To understand whether the association between Epstein–Barr Virus (EBV) and gastric cancer (GC) has any role in loss of surface expression of human lymphocyte antigen (HLA) class I, the authors analyzed locus-specific transcriptional expression of HLA-A, HLA-B, HLA-C, and HLA-E along with other HLA-associated molecules (β2-microglobulin [β2M], cellular latent membrane protein [LMP], and transporter associated with antigen presentation [TAP]) in EBV-associated, primary GC (EBVaGC) and EBV-negative GC (EBVnGC) tissues.
Approximately 20 EBVaGC tissues and 40 EBVnGC tissues and their corresponding normal tissues were used in the study. The presence of EBV in GC was established by EBV-encoded small RNA in situ hybridization analysis and BamHI W polymerase chain reaction (PCR) analysis. Transcriptional expression of viral LMP2A and several HLA class I genes were analyzed by reverse transcriptase (RT)-PCR. Surface expression levels of HLA class I proteins in cancer samples along with their normal counterparts also were quantified by flow cytometry.
The RT-PCR data suggested selective down-regulation of the HLA-A/HLA-B locus along with over-expression of HLA-E transcripts in EBVaGC (P < .05). This was confirmed further by the flow-cytometric studies using antibodies to HLA-ABC and HLA-E. Among the accessory molecules, LMP7 transcript was down-regulated in a number of EBVaGC tissues compared with EBVnGC.
Several studies have revealed that the γ-herpesvirus Epstein–Barr virus (EBV) is associated with 5-20% of gastric carcinomas (GCs) worldwide.1 The virus has been linked to the development of a number of malignancies, including Burkitt lymphoma (BL), Hodgkin disease (HD), and nasopharyngeal carcinoma (NPC).2 However, it has been observed that the carcinogenesis process of EBV-associated primary GC (EBVaGC) is quite different from that of EBV-negative GC (EBVnGC). The clinicopathologic features of EBVaGC are distinct and include a male preference, frequent accompaniment of atrophic gastritis, predominant involvement of the proximal stomach, and moderately differentiated tubular or poorly differentiated solid type of histopathology.3 In EBVaGC, it has been established that the viral gene expression is restricted to latency I genes quite similar to BL, such as EBV-encoded small RNA 1 (EBNA1), EBV-encoded RNA (EBER), BARF0, BARF1, and latent membrane protein 2A (LMP2A).4 The presence of EBER also has been demonstrated in malignant gastric epithelial cells by in situ hybridization (ISH).5
Many reports have indicated that there are defects in the HLA class I-associated antigen processing and presentation pathway in EBV-associated BL and nasal natural killer (NK) cell/T-cell carcinoma.6, 7 Furthermore, there is evidence of the interference of certain viral antigens in the locus-specific and functional expression of HLA class I antigens in various malignancies.8, 9 Surprisingly, to our knowledge little is known regarding HLA class I antigen expression in EBVaGC, although it has been proposed that the down-regulation or complete loss of HLA class Ia antigen expression is a mechanism to escape host immunosurveillance and, thus, tumor progression.10 The triggering of specific cytotoxic T-lymphocyte (CTL) response (largely CD8-positive) is dependent on the appropriate presentation of viral or tumor-specific antigens in the context of proper HLA class Ia molecules, giving rise to the first step of immune defense.11 HLA class Ia molecules that are expressed on the surface of nearly all nucleated cells are composed of a polymorphic transmembrane heavy chain and a monomorphic light chain called β2 microglobulin (β2m). The heavy-chain polypeptides are encoded by 3 closely linked loci, HLA-A, HLA-B, and HLA-C. Many alleles are assigned to a particular locus.12 Immunohistochemical (IHC) studies in different types of solid tumors have demonstrated defects in HLA class Ia expression.13 Moreover, selective down-regulation of the HLA class I A or B locus also has been observed in GC, colon carcinoma, and laryngeal carcinoma.14, 15 However, most IHC studies of functional HLA expression on tumor samples have been performed with either monomorphic-determinant and/or allele-specific antibodies directed against HLA class Ia, because locus-specific antibodies are not available commercially. IHC studies based on absolute or mere partial staining of HLA molecules that use monomorphic antibodies may not provide an accurate indication of the functional peptide presentation of HLA class I molecules. Recently, with the advent of HLA allelic gene sequences, it has become possible to identify the conserved region of the highly polymorphic HLA genes and to design the primer of each locus separately to investigate HLA status.16 This approach has been used recently to investigate HLA status in solid tumors and hematologic malignancies.17–19
Therefore, we used a semiquantitative and quantitative real-time reverse transcriptase-polymerase chain reaction (RT-PCR)-based approach for transcriptional analysis to study the locus-specific expression of HLA-A, HLA-B, and HLA-C in EBVaGC. Several other molecules, such as transporter associated with antigen presentation (TAP) and LMP, that have been associated with HLA class Ia antigen presentation20 also were analyzed for transcriptional expression. It is noteworthy that malignant cells can escape CTL-mediated immune response by down-regulating HLA class Ia expression; however, then, they may become susceptible to NK cell-mediated lysis.10 In this context, we studied the transcriptional and translational expression of nonclassic HLA-E molecules, which are used as a secondary escape mechanism in several viral infections and in tumor progression.
A number of reports have suggested that the expressions of insulin-like growth factor 1 (IGF1) and IGF2 function as autocrine/paracrine growth factors and are potent stimuli for tumor cells of various origin.21 The biologic responses of IGF1 and IGF2 are transmitted through the IGF1 receptor (IGF1R), which is a tyrosine kinase transmembrane receptor with expression that has been observed in several types of tumors.22 It has been reported that the IGF2 and IGF1R genes are overexpressed in GC.23 Moreover, increased levels of IGF1 in primary tissue from EBVaGC also have been reported using PCR.24 In vitro infections of GC and NPC cell lines with recombinant EBV have demonstrated up-regulation of IGF1 expression by the EBER small RNA molecule.24, 25 Therefore, in the current study, we examined the expression of the IGF1, IGF2, and IGF1R genes using real-time RT-PCR.
MATERIALS AND METHODS
Patients and Tissue Samples
All human samples were handled according to the ethical guidelines followed in Cancer Center and Welfare Home and Research Institute, Thakurpukur (Kolkata, India). Endoscopic tissue samples from patients with different types and grades of GC were collected from the tumor and adjacent normal mucosa and were processed for RNA, DNA, and cell surface analysis as described previously.18
In this study, the human B cell lines Raji and BJAB were used as positive and negative controls, respectively, for the detection of EBV. The Raji cell line also was used as a positive control for HLA-A, HLA-B, HLA-C, HLA-E, TAP1, LMP7, and β2m expression, and a myeloid leukemia cell line (K562) was used as an appropriate negative control.
DNA and RNA Isolation
DNA and RNA from each tissue sample were isolated by using Tripure™ RNA isolation reagent (Roche, Mannheim, Germany). DNA from each sample was dissolved in 100 μL Tris-HCL/ethylenediamine tetraacetic acid buffer, pH 8.0, and the corresponding RNA pellet was dissolved in 20 μL diethyl procarbonate-treated water with mild incubation at 50°C. The RNA preparation was treated with Rnase-free DNase I using DNA-free™ (Ambion, Inc., Austin, TX) to eliminate DNA contamination.
Detection of EBV with PCR and ISH Analyses
To detect the viral genome, PCR amplification was performed against the repeat region of the EBV with BamHI W in 20 μL of reaction buffer. The reaction mixture was amplified in a Mastercycler (Eppendorf AG, Hamburg, Germany) for 30 cycles. The primer sequences and conditions have been described previously.26 The PCR products were analyzed by 2.5% agarose gel electrophoresis using Syber green I dye staining (Roche).
To localize the EBV genome within the malignant cells, ISH analysis of EBER5 was performed using EBV (EBER) PNA probe (DakoCytomation, Glosstrup, Denmark). Tissue sections (4 μm) were deparaffinized and rehydrated in graded alcohol. After proteinase K digestion, sections then were hybridized with an fluorescein isothiocyanate-conjugated, EBER-specific PNA probe at 55°C for 2 hours. Further treatments on tissue sections were done according to the manufacturer protocol using the PNA ISH detection kit (DakoCytomation). The slides were counterstained with a light hematoxylin, and positive signals were recognized as dark blue/black under the light microscope at × 100 or × 400 original magnification.
RT-PCR of Target Genes
Complimentary DNA (cDNA) was prepared for RT-PCR analysis as described previously.18 Briefly, 3 to 5 μg of total RNA were reverse transcribed in a 20-μL reaction volume using Superscript II (Life Technologies, Bethesda, MD). To detect different gene expression, 3 or 4 μL of cDNA were subjected to PCR in a 20-μL reaction volume using 1 U Platinum Taq polymerase (Life Technologies).
Expression analysis of EBV-specific gene LMP2A was performed only in EBVaGC samples using primers and conditions described previously.26 The primers and the conditions of PCR for HLA class I (locus A, B, C, and E), TAP1, LMP7, and β2m were the same as described in previous literatures.18, 27 The β-actin gene was used as an internal control. The amplified products were separated and visualized after staining the agarose gels. With this method, the presence or absence of a gene of interest is determined at the transcript level within the tumor sample compared with its autologous, normal counterpart.
In total, 2 to 4 μg of RNA were reversed transcribed using a high-capacity cDNA archive kit (Applied Biosystems, Foster City, CA) in a 100-μL reaction volume according to the manufacturer's protocol. Real-time RT-PCR analysis was performed using TaqMan 2 × Universal PCR Master Mix (Applied Biosystems) and an ABI Prism 7500 sequence-detection system (Applied Biosystems). Each reaction was performed in triplicate using 100 ng of total RNA. Oligonucleotide primers and probes were obtained from Assay On Demand Service (TaqMan gene expression assay 20X; Applied Biosystems) with the following assay identification numbers: HLA-A, HS00740413_g1; HLA-B, Hs00818803_g1; HLA-E, Hs00428366_m1; IGF 1, Hs00153126_m1; IGF 2, Hs00171254_m1; IGF 1R, Hs00609566_m1 (as targets)' and glyceraldehyde 3-phosphate dehydrogenase, Hs9999905_m1 (as endogenous controls). For each gene, a nontemplate control was performed under the same condition to determine the baseline and threshold values for the analysis. Normal, healthy gastric tissues were used as calibration samples for all analyses. The relative quantification of a target gene in a sample compared with a calibrator was analyzed by using ABI sequence-detection software (version 1.2.2; Applied Biosystems) and applying a 2ΔΔCt method.
Flow Cytometry of Cell-Surface HLA Class I Expression
Single cell suspensions were prepared from the operative tumor tissues and their normal counterparts and were analyzed for HLA class I surface expression according to the previously described protocol.17 Briefly, samples were homogenized into single cells, and approximately 3 × 105 cells were treated with saturating amounts of mouse monoclonal antibody specific for monomorphic epitopes on HLA class I molecules coupled with a fluorescence tag (clone G46-2.6; BD Biosciences, Franklin Lakes, NJ) or their isotype-matched control antibody (mouse immunoglobulin G1,κ; BD Biosciences) according to the manufacturer's protocol. Similarly, using the antibody MEM/E07 (a gift of Dr. V. Horejsi, Czech Republic), the surface expression of HLA-E was analyzed.28 Fluorescence intensities of 104 intact cells (gated on forward and side light-scatter parameters) were measured on a flow cytometer (FACS Caliber; Becton Dickinson, San Jose, CA).
The associations of various transcripts with EBV in GC tissues were determined by using a chi-square test with Yates correction in a 2 × 2 table with 1 degree of freedom (SAS software, version 6.12; SAS Institute, Inc., Cary, NC). P values < .05 at the 95% confidence interval (95% CI) were considered statistically significant.
Detection of EBV in GC
To detect EBV-associated GC, we performed EBER ISH analysis on the paraffin embedded tissue sections (Fig. 1A–C) along with PCR-based amplification of the BamHI W repeat region (Fig. 1D). In EBVaGC samples, the viral genome was detected within the malignant cells but was not detected significantly in normal mucosa against EBER (Fig. 1B, C). Normal gastric mucosa and lymphoid stroma were negative for EBER staining. For the RT-PCR-based expression study of different genes, we used approximately 60 cancerous tissue samples and their normal counterparts. There were 20 EBVaGC and 40 EBVnGC tissue samples. Expression of LMP2A was detected in limited numbers of EBVaGC samples (Fig. 1D), indicating the episomal existence of EBV genome in these GCs.5
Expression of HLA Class Ia Genes
Locus-specific HLA-A, HLA-B, and HLA-C expression at the transcriptional level was detected in both EBVaGC and EBVnGC tissue samples (Fig. 2A). For cells that expressed equivalent amounts of β-actin, a significant difference in the level of HLA class Ia loci expression was observed. Samples from normal counterparts expressed all of the class I HLA-A, HLA-B, and HLA-C genes. In the majority of the GC samples, HLA-C transcript was expressed, whereas HLA-A and HLA-B transcripts could be detected in only 56% of samples (Table 1). Flow cytometric analysis of HLA class Ia cell surface expression indicated that the mean fluorescence intensities of EBVaGC were less (P < .05) compared with the intensities of EBVnGC samples and normal tissues (Fig. 2C). This is in agreement with the locus-specific, transcriptional down-regulation of HLA class Ia genes (Fig. 2A). Moreover, we observed a good correlation between the locus-specific down-regulation of HLA-A and HLA-B with EBV infection in GC samples (P = .008 at the 95% CI). In the EBVaGC samples, 70% were negative for both HLA-A and HLA-B, whereas only 30% of EBVnGC samples were negative for both HLA-A and HLA-B. To confirm this locus-specific down-regulation of transcript in EBVaGC, we also studied relative quantification of the HLA-A and HLA-B genes by using real time RT-PCR analysis (Fig. 2B).
Table 1. RT-PCR Analysis of HLA Class Ia, HLA-E, Transporter Associated with Antigen Presentation 1, and Low Molecular Weight Protein 7 in Gastric Cancer
The results were scored as positive (+) when the reverse transcriptase-polymerase chain reaction product was present and negative(−) when the product was absent compared with autologous normal gastric tissue.
EBVaGC (20 patients)
EBVnGC (40 patients)
Expression of HLA-E Gene
Several investigators have demonstrated that the expression of a few nonclassical HLA molecules help in immune evasion in malignant cells. Therefore, we examined the presence of the HLA-E transcript in GC samples (Fig. 2A). Expression was observed in 48% of GC samples and in all normal samples in the study. It is noteworthy that the HLA-E transcript was detected in 75% of EBVaGC samples compared with 35% of EBVnGC samples (P = .008 at the 95% CI) (Table 1). A real-time, relative quantification study of HLA-E indicated that expression was 3.5 times greater in EBVaGC samples compared with EBVnGC samples (Fig. 2B). Flow cytometric studies of HLA-E surface expression confirmed the up-regulation of HLA-E in EBVaGC (Fig. 2C). Table 2 depicts the correlation between the expression of classic HLA-A and HLA-B with the nonclassic HLA-E in GC. Expression of HLA-E was observed in 66% of samples in which HLA-A and HLA-B expression was absent, whereas the simultaneous expression of HLA-E and HLA-A/B was observed in 36% of GC samples.
Table 2. Correlation between HLA-E and HLA-/B Expression in Gastric Cancer Samples
The results were scored as positive (+) when the reverse transcriptase-polymerase chain reaction product was present and negative (−) when the product was absent compared with autologous normal gastric tissue.
Numbers in parenthesis indicate the number of patients with Epstein–Barr virus-associated primary gastric cancer.
There are several associated genes that are expressed simultaneously along with HLA class I for antigen presentation and proper immune control through CTL responses. Therefore, we studied transcriptional expression of the TAP1, LMP7, and β2m genes within the samples (Table 1; Fig. 3). Irrespective of EBV association, the β2m gene was expressed in all tissues at the transcriptional level. However, our data indicated that there was no correlation between EBV association and TAP1 down-regulation, because TAP1 transcript was absent in 45% of EBVaGC samples compared with 40% of EBVnGC samples. However, the LMP7 gene, which is associated with HLA proteasome complex, was absent in 65% of EBVaGC samples compared with 25% of EBVnGC samples, indicating a significant correlation of EBV in the regulation of LMP7 (P = .007 at the 95% CI).
Expression of IGF Family Genes
Real-time RT-PCR-based, relative quantification was used to analyze expression of the IGF1, IGF2, and IGF1R genes in normal and primary GC tissues. It was observed that, compared with normal samples and EBVnGC samples, relative expression of the IGF1 gene was much greater in EBVaGC samples (Fig. 4A). Similarly, compared with normal tissues, relative expression of the IGF2 and IGF1R genes was much greater in GC samples, although there was no significant difference in the relative quantities of transcript between the EBVaGC samples and the EBVnGC samples (Fig. 4B, C).
The results of the current study demonstrated that, in EBVaGC tissues, there is differential, locus-specific, transcriptional expression of HLA class I genes. In contrast, previous reports based on IHC using monomorphic antibody against total HLA class Ia molecules showed that both EBV-positive and EBV-negative GC expressed HLA class I molecules to the same extent.29 In the current study, we observed a high degree of association between EBV and locus-specific down-regulation of HLA class Ia in GC tissue samples (P < .05 at 95% CI). Moreover, we observed greater HLA-E expression levels in HLA-A and HLA-B down-regulated EBVaGC samples (P = .008 at the 95% CI). It is noteworthy that recent studies of HLA class I expression using cytometric analysis in GC cell lines indicated low levels of HLA surface molecules in EBV-positive cell lines, such as SNU-719, corroborating in vitro (cell line) and in vivo (primary tissue) observations.27, 30
Previously, it was shown that EBV-associated BL cells expressed low levels of HLA-associated genes, such as TAP1 and TAP2, which have a crucial role in HLA class I maturation.31 Defects of the proteasome complex LMP and TAP also have been reported in a number of GC cell lines.27 In the current study, the absence of TAP1 transcript was observed in both EBVaGC samples and EBVnGC samples; whereas the absence of LMP7 transcript was more prevalent in EBVaGC samples (65%) compared with EBVnGC samples (25%; P = .007 at the 95% CI). Transcription of β2m was detected in all GC samples, irrespective of their EBV status, a finding that is in agreement with a number of other reports in various cancers.32, 33
EBV-harboring NK cell/T-cell lymphoma, BL, HD, and NPC may suppress local immune response to the infiltrating T cell by up-regulating cytokines and cellular growth factors, such as IGF1.7, 24, 25, 34–36 It is noteworthy that IGF1 has been implicated in the modulation of HLA class Ia expression and in the inhibition of apoptosis in glioma cells.37 Our RT-PCR data for IGF1 indicated greater expression of the transcript in EBVaGC samples (P < .005), which is in agreement with a previous report.24 Although IGF2 and IGF1R expression were relatively greater in GC samples, there was no marked difference between EBVaGC and EBVnGC samples. In contrast, EBV infection of different BL cell lines revealed that viral infection led to a decrease in IGF1R expression.38
Viruses such as human papillomavirus (HPV), human cytomegalovirus (CMV), and human immunodeficiency virus (HIV) use various mechanisms to modify the maturation, assembly, and export of HLA class I molecules.8, 9 Certain viral proteins, including E5 of HPV type 16, Nef of HIV, and US3/UL40 proteins of CMV, have evolved to selectively down-regulate HLA-A and HLA-B, which are the main presenters of peptides to CTLs, but not HLA-C or other nonclassic HLA molecules.39–41 EBV also adopts several kind of latency patterns by expressing only a few nonimmunogenic viral proteins in different pathogenic conditions and in healthy conditions to evade the CTL response.42, 43 In view of the contradictory reports regarding the presence of HPV in GC,5, 44, 45 it would be interesting to identify homologous EBV-encoded proteins that mediate similar locus-specific down-regulation of HLA class Ia genes. Incidentally, it has been reported that there was no significant difference in the prognosis between patients with EBV-positive GC and EBV-negative GC, although EBVaGCs were infiltrated significantly with CD8 T cells, NK cells, and dendritic cells.3 Therefore, it is tempting to speculate that the locus-specific down-regulation of HLA-A and HLA-B and the expression of HLA-E and HLA-C observed in the current study may explain this apparent discrepancy between tumor-infiltrating lymphocytes and prognosis in patients with GC. One possible explanation for the viral-induced, locus-specific down-regulation of HLA class I genes is the interaction of viral LMP2 and cellular nuclear factor κB (NF-κB). In the current study, viral LMP2A expression was observed in some EBVaGC samples, although the expression level was considerably low compared with LMP2A expression in the positive control Raji cells. The κB motif of enhancer A element of the HLA class I gene is the binding site of NF-κB/Rel family transcription factors and is highly conserved and present only in HLA-A and HLA-B gene promoters.46 However, because of the lack of NF-κB binding sites on other HLA class I gene promoters, such as HLA-C, HLA-E, and HLA-F, they are not regulated by NF-κB.47 Recently, it was demonstrated that EBV-encoded LMP2A expressed in NPC and GC cell lines down-regulated cellular NF-κB.48 Analogous to other malignancies, aberrant methylation of HLA class Ia gene promoters may lead to the loss of expression.49 Of course, it would be interesting to investigate whether EBV modulates the methylation of HLA genes, as reported for other cellular genes.50
GC is an aggressive tumor of multifactorial etiology that, although it is rare in Western parts of the world, poses a significant mortality problem in South Asia, Korea, and Japan. Improved therapies are an urgent requirement; and, toward that end, immunotherapeutic methods are being developed in several centers. Of course, such strategies will depend on the immune competence of the target tumor and, in particular, its locus-specific HLA class I expression. Because of the observation that the majority of CTLs recognize peptides presented by HLA-A and HLA-B, whereas cells are protected from NK cell cytotoxicity by HLA-E and HLA-C expression,51–53 it is necessary to have prior knowledge about locus-specific gene/protein expression to produce a rational immunotherapeutic design.
The authors thank Dr. V. Horejsi (The Institute of Molecular Genetics, Academy of Sciences of the Czech Republic, Czech Republic) for the gift of the antibody MEM/E07 to human lymphocyte antigen-E. They also thank Dr. Supti Mukherjee (Cancer Center Welfare Home and Research Institute, Thakurpukur, Kolkata) and Dr. Debasish Banerjee (Rama Krishna Mission Seva Pratisthan [RKMSP], Kolkata) for providing their help and expertise in histopathologic analyses and in situ hybridization studies.