Hepatocellular carcinoma (HCC), intra- and extrahepatic cholangiocarcinoma (CCA) and gallbladder carcinoma (GBC) are often diagnosed at an advanced stage.1–3 Although some patients can be cured by surgical resection or liver transplantation, palliative procedures are applied in most patients and prognosis remains poor. Immunotherapy using cancer vaccines may represent a novel approach to improve outcome in patients with hepato-biliary cancers.4
Cancer testis (CT) antigens are attractive targets for immunotherapy. So far, more than 100 CT genes have been identified, which belong to more than 44 distinct families.5 Although CT antigens have been implicated in normal development and tumorigenesis, their functions remain largely unknown. Because of their immunogenicity, CT antigens or peptides derived from CT antigens are candidates to be used for cancer vaccination. Effective application however depends on CT antigen expression in tumor cells.6 The expression of numerous CT genes such as members of the MAGE-A, -B and -C families, SSX-1, SSX-2, SSX-5, NY-ESO-1, CTp11, HCA587, CAGE and TSPY, among others, has been demonstrated at the RNA level in HCC, although for some CT genes the frequency of expression varied considerably between studies (reviewed in Ref. 4). Less information is available with respect to CT gene expression in carcinomas of the biliary tract. MAGE-1, NY-ESO-1 and MAGE-A3 mRNA was found in a subset of CCA.7, 8 Although it is more likely that CT antigen expression rather than CT gene expression at the RNA level is a reliable predictor of effective vaccination in a given tumor/patient, only few studies using antibodies against MAGE-C1/CT-7, MAGE-A4 and NY-ESO-1 have analyzed CT antigen expression in hepato-biliary carcinomas at the protein level so far.8–12
Recently, an antibody against MAGE-C2/CT-10, a novel CT antigen, has been generated.13, 14 The MAGE-C2/CT-10 gene shows significant homology with the MAGE-C1/CT-7 gene and both genes map in close proximity to chromosome Xq27.13 MAGE-C2/CT-10 mRNA has initially been identified in a melanoma cell line, but was also detected in a large proportion of HCC.15, 16
In our study, expression of MAGE-C2/CT-10, as well as MAGE-A4, MAGE-C1/CT-7, GAGE and NY-ESO-1 expression, was investigated in 146 HCC, 50 CCA and 32 GBC using immunohistochemistry on tissue microarrays (TMAs) and large sections. CT antigen expression was correlated with clinicopathological parameters, patient outcome, expression of MHC Class I antigen, the numbers of intratumoral CD4+, CD8+ and FOXP3+ regulatory T cells and the presence of CD163+ antigen-presenting cells. We provide evidence that MAGE-C2/CT-10, GAGE and MAGE-C1/CT-7 proteins are expressed in a substantial subset of HCC and are therefore a potential target for vaccination immunotherapy in patients with advanced hepatocellular carcinoma.
Material and methods
A total of 228 carcinomas of the liver and the biliary tract were retrieved from the surgical pathology files of the Department of Pathology, University Hospital Zürich, Switzerland, and the Institute of Pathology, University of Regensburg, Germany, covering a period of 12 years (1993–2005). Tumors were classified according to the WHO Classification of the Digestive System (2000). The series included a total of 146 HCC, 13 intrahepatic CCA, 37 extrahepatic CCA and 32 GBC. All patients were treated with surgical resection or liver transplantation. The following clinicopathological parameters were studied in patients with HCC: patient age and gender, tumor size, grade and stage, underlying liver disease and overall survival. The clinicopathological characteristics of the patients with HCC are provided in Table II. The mean patient age was 63 years (range 36–76 years) and 68 years (range 45–90 years) in the CCA and GBC group, respectively. The CCA group showed a slight male preponderance: 27 (54%) male versus 23 (46%) female patients, whereas the GBC group consisted mostly of women (26/32, 81%). The study was approved by the local ethics committees of Zurich (Kantonale Ethikkommission Zurich, StV 26-2005) and Regensburg.
Formalin-fixed, paraffin embedded tumor tissues were used to construct 4 TMAs with cores of 146 HCC, 13 intrahepatic CCA, 37 extrahepatic CCA, 32 GBC and 20 normal liver tissues. The TMAs have been constructed as described previously.17 Two tissue cores per tumor with a diameter of 0.6 mm were punched out of the donor block and transferred to the recipient block.
Histology and immunohistochemistry
Three-micron-thick sections of TMA blocks and formalin-fixed, paraffin-embedded tissues were mounted on glass slides (SuperFrost Plus; Menzel, Braunschweig, Germany), deparaffinized, rehydrated and stained with hematoxylin–eosin using standard histological techniques. For immunohistochemical staining of both TMA and large sections, the Ventana Benchmark automated staining system (Ventana Medical Systems, Tucson, AZ) and Ventana reagents were used. After deparaffinization in xylene, slides were rehydrated in decreasing concentrations of ethanol. Endogenous peroxidase was blocked using the Ventana endogenous peroxidase blocking kit after a rinse with distilled water. For antigen retrieval, slides were heated with cell conditioning solution (CC1, Ventana) according to the manufacturer's instructions. For the detection of the MAGE-A4 protein, the 57B monoclonal antibody (1:50, kindly provided by Dr. G.C. Spagnoli, University of Basel, Switzerland) was used that recognizes most of the MAGE-A family members, but predominantly the MAGE-A4 protein in paraffin-embedded sections. Primary antibodies against MAGE-C2/CT-10 (clone LX-CT10.5, 1:100),14 MAGE-C1/CT-7 (clone CT7-33, 1:80, Dakocytomation, Barr, Switzerland), GAGE (clone 26, reacts with GAGE-3, -4, -5, -6 and -7B proteins; 1:2,000; BD Transduction Laboratories, San Jose, CA), NY-ESO-1 (clone E978, 1:50, Zymed Laboratories, South San Francisco, CA), CD4 (clone 1F6, 1:30, Novocastra, Newcastle Upon Tyne, UK), CD8 (clone C8/114B, 1:100, DAKO), FOXP3 (1:50, Abcam Ltd, Cambridge, UK), CD163 (clone 163C01/10D6, 1:100, Neomarkers, Fremont, CA), HLAII (clone IQU9, 1:50, Novocastra), Ki-67 (clone MIB-1, 1:20, DAKO) and β-2 microglobulin (clone B2, 1.200; Research Diagnostics, Concord, MA), which is part of the major histocompatibility complex (MHC) Class I molecule, were applied adjusted to the Ventana Benchmark system after performing titrations. iVIEW-DAB was used as chromogen. Immunoreactivity was nuclear for MAGE-C2/CT-10, cytoplasmic for MAGE-C1/CT-7 and both nuclear and cytoplasmic for GAGE and NY-ESO-1. For CT antigens, only strong nuclear and/or cytoplasmic staining, as it was observed in testicular tissue as a positive control, in at least 10% of cells was scored as positive. This relatively conservative cut-off was defined to exclude false-positivity in the TMA analysis. Large sections of normal liver tissue (n = 15) were consistently negative for MAGE-C2/CT-10, MAGE-C1/CT-7, GAGE, NY-ESO-1 and MAGE-A4. Ki-67 labeling index (LI) was defined as percentage of positive nuclei per 100 tumor cells and was determined on TMA tumor tissue cores. CT antigen staining on large sections was performed on one representative paraffin block. Intratumoral CD4+-, CD8+- and FOXP3+-positive T-lymphocytes were counted in 10 HPF (400×).
To test for differences between groups, the Mann–Whitney and Kruskal–Wallis tests were used. Contingency table analysis and 2-sided Fisher's exact tests were used to study a possible association between CT antigen expression based on TMA core positivity and clinicopathologic parameters. Overall survival was analyzed with the Kaplan–Meier method. Differences between survival curves were evaluated by 2-sided log-rank statistics. Overall survival was measured from time of surgery. A multivariate Cox regression model was also performed to analyze the prognostic impact of CT antigen immunoreactivity (MAGE-C1/CT-7 and GAGE). The following covariates were included: age at diagnosis, grade and pT stage. The assumption of proportional hazards was tested using log-neg-log plots. A p value <0.05 was regarded as significant. Analyses were performed with SPSS (SPSS, Chicago, IL) and GraphPad Prism software (GraphPad Software, San Diego, CA). The closed testing procedure was used in case of multiple comparisons.
CT antigen expression in HCC
CT antigen expression was first analyzed on TMA tumor tissue cores of 146 HCC (Fig. 1a, Table I). At least 1 CT antigen was expressed in 40% of HCC. Fifteen percent of HCC was positive for 2 or more CT antigens. Among the 5 CT antigens analyzed, MAGE-C2/CT-10 positivity was observed most frequently (34%). MAGE-C1/CT-7, GAGE and NY-ESO-1 were expressed in 12, 11 and 2% of HCC, respectively. All HCC were negative for MAGE-A4.
Table I. Cancer/Testis Antigen Expression in Hepatocellular, Cholangio- and Gallbladder Carcinomas
Number of tumors/frequency (%)
Hepatocellular carcinoma (n = 146)
Cholangiocarcinoma (n = 50)
Gallbladder carcinoma (n = 32)
At least one CT antigen
To analyze the staining pattern of CT antigens in more detail, large sections of a subset of core-positive HCC (MAGE-C2/CT-10: n = 9, MAGE-C1/CT-7: n = 3, GAGE: n = 7 and NY-ESO-1: n = 1) were also stained with the respective antibodies. All core-positive HCC were also positive for the respective antibodies on large sections, but revealed a heterogenous staining pattern with single positive tumor cells or small groups of tumor cells or uniform immunoreactivity of all tumors cells (Fig. 1b). For MAGE-C2/CT-10, there were 3/9 HCC with positivity in less than 5%, 4/9 HCC with positivity in 6–25%, 1/9 HCC with positivity in 26–50% and 1/9 HCC with positivity in >75% tumor cells.
CT antigen expression in CCA and GBC
CT antigen expression was analyzed on TMAs with tumor tissue cores of 13 intrahepatic CCA, 37 extrahepatic CCA and 32 GBC (Fig. 1, Table I). All 50 intra- and extrahepatic CCA were negative for MAGE-C2/CT-10, MAGE-C1/CT-7, GAGE, NY-ESO-1 and MAGE-A4. MAGE-C2/CT-10 immunoreactivity was found in 4 out of 32 GBC (13%), of which 1 (3%) also stained positive for both NY-ESO-1 and GAGE (Fig. 1). All GBCs were negative for MAGE-C1/CT-7 and MAGE-A4. Large tissue sections of intra- and extrahepatic bile ducts (n = 5) and the gallbladder (n = 5) were negative for all 5 CT antigens analyzed.
Expression of MHC Class I proteins in HCC, CCA and GBC
To be recognized by CD8+ cytotoxic T cells, CT antigens need to be presented together with MHC Class I molecules. Therefore, the expression of the MHC Class I-associated protein β2-microglobulin was analyzed in the tumor cells. β-2-Microglobulin immunoreactivity was observed in 126/146 (86%) of HCC. β-2-Microglobulin expression was detected in 42/50 (84%) MAGE-C2/CT-10-positive, 14/17 (92%) MAGE-C1/CT-7-positive, 15/16 (94%) GAGE-positive and 3/3 (100%) NY-ESO-1-positive HCC, respectively. β2-Microglobulin immunoreactivity was found in 37/50 (74%) and 28/32 (88%) of CCA and GBC, respectively. All CT antigen-positive GBC (4/4, 100%) were also positive for β-2-microglobulin.
Intratumoral immune cells in CT antigen-positive HCC
The numbers of intratumoral CD4+ T cells, CD8+ T cells and FOXP3+ regulatory T (Treg) cells were determined on large tissue sections of MAGE-C2/CT-10-positive (n = 10), MAGE-C1/CT-7-positive (n = 3), GAGE-positive (n = 7) and randomly selected CT antigen-negative HCC (n = 9) (Fig. 2). Comparable numbers of CD4+ T cells were found in MAGE-C2/CT-10-positive, MAGE-C1/CT-7-positive, GAGE-positive and CT antigen-negative HCC. In contrast, the number of CD8+ T cells was significantly increased in MAGE-C1/CT-7-positive HCC in comparison to CT antigen-negative HCC (p = 0.003). MAGE-C2/CT-10 and GAGE positivity was not associated with increased CD8+ T cell numbers. Interestingly, larger numbers of FOXP3+ Treg cells were found in MAGE-C2/CT-10-positive, MAGE-C1/CT-7-positive and GAGE-positive HCC than in CT antigen-negative HCC (p = 0.004). Both CT antigen-positive and -negative HCC contained numerous CD163+ macrophages, which were also positive for MHC Class II proteins (HLA-DP, DQ and DR).
CT antigen expression and clinicopathological parameters in HCC
A potential association between CT antigen expression and clinicopathological parameters was analyzed for MAGE-C2/CT-10, MAGE-C1/CT-7 and GAGE (Table II). NY-ESO-1-positive HCCs were excluded from this analysis due to their small number in our cohort. Protein expression of each CT antigen was analyzed for a possible correlation with the following parameters: patient age and gender, tumor stage, histologic grade, size, proliferative activity, the type of underlying liver disease and patient overall survival. For none of the CT antigens analyzed, a correlation was found between CT antigen expression and patient age, gender, tumor stage, histologic grade, size and the type of liver disease (Table II). Expression of at least 1 CT antigen was associated with a higher proliferation rate (Ki-67 LI) of tumor cells (p < 0.01), which was however not found for individual CT antigens. Univariate survival analysis with 134 patients, for whom follow-up information was available, revealed that MAGE-C1/CT-7 and GAGE expression in tumor cells was associated with reduced overall survival (Fig. 3). In a multivariate Cox regression model (MAGE-C1/CT-7, GAGE, age at diagnosis, grade, pT stage), neither MAGE-C1/CT-7 nor GAGE expression was an independent prognostic factor for overall survival (data not shown).
Table II. Cancer/Testis Antigen Expression and Clinicopathological Characteristics in Hepatocellular Carcinomas
Hemochromatosis, Alagille's Syndrome. # Information on tumor stage and size were available for 140 and 138 hepatocellular carcinomas, respectively.
Here, we demonstrate that at least 1 of the 4 CT genes MAGE-C2/CT-10, MAGE-C1/CT-7, GAGE and NY-ESO-1 is expressed at the protein level in a substantial proportion of human HCC (40%). MAGE-C2/CT-10 protein is expressed with the highest frequency in HCC. This finding may be of clinical significance, since previous studies have shown that MAGE-C2/CT-10 is able to induce specific immune responses in patients. Cytotoxic T lymphocytes directed against MAGE/CT-10 epitopes have been found in melanoma patients.18–20 Furthermore, antibodies directed against MAGE-C2/CT-10 were detected in melanoma and HCC patients.13, 15 These results suggest that MAGE-C2/CT-10 is a candidate for peptide vaccination in patients with HCC.
We detected GAGE and MAGE-C1/CT-7 protein expression in 11 and 12% of HCC, whereas mRNA expression was reported with a frequency of 40 and 48%, respectively.21–23 NY-ESO-1 and MAGE-A4 mRNA was reported in 0–43% and 20% of HCC, whereas we observed protein expression in only 2 and 0% of HCC, respectively.22, 24, 25 Similar discrepancies between protein expression analyzed by immunohistochemistry and mRNA expression studied by RT-PCR have also been reported for other cancer types and CT genes,9, 14, 26, 27 including breast carcinoma, in which NY-ESO-1 expression was reported in up to 42% at the mRNA level using RT-PCR, whereas we have recently observed NY-ESO-1 protein expression in only 2% of carcinomas.28 These discrepancies may be due to higher sensitivity of PCR-based methods, intratumoral heterogeneity or post-transcriptional regulation of CT expression. Future studies are necessary to determine whether CT gene expression at the RNA or protein level is a better predictor for effective vaccination therapy.
For MAGE-C1/CT7 mRNA, expression rates of 48% have been reported.9, 23 We have only detected 12% immunopositive HCC. This is in contrast to Jungbluth et al. who described MAGE-C1/CT7 expression in 60% of HCC, using the same antibody. Our result may slightly underestimate the real prevalence of MAGE-C1/CT-7 protein expression in HCC, because we used the TMA approach and a conservative cut-off for the definition of CT antigen positivity. Our large section analysis and the previous reported data by Jungbluth et al. demonstrate a high intratumoral heterogeneity of MAGE-C1/CT-7 protein expression in HCC. Jungbluth et al. have analyzed 20 HCCs. He observed a focal expression in less than 5% of the tumor cells in 4 of 20 MAGE-C1/CT-7-positive tumors (20%). Another 4 HCCs showed expression in less than 25% of the tumor cells. Given our conservative 10% cut-off to define positive HCC on TMA cores, we postulate that there are about 20–30% of HCC with MAGE-C1/CT-7 expression, justifying potential vaccination approaches.
In our study, only the expression of MAGE-C1/CT-7 and GAGE protein was associated with reduced overall survival in HCC patients. A prognostic significance of MAGE-C1/CT-7 and GAGE protein expression has also been shown for multiple myeloma and lung, pancreatic and ovarian carcinoma.13, 29–32 We have recently shown that the subcellular localization of MAGE-C1/CT-7 in the nucleus or cytoplasm of the tumor cells is of prognostic relevance in multiple myeloma.31 MAGE-C2/CT-10 expression had no prognostic significance in our cohort of HCC, whereas a poor survival was observed in advanced MAGE-C2/CT-10-positive urothelial carcinoma of the urinary bladder.33 Because of the small number of NY-ESO-1-positive HCC, we were not able to analyze a possible association between NY-ESO-1 expression and prognosis. The observed frequency of NY-ESO-1 immunopositivity in our study (2%) was in the same range as in 2 previous studies, which reported NY-ESO-1 protein expression in 3 and 7% of HCC, respectively.11, 12 In these studies, large sections of HCC were used to study the prevalence of NY-ESO-1 expression. This similarity of the results confirms previous studies, showing that the TMA approach is suitable to obtain reliable expression data even for antigens with a high intratumoral heterogeneity.34
Destruction of tumors by an adaptive immune response is largely mediated by tumor cell-specific CD8+ cytotoxic T lymphocytes as effector cells, which recognize target antigens in the context of MHC Class I molecules. We therefore analyzed MHC Class I expression in CT antigen-positive HCC. A high correlation was observed between CT antigen positivity and MHC Class I expression: 84% of MAGE-C2/CT-10-positive, 94% of GAGE-positive, 92% of MAGE-C1/CT-7-positive and 100% of NY-ESO-1-positive HCC coexpressed β-2-microglobulin as a marker for MHC Class I proteins suggesting that CT antigens can be efficiently presented on the surface of tumor cells.
Increased numbers of CD4+ CD25+ regulatory T cells (Treg) have been observed in the tumor, ascites and peripheral blood of patients with HCC35–38 and are associated with a shortened survival.21 We therefore also determined the number of intratumoral Treg cells, which mediate immunotolerance by suppressing self-antigen-reactive T cells, using the transcription factor forkhead box p3 (FOXP3).39 The number of FOXP3+ cells was significantly increased in CT antigen-positive HCC suggesting a positive correlation between CT antigen expression and Treg cells. This observation indicates the possibility that the observed shortened survival of patients with CT antigen-positive HCC may be at least partly due to inhibition of the local immune response by Treg cells. Increased numbers of FOXP3+ cells were not associated with increased numbers of CD4+ cells. This result is most likely due to the fact that FOXP3+ cells represent only a small proportion of CD4+ T cells. Consistent with this interpretation, the absolute numbers of FOXP3+ cells were about 5 times lower than the numbers of CD4+ cells (Fig. 2).
Importantly, MAGE-C2/CT-10, MAGE-C1/CT-7, GAGE and NY-ESO-1 protein were not expressed in CCA and only MAGE-C2/CT-10 immunoreactivity was detectable in 13% of gall bladder carcinoma, including 1 tumor, which was also positive for GAGE and NY-ESO-1. Although Utsunomyia et al. reported MAGE-A4 staining (antibody 57B) in some intrahepatic CCA without providing exact numbers,8 our results suggest that CT antigen expression is rare in CCA.
In summary, our results provide evidence that MAGE-C2/CT-10 is a good candidate for peptide vaccination in patients with HCC. MAGE-C1/CT-7 and GAGE may represent additional immunological targets. Future studies will investigate antibody and cellular immune responses in patients with CT antigen-positive HCC to evaluate their immunogenicity in vivo.
The authors thank M. Storz, S. Behnke and M. Bawohl for excellent technical assistance. We thank G.C. Spagnoli for providing the 57B antibody.