The Coincidence of Chromosome 15 Aberrations and β2-Microglobulin Gene Mutations Is Causative for the Total Loss of Human Leukocyte Antigen Class I Expression in Melanoma

Melanoma cells can be effectively eradicated in vivo by the cytotoxic activity of HLA class I–restricted tumor antigen-specific CD8⁺ T cells as recently shown in clinical trials of adoptive T-cell transfer (1, 2). However, these studies also showed that adoptively transferred T cells infiltrating the metastasis not only act as cytotoxic immune effectors but also as immune editors (3, 4) selecting for tumor cell variants that escape immune surveillance by down-regulation of antigen expression (1). Besides antigen down-regulation, other properties of melanoma cells that interfere with T-cell effector function have been described, such as release of immunosuppressive cytokines (5, 6), release of Fas ligand–bearing microvesicles (7), and alterations in the surface presentation of HLA class I molecules (8–10).

HLA class I molecules are heterodimeric noncovalently associated complexes consisting of the constant β2-microglobulin (β2m) light chain and the variable HLA heavy α chain. Irreversible alterations in the HLA class I phenotype of melanoma cells, such as loss of a single HLA allele, an HLA locus, or an HLA haplotype, are generally caused by mutations affecting HLA heavy chain genes located on chromosome arm 6p (11, 12). In contrast, total loss of HLA class I expression in melanoma is mainly due to mutations affecting the β2m gene (8–10, 13), mapping to chromosome region 15q21 (14). Interestingly, such alterations have repeatedly been detected in tumors of melanoma patients after immunotherapy (15–17).

Analysis of the molecular mechanisms of β2m deficiency in different melanoma cell lines revealed that, in some cases, large deletions within the β2m gene prevented transcription whereas, in the majority, a mutated nonfunctional gene product was expressed originating from nucleotide microdeletions or transitions in exons I and II of the β2m gene (16–18). Interestingly, the coincidence of two different β2m gene mutations within one melanoma cell line has never been described. However, total loss of β2m expression can only result from two mutational events affecting both alleles. Previous studies suggested that loss of one β2m gene contributes to β2m deficiency (19, 20). We therefore asked whether and how abnormalities of chromosome 15, to which the β2m gene maps, contribute to the loss of β2m expression in HLA class I–deficient melanoma cells.

To answer this question, we did analysis at the gene and chromosome levels on cell lines derived from HLA class I–negative metastatic tissue specimens of different melanoma patients, thereby ensuring that we concentrated our studies on in vivo shaped tumor cell phenotypes. As expected, mutations at the gene level could be detected in all cell lines. Loss of heterozygosity (LOH) studies additionally pointed to extensive deletions within chromosome 15q and, interestingly, subsequent cytogenetic analysis revealed the presence of chromosome 15 rearrangements or translocations in all cases analyzed. This strongly suggests that in β2m-negative melanoma, chromosome 15 instability critically contributes to the loss of β2m expression.

Tumor tissue and cells. Metastatic melanoma lesions were obtained from patients Mel249, Mel499, Mel505, and Mel592 with no history of chemotherapy or immunotherapy. Tissues were processed within 30 minutes following surgical removal. Each specimen was divided into three parts under sterile conditions. One third was fixed in Bouin's solution or neutral buffered formalin for histopathologic analysis. One third of unfixed tissue was snap frozen in liquid nitrogen and stored at −80°C. The remaining tissue was used to generate primary cultures of melanoma cells as previously described (21). Melanoma cells (Mel249, Mel499, Mel505, Mel592, Ma-Mel-86a, and UKRV-Mel-15a) were maintained in RPMI 1640/2 mmol/L glutamine (PAA Laboratories, Cölbe, Germany) supplemented with 10% FCS (PAA Laboratories), 100 units/mL penicillin, and 100 μg/mL streptomycin. NKL cells (22), kindly provided by M.J. Robertson (Indiana University School of Medicine, Indianapolis, IN), were cultured in melanoma cell medium supplemented with interleukin 2 (200 units/mL). Patients' peripheral blood mononuclear cells were separated from heparinized blood by Ficoll-Hypaque (Pharmacia, Uppsala, Sweden) density gradient centrifugation (400 × g for 30 minutes) and kept frozen under viable conditions until use. Blood and tumor donations from patients were approved by the Institutional Review Broad and informed consent was given by all patients.

Immunohistochemistry. Immunohistochemical staining of cryostat tissue sections by the alkaline phosphatase-anti–alkaline phosphatase method was done as previously described (23). The anti–HLA class I monoclonal antibody (mAb) W6/32 was purchased from American Type Culture Collection (Rockville, MD). The anti–HLA class I mAb TP-25 (24), the anti-β2m mAb NAMB-1 (24), and the anti–high molecular weight melanoma-associated antigen mAb 763.74T (25) were kindly provided by Dr. Soldano Ferrone (Department of Immunology, Roswell Park Cancer Institute, Buffalo, NY). The anti–HMB-45 antigen mAb and all reagents for alkaline phosphatase-anti–alkaline phosphatase technique were purchased from DakoCytomation (Glostrup, Denmark). Negative control experiments were done by incubating sections with irrelevant isotype-matched mouse immunoglobulin and by omitting the primary antibody.

Flow cytometry. For detection of surface HLA class I expression on established tumor cell lines, IFN-γ-treated (200 units/mL, 48 hours; R&D Systems, Wiesbaden, Germany) and nontreated cells were labeled with the primary mAb W6/32 followed by staining with the secondary FITC-conjugated goat anti-mouse immunoglobulin (H + L; Dianova, Hamburg, Germany). Control cells were solely incubated with FITC-immunoglobulin. Cells were fixed in 3% formaldehyde and their fluorescence was measured in a FACS Calibur apparatus using the CellQuest software (Becton Dickinson, Franklin Lakes, NJ) for data analysis.

Cytotoxicity assay. Melanoma cells (1 × 10⁶) were labeled with 50 μCi Na₂⁵¹CrO₄ for 1 hour. After washing, labeled target cells were incubated with NKL effector cells for 5 hours at different effector-to-target ratios. Supernatants were analyzed for the presence of radioactivity by gamma counting. Percentage of specific lysis was calculated as [(experimental release − spontaneous release) / (maximum release − spontaneous release)] × 100. Spontaneous release was <10% to 20% of maximum release. Experiments were done in triplicates.

Transfection. Melanoma cells were transfected with a β2m expression plasmid, kindly provided by Dr. Soldano Ferrone, using the Lipofectamine transfection reagent (Invitrogen, Karlsruhe, Germany). Transfections were done according to the protocol of the manufacturer: melanoma cells (2 × 10⁵ per well in a six-well plate) were plated 1 day before transfection and then treated with a mixture of 5 μL Lipofectamine and 1 μg DNA. After 72 hours, Mel499, Mel505, and Mel592 cells were analyzed for transient surface expression of HLA class I molecules by flow cytometry as described above. Where indicated, IFN-γ (200 units/mL) was added to the cells during the last 48 hours of incubation. In the case of Mel249 cells, geneticin (800 μg/mL) was added to the medium 2 days after plasmid transfection. Six days after plasma transfection, cells were harvested and incubated for 48 hours either in the absence or presence of IFN-γ (200 units/mL) for subsequent analysis of HLA class I expression.

To obtain additional information about the efficieny of transfection, melanoma cells were once transfected in parallel with the pEGFP-N3 expression plasmid (Clontech Laboratories, Heidelberg, Germany) encoding the enhanced green fluorescent protein and the β2m expression plasmid. Flow cytometry analysis after 72 hours showed that the number of enhanced green fluorescent protein–expressing melanoma cells was in the range of the number of HLA class I–expressing cells, suggesting that all β2m-transfected cells reexpressed HLA class I molecules at the cell surface.

PCR. For RT-RCR, total RNA was isolated from tumor cells with the RNeasy kit (Qiagen, Hilden, Germany) and polyadenylate RNA was reverse transcribed into cDNA using the first strand cDNA kit (Roche Diagnostics, Mannheim Germany) following the instructions of the manufacturer. Specific amplification of β2m cDNA (sense primer, 5′-cgagatgtctcgctccgtgg-3′; antisense primer, 5′ ataacctctagaacctccatgatgctgcttaca-3′) was carried out in a 30-cycle PCR using the proofreading polymerase Expand High Fidelity (Roche Diagnostics). PCR products were run on a 1% agarose gel, stained with ethidium bromide, and visualized. Reverse transcription-PCR products were cloned into pCR2.1 (Invitrogen) and sequencing of the insert was done by MWG Biotech (Ebersberg, Germany).

Genomic DNA was isolated from tumor cells with the QIAamp kit (Qiagen) following the instructions of the manufacturer. Amplification of exon I of the β2m gene from genomic DNA (sense primer, 5′-ctctaacctggcactgcgtcg-3′; antisense primer, 5′-ttggagaagggaagtcacggag-3′) was carried out in a 35-cycle PCR using the Taq polymerase (Bioron, Ludwigshafen, Germany). PCR products were run on a 1% agarose gel, stained with ethidium bromide, and visualized.

Microsatellite analysis. DNA from matching pairs of melanoma cells (Mel505, Mel592, and Mel249) and peripheral blood mononuclear cells was isolated using standard protocols (for Mel499, no peripheral blood mononuclear cells were available). LOH analysis of chromosome 15 was carried out by PCR amplification of five microsatellite markers (D15S117, D15S126, D15S1015, D15S818, and D15S642) using appropriate Cy5-labeled primer sets. Microsatellite sequences were amplified according to the following PCR protocol: initial denaturation step (5 minutes at 94°C); 30 cycles of 1 minute at 94°C, 1 minute at 52°C (D15S126 and D15S1015) or 57°C (D15S117, D15S818, and D15S642), and 1 minute at 72°C; and final extension step (10 minutes at 72°C). PCR products were analyzed using an ALF express II device (Amersham Pharmacia, Uppsala, Sweden). After separation of the fluorescence-labeled DNA fragments in an 8% polyacrylamide gel, evaluation of the laser-detected DNA molecules was done using the Fragment Analyzer software (Amersham Pharmacia).

Fluorescence in situ hybridization. Metaphase spreads were prepared from each melanoma cell line by addition of colcemide, followed by hypotonic treatment and final fixation in Carnoy's fixative (methanol/acetic acid, 3:1) according to standard procedures. The air-dried metaphase-spread preparations were stored at −20°C until further use. A SpectrumGreen-labeled chromosome 15q painting probe was used in combination with a SpectrumOrange-labeled probe (Vysis, Downers Grove, IL) staining the subtelomeric region of chromosome 15q. In a second setup, we simultaneously hybridized a centromere 15–specific, SpectrumGreen-labeled sample and the SpectrumOrange-labeled telomere 15q–specific DNA probe to the metaphase spreads. All fluorescence in situ hybridization (FISH) probes were purchased from Vysis. The metaphase slides and the FISH probes were simultaneously denaturated for 5 minutes at 75°C on a heating plate. Hybridization was allowed overnight in a humid chamber at 37°C. The coverslip was removed and the slides were washed twice in 0.05× SSC at 42°C, shortly rinsed in 2× SSC/0.3% NP40 (Sigma, Deisenhofen, Germany). Finally, the slides were mounted with Vectashield/4′,6-diamidino-2-phenylindole (Vector Laboratories, Burlingame, CA). Hybridization signals were analyzed using an epifluorescence microscope (Zeiss, Oberkochen, Germany) equipped with appropriate filter sets. A minimum of 10 metaphase spreads were analyzed per case.

Multicolor FISH. Multicolor FISH, introduced in 1996 by Speicher et al. (26), allows the investigation of complex karyotypes by assigning each chromosome pair a unique color code. This is achieved by simultaneous hybridization of combinatorial-labeled chromosome-specific DNA probes. We applied multicolor FISH to unveil chromosomal rearrangements with participation of chromosome 15 in the cell lines Mel505 and Mel592. Hybridization of the multicolor FISH probe (Vysis) to the metaphase spreads was done as previously described (27). Hybridization signals were analyzed with a fluorescence microscope equipped with a filter set (4′,6-diamidino-2-phenylindole, SpectrumAqua, SpectrumGold, SpectrumGreen, SpectrumRed, and FarRed) and multicolor FISH capture and classification software (SmartCapture and SpectraVysion, Vysis).

Loss of HLA class I expression in melanoma metastases. When tumor tissues from the four melanoma patients, Mel249, Mel499, Mel505, and Mel592, were analyzed by immunohistochemistry, strong expression of the high molecular weight melanoma-associated antigen and HMB-45 proteins could be detected, revealing their melanocytic origin. Interestingly, tumor cells could not be stained either with the antibody W6/32, which binds to a structural epitope formed by the β2m light/HLA heavy chain complexes, or by an antibody which merely detects the β2m molecule (Fig. 1 and data not shown). This strongly suggested that the tumor cells lost surface presentation of HLA class I molecules due to mutations affecting the β2m gene. To analyze this, primary cell lines were established from the same tissue samples. The short-term cultured tumors cells maintained their HLA class I–deficient phenotype even when treated with IFN-γ, an inducer of HLA expression (Fig. 2A). Loss of HLA class I expression renders tumor cells susceptible to recognition by natural killer cells, the cytolytic effectors of the innate immune system. To address this, we coincubated the well-growing cell lines Mel249 and Mel592 with the natural killer cell line NKL. Indeed, the HLA class I–deficient cells, in contrast to HLA class I–positive melanoma cells, were effectively killed by NKL cells (Fig. 2B).

To verify that loss of β2m expression was indeed causative for the total loss of HLA class I surface presentation, melanoma cells were transfected with a β2m expression plasmid. An induction of HLA class I expression could clearly be detected on the population of β2m-transfected cells of all four cell lines. In the presence of IFN-γ, this expression level could further be increased on Mel499 and Mel505 cells (Fig. 2A). Based on these results, we asked which mutational events at the β2m gene and/or chromosome 15 levels led to β2m deficiency.

Analysis of β2m expression at the RNA level. When tumor cells were analyzed for transcription of the β2m gene by reverse transcription-PCR, β2m-specific DNA could be amplified from Mel249 and Mel499 cells (Fig. 3A). Whereas the cDNA product from Mel249 cells apparently corresponded in size to that obtained from positive control cells, two different faint bands, both of higher molecular weight, were obtained from Mel499 cells. Sequence analysis of the PCR products revealed a 2-bp microdeletion in codon 62 (exon II) of the β2m gene for Mel249 cells, causing a frameshift (Fig. 3B) in the gene product. The different β2m-specific cDNAs amplified from Mel499 cells were characterized by the insertion of intron I sequences of different lengths, 27 and 407 bp, respectively, between exon I and exon II. The defect in splicing of the β2m primary transcript was due to a T-to-A transition at position two of the intron I sequence (Fig. 3C), leading to a destruction of the conserved GU element in the donor splice site and to the usage of downstream located cryptic donor splice sites. In contrast to Mel249 and Mel499 cells, no β2m cDNA could be amplified from Mel505 and Mel592 cells, suggesting that either β2m transcription was impaired or that the β2m gene was at least partially deleted. To analyze this, we amplified exon I of the β2m gene from genomic DNA. In accordance to the data obtained by reverse transcription-PCR, a clear β2m-specific DNA product could be amplified only from Mel249 and Mel499 cells but not from Mel505 and Mel592 cells (Fig. 3D), pointing to a deletion of genomic β2m sequences.

Deletion mapping. To determine if loss of chromosome 15 material contributes to the β2m deficiency of the tumor cells, we analyzed the status of heterozygosity of five microsatellite markers, one marker located proximal [D15S117 (15q21.1)] and four markers [D15S126 (15q21.3), D15S1015 (15q22.3), D15S818 (15q24.2), and D15S642 (15q26.3)] located distal to the β2m gene (15q21). Genomic DNA isolated from autologous peripheral blood mononuclear cells served as the matching control. This analysis was done on Mel249, Mel505, and Mel592 cells (no peripheral blood mononuclear cells from donor Mel499). As shown in Fig. 4, LOH for all informative (heterozygote) markers could be detected in Mel249 and Mel505 cells, pointing to an extensive loss of genetic material from one parental chromosome 15 encompassing the β2m gene. Mel592 cells exhibited LOH for the two markers flanking the β2m region but retained heterozygosity for marker D15S818 located distal to the β2m gene.

Detection of chromosome 15 abnormalities. The loss of allelic markers indicated that broad-range deletions within chromosome 15 occurred in all cell lines investigated by LOH analysis. To analyze this on a molecular cytogenetic level, we did FISH studies on metaphase chromosomes employing two combinations of fluorescent labeled DNA probes, either probes specific for the chromosome arm 15q (green) and the telomere of 15q (red) or probes specific for the chromosome 15 centromere (green) and the telomere of 15q (red). Metaphase analysis showed a near-triploid karyotype for Mel249, Mel499, and Mel505 cells and a highly variable karyotype for Mel592 with chromosome number ranging from 86 to 102 per cell. All cell lines were characterized by structural aberrations of chromosome 15 (Fig. 5). In Mel249 cells, FISH analysis revealed one apparently normal chromosome 15, one translocation involving chromosome 15 material, and one intrachromosomal rearrangement of chromosome 15. In Mel505 cells, three copies of a rearranged chromosome 15 could be detected in all metaphases analyzed. These chromosomes were characterized by the presence of 15q telomere-specific signals at both chromosome ends comparable to the intrachromosomal rearrangement observed in Mel249 cells. With respect to the data obtained by PCR and LOH analysis, we suggest that all these abnormal chromosomes in Mel 505 cells originate from the same parental chromosome 15 characterized by a deletion in the β2m region. Besides the abnormal chromosome, translocations of small chromosome 15 fragments to other chromosomes could be detected.

A different pattern of chromosome 15 abnormalities was observed in Mel592 cells. These cells carried apparently normal versions of chromosome 15 and translocations containing chromosome 15 material. The translocated fragments consisted of telomere 15q–specific signals as well as centromere-specific signals each fused to another chromosome (see multicolor FISH analysis). Cell line Mel499 revealed two apparently normal copies of chromosome 15 and two translocations to other chromosomes, both bearing telomere-specific signals.

Analysis of chromosome 15 composition by multicolor FISH. To elucidate interchromosomal and intrachromosomal rearrangements of chromosome 15, we did multicolor FISH analysis on Mel505 and Mel592 cells. As seen in dual-FISH analysis, Mel505 revealed a near-triploid karyotype with three copies of a rearranged chromosome 15. Multicolor FISH data indicated that these three rearranged chromosomes contained solely chromosome 15 material (Fig. 6). Besides this intrachromosomal rearrangement, we found translocations of chromosome 15 material to chromosomes 9, 11, and 12 [translocation t(11;15) not contained in the metaphase shown in Fig. 6]. Mel592 revealed translocations involving chromosome 15 in every metaphase analyzed. Here, chromosome 15 material most frequently translocated to chromosomes 13, 17, and 21. Besides these translocations, two apparently normal copies of chromosome 15 were visible in every metaphase analyzed.

Subsequent analysis in three additional β2m-deficient melanoma cell lines from different patients also revealed chromosome 15 abnormalities associated with extensive loss of genetic material (data not shown), indicating that chromosome 15 instability critically contributes to the loss of β2m expression.

Melanoma tissues often show an intense infiltration of T cells, pointing to a close interaction between tumor cells and immune effectors. Recent studies of adoptive T-cell transfer showed that, in principle, such T cells have the capability to destroy tumor cells in vivo (1, 2). On the other hand, T-cell activity might select for the outgrowth of tumor cells that are no longer targets of host immune effectors due to regulatory and mutational events affecting genes involved in antigen processing and presentation.

In case of melanoma, alterations in the HLA class I phenotype can be detected frequently in tumor tissues and cell lines (28, 29). A low HLA class I expression profile of melanoma cells is known to be caused by reversible regulatory defects leading to the coordinate down-regulation of genes encoding components of the antigen processing and presenting machinery (30, 31). On the other hand, irreversible alterations in the HLA class I phenotype are mainly caused by mutational events affecting the HLA heavy chain genes located on chromosome arm 6p (11, 12, 28). Mutations being responsible for the irreversible total loss of HLA class I expression in melanoma have been identified in two different genes encoding transporter associated with antigen presentation 1 (32) and β2m (9, 10, 13).

We analyzed the four cell lines Mel249, Mel499, Mel505, and Mel592, derived from the HLA class I–negative metastatic tissue specimens of different melanoma patients, for the molecular mechanisms mediating HLA class I loss and detected β2m gene mutations in all candidates. Mel249 cells were characterized by a frameshift mutation consisting of a 2-bp microdeletion in codon 62 of exon II of the β2m gene. The majority of β2m gene mutations detected in different tumor types thus far is concentrated in exon I and exon II, but a repetitive sequence of CT dinucleotides in exon I has been described as a mutational hotspot region of the β2m gene (17, 19).

A novel type of mutation was detected in Mel499 cells, which were characterized by a T→A transition in the second nucleotide of intron I. This mutation destroyed the GU donor splice consensus site of intron I. Therefore, the pre-mRNA of the β2m gene was differently spliced, using two cryptic sequences within intron I as donor splice sites, whereas the acceptor splice site was maintained. As a consequence, two different β2m-specific cDNAs could be amplified from Mel499 cells, one containing an insert of the first 27 bp of intron I between exon I and exon II, the second containing an insert of the first 407 bp. Whereas a mutation in the donor splice site has not been described thus far, Hicklin et al. (18) characterized the mutation of an acceptor splice site in one melanoma cell line, also leading to a differently spliced β2m mRNA.

In contrast to Mel249 and Mel499 cells, no β2m-specific cDNA could be amplified from Mel505 and Mel592 cells. Even PCR amplification of the β2m exon I sequence from genomic DNA of Mel505 and Mel592 cells failed, pointing to large deletion of genomic β2m sequences probably comparable to that previously described by D'Urso et al. (33) for the FO-1 melanoma cell line.

Interestingly, in each of the β2m-deficient melanoma cell lines analyzed thus far, only a single type of mutation in the β2m gene was characterized. The coexistence of two different gene mutations, affecting both β2m alleles, has not been detected in melanoma but was recently shown to be of relevance for tumors of the microsatellite mutator phenotype, such as colorectal cancer in which an inactivation of DNA mismatch repair genes leads to an accumulation of gene mutation (34, 35). Because melanoma is proposed to be a microsatellite mutator phenotype–negative tumor, we asked whether chromosome 15 instability contributes to β2m deficiency.

To answer this, we subsequently did analysis at the chromosome 15 level of Mel249, Mel499, Mel505, and Mel592 cells. To determine if loss of chromosome 15 material occurred within these cells, we first analyzed the status of heterozygosity of five microsatellite markers of chromosome arm 15q, one of them flanking the β2m gene on the proximal chromosome arm and the others located on the distal arm. In Mel249 and Mel505 cells, LOH was detectable for all informative markers. Although it cannot be ruled out that deletions were distributed on both parental chromosomes, it is more likely that this LOH pattern was due to an extensive loss of genetic material from one parental chromosome 15.

Loss of microsatellite DNA was also observed in Mel592 cells. However, these cells still exhibited retention of heterozygosity for marker D15S818 (15q24.2) located distal to the β2m gene, suggesting that the deletion in Mel592 cells was less extended than those in Mel249 and Mel505 cells but clearly encompassed the β2m gene. This assumption was corroborated by FISH analysis employing differently fluorescence-labeled DNA probes. Analysis on Mel592 metaphase chromosomes revealed the presence of two different chromosome 15 signals: one originating from the apparently intact chromosome 15, the second derived from translocated chromosome 15 material. The translocated fragment consisted not only of 15q telomere and proximal 15q material but also of centromere-specific DNA. Multicolor FISH studies revealed that in different metaphases, this fragment was translocated to different chromosomes: t(13;15), t(15;17), and t(15;21).

FISH analysis of Mel249 and Mel505 cells showed that an intrachromosomal rearrangement of the same type was observed in both cells: an abnormal chromosome 15 containing a 15q telomere signal at both chromosome ends. All Mel505 metaphases only contained this abnormal chromosome characterized by a deletion encompassing the β2m gene. In contrast, in Mel249 cells, the rearranged chromosome 15 coexisted with apparently normal chromosome 15, the last most probably containing the mutated β2m gene.

Recently, multicolor FISH and 4′,6-diamidino-2-phenylindole banding analysis of chromosomal aberrations in seven metastatic melanoma cell lines showed that >50% of these were characterized by a breakpoint cluster/region on chromosome arm 15q21-26 (36). However, chromosome 15 abnormalities are not known as early genetic imbalances in melanoma; they might occur at later tumor stages. In this regard, two pathways of melanoma karyotype evolution based on the meta-analysis of karyotype data from 340 tumors were recently proposed by the group of Felix Mitelmann (37, 38): one pathway is initiated with a gain of the chromosomal region 6p, the second with a loss of chromosome 3. Within this model, chromosome 15 imbalances are described as a late change. However, as they coincide with β2m gene mutations, they might have an effect on the interaction of the tumor cells with the host immune effectors even in this late phase. Melanoma cells which have lost HLA class I expression are no longer susceptible to CTL activity but might be targets of natural killer cells. Therefore, it might be of interest to mobilize both effector populations against the tumor during immunotherapy.

We thank Ivana Zanette for her excellent technical assistance in immunohistochemistry.