Complete loss of HLA class I antigen expression on melanoma cells: A result of successive mutational events†
Article first published online: 12 DEC 2002
Copyright © 2002 Wiley-Liss, Inc.
International Journal of Cancer
Volume 103, Issue 6, pages 759–767, 1 March 2003
How to Cite
Paschen, A., Méndez, R. M., Jimenez, P., Sucker, A., Ruiz-Cabello, F., Song, M., Garrido, F. and Schadendorf, D. (2003), Complete loss of HLA class I antigen expression on melanoma cells: A result of successive mutational events. Int. J. Cancer, 103: 759–767. doi: 10.1002/ijc.10906
This article is dedicated to Prof. Harald zur Hausen on the occasion of his retirement as head of the German Cancer Research Center, with gratitude and appreciation for 20 years of leadership.
- Issue published online: 3 JAN 2003
- Article first published online: 12 DEC 2002
- Manuscript Accepted: 31 OCT 2002
- Manuscript Revised: 29 OCT 2002
- Manuscript Received: 4 JUL 2002
- Verein zur Förderung der Krebsforschung in Deutschland
- European Commission. Grant Numbers: OISTER-QLG1-CT-2002-0068, ESTDAB-QLRT-2000-01325
- Fondo de Investigaciones Sanitarias, Servicio Andaluz de Salud, Spain
- immune escape;
- HLA loss;
- chromosomal aberration
Alterations in the surface expression of HLA class I molecules have been described as a strategy of tumors to evade recognition by cytotoxic T cells. We detected complete loss of HLA class I antigen presentation for 2 tumor cell lines from 1 melanoma patient, the first originated from a regional lymph node lesion diagnosed simultaneously with the primary tumor and the second established 8 months later from a metastatic pleural effusion sample. Antigen presentation was not inducible with IFN-γ but could be restored after transfection of tumor cells with b2m cDNA, indicating a defect in b2m expression. Analysis of the nature of this defect revealed that it originated from at least 2 mutational events affecting both copies of the b2m gene: a microdeletion of 498 bp in one b2m gene, including its entire exon 1, and a macrodeletion involving the entire copy of the second b2m gene. Microsatellite analysis pointed to the macrodeletion by demonstrating LOH for several specific markers on the long arm (q) of chromosome 15. Structural imbalance of 15q was verified by FISH. FISH studies also indicated the coexistence of a structurally abnormal variant of chromosome 15q with 2 apparently entire chromosomes 15q harboring the homozygous b2m microdeletion. Block of b2m expression in tumor cells builds a barrier to immunotherapy of cancer patients, and its early incidence should be of major consideration in the development and design of immunotherapeutic strategies. © 2002 Wiley-Liss, Inc.
Evidence that melanoma-associated antigens can be specifically recognized by the immune system led to the development and clinical application of several tumor-specific vaccination strategies. Most of these approaches focused on the induction of a CD8+ CTL-mediated immune response since CTLs have been postulated to be the main mediators of tumor immunity due to their capability to directly lyse the tumor cell. However, when evaluated in clinical phase I/II studies, these approaches exhibited therapeutic effects only in a minority of melanoma patients.1 This failure might to a certain extent be due to the fact that tumors have evolved a number of different strategies to escape recognition by T cells. Since CTLs recognize tumor antigens in the context of HLA class I molecules, alteration of HLA surface expression is a frequent mechanism used by tumors to evade T-cell control.2, 3, 4 Different altered HLA class I phenotypes, including total loss or downregulation of HLA class I antigens, HLA haplotype, HLA locus or HLA allele, have been described for tumors originating from different tissues; and multiple molecular mechanisms have been identified as responsible for these changes.3, 4
Total lack of HLA class I antigen expression is frequently observed for tumor tissue and cultured tumor cell lines of malignant melanoma.3, 4 The molecular mechanisms causing this defect are heterogenous and can be of a reversible or an irreversible nature. An apparent HLA class I-negative phenotype, reversible by IFN-γ, can be due to regulatory defects leading to transcriptional downregulation of different components of the antigen-processing and presentation machinery.5, 6, 7 In contrast, several mutations located in genes encoding structural components of this machinery induce irreversible reduction or total loss of HLA class I antigen expression.4, 5 Alterations in the b2m gene have been characterized as being mainly responsible for the total lack of antigen surface presentation on melanoma cells.8, 9, 10, 11b2m, a highly conserved gene mapping to the long (q) arm of chromosome 15 at band q22, encodes a protein of 12 kDa (light chain) that associates with the polymorphic HLA class I heavy chains.12 The intracellular assembly of both components is a prerequisite for the surface presentation of HLA class I antigens. Total loss of antigen expression is dependent on mutational events involving both copies of the b2m gene. So far, studies defining a defect in b2m gene expression have not directly addressed the question of whether the characterized mutation was hemizygous (accompanied by deletion of nonmutated b2m) or homozygous and which molecular mechanisms contributed to the lack of b2m synthesis.
In this work, we demonstrate that total loss of HLA class I antigen expression on tumor cells from a melanoma patient originated from at least 2 mutational events, a micro- and a macrodeletion, both affecting the b2m genes. These mutations occurred without application of any therapeutic regimen, and we postulate that their accumulation protected the tumor cells from CTL recognition and induced marked progression of malignant disease in this melanoma patient.
MATERIAL AND METHODS
Human melanoma cell lines UKRV-Mel-2a/-2b (HLA-A*0101, -A*0202, -B*0801, -B*1801) and Ma-Mel-37b (HLA-A1, -A11, -B37, -B51) and human cervical cancer cell line HeLa were cultivated in RPMI medium (Biochrom, Berlin, Germany) supplemented with 10% FCS (GIBCO BRL–Life Technologies, Karlsruhe, Germany), 2 mM glutamine (Biochrom) and 100 U/100 μg/ml penicillin/streptomycin (Biochrom) at 37°C in a humidified atmosphere with 5% CO2.
PCR on genomic DNA
High m.w. genomic DNA was prepared from tumor cells, applying a standard phenol-chloroform method. DNA was used in PCR to specifically amplify sequences of the HLA-A and HLA-B loci, as described previously.13 In addition, b2m-specific amplification of genomic DNA sequences was carried out in a 35-cycle PCR using the primers b2m-657 (5′-ataagaggatccggaggaacttcttggcacag-3′), b2m+697 (5′-ataagactcgagaaccgctttgtatcacagcc-3′) and the Expand high-fidelity enzyme mix (Roche, Mannheim, Germany). After BamHI and XhoI restriction, b2m fragments were cloned into expression plasmid pCR3.1 (Invitrogen, Karlsruhe, Germany), followed by sequence analysis of the inserts. All samples were sequenced on an automated 373 DNA sequencer (Applied Biosystems, Warrington, UK). Data obtained by sequencing were analyzed using Match Tools version 1.0 Sequencing Analysis software (Applied Biosystems).
Total cellular RNA was purified with a micro-RNA isolation kit (Stratagene, La Jolla, CA). First-strand cDNA was synthesized with 2 μg of total RNA incubated at 42°C for 40 min in 20 μl of a solution containing 500 μM of each dNTP, 50 mM TRIS-HCl (pH 8.3), 25 mM MgCl2, 10 mM dithiothreitol, 2 μM solution of oligo (dT) primers, 20 U RNasin (Promega, Madison, WI) and 200 U of Moloney murine leukemia virus reverse transcriptase (Promega). Specific amplification of b2m cDNA (sense primer 5′-ataaccaagctttcctgaagctgacagcattcg-3′, antisense primer 5′-ataacctctagaacctccatgatgctgcttaca-3′) was carried out in a 30-cycle PCR at an annealing temperature of 55°C using the Taq polymerase (Qiagen, Hilden, Germany).
Southern blot hybridization
Genomic DNA (10 μg) was first digested with HindIII, EcoRI and SacI. Each digestion was done for at least 16 hr at 37°C with 40 U of enzyme. After electrophoresis, DNA was blotted onto nylon membranes and hybridized overnight at 65°C in hybridization buffer [5 × SSC, 0.1% SDS, 5% (w/v) dextran sulfate] to a fluorescein-labeled DNA probe using the Gene Images kit (Amersham, Arlington Heights, IL). The probe consisted of a fragment of exon 2 of the b2m gene generated by PCR. Membranes were washed in 2 × SSC, 0.1% SDS at room temperature (45 min) and in 0.2 × SSC, 0.1% SDS at 55°C (30 min) and then exposed with an intensifying screen.
Analysis of HLA surface expression
Surface HLA class I and II expression on cultured cells was determined by indirect immunofluorescence using the appropriate anti-class I MAb and FITC-labeled rabbit antimouse Ig (Fab) secondary fragments (Sigma, St. Louis, MO). Fluorescence was analyzed with a FACSort flow cytometer (Becton Dickinson, Mountain View, CA). A total of 104 cells were analyzed for each immunofluorescence profile. As a negative control, cells were incubated with an irrelevant MAb. To increase HLA expression, cell lines were treated with recombinant IFN-γ (Amersham, Aylesbury, UK) for 48 hr. Cell-surface HLA class I antigen expression was then examined by immunofluorescence as described above. We used the following MAbs to examine HLA class I expression on cultured cells: W6/32 against HLA class I heavy chain/b2m complex;14 GRH1 and GRB1 against b2m and HLA-DR, respectively, produced in our laboratory;15 A131, defining an HLA-A locus-specific determinant;16 and YTH-76, defining an HLA-B locus-specific determinant.17
LOH analysis was performed by PCR amplification of 4 highly polymorphic microsatellite sequences. Markers were chosen on the basis of their heterozygosity (PIC value >0.7) and location within the chromosome. Two of the markers studied, D15S11 and D15S122, are located at 15q11-15q13. The other 2 markers used in this work were D15S126 and D15S117, located at 15q21.1-15q21.3, flanking the b2m gene. The fluorescence microsatellite assay was performed as follows: PCRs were performed in a total volume of 15 μl containing 60 ng of each DNA sample, 1 × PCR buffer, 5 μM each of unlabeled primer and 5′ end-labeled primers with fluorescent dyes, 0.5 U Taq DNA polymerase and 250 μM of each dideoxynucleotide. Each PCR was carried out as a single assay and then pooled for electrophoresis. To precisely size the alleles, a portion of the PCR products was divided into aliquots and combined with dextran blue dye, formamide and GeneScan TAMRA internal size marker (Applied Biosystems). Samples were denatured at 94°C for 3 min, snap-cooled on ice and loaded on an ABI 310 automated capillary sequencer. Specific GeneScan and Genotyper software (Applied Biosystems) was used to size, quantify and compare normal and tumor amplicon patterns for each marker. LOH was calculated by determining the ratio of the 2 allele pairs corrected for differences in the amplification efficiencies of the normal/tumor DNA samples. LOH was considered to be present if the rate of reduction of the height of the allele in the tumor was >25%.18
In situ hybridization
FISH was performed with a WCP probe for the q arm of chromosome 15 (WCP15q; Vysis, Downers Grove, IL). Metaphases were prepared from the subconfluent cell line culture treated with colcemid (KaryoMax, Life Technologies). One or 2 drops of the cell suspension were added on slides and allowed to air-dry. Films were frozen wrapped in foil paper and stored at −20°C until use. After thawing, slides were hybridized according to the manufacturer's recommendation. Chromosomes were counterstained with DAPI and analyzed by fluorescence microscopy (Eclipse 400 fluorescence microscope; Nikon, Tokyo, Japan). A total of 25 metaphases were analyzed, the proportion and number of cell fluorescent signals being measured.
UKRV-Mel-2b cells were transfected using the FuGENE6 Transfection Reagent (Roche). Briefly, 3 μl of FuGENE6 reagent were mixed with 1 μg of b2m encoding plasmid DNA (kindly provided by Dr. S. Ferrone, Buffalo, NY) in 100 μl of RPMI medium and incubated at room temperature for 15 min. The DNA–FuGENE6 mixture was then added to UKRV-Mel-2b cells (2 × 105/well of a 6-well plate, thawed 1 day before transfection). After 48 hr, cells were analyzed for surface expression of HLA class I antigens by immunofluorescence, as described above.
Paraffin-embedded tumor tissue was analyzed for b2m, HLA class I and HLA-DR expression using the mouse MAbs GRH1 (recognizing free and HLA class I heavy chain-associated b2m),9, 15 HC-10 (binding to free HLA-A, -B and -C heavy chains)9 and GRB1 (against HLA-DR),9, 15 respectively. Staining was performed as described previously.19 Briefly, 5–7 μm paraffin sections were deparaffinized with xylene, progressively rehydrated in decreasing concentrations of ethanol and immersed in trypsin solution (0.1% trypsin, 0.1% CaCl in TBS) for 20 min. After preincubation with 10% normal rabbit serum, sections were incubated with the primary antibody followed by treatment with peroxidase-labeled rabbit antimouse Ig. Sections were then incubated with the peroxidase substrate 3,3′-diaminobenzidine tetrahydrochloride solution. Control tissue samples were treated by the same protocol but omitting incubation with the primary antibody.
T-cell stimulation assay
A peptide-specific T-cell line against the HLA-A1-restricted 146SSDYVIPIGTY156 tyrosinase epitope20 was generated from thawed PBMCs of a healthy donor by following a previously described protocol.21 T-cell stimulation by peptide-loaded b2m transfected or nontransfected UKRV-Mel-2b cells was analyzed as follows: UKRV-Mel-2b cells (1 × 104/well of a 96-well plate) were transfected with b2m encoding plasmid DNA (0.1 μg/well) as described above or left nontransfected. Control cell line Ma-Mel-37b, constitutively expressing HLA-A1, was left nontransfected. Two days later, cells were pulsed with the synthetic tyrosinase peptide epitope (50 μg/ml) for 2 hr. After washing, tumor target cells were incubated with 1 × 104 peptide-specific responder T cells in a volume of 200 μl culture medium/well. Twenty-four hours later, 100 μl of the supernatant of each well were analyzed for IFN-γ secretion by stimulated responder T cells in a standard capture ELISA. The ELISA was set up using the capture MAb 1-D1K and the biotinylated MAb 7-B6-1 for detection according to the manufacturer's protocol (Mabtech, Nacka, Sweden).
Lack of HLA class I expression on tumor cells originated from different tumor tissues at different time points of the clinical course of patient UKRV-Mel-2
Patient UKRV-Mel-2, a 38-year-old woman, presented in January 1992 with a primary melanoma on the right upper arm and a first regional lymph node metastasis. The primary tumor exhibited regions of partial regression pointing to destructive immunologic reactions, probably mediated by tumor-reactive T cells. The tumor mass was surgically removed, and the patient received several cycles of combined chemo-immunotherapies, under which she developed progressive disease. The patient died only 9 months after tumor diagnosis in September 1992 (Fig. 1).
Because of the tremendous disease progression, we examined whether the tumor cells might have developed escape mechanisms to prevent their immunologic destruction. Two cell lines from the patient were analyzed in regard to their surface expression of HLA class I molecules. UKRV-Mel-2a was established at diagnosis from the regional lymph node metastasis, and UKRV-Mel-2b derived from a pleural effusion 8 months after diagnosis. Neither UKRV-Mel-2a nor UKRV-Mel-2b cells could be stained with monomorphic or locus-specific anti-HLA class I antibodies, indicating complete loss of antigen expression that was not even detectable after IFN-γ treatment (Fig. 2). In contrast, expression of HLA-DR molecules could be induced in the presence of IFN-γ (Fig. 2).
To demonstrate that the HLA class I-negative phenotype was not restricted to in vitro cultured tumor cells, we analyzed sections from paraffin-embedded tumor tissue of a lymph node metastasis (diagnosed simultaneously with the primary tumor) for the expression of HLA class I antigens (Fig. 3). Immunohistologic analysis revealed that tumor cells expressed HLA class I heavy chains but did not synthesize the b2m light chain. Since HLA class I antigens constitute a noncovalent complex of HLA class I heavy chains and the b2m constant light chain, this led to the assumption that free heavy chains accumulated in the cell cytosol and were not transported to the cell surface due to a lack of b2m expression.
To confirm this hypothesis, we transiently transfected UKRV-Mel-2b cells with b2m wild-type cDNA. Indeed, transfection restored surface presentation of HLA class I molecules, as demonstrated by FACS analysis (Fig. 4a). Functional restoration of HLA-A1 expression after b2m transfection of UKRV-Mel-2b tumor cells was demonstrated by activation of an in vitro generated, peptide-specific T-cell line (Fig. 4b). T cells recognizing the tyrosinase 146SSDYVIPIGTY156 epitope20 in the context of the HLA-A1 restriction molecule were activated only after incubation with b2m transfected, peptide-loaded UKRV-Mel-2b cells, whereas nontransfected, peptide-loaded target cells did not induce cytokine secretion. The fact that UKRV-Mel-2b cells, which express the tyrosinase antigen,22 were recognized only after peptide loading was most probably due to a low-affinity phenotype of the self-antigen-restricted T-cell line, though a tumor cell defect in antigen processing could not be excluded.
Detection of a microdeletion in 1 copy of the b2m gene
To clarify the defect in b2m synthesis, we tested the tumor cells for b2m RNA expression, concentrating our analysis on the UKRV-Mel-2b cell line. Total RNA from tumor cells was reverse-transcribed to cDNA, followed by PCR with specific primers for b2m cDNA amplification (Fig. 5). In contrast to the control, no PCR fragment was detectable for UKRV-Mel-2b cells, indicating that neither of the 2 b2m genes was correctly transcribed. To define the molecular nature of this defect, we performed PCR amplifications on b2m genomic DNA. Overlapping fragments covering different regions of the b2m gene were analyzed for their size in an agarose gel (data not shown). A PCR fragment, spanning sequences from the 5′-untranslated region of the b2m gene to the first intron, exhibited a deletion of several hundred base pairs compared to the control (Fig. 6a,b). Sequence analysis of this PCR fragment confirmed a loss of 498 bp (from nucleotide −426 to +72), including the whole exon 1 of the b2m gene. Southern blot analysis on genomic DNA corroborated the PCR result (Fig. 6a,c). Interestingly, UKRV-Mel-2b cells exhibited only the mutated chromosomal b2m DNA fragment carrying the microdeletion and no product of normal size. This observation suggested either the presence of a homozygous mutation affecting both b2m genes or the existence of a hemizygous microdeletion accompanied by total loss of the second copy of the b2m gene. Molecular mechanisms that could lead to this finding are presented in Figure 7.
Detection of LOH for specific markers on chromosome 15q
UKRV-Mel-2b cells were investigated for chromosome 15 aberrations by microsatellite analysis, determining the status of heterozygosity of 4 polymorphic markers located in different regions on 15q. The PCR-based analysis was performed on DNA from UKRV-Mel-2b cells in comparison to control DNA extracted directly from autologous normal tissue obtained together with metastatic lymph node material taken at the very beginning of the patient's medical history (Fig. 8). In contrast to the control, 2 of 2 markers (D15S126, D15S117) located near the b2m region (15q22) displayed hemizygosity/homozygosity in UKRV-Mel-2b, indicating LOH for these highly polymorphic sequences, whereas 2 markers (D15S11, D15S122) located proximal to the centromere of chromosome 15 still showed ROH (Fig. 8). Detection of LOH and ROH for specific markers on 15q ruled out a single interstitial deletion of the wild-type b2m gene, a gene conversion and total loss of 1 chromosome 15, respectively.
Detection of structural imbalance of chromosome 15
To further characterize structural rearrangements leading to LOH for specific markers on chromosome 15, FISH on tumor metaphase cells was performed using WCP for the long arm of chromosome 15 (Fig. 9a). Two different types of fluorescence signal could be detected, indicating that the painting probe hybridized to 15q chromosomes of different length: 1 signal corresponded to an apparently entire chromosome 15q, whereas the second signal occurred as a result of WCP hybridization to an abnormal 15q, which appeared foreshortened. Interestingly, metaphase cells exhibited 3 fluorescence signals (Fig. 9a).
In view of the results obtained by PCR/Southern blot and LOH analysis, we postulate that (i) the apparently entire chromosome 15q contains the b2m microdeletion and (ii) 2 apparently entire chromosomes 15q coexist with a deletion variant in the tumor cells. This coexistence either can be due to duplication of the chromosome harboring the microdeletion and partial loss of the second chromosome 15q (reflecting trisomy 15, Fig. 9b) or can originate from transfer of sequences from the mutated chromosome 15q (harboring the microdeletion) to the second chromosome 15q (mitotic recombination) accompanied by deletion and translocation of partial sequences of the second chromosome (Fig. 9c).
In our study, 2 mutations involving both genomic copies of the b2m gene caused the HLA class I-negative phenotype of tumor cells from melanoma patient UKRV-Mel-2: (i) a microdeletion of 498 bp in 1 copy of the b2m gene including exon 1 and part of its flanking upstream regions (Fig. 6) and (ii) a macrodeletion including the entire copy of the second b2m gene (Figs. 8, 9).
Local mutations in the b2m gene have previously been demonstrated to be responsible for the total loss of antigen presentation in melanoma.8, 9, 10, 11 Most of the analyzed defects originated from nucleotide transitions or deletions of 1 or a few base pairs in exon 1 or exon 2 that blocked the synthesis of an intact b2m molecule.9, 10, 11 In general, these studies did not ask whether the characterized b2m mutation was homozygous or hemizygous. This work clearly demonstrates that the blockade of b2m expression in UKRV-Mel-2 cells could be traced back to successive mutational events including a micro- and a macrodeletion. LOH analysis pointed to an extensive deletion of sequences from 1 chromosome 15q (Fig. 8), and subsequent FISH on metaphase chromosomes confirmed the structural imbalance for chromosome 15q. FISH also indicated that the structurally altered chromosome variant coexisted with 2 apparently entire copies of chromosome 15q per cell. Our data allowed us to conclude that these apparently intact chromosomes 15q contained the homozygous microdeletion, though we could not distinguish between chromosome duplication and mitotic recombination as the causative mechanism for homozygosity (Fig. 8). Thus, structural imbalance of chromosome 15q together with a microdeletion in 15q are responsible for the total loss of HLA class I antigen expression on tumor cells. Up to now, studies characterizing chromosomal abnormalities in melanoma as causes of defects in the expression of classical HLA class I antigens mainly focused on chromosome 6. Deletions at 6p21, where the HLA region has been mapped, are responsible for loss of HLA haplotype or HLA allele.22, 23, 24, 25
Loss of surface presentation of HLA class I antigens on melanoma cells is not exclusively dependent on mutations affecting the b2m genes. A 1 bp deletion in the gene encoding the TAP1 protein was identified as being responsible for the lack of HLA class I expression.6 Besides these irreversible defects, different molecular mechanisms of reversible transcriptional downregulation of HLA class I antigen expression have been described, including silencing of HLA class I gene transcription by hypermethylation of the corresponding promoters and altered binding of regulatory factors to the HLA class I heavy chain enhancer element.7, 26
Total lack of HLA class I antigen presentation enables the tumor to escape immune recognition by CTLs, which might explain the marked clinical evolution of patient UKRV-Mel-2. All tumor samples analyzed in our study, the tissue cells from an early lymph node metastasis as well as the cells from both tumor lines, exhibited an HLA class I-negative phenotype, though the samples were taken at different time points from different locations. This might indicate an advantage of the cellular phenotype for growth and dissemination in an immunocompetent organism. Indeed, b2m transfection functionally restored HLA class I surface presentation on UKRV-Mel-2b cells since transfectants, in contrast to nontransfected cells, were recognized by specific T cells. These experimental data contribute to the theory that the selection of tumor variants with altered MHC expression might be the result of a T cell-dependent process.27
In addition to the total loss of HLA class I expression, other escape mechanisms might have contributed to the aggressive metastatic behavior of the tumor in patient UKRV-Mel-2. CTLs mediate tumor cell killing in either a Fas- or a perforin-dependent manner. Tumor escape mechanisms that alter both functions have been described: (i) downregulation of Fas death receptors and defects in death receptor signaling28 and (ii) blockade of the perforin pathway through overexpression of serine protease inhibitors.29 In addition, these evasion strategies can be enforced by simultaneous secretion of immunosuppressive cytokines as well as downregulation of tumor antigen expression.30, 31, 32, 33, 34 The information on immune evasion strategies developed by tumors clearly indicates that cancer cells successively acquire various mutations to survive, disseminate and metastasize in an immunocompetent host.
Several data gathered so far suggest that these immune evasion tumor cell phenotypes might be selected during vaccination since the escape mechanisms represent a barrier to T cell–mediated lysis.9, 13 This patient, however, demonstrates that evasion strategies can develop before the clinical application of any therapeutic regimen. Therefore, the possibility of early development of HLA-loss tumor variants should be taken into consideration when immunologic treatment strategies are designed.
We thank A. Concha (Department of Pathology, Hospital Universitario Virgen de las Nieves) for performing immunologic tissue staining, J. Mueller-Berghaus (Skin Cancer Unit) for helpful discussion and M. Vazansky (Skin Cancer Unit) for critically reading the manuscript. This work was supported by the Verein zur Förderung der Krebsforschung in Deutschland and the European Commission (OISTER-QLG1-CT-2002-0068 to A.P., A.S., M.S. and D.S.) and by the Fondo de Investigaciones Sanitarias (Servicio Andaluz de Salud, Spain) and the European Commission (ESTDAB-QLRT-2000-01325 to R.M.M., P.J., F.R.C. and F.G.).
- 15Production and characterization of monoclonal antibodies against leukemic cells. Immunologia 1986; 5: 51–9., , , , .
- 33A novel autocrine pathway of tumor escape from immune recognition: melanoma cell lines produce a soluble protein that diminishes expression of the gene encoding the melanocyte lineage melan-A/MART-1 antigen through down-modulation of its promoter. J Immunol 2001; 167: 1204–11., , , , , , , , , , .