Expression spectrum and methylation-dependent regulation of melanoma antigen-encoding gene family members in pancreatic cancer cells

Original Paper

Pancreatology 2002;2:146–154

Expression Spectrum andMethylation-Dependent Regulation ofMelanoma Antigen-Encoding Gene FamilyMembers in Pancreatic Cancer Cells

Tillmann Berta,1 Nikolaus Lubomierskia,1 Susanne Gangsaugeb

Karin Müncha Hartmut Printza Nicole Prasnikara Christian Robbela

Babette Simona

aDepartment of Internal Medicine, Philipps-University, Marburg, and bDepartment of Surgery, University Hospital,Ulm, Germany

Received: July 30, 2001Accepted: December 27, 2001

Babette Simon, MDDepartment of Internal MedicineDivision of Gastroenterology, Philipps UniversityD–35033 Marburg (Germany)Tel. +49 6421 286 2721, Fax +49 6421 286 2799, E-Mail [email protected]

ABCFax + 41 61 306 12 34E-Mail [email protected]

© 2002 S. Karger AG, Basel and IAP1424–3903/02/0022–0146$18.50/0

Accessible online at:www.karger.com/journals/pan

Key WordsTumor antigens W MAGE W GAGE W Pancreatic cancer W

Methylation

AbstractHuman MAGE and GAGE genes encode tumor-specificantigens presented by HLA I molecules recognized ontumor cells by cytolytic T lymphocytes. To determine ifpancreatic cancer patients would be suitable for MAGE-or GAGE-based immunotherapy, the expression fre-quency of MAGE-A1, -A2, -A3, -A4, -A6 and GAGE1–8genes was assessed in 15 pancreatic tumor cell lines and23 pancreatic tumor specimens using reverse transcrip-tion-polymerase chain reaction (RT-PCR). In 67% of thecell lines at least one of the MAGE-A genes was detected,53% revealed concomitant expression of two or moregenes. GAGE1–8 expression was detected in 47% of thecell lines. In the primary pancreatic tumors, MAGE-Aanalysis revealed exclusive MAGE-A1 and MAGE-A2gene expression in 26 and 30% of the specimens, respec-

tively, independent from clinicopathologic factors. Treat-ment of MAGE-A expression-negative pancreatic tumorcells with the demethylating agent 5-aza-2)-deoxycytid-ine could activate MAGE-A1, MAGE-A2, MAGE-A3,MAGE-A4 and GAGE transcription suggesting silencingdue to promoter methylation. Interestingly, a metastaticlesion to the liver revealed concomittant expression ofMAGE-A1, -A2, -A3 and -A6 consistent with a more pro-nounced genome-wide hypomethylation in metastases.Therefore, a subset of pancreatic cancer patients couldbe eligible for active, specific immunotherapy directedagainst MAGE-A antigens and demethylating agentscould increase the number of candidate patients.

Copyright © 2002 S. Karger AG, Basel and IAP

Introduction

Treatment of pancreatic cancer has been unsatisfacto-ry up to now and an attractive alternative would be thedevelopment of vaccines to boost the host immune re-sponse. The MAGE and GAGE genes encode tumor anti-gens that are recognized on melanoma cells by autologouscytolytic T lymphocytes (CTL) in a major histocompati-bility complex (MHC) class-I restricted fashion [1–4]. The1 Equal contributors.

MAGE and GAGE Gene Expression inPancreatic Cancer

Pancreatology 2002;2:146–154 147

MAGE family contains 18 genes divided into three geneclusters on chromosome X. The MAGE-A genes arelocated in cluster A in the Xq28 region, additional mem-bers comprising four MAGE-B genes in Xp21, and a thirdcluster called C in Xq26–27 [5–8]. The unrelated GAGEfamily with a MAGE-type expression profile comprises aseries of 8 closely related genes clustered in the Xp11.2-p11.4 region [3]. The function of the MAGE and GAGEproteins is unknown at present.

A unique characteristic of these antigens is their exclu-sive expression in malignant tumor tissues with the excep-tion of testis and placenta [1, 9]. Besides melanomas,MAGE genes A1, A2, A3, A4 and A6 are significantlyexpressed in a number of tumors of different histologicaltypes, such as lung, breast, brain, colorectal and gastriccancers, as well as on Hodgkin Reed Sternberg cells [10–16]. MAGE-A1 and MAGE-A3 code for tumor antigensthat are recognized by CTLs from HLA-A1+ andCw*1601 or -A1+ and -A2+ melanoma patients, respec-tively [17, 18]. MAGE-A4 and MAGE-A6 also encodepotential tumor antigens on HLA-A1 [9, 19]. Recently, anew HLA-B*3701-restricted epitope, encoded by homolo-gous regions of the MAGE-A1, -A2, -A3, and -A6 geneswas reported [2]. The specific autoimmunity by cytotoxicT lymphocytes in the context of HLA class I restrictioncan be elicited by these antigens in vitro [18] and in vivo[4, 20–24] and the selective expression in a broad spec-trum of neoplasms makes them ideal candidates for spe-cific cancer immunotherapy. The absence of expression ofencoding genes in normal tissues ensures strict tumoralspecificity of immune responses and no adverse effect isexpected in testis since germline cells do not carry majorhistocompatibility complex I molecules [25]. SinceMAGE and GAGE expression had not yet been addressedin pancreatic cancer cells, we analyzed their expressionprofile to evaluate whether MAGE-type tumor-specificantigens might be useful for specific antitumor vaccinesin the treatment of pancreatic cancer patients.

Material and Methods

Cell Lines and Tissue SamplesHuman ductal pancreatic carcinoma cell lines AsPC-1, BxPC-3,

Capan-2, HPAF, MIAPaCa-2, Panc-1, PC-2, PC-3, PC-44, PaTuII,8902, 8988S, 8988T and two neuroendocrine pancreatic tumor celllines BON and QGP-1 were cultured as described [26, 27]. The mela-noma MZ2-MEL 3.0 and sarcoma LB23 SARC cell lines served ascontrols for differential MAGE-A gene expression [8], the gastric car-cinoma cell line KATOIII as positive control for GAGE gene expres-sion [14]. The tumor tissue samples (18 pancreatic ductal adenocarci-nomas, 1 liver metastasis from a pancreatic ductal adenocarcinoma,

1 mucinous cystadenocarcinoma, 2 pancreatic neuroendocrine tu-mors, 1 pancreatic acinus cell carcinoma) were obtained after surgi-cal resection and immediately snap frozen in liquid nitrogen prior toRNA extraction. Frozen section analysis confirmed that the tumorswere essentially free of adjacent normal tissue.

5-Aza-CdR Treatment of CellsThe cell lines were grown in flasks wrapped in aluminium foil to

protect from light. Medium containing 1 ÌM 5-aza-2)-deoxycytidine(5-Aza-CdR) (Sigma Chemical Corp., St. Louis, Mo., USA) was add-ed for 96 h and cells were harvested by trypsinization and subjectedto RT-PCR for MAGE-A and GAGE gene expression analysis.

RT-PCR and Sequence AnalysisTotal RNA was purified from cell lines as described [26]. Extrac-

tion of the primary pancreatic tumor specimens was performed withthe QuickPrep Micro mRNA Purification Kit (Pharmacia). Firststrand cDNAs were synthesized from 2.5 Ìg of total RNA with 100 Uof Moloney leukemia virus reverse transcriptase in a total volume of20 Ìl (37°C, 60 min), and 1 Ìl subjected to PCR amplification usingpairs of oligonucleotide primers highly specific for each MAGE geneas follows: MAGE-A1: sense 5)-CGGCCGAAGGAACCTGACC-CAG-3) (CHO-14), antisense 5)-GCTGGAACCCTCACTGGGTT-GCC-3) (CHO-12) [8, 10]; MAGE-A2: sense 5)- AAGTAGGACC-CGAGGCACTG-3) (CDS-9), antisense 5)-AAGAGGAAGAAGCG-GTCTG-3) (CDS-7) [28]; MAGE-A3: sense 5)-TGGAGGACCAGA-GGCCCCC-3) (AB-1197), antisense 5)-GGACGATTATCAGGAG-GCCTGC-3) (BLE-5) [8, 9]; MAGE-A4: sense 5)-ACCAAGGAGA-AGATCTGCCAG-3) (MFSf-1), antisense 5)-GTCGCCCTCCAT-TGCAATTGT-3) (M41Sr) [29]; MAGE-A6: sense 5)-TGGAGGAC-CAGAGGCCCCC-3) (AB-1197), antisense 5)-CAGGATGATTAT-CAGGAAGCCTGT-3) (EDP-28) [8]. GAGE-1, -2, -3, -4, -5, -6, -7, -8expression was determined by PCR amplification reaction with theoligonucleotide primer pair: sense 5)-GACCAAGACGCTACGT-AG-3) (VDE43), antisense 5)-CCATCAGGACCATCTTCA-3)(VDE24); specific GAGE-1, -2, -7 gene expression was evaluatedusing oligonucleotide primer pair: sense 5)-GCGGCCCGAGCAGT-TCA-3) (VDE44) and VDE24 [15]. Amplification was performed asdescribed [8]. The PCR products were size fractionated on 2% ethid-ium-bromide-stained agarose gels. PCR amplification of tumor tis-sue derived cDNAs and 5-Aza-CdR experiments, as well as GAGEspecific PCR amplification were performed in the presence of [·-33P]dCTP and analyzed by 6% nondenaturing PAGE. MAGE andGAGE primers were located in different exons to prevent false-posi-tive PCR products due to contaminating genomic DNA. PCRamplifications were done in duplicate. The identity of the PCR prod-ucts was confirmed by DNA sequence analysis. For verification ofRNA integrity, control RT-PCR amplifications using primers for ß-actin was performed as described [30].

Results

MAGE-A and GAGE Gene Expression in PancreaticTumor Cell LinesA panel of 13 ductal and 2 neuroendocrine pancreatic

tumor cell lines was investigated by RT-PCR for expres-sion of the MAGE-A and GAGE gene family members at

148 Pancreatology 2002;2:146–154 Bert/Lubomierski/Gangsauge/Münch/Printz/Prasnikar/Robbel/Simon

Fig. 1. Expression analysis of MAGE-A and GAGE genes in represen-tative pancreatic tumor cell lines. Reverse transcription and PCRamplification of MAGE-A1, MAGE-A2, MAGE-A3 MAGE-A4,MAGE-A6 (A) and the GAGE genes (B) revealed specific PCR prod-ucts as indicated on the left side. The size of the specific PCR productis indicated on the right side in basepairs (bp). MZ2-MEL 3.0(MZ2.M), melanoma cell line (positive control for MAGE-A1, -A2,-A3, and -A6, negative for MAGE-A4). LB23SARC sarcoma cell line(positive control for MAGE-A4, negative for MAGE-A1, -A2, -A3 and-A6 ). The gastric cancer cell line KATOIII served as positive controlfor GAGE1–8 expression. GAGE gene amplification using 33P-labeled dCTP was performed with 2 pairs of specific primers.VDE43 and VDE24 amplified all GAGE genes if present, VDE44and VDE24 exclusively amplified GAGE-1, -2 and -7. MIAP = Pan-creatic carcinoma cell line MIAPaCa-2. M = Migration of molecularweight marker pUC Mix Marker 8 (MBI, Fermentas). Co = PCRreaction without cDNA. C ß-Actin amplification served as controlfor cDNA quality.

the mRNA level. Considering the diversity of the MAGE-A genes, we used specific PCR primers that properly dis-tinguished the various genes. The melanoma-derived cellline MZ2-MEL 3.0 served as positive control for MAGE-A1, -A2, -A3 expression and as negative control forMAGE-A4 expression (fig. 1A, lane 4), while the sarcomacell line LB23 SARC provided a positive control forMAGE-A4 expression and a negative control for the otherMAGE-A genes (fig. 1A, lane 5). RT-PCR revealeda specific MAGE-A1 PCR product of 421 bp in 8 of15 (53%) pancreatic tumor cell lines (table 1; fig. 1A).MAGE-A1 expression was relatively low except for the 3cell lines BxPC3, 8902 and Panc1. The MAGE-A2 PCRproduct was detected as a 230 bp fragment in 6 of 15(40%) cell lines (table 1; fig. 1A), revealing in most of thecases concomitant expression with MAGE-A1 except forcell line Capan-2. Since the primer pairs were located indifferent exons to identify any occasional false-positivePCR products due to amplification of genomic DNA, thelarger 300 bp cDNA fragment corresponded to MAGE-A2genomic sequence. MAGE-A3 amplification revealed aspecific RT-PCR fragment of 725 bp in 8 of 15 (53%) pan-creatic carcinoma cell lines (table 1, fig. 1A), althoughfaint in about half of them. The 805-bp PCR product afteramplification with the MAGE-A3 specific primer pairrevealed a PCR product due to amplification of MAGE-A3 genomic sequence. All cell lines revealed no MAGE-A4RT-PCR product except 2, Panc-1 and QGP-1, whereMAGE-A4, however, was only scarcely detectable inPanc-1 cells as a 726-bp fragment (table 1; fig. 1A).MAGE-A6 amplification revealed a specific 728-bp frag-ment in 4 of 15 pancreatic carcinoma cell lines (27%) (ta-ble 1; fig. 1A). The ductal pancreatic carcinoma cell linePanc-1 was the only one that revealed concomitant ex-pression of all MAGE-A genes analyzed. We further evalu-ated whether the MAGE-related GAGE genes were alsoexpressed in pancreatic cancer cells. Amplification of acommon nucleotide sequence of the 8 known GAGEgenes revealed a specific 201 bp PCR fragment in 7 of 15pancreatic carcinoma cell lines (47%), although weakly insome of them, indicating expression of at least one of theGAGE genes (table 1; fig. 1B). The gastric cancer cell lineKATOIII served as positive control (fig. 1B, lane 2). Am-plification of a nucleotide sequence specific to the GAGE-1, -2 and -7 genes detected a specific 244-bp PCR productin 7 of 15 cell lines (47%) (table 1; fig. 1B), although veryweak in 4 of them. Amplification of ß-actin confirmedRNA integrity (fig. 1C).

PaCa



Table 1. MAGE-A and GAGE geneexpression in pancreatic tumor cell lines

MAGE-A1 MAGE-A2 MAGE-A3 MAGE-A4 MAGE-A6 GAGE 1–8 GAGE 1/2/7

8988T – – + – – + +8902 + + + – + – (+)8988S – – – – – – –PC-3 – – – – – – –PC-44 + – (+) – – + +PC-2 – – – – – + –MIAPaCa-2 – – – – – – –Panc-1 + + + (+) + + +AsPC-1 – – – – – – –PaTuII + – (+) – – + (+)HPAF + – (+) – – + (+)Capan-2 – + – – – – –BxPC-3 + + + – + + (+)

PNETBON (+) + – – – – –QGP-1 + (+) + (+) (+) – –

+ = Expression present; (+) = very weak/scarce expression, – = no expression; PaCa =ductal pancreatic carcinoma cell lines; PNET = pancreatic neuroendocrine tumor cell lines.

Differential MAGE-A Gene Expression in SurgicalPancreatic Tumor SpecimensIn this study, a total of 22 primary pancreatic tumors

and one metastatic lesion to the liver were analyzed forMAGE-A1, -A2, -A3 and -A6 expression using RT-PCRamplification with oligonucleotide primers specific foreach MAGE-A gene (table 2; fig. 2). In contrast to the celllines, the primary pancreatic carcinomas of ductal originexclusively revealed expression of MAGE-A1 and/orMAGE-A2 as demonstrated in 6 of 18 (33%) tumors,although their expression level was rather low. MAGE-A1expression was detected in 17% (3/18), and MAGE-A2 in28% (5/18) of the tumor specimens analyzed. Interestingly,the metastasis to the liver of a ductal pancreatic carcinomarevealed an exception showing concomitant strong expres-sion of MAGE-A1 and MAGE-A2, and to a lower extentalso of MAGE-A3 and MAGE-A6. Unfortunately, only oneliver metastasis specimen was available for analysis sincepatients with metastatic disease are usually not operated.The pancreatic acinus cell carcinoma (PC517) also re-vealed exclusive amplification of MAGE-A1 and MAGE-A2. One of the two neuroendocrine pancreatic tumors(PC536, PC495) revealed MAGE-A1 expression, while themucinous cystadenocarcinoma (PC467) was MAGE-Anegative. In the primary ductal pancreatic carcinomasMAGE-A gene expression appeared independent from theclinicopathological factors of the patients (table 2).

Demethylation-Dependent Upregulation of MAGE-Aand GAGE Genes in Pancreatic Tumor CellsTo analyze the mechanisms of MAGE-A and GAGE

silencing and activation in pancreatic tumor cells, tworepresentative ductal (PC-44, MIAPaCa-2) and neuroen-docrine pancreatic tumor cell lines (BON, QGP-1) withlow or none MAGE-A1 and MAGE-A2 gene expressionwere analyzed for MAGE-A and GAGE transcriptionalregulation using the DNA hypomethylating agent 5-aza-2)-deoxycytidine (5-Aza-CdR). Treatment with the deme-thylating agent 5-Aza-CdR for 96 h in a concentration of1 ÌM clearly induced MAGE-A1 and MAGE-A2 expres-sion in the pancreatic carcinoma cell line PC-44 as com-pared to untreated cells (fig. 3A, lanes 5 and 6). BON cellsrevealed a weak specific MAGE-A2 PCR product of230 bp, but hardly MAGE-A1 expression (fig. 3A, lane 2),while QGP-1 cells showed weak MAGE-A1 expressionand hardly MAGE-A2 amplification (fig. 3A, lane 4).Treatment with 5-Aza-CdR was capable of upregulatingMAGE-A1 as well as MAGE-A2 expression in both neu-roendocrine tumor cell lines (fig. 3A, lanes 1 and 3). Acti-vation of expression following 5-Aza-CdR treatment wasalso observed for MAGE-A3 and MAGE-A4 in QGP-1cells but not in BON cells, while GAGE1/2/7 andGAGE1–8 expression could be upregulated in both celllines (fig. 3B, lanes 2–6). In contrast, although only weaklyexpressed, MAGE-6 expression appeared downregulated

PC1


Fig. 2. Expression profile of MAGE-A genesin representative primary pancreatic tu-mors. RT-PCR analysis of MAGE-A1,MAGE-A2, MAGE-A3 and MAGE-A6 ex-pression from pancreatic tumor tissues usingMAGE-specific primers and incorporationof 33P-labeled dCTP. The pancreatic ductalcarcinoma cell line BxPC-3 served as posi-tive control for MAGE-A gene expression.Numbers indicate independent tumor sam-ples characterized in table 2. Co = PCR reac-tion without cDNA. ß-Actin was amplifiedas control for cDNA quality.

Table 2. Clinicopathological data and MAGE-A gene expression in primary pancreatic tumors

Patient Sex Age Histology TNM UICC Grading Expression of MAGE-A

–1 –2 –3 –6

F 72 AdenoMhep T3N1Mhep IV G3 + + + +PC33 M 66 Adeno T2N1M0 III G2 – – – –PC38 M 67 Adeno T2N1M0 III G3 – – – –PC63 M 63 Adeno T2N0M0 I G2 – + – –PC95 M 76 Adeno T2N1M0 III G2 – – – –PC473 F 60 Adeno T3N1M0 III G3 – + – –PC475 F 57 Adeno T1N0M0 I G2–3 – – – –PC467 F 61 MuC n.a. n.a. n.a. – – – –PC494 F 78 Adeno T2N1M0 III G2 – + – –PC495 M 60 NET – – G1 – – – –PC497 M 53 Adeno T3N1M0 III G3 – – – –PC511 M 57 Adeno T3N1M0 III G3 – – – –PC513 M 42 Adeno T3N1Mhep IV G2 – – – –PC516 M 68 Adeno T3N1Mhep IV G2 – – – –PC529I M 61 Adeno T3N1M0 III G3 – – – –PC539II F 63 Adeno T3N0M0 II G2 + + – –PC536II M 76 NET – – n.a. + – – –PC512I F 53 Adeno T3N0M0 II G2 + – – –PC517II F 57 Acinus T2N0M0 I G2 + + – –PC553I F 61 Adeno T3N1M0 III G2 – – – –PC560 M 51 Adeno T3N1M0 III G2 – – – –PC565 F 75 Adeno T4N1M0 III G1 + + – –PC524I M 72 Adeno T3N1M0 III G2 – – – –

AdenoMhep = Hepatic metastasis of a ductal adenocarcinoma of the pancreas; Adeno = primary ductal adenocarci-noma of the pancreas; MuC = mucinous cystadenocarcinoma of the pancreas; NET = neuroendocrine tumor of thepancreas, Acinus = acinus cell carcinoma of the pancreas; n.a. = not available. (+) = expression; (–) = no expression.



Fig. 3. Methylation-dependent expression ofMAGE-A and GAGE genes in pancreatic tu-mor cells. Gene expression was analyzed be-fore and after 5)-aza-2-deoxycytidine treat-ment. Treated cells included two neuroendo-crine pancreatic tumor cell lines (BON,QGP-1) and two pancreatic carcinoma celllines (PC-44, MIAPaCa-2). Cells were incu-bated for 96 h in the presence (+) or absence(–) of 1 ÌM 5)-aza-2-deoxycytidine (DAC).Total RNA was extracted and submitted toMAGE-A and GAGE gene specific RT-PCRusing 33P-labeled dCTP. PosCo, BxPC-3positive control for MAGE-A1, MAGE-A2,MAGE-A3, MAGE-A6, GAGE expression,Panc-1 cells positive control for MAGE-A4expression. Actin was amplified as controlfor cDNA quality in the presence and ab-sence of 5)-aza-2-deoxycytidine. The ß-actincontrols for BxPC-3 and Panc-1 are demon-strated in figure 1C. negCo = PCR reactionwithout cDNA.

in QGP-1 cells after demethylation (fig. 3B). Except forMAGE-6, the present data are consistent with MAGE-Aand GAGE upregulation due to demethylation in neu-roendocrine and ductal pancreatic tumor cells. However,neither MAGE-A1 nor MAGE-A2 expression could bereinstalled by demethylation in the ductal pancreatic car-cinoma cell line MIAPaCa-2 (fig. 3A, lanes 7 and 8), norMAGE-A3 and MAGE-A4 expression in the neuroendo-crine BON cell line (fig. 3B, lanes 2 and 3) suggesting fur-ther mechanisms for MAGE gene silencing besides genemethylation.

Discussion

The 5-year survival rate of ductal pancreatic cancer islow and has not improved very much during recent years[31]. Also, treatment options for neuroendocrine pan-creatic tumors remain unsatisfactory [32]. Thus, alterna-tive treatment modalities would be desirable. Antigen-specific immunotherapy has been suggested by the dis-covery of the MAGE gene family and related tumor anti-gens that are exclusively expressed in malignant tissue,except for male germline cells and placenta [8]. Of all pan-creatic tumor cell lines 67% expressed at least one of the


MAGE-A genes analyzed. Approximately 54% of the duc-tal pancreatic carcinoma cell lines revealed MAGE-A1and/or MAGE-A3 gene expression, respectively, while thepositive rate of either MAGE-A2 (31%) or MAGE-A6(23%) was lower. Moreover, 54% of the ductal pancreatictumor cell lines revealed GAGE1–8 gene expression,although weak. Of note, the primary ductal pancreaticadenocarcinomas exclusively expressed MAGE-A1 and/or MAGE-A2 in 33% of the specimens, while no MAGE-A3 nor MAGE-A6 expression could be detected. Overall,the present study provides evidence for MAGE-A andGAGE gene expression in ductal and neuroendocrine pan-creatic tumor cells indicating that immunization couldpotentially become an additional treatment modality in asubset of pancreatic tumor patients. MAGE-A antigensencode peptide antigens presented in association withHLA class 1 molecules recognized by CTLs and clinicaltrials of targeted immunotherapy using MAGE gene prod-ucts are presently ongoing using peptide loaded dendriticcells or modified peptide vaccination with promisingresults [9, 18–24]. However, at present, there are no reli-able commercially antibodies available and protein ex-pression studies have to be awaited.

The lower incidence of MAGE-A gene expression inthe primary carcinomas could be explained by analyzingcultured versus primary tumor cells. MAGE-type CpGislands are highly methylated in normal somatic tissuesand specific activation as a result of demethylation hasbeen demonstrated for MAGE-A1, MAGE-A2, MAGE-A3, MAGE-A4, and GAGE genes [7, 33–36]. Since de-methylation of the MAGE-type promoters seems to be arandom consequence of genome-wide demethylation itwould be conceivable, that the process of overall genomicdemethylation and thus activation of MAGE and GAGEgene expression proceeds faster in cultured tumor cellsthan in primary tumors. This could also explain the coex-pression pattern of MAGE-A and GAGE genes in varioustumor cell lines.

Previous studies suggested, that MAGE-A1 inductionby the DNA hypomethylating agent decitabine (5-Aza-CdR) could not be reached with the same ease in everycell line [34, 35, 37, 38]. In our study, we present evidencethat in ductal as well as neuroendocrine pancreatic tumorcells MAGE-A expression is indeed controlled by DNAmethylation. Since only 23% and 27% of all primary pan-creatic tumors analysed were MAGE-1 or MAGE-2 posi-tive, respectively, pretreatment with decitabine in initial-ly MAGE-negative or low expression tumors could in-crease the number of patients who might be candidatesfor MAGE-specific immunotherapy. Drugs such as deci-

tabine have already been subjected to intense clinicalscrutiny in other tumor entities [39]. However, the corre-lation appears not absolute, since the pancreatic carcino-ma cell line MIAPaCa-2 did express neither MAGE-A1nor MAGE-A2 after demethylation. This is consistentwith a report of heavily demethylated melanoma cell lineswithout MAGE-A1 gene expression [33], suggesting thatglobal demethylation affects the genome randomly, andtherefore does not always lead to demethylation ofMAGE-A gene promoter regions. Also, lack of appropriatetranscription factors in these cells could account for thesefindings. Since pancreatic carcinoma specimens exclu-sively revealed MAGE-A1 and/or MAGE-A2 expression,one could hypothesize that the MAGE-A1 and MAGE-A2promoters are preferentially tissue-specific demethylated.Demethylation of QGP-1 cells with restoration of MAGE-A1, MAGE-A2, MAGE-A3 and MAGE-A4 expression wassurprisingly accompanied by the reduction of MAGE-A6expression level. Inverse methylation-dependent expres-sion in QGP-1 cells had recently been reported forCDKN2A/p16 and CDKN2D/p14, suggesting a direct re-lationship between CDKN2A exon 1· CpG island methyl-ation and CDKN2D/p14 expression due to faciliatedchromatin compaction around the promoter thus favor-ing initiation of CDKN2A/p16 transcription [27]. A simi-lar mechanism could account for the differential MAGE-A gene expression pattern observed after 5-Aza-CdRtreatment due to the proximity of the MAGE-A genes inthe q-terminal region of chromosome X. Our data did notexhibit any clear correlation between MAGE-A1 orMAGE-A2 gene expression and clinicopathologic data,such as lymph node and hepatic metastasis nor diseasestage in the primary pancreatic carcinomas. However,concomitant strong MAGE-A1, MAGE-A2, MAGE-A3and MAGE-A4 gene expression was observed in a hepaticmetastasis, suggesting a correlation of MAGE-A gene acti-vation with metastatic spread. This is consistent with pre-vious reports of significantly higher MAGE gene expres-sion in metastatic lesions than in the primary lesions inother tumors, such as melanomas, breast and colorectalcarcinomas [10, 12, 34]. Furthermore, a correlation be-tween the presence of liver metastases and MAGE geneexpression had been reported [14]. However, in our study,MAGE-A gene expression could not be detected in the pri-mary pancreatic carcinomas of 2 patients with knownhepatic metastases. This could be explained by a lownumber of cells in the primary tumor with the tendency tometastasize to the liver. On the other hand, MAGE-A geneexpression could be dependent on homing of the meta-static pancreatic carcinoma cell in the liver. The mecha-



nisms of MAGE-A gene expression at late stages of can-cers are unclear at present and it is open to questionwhether MAGE-A gene expression in malignant tissues issimply a consequence of the demethylation process occur-ring in many tumors. Consistent with the role proposedfor demethylation in MAGE-A1 activation, genome widehypomethylation is usually more pronounced in metas-tases than in the primary tumors [40–42]. Conversely, noexpression of MAGE-A1 has been found in acute leuke-mias, a malignancy retaining high level of methylation[43, 44]. However, one metastasis is too small to draw anysolid conclusions and studies addressing MAGE-A expres-sion in a larger series have to be awaited to confirm anyrelevant association.

Molecular strategies to detect very small numbers ofcancer cells have recently been reported by MAGE-A mul-timarker RT-PCR assay in patients with MAGE-A ex-

pressing primary tumors [45, 46]. The ÌMAGE-A [46]mRNA assay detected circulating tumor cells in the bloodof melanoma, breast cancer and colorectal cancer pa-tients. Thus, evaluation of MAGE-A gene expression inpancreatic tumors could not only open a window for alter-native therapeutic strategies, but also for a promisingdiagnostic tool indicating metastatic spread.

Acknowledgments

The authors thank Prof. Rudolf Arnold for continuous support.We are indebted to Dr. Francis Brasseur and Dr. Daniele Godelaine(Ludwig Institute for Cancer Resarch, Brussels, Belgium) for the gen-erous gift of the human cell lines MZ2-MEL 3.0 and LB23 SARC,and the MAGE specific primer sequences. This research was support-ed by a grant from the Alfred und Ursula Kulemann Stiftung to B.S.

References

1 Boon T, Cerottini JC, Van den Eynde B, Vander Bruggen P, Van Pel A: Tumor antigens rec-ognized by T lymphocytes. Annu Rev Immu-nol 1994;12:337–365.

2 Tanzarella S, Russo V, Lionello I, Dalerba P,Rigatti D, Bordignon C, Traversari C: Identifi-cation of promiscuous T-cell epitope encodedby multiple members of the MAGE family.Cancer Res 1999;59:2668–2674.

3 De Backer O, Arden KC, Boretti M, Van-tomme V, De Smet C, Czekay S, Viars CS, DePlaen E, Brasseur F, Chomez P, Van den EyndeB, Boon T, van der Bruggen P: Characteriza-tion of the GAGE genes that are expressed invarious human cancers and in normal testis.Cancer Res 1999;59:3157–3165.

4 Russo V, Tanzarella S, Dalerba P, Rigatti P,Rovere P, Villa A, Bordignon C, Traversari C:Dendritic cells acquire the MAGE-3 humantumor antigen from apoptotic cells and inducea class I-restricted T cell response. Proc NatlAcad Sci USA 2000;97:2185–2190.

5 Rogner UC, Wilke K, Steck E, Korn B, PoustkaA: The melanoma antigen gene (MAGE) familyis clustered in the chromosomal band Xq28.Genomics 1995;29:725–731.

6 Muscatelli F, Walker AP, De Plaen E, StaffordAN, Monaco AP: Isolation and characteriza-tion of a new MAGE gene family in the Xp21.3region. Proc Natl Acad Sci USA 1995;92:4987–4991.

7 Lucas S, De Smet C, Arden KC, Viars CS,Lethe B, Lurquin C, Boon T: Identification of anew MAGE gene with tumor-specific expres-sion by representational difference analysis.Cancer Res 1998;58:743–752.

8 De Plaen E, Arden K, Traversari C, Gaforio JJ,Szikora JP, De Smet C, Brasseur R, Chomez P,De Backer O, Caveneew X, Boon T: Structure,chromosomal localization and expression of 12genes of the MAGE family. Immunogenetics1994;40:360–369.

9 Gaugler B, Van den Eynde B, Van der BruggenP, Pomero P, Gaforio S, De Plaen E, Lethe B,Brasseur F, Boon T: Human gene MAGE-A3codes for an antigen recognized on a melanomaby autologous cytolytic T lymphocytes. J ExpMed 1994;179:921–930.

10 Brasseur F, Rimoldi D, Lienard D, Lethe B,Carrel S, Arienti F, Suter L, Vanwijck R, Bour-lond A, Humblet Y, Vacca A, Conese M, La-haye, T, Degiovanni G, Deraemaecker R,Beauduin M, Sastre X, Salamon E, Dreno B,Jager E, Knuth A, Chevreau C, Siciu S, Lacha-pelle JM, Pouillart P, Parmiani G, Lejeune F,Cerottini JC, Boon T, Marchand M: Expres-sion of MAGE genes in primary and metastaticcutaneous melanoma. Int J Cancer 1995;63:375–380.

11 Weynants P, Bernard L, Brasseur F, MarchandM, Boon T: Expression of MAGE genes bynon-small-cell lung carcinomas. Int J Cancer1994;56:826–829.

12 Russo V, Traversari C, Verrecchia A, Motto-lese M, Natali PG, Bordignon C: Expression ofthe MAGE gene family in primary and meta-static human breast cancer: Implications fortumor antigen-specific immunotherapy. Int JCancer 1995;64:216–221.

13 Sahin U, Koslowski M, Tureci O, Eberle T,Zwick C, Romeike B, Moringlane JR,Schwechheimer K, Feidn W, Pfreundschuh M:Expression of cancer testis genes in humanbrain tumors. Clin Cancer Res 2000;6:3916–3922.

14 Hasegawa H, Mori M, Haraguchi M, Ueo H,Sugimachi K, Akiyoshi T: Expression spectrumof melanoma antigen-encoding gene familymembers in colorectal carcinoma. Arch PatholLab Med 1998;122:551–554.

15 Li J, Yang Y, Tatsuro F, Kinya B, Ueo H, MoriM, Akiyoshi T: Expression of BAGE, GAGE,and MAGE genes in human gastric carcinoma.Clin Cancer Res 1996;2:1619–1625.

16 Chambost H, Van Baren N, Brasseur F, Gode-laine D, Xerri L, Landi SJ, Theate I, Plumas J,Spagnoli GC, Michel G, Coulie PG, Olive D:Expression of MAGE-A4 in Reed-Sternbergcells. Blood 2000;11:3530–3533.

17 Traversari C, Van der Bruggen P, Luescher IF,Lurguin C, Chomez P, Van Pel A, De Plaen E,Amar-Costesec A, Boon T: A nonapeptide en-coded by human gene MAGE-A1 is recognizedon HLA-A1 by cytolytic T lymphocytes di-rected against tumor antigen MZ2-E. J ExpMed 1992;176:1453–1457.

18 Van der Bruggen P, Traversari C, Chomez P,Lurquin C, De Plaen E, Van den Eynde B,Knuth A, Boon T: A gene encoding an antigenrecognized by cytolytic T lymphocytes on ahuman melanoma. Science 1991;254:1643–1647.

19 Kawakami Y, Eliyahu S, Delgado CH, RobbinsP, Riroltini L, Toplalian SL, Miki T, Rosen-berg SA: Cloning of the gene coding for ashared human melanoma antigen recognizedby autologous T cells infiltrating into tumor.Proc Natl Acad Sci USA 1994;91:3515–3519.


20 Kawakami Y, Eliyahu S, Delgado CH, RoblinsPF, Sakaguchi K, Appella E, Yannelli JR, Ad-ena G, Miki T, Rosenberg SA: Identification ofa human melanoma antigen recognized by tu-mor-infiltrating lymphocytes associated within vivo tumor rejection. Proc Natl Acad SciUSA 1994;91:6458–6462.

21 Rosenberg SA, Yang JC, Schwartzentruber DJ,Hwu P, Marincola FM, Popalian SL, RestifoNP, Dudley ME, Schwarz SL, Spiess PJ, Wun-derlich JR, Parkhurst MR, Kawakami Y, SeippCA, Einhorn JH, White DE: Immunologic andtherapeutic evaluation of a synthetic peptidevaccine for the treatment of patients with meta-static melanoma. Nat Med 1998;4:321–327.

22 Marchand M, Weynants P, Rankin E, ArientiF, Belli F, Parmiani G, Cascinelli N, BourlondA, Vanwijck R, Humblet Y, Canon JL, LaurentC, Naeyaert JM, Plagne R, Deraemaeker R,Knuth A, Jaeger E, Brasseur F, Herman J, Cou-lie PG, Boon T: Tumor regression responses inmelanoma patients treated with a peptide en-coded by gene MAGE-3. Int J Cancer 1995;63:883–885.

23 Nestle FO, Alijagic S, Gilliet M, Sun Y, GrabbeS, Dummer R, Burg G, Schadendorf D: Vaci-nation of melanoma patients with peptide-ortumor lysate-pulsed dendritic cells. Nat Med1998;4:328–332.

24 Nishiyama T, Tachibana M, Horiguchi , Naka-mura K, Ikeda Y, Takesako K, Murai M: Im-munotherapy of bladder cancer using autolo-gous dendritic cells pulsed with human lym-phocye antigen-A24-specific MAGE-3 peptide.Clin Cancer Res 2000;7:23–31.

25 Takahashi K, Shichijo S, Noguchi M, HirohataM, Itoh K: Identification of MAGE-1 andMAGE-4 proteins in spermatogonia and pri-mary spermatocytes of testis. Cancer Res 1995;55:3478–3482.

26 Simon B, Bartsch D, Barth P, Prasnikar N,Münch K, Blum A, Arnold R, Göke B: Fre-quent abnormalities of the putative tumor sup-pressor gene FHIT at 3p14.2 in pancreatic car-cinoma cell lines. Cancer Res 1998;58:1583–1587.

27 Lubomierski N, Kersting M, Bert T, MuenchK, Wulbrand U, Schuermann M, Bartsch D,Simon B: Tumor suppressor genes in the 9p21gene cluster are selective targets of inactivationin neuroendocrine gastroenteropancreatic tu-mors. Cancer Res 2001;61;5905–5910.

28 De Smet C, Lurquin C, Van der Bruggen P, DePlaen E, Brasseur F, Boon T: Sequence and pat-tern of expression of human gene MAGE-A2.Immunogenetics 1994;39:121–129.

29 Shichijo S, Hayashi A, takamori S, TsunosueR, Hoshino T, Sakata M, Kuramoto T, OizumiK, Itoh K: Detection of MAGE-4 protein inlung cancers. Int J Cancer 1995;64:158–165.

30 Simon B, Weinel R, Höhne M, Watz J,Schmidt J, Körtner G, Arnold R: Frequentalterations of the tumor suppressor genes p53and DCC in pancreatic adenocarcinoma. Gas-troenterology 1994;106:1645–1651.

31 Simon B, Printz H: Epidemiological trends inpancreatic neoplasias. Dig Dis 2001;19:6–14.

32 Arnold R, Simon B, Wied M: Treatment ofneuroendocrine GEP tumors with somatosta-tin analogues. Digestion 2000;62:84–91.

33 De Smet C, De Backer O, Faraoni I, Lurquin C,Brasseur F, Boon T: The activation of humangene MAGE-A1 in tumor cells is correlatedwith genome-wide demethylation. Proc NatlAcad Sci USA 1996;93:7149–7153.

34 Mori M, Inoue H, Mimori K, Shibuta K, BabaK, Nakashima H, Haraguchi M, Tsuji K, UeoH, Barnard GF, Akiyoshi T: Expression ofMAGE genes in human colorectal carcinoma.Ann Surg 1996;224:183–188.

35 De Plaen E, Naerhuyzen B, De Smet C, SzikoraJP, Boon T: Alternative promoters of geneMAGE-A4a. Genomics 1997;40:305–313.

36 De Smet C, Lurquin C, Lethe B, Martelange V,Boon T: DNA Methylation is the primarysilencing mechanism for a set of germ line- andtumor-specific genes with a CpG-rich promot-er. Mol Cell Biol 1999;19:7327–7335.

37 Jansson BLM, Van de Locht LTF, Fourkour A,De Smet C, Mensink EJBM, Muijen GNP, DeVries TJ: Transcription of the MAGE-1 geneand the methylation status of its Ets bindingpromoter elements. Melanoma Res 1999;9:213–222.

38 Serrano A, Garcia A, Abril E, Garrido F, Riuz-Cabello F: Methylated CpG points identifiedwithin MAGE-1 promoter are involved in generepression. Int J Cancer 1996;68:464–470.

39 Wijermans P, Lubbert M, Verhoef G, Bosly A,Ravoet C, Andre M, Ferrant A: Low-dose 5aza-2'deoxycytidine, a DNA hypomethylatingagent, for the treatment of high-risk myelodys-plastic syndrome: Multicenter phase II study inelderly patients. J Clin Oncol 2000;18:956–962.

40 Gama-Sosa M, Slagel VA, Trewyn RW, Oxen-handler R, Kuo KC, Gehrke CW, Ehrlich M:The 5-methylcytosine content of DNA fromhuman tumors. Nucl Acids Res 1983;11:6883–6894.

41 Bedford MT, van Helden PD: Hypomethyla-tion of DNA in pathological conditions of thehuman prostate. Cancer Res 1987;47:5274–5276.

42 Liteplo RC, Kerbel RS: Reduced levels of DNA5-methylcytosine in metastatic variants of thehuman melanoma cell line MeWo. Cancer Res1987;47:2264–2267.

43 Chambost H, Brasseur F, Coulie P, De Plaen E,Stoppa AM, Baume D, Mannoni P, Boon T,Maraninchi D, Olive D: A tumor-associatedantigen expression in human haematologicalmalignancies. Br J Haematol 1993;84:524–526.

44 Pfeifer GP, Steigerwald S, Boehm LJ, Drahovs-ky D: DNA methylation levels in acute humanleukemia. Cancer Lett 1988;39:185–192.

45 Miyamoto A, Ujiwara Y, Sakon M, Nagano H,Sugita Y, Kondo M, Eguchi H, Dono K, Ume-shita K, Nakamori S, Mondem M: Develop-ment of a multiple-marker RT-PCR assay fordetection of micrometastases of hepatocellularcarcinoma. Dig Dis Sci 2000;45:1376–1382.

46 Miyashiro I, Kuo C, Huynh K, Iisa A, MortonD, Bilchik A, Giulian A, Hoon DS: Molecularstrategy for detecting metastatic cancers withthe use of multiple tumor-specific MAGE-Agenes. Clin Chem 2001;47:505–512.

Documents

Expression spectrum and methylation-dependent regulation of melanoma antigen-encoding gene family members in pancreatic cancer cells