment might increase but 120 kDa ITIH4 might not increase in the NASH group.
Pro-inflammatory cytokines such as interleukin (IL)-6 and tumor necrosis factor alpha (TNFα) have been reported to be linked to progression from NASH to HCC [
36]. Macrophage infiltration of the expanded adipose tissue that results from weight gain promotes the production and secretion of IL-6 and TNFα by adipocytes [
37]. Because ITIH4 is positively regulated by IL-6 [
38,
39], it is plausible that the elevation in hepatic ITIH4 synthesis is induced by long-term exposure to IL-6 originating from immune cells as well as adipocytes. Furthermore, we demonstrated that HCC patients with NAFLD exhibited higher expression of serum ITIH4 than did virus-related HCC patients. Although no report has analyzed serum ITIH4 levels in NAFLD-associated HCC patients, decreased serum ITIH4 levels have been previously reported in HCC patients with hepatitis virus-related cirrhosis [
34,
40]. In our analysis, the virus-related HCC patients who had F3 or 4 fibroses showed lower expression of serum ITIH4 than the HCC with NAFLD patients with F3 or 4 fibroses (Additional file
3: Table S2). Therefore, ITIH4 appears to be associated with etiology-dependent and fibrosis-independent carcinogenesis. Higher expression of ITIH4 may be led by HCC development via persistent systemic inflammation due to obesity or metabolic syndrome, but not via complete cirrhosis caused by hepatitis virus infection. Altered expression of ITIH4 at protein level in sera was found in a number of cancers as well as obstructive pulmonary disease or bacterial and viral infections [
41‐
45]. Although serum ITIH4 levels could be influenced by various inflammatory conditions, the fact that ITIH4 was highly synthesized in cancer lesion of NAFLD pigs to generate MPC in serum strongly supports that serum ITIH4 is closely associated with NAFLD associated HCC development. Previous report evaluated that genetically manipulated mouse model that showed activated IL-6 signaling and inactivated TGF-β signaling spontaneously developed liver tumors and steatosis [
46]. In these mice, tumor formation was inhibited by downregulation of the IL-6 pathway by Itih4 deletion. Thus, ITIH4 likely plays a significant role in the pathway that is involved in TGF-β- and fibrosis-independent carcinogenesis. Because obesity and steatosis without cirrhosis has also been recognized to associate with carcinogenesis of NAFLD patients [
3], ITIH4 could be a specific biomarker reflecting the pathogenesis in NAFLD as well as metabolic syndrome-related HCC.
Certain limitations and special circumstances of this study must be acknowledged. First, statistical consideration is weak in this study, because of small sample size in the pig model. By contrast, utilizing pig as a model system is advantageous because we can trace genetically matched and serial serum samples on the same experimental subject. This cannot be realized using rodent models. Second, in the current study, we adopted a blue-native gel-based proteomic approach not by shotgun proteomics. In depth proteomic analysis in combination with shotgun proteomics will be a promising strategy in understanding pathogenesis of NAFLD associated HCC in the future study.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.