Clinical utility of serum fucosylated hemopexin in Japanese patients with hepatocellular carcinoma

Aim:  Hepatocellular carcinoma (HCC) is a common clinical problem all over the world. Fucosylated hemopexin (Fuc‐Hpx) is a newly reported glycoprotein for the diagnosis of HCC, however, its clinical implications are unclear. The aim of this study was to elucidate the clinical utility of Fuc‐Hpx in Japanese patients with HCC.


INTRODUCTION
H EPATOCELLULAR CARCINOMA (HCC) is the fifth most common cancer and its very poor prognosis makes it the third leading cause of cancer death worldwide. 1,2 HCC accounts for over 90% of common primary liver cancer in Japan. More than 80% of HCC cases develop in patients suffering from long-lasting viral hepatitis. Recently, rising rates of diabetes, obesity and non-alcoholic steatohepatitis (NASH) have become increasingly important risk factors of future HCC incidence trends globally, particularly in developed countries. 3,4 Although HCC without hepatitis virus infection, which is difficult to survey, is increasing and the percentage of cases with viral hepatitis is decreasing in Japan, the majority of HCC patients (>80%) still suffer from either hepatitis C or hepatitis B virus (HBV) infection. 5 Many of these patients were under surveillance programs for the diagnosis of HCC, resulting in smaller tumor size at diagnosis.
While modalities of imaging diagnoses have been improving and therapeutic options have progressed, a major problem in HCC surveillance is the lack of reliable biomarkers. 4 a-Fetoprotein (AFP) is the best available biomarker with high sensitivity for HCC surveillance, but the low specificity of AFP led the American Association for the Study of Liver Diseases Practice Guideline Committee to recommend that surveillance has to be based on ultrasound (US) examination. 6 Des-g-carboxy prothrombin (DCP) is used widely as a HCC biomarker in Japan, but it is not popular in other countries including the USA. DCP is more closely related to tumor size with high sensitivity in the diagnosis of large HCC than AFP, but the sensitivity is low for the diagnosis of small HCC. 7 It is known that the fucosylation of glycoprotein often emerges during carcinogenesis. [8][9][10][11][12][13][14][15] The fucosylated AFP (AFP-L3) was highly specific and correlated with biological malignancy and prognosis of HCC patients. [16][17][18][19] Recent glycan analysis demonstrated the increasing fucosylation of serum glycoproteins, not only AFP but also haptoglobin, fetuin A, hemopexin (Hpx), kininogen, a-1 antitrypsin and Golgi protein 73 (GP73) with the development of HCC. 8,15 Hemopexin is a 60-kDa glycoprotein that is one of the acute-phase reactant proteins. Besides its classical functions, such as binding and transportation of free heme in peripheral blood, a wide range of other properties of the hemopexin molecule have been described, such as antioxidant activity. 20 Hpx is produced in the liver and secreted in serum. A report from the USA demonstrated that the fucosylated form of hemopexin (Fuc-Hpx) was a good serum marker for HCC and its capacity for the diagnosis of HCC was superior to that of AFP. 8,9,21 However, the profile of glycosylation is known to be different by age, race or country of residence. 22 In addition, HCC surveillance has become popular, so the size of HCC at diagnosis is smaller in Japan than in other countries. 23,24 Thus, the aim of this study is to evaluate the clinical utilities of Fuc-Hpx in Japanese HCC patients.

Human subjects
H UMAN SERUM SAMPLES were obtained from patients with newly developed HCC (n = 331), chronic hepatitis (CH, n = 85) or liver cirrhosis (LC, n = 45), who were admitted to Okayama University Hospital between 2002 and 2009, as well as from healthy volunteers (n = 22). The serum was collected at the time of admission, meaning that no intervention had been performed. The characteristics of the patients are summarized in Table 1. Healthy subjects did not have a past history of liver disease, cancer, or metabolic or hormonal disorder that required medication. Age is shown as median and interquartile range. The median age of HCC patients was older than that of others (P < 0.001). For etiology, patients with hepatitis B virus surface antigen positivity were classified as having HBV, and those with hepatitis C virus antibody were classified as hepatitis C virus (HCV). Alcohol-induced liver injury, NASH, autoimmune hepatitis or liver disease of unknown origin were classified as others. Over 80% of the patients suffered from viral infection in both the HCC and the non-HCC groups, and HCV infection was more prevalent in HCC patients than in non-HCC patients (73% vs 49%, P < 0.001). The changes of Fuc- Hpx between before and after curative treatments of HCC were examined in 21 cases. Nine cases were treated by local curative treatments (five surgical resection and four radiofrequency ablation). The others were treated by liver transplantation. Hepatocellular carcinoma tissue samples and the corresponding adjacent liver tissue samples were obtained from 12 patients who received liver transplantation. Informed consent was obtained from all patients, and the study protocol conformed to the ethical guidelines of the World Medical Association Declaration of Helsinki and was approved by our institutional review board.

Diagnosis of HCC
In accordance with the AASLD 2005 Practice Guidelines, we confirmed the diagnosis of HCC by at least two dynamic imaging modalities. Typical findings were confirmed as hyperattenuation at the arterial phase and hypoattenuation at the portal phase in dynamic computed tomography (CT) or magnetic resonance imaging (MRI), and tumor staining on angiography. The nodules without these findings were diagnosed by histological examination via US-guided, fine-needle biopsy. Stage was based on the General Rules for the Clinical and Pathological Study of Primary Liver Cancer. The diagnosis of CH and LC was based on liver histology, or clinical and laboratory data including the findings of ultrasound, CT or MRI.

Sample preparation from human liver tissues
Human liver samples were extracted from 50 mg of frozen tissues. Briefly, samples were homogenized with 250 mL reagent mixed CelLytic-MT (Sigma-Aldrich, St Louis, MO, USA) containing protease inhibitor. The lysed samples were centrifuged for 10 min at 4°C, 12 000-20 000 g, to pellet the tissue debris. The supernatant was harvested in a clean tube and used for the following studies. Protein concentration in each sample was measured by the Bradford method.

Measurement of Hpx
Serum Hpx concentrations were measured by enzymelinked immunosorbent assay (ELISA). We used the AssayMax Human Hemopexin ELISA kit (AssayPro, St Charles, MO, USA). The samples were measured in duplicate according to the manufacturer's instructions. A microplate reader (Model 680; Bio-Rad Laboratories, Tokyo, Japan) was used for reading absorbance at 450 nm.

Lectin ELISA for Fuc-Hpx
We performed lectin ELISA for quantitative analysis of Fuc-Hpx in accordance with the method reported by Metha et al. with some modification. 9 Briefly, the rabbit antihuman hemopexin antibody (AssayPro) was incubated with 10 mmol/L sodium periodate to remove the fucosylation of the captured antibody at 4°C for 1 h under dark conditions. An equal volume of ethylene glycol was added and the oxidized antibody was diluted to a concentration of 10 mg/mL with sodium carbonate buffer (pH 9.5). Antibody (1 mg) was added to each well of the ELISA plate and incubated overnight at 4°C. The plate was washed five times with 0.1% Tween-20/ phosphate-buffered saline 7.4 (PBS-T) and then blocked overnight with 3% bovine serum albumin/phosphatebuffered saline (PBS).
For analysis, 50 mL of serum was diluted in 50 mL of PBS with 1 mL of Immunoglobulin Inhibiting Reagent (Bioreclamation, Westbury, NY, USA) and incubated at room temperature for 45 min. The samples were added to the plate and incubated at 37°C for 1 h, followed by washing with lectin incubation buffer (10 mM Tris pH 8.0, 0.15 M NaCl, 0.1% Tween-20) five times. After that, AAL lectin (Vector Laboratories, Burlingame, CA, USA) diluted 250 times by lectin incubation buffer was applied and incubated at room temperature for 1 h. After five washes with PBS-T, AP-streptavidin (Vector Laboratories) diluted 1000 times by PBS was applied and incubated at room temperature for 1 h. After washing five times, color was developed using phosphatase substrate (KPL, Baltimore, MD, USA) and the optical density (OD) at 630 nm was measured. The concentration is expressed as arbitrary unit (AU) based on the relative concentration against a standard HCC sample and normal stock serum. The control curve of lectin-ELISA is shown in Figure 1.

Statistical analysis
The JMP ver. 8.02 software (SAS Institute, Cary, NC, USA) was used for the analyses. Continuous variables are shown as median and interquartile range. The Wilcoxon rank sum test was used to compare the continuous data and the c 2 -test was used to compare categorical data. Statistical significance was set at P < 0.05. Univariate analysis was performed in allpatients except healthy volunteers to identify the potential factors correlated with Fuc-Hpx in liver diseases. Variables at P < 0.05 in the univariate analysis were further analyzed to identify independent factors correlated with Fuc-Hpx by multivariate analysis. The variables used in the analysis

T O DETERMINE THE effect of Hpx concentration on
Fuc-Hpx level, we measured both Hpx and Fuc-Hpx expressions in 18 samples simultaneously (Fig. 2). No correlation was observed between Hpx and Fuc-Hpx (P = 0.89). The level of Hpx was not significantly different between the non-HCC (median, 648 AU/mL; range, 488-750) and HCC groups (median, 772 AU/mL; range, 483-1022; P = 0.16), whereas Fuc-Hpx level was higher in the HCC group (median, 6.8 AU/mL; range, 4.9-11.0) than in the non-HCC group (median, 2.6 AU/mL; range, 0.9-4.8; P < 0.001). Because total Fuc-Hpx level was closely correlated with the percentage of Fuc-Hpx (R 2 = 0.6, P < 0.001) and no difference of AUROC of total and percentage of Fuc-Hpx was observed in this study population (0.84 and 0.77, respectively), we used total Fuc-Hpx level in the following analysis.

Serum Fuc-Hpx level in liver diseases
To confirm the Fuc-Hpx expression in various liver diseases, we measured it in large populations. The median value in the HCC group (n = 331) was 7.6 AU/mL (range, 5.6-10.8), which was significantly higher than that of the non-HCC group (n = 130; median, 4.6 AU/ mL; range, 2.5-7.1; P < 0.001). A progressive increase of Fuc-Hpx was observed from that of healthy controls (median, 3.8 AU/mL; range, 0.1-5.8) through CH (median, 3.7 AU/mL; range, 1.9-6.2) to LC (median, 6.1 AU/mL; range, 4.1-8.9). Significant difference was  observed between the HCC group and LC (P = 0.02) or CH group (P < 0.001), and between LC and CH groups (P = 0.002), but no difference was observed between the CH group and healthy subjects (Fig. 3). We examined Fuc-Hpx level in patients with or without HCC with the same liver function. The median was 7.7 AU/mL (range, 5.4-10.5) in the HCC group, which was significantly higher than that in the non-HCC group (median, 3.9 AU/mL; range, 2.1-6.7; P < 0.001) in Child-Pugh grade A patients. In Child-Pugh grade B/C patients, no difference was observed between the groups, and the median was 7.8 AU/mL (range, 6.2-11.1) and 6.6 AU/mL (range, 5.5-11.2) in the HCC group and in non-HCC group, respectively. We measured to compare Fuc-Hpx levels in 21 HCC cases before and after curative therapy. Fuc-Hpx levels in all nine cases but one who received local curative treatments did not decrease after the treatments. The median Fuc-Hpx levels before and after the treatments were 5.23 and 6.77 AU/mL, respectively. On the other hand, in nine out of 12 HCC cases who received liver transplantation, the median Fuc-Hpx level significantly decreased from 10.2 to 4.87 AU/mL (P = 0.02). Significant difference was observed between local curative treatment and liver transplantation (P = 0.001).

Factors correlated with serum Fuc-Hpx
We evaluated the relationship between serum Fuc-Hpx and clinical parameters in patients with liver diseases ( Table 2). Fuc-Hpx in elderly patients and HCV-infected patients was high. High AST (340 IU/L) and T-Bil (31.0 mg/dL), and low Plt (210 ¥ 10 4 /mL), PT (<100%) and Alb (23.5 g/dL), were also correlated with high serum Fuc-Hpx level. In addition, the presence of HCC was significantly associated with high Fuc-Hpx   (P < 0.001). On multivariate analysis, low Alb, low PT and the presence of HCC were significantly correlated with high Fuc-Hpx (P = 0.013, P = 0.001, and P < 0.001, respectively). The relationship between Fuc-Hpx and tumor factors in combination with three variables that showed correlation with Fuc-Hpx on multivariate analysis was examined in HCC patients ( Table 3). None of the tumor factors such as tumor size, tumor number, Vp or stage was correlated with Fuc-Hpx level. Fuc-Hpx was high in patients with high DCP (340 mAU/mL), while AFP and AFP-L3 were not correlated with Fuc-Hpx. On multivariate analysis, Alb was the only factor correlated with serum Fuc-Hpx level (P = 0.027).

Utility of Fuc-Hpx for the diagnosis of HCC
The accuracy, sensitivity and specificity of Fuc-Hpx for the diagnosis of HCC were 69%, 71% and 63% at a cut-off of 5.95 AU/mL, respectively ( Table 4). The diagnostic accuracies of AFP and DCP in the same serum samples were 56% and 58%, sensitivities were 46% and 47%, and specificities were 87% and 91% at cut-offs of 20 ng/mL and 40.0 mAU/mL, respectively. The receiver-operator curve (ROC) of three individual markers is shown in Figure 4. The AUROC of Fuc-Hpx for the diagnosis of HCC was 0.739, which was inferior to that of AFP (0.791) but superior to DCP (0.723).
The levels of AFP and DCP gradually increased as the stage progressed, but no correlation was observed between Fuc-Hpx and the stage. The sensitivity of Fuc-Hpx was superior to that of the others in both stage I and stage II or more patients. The clinical utility of Fuc-Hpx was equivalent in both stage I and stage II or more patients as well as AFP. AUROC was statistical significantly superior to DCP in stage I.   The sensitivities of Fuc-Hpx + AFP and Fuc-Hpx + DCP were 84% and 74%, respectively, and the specificities were 66% and 71%, respectively. Sensitivity was improved, whereas specificity was not improved by combination with AFP or DCP.

Fuc-Hpx expression in liver tissue
The expression of Fuc-Hpx in HCC tissue was higher than that in adjacent non-cancerous liver tissue in four out of 12 HCC patients, almost equal in one patient and lower in seven patients. Median Fuc-Hpx level in HCC tissue was 6.5 AU/mL and 7.0 AU/mL in adjacent noncancerous tissue. The difference between them was not statistically significant (Fig. 5).

S EVERAL TUMOR MARKERS of HCC have been iden-
tified, but there is no evidence indicating that the detection of HCC by these markers precedes clinical imaging diagnosis. However, the diagnostic accuracy of the radiological tools is dependent on tumor size and this approach is expensive. 25 Moreover, US examination is affected by the skill of individual operators. Therefore, it is necessary to find non-invasive, reliable markers for detecting or predicting HCC.
The expression of Fuc-Hpx increased according to the progression of liver disease from hepatitis, cirrhosis, to HCC, and albumin, PT and the presence of HCC were major factors to determine the expression level. However, we did not observe any correlations between Fuc-Hpx and tumor factors such as tumor size or tumor number. The result is quite different from those of con-ventional tumor markers such as AFP and DCP. From the analysis of the expression in liver tissues, Fuc-Hpx was produced not only in HCC but also in noncancerous tissue, meaning that Fuc-Hpx might be a biomarker for hypercarcinogenic liver rather than a marker for tumor-bearing liver.
Recently, glycoproteomics and glycomics have been focused on as a post-genomic research field to find  Closed circles indicate that Fuc-Hpx was lower in cancerous tissue than in non-cancerous tissue. Closed triangles indicate that Fuc-Hpx expression was higher in cancerous tissue. Closed squares indicate that the expression was at the same level in both tissues. diagnostic markers. 26,27 Glycosylation is involved in both physiological and pathological events, such as cell growth, migration, differentiation and tumor invasion. In particular, fucosylation of N-glycan is well known as one of the changes during carcinogenesis of various cancers. 8,14,15,28 There are several putative mechanisms of elevation of fucosylated proteins in cancers. A tumor marker of HCC, AFP-L3, was produced by core fucosylation of AFP by a-1,6fucosyltransferase (Fut8), which is overexpressed in advanced liver diseases. However, high expression of Fut8 was also observed in non-cancerous liver cirrhotic tissues as well as HCC tissues. 29 a-1,6-Fucosylated proteins are normally rare in the blood and are enriched in the bile by proper balance of two secretion pathways of glycoproteins; one is sorting to an apical surface of hepatocytes followed by secretion into bile ducts and the other is sorting to the basolateral surface followed by secretion into blood vessels. 13 If hepatocytes become depolarized in hepatocarcinogenesis, these normal secretion pathways cannot work and, thus, fucosylated proteins are elevated in the blood. 30 Several reports have been published dealing with the utility of Fuc-Hpx for the diagnosis of HCC. 8,9,21 They reported that Fuc-Hpx is superior to AFP, which has been a standard marker for the detection of HCC.
Comunale et al. reported that the sensitivity and specificity of Fuc-Hpx for the diagnosis of HCC were high (both 92%) and the AUROC for Fuc-Hpx was 0.951. 9 In our study, the diagnostic ability in Japanese patients was inferior to the data described above. In a previous report, they analyzed 72 HCC patients and 280 patients without HCC; however, 248 out of 280 were noncirrhotic patients including 20 healthy controls. 9 The AUROC decreased to 0.8665 when only cirrhotic patients were used as controls. We did not include healthy controls for AUROC analysis so the difference of the liver function in non-HCC patients might be one of the reasons for the difference of AUROC between the studies. In addition, the race was different, the median age was higher and the etiology was different; hepatitis virus infection was a major cause of liver injury in our research, while alcoholism was the main etiology in previous reports. Although it is not clear whether these differences affect the diagnostic utility, it is possible that albumin and PT, which are factors correlated with Fuc-Hpx expression, are different between the studies, which were not precisely indicated in other reports. Despite the differences, Fuc-Hpx expression in HCC patients was high in both studies, indicating that Fuc-Hpx is an effective biomarker for HCC.
Although serum Fuc-Hpx increased in HCC patients, the expression level was not correlated with any tumor factors. Furthermore, Fuc-Hpx levels did not decrease except one case by surgical resection or radiofrequency ablation. On the other hand, in nine out of 12 cases who received liver transplantation, which replaced the hypercarcinogenic liver with normal liver, Fuc-Hpx level decreased by the treatment. The result indicated that the major source of Fuc-Hpx in blood is noncancerous liver tissue although it might be secreted from HCC by the mechanism described above. Scarce correlation with tumor factors is a disadvantage as a conventional tumor marker. Generally, the annual incidence of HCC from LC is known to be 4-8%. On the other hand, the recurrence rate of HCC is reported at an annual rate of 20%, indicating tumor-bearing liver is hypercarcinogenic. We conjectured that the difference of Fuc-Hpx between LC and HCC might correspond to the hypercarcinogenic status mentioned above. We inferred that a high level of Fuc-Hpx might not be shown in HCC but could be shown in hypercarcinogenic liver. In the present study, we could not confirm how effective Fuc-Hpx was as a hypercarcinogenic marker because it was not examined prospectively. If we assumed that the AUROC for the diagnosis of HCC was a surrogate marker of hypercarcinogenicity, the ability of Fuc-Hpx (0.73) was higher than that of Alb, Plt count and Child-Pugh grade (0.53, 0.66 and 0.67, respectively) In this study, we demonstrated that Fuc-Hpx could be an effective biomarker of HCC. Future prospective research is necessary to verify the utility of Fuc-Hpx as a marker for hypercarcinogenic liver.