Selective Histocompatibility Leukocyte Antigen (Hla)-A2 Loss Caused by Aberrant Pre-mRNA Splicing in 624mel28 Melanoma Cells

Recombinant human IFN-γ was obtained from Hoffmann-La Roche, Inc.

The human melanoma cell lines 624MEL28 and 624MEL38 17 and the B lymphoid cell line Raji were cultured at 37°C in a 5% CO₂ atmosphere in RPMI 1640 medium (GIBCO BRL) supplemented with 10% Serum plus (Hazelton Biologics, Inc.).

The mAb W6/32, which recognizes a monomorphic determinant expressed on β₂-μ–associated HLA-A,B,C heavy chains; the mAb LGIII-220.6, which recognizes a determinant preferentially expressed on β₂-μ–associated HLA-A heavy chains; the mAb H2-89.1, which recognizes a determinant preferentially expressed on β₂-μ–associated HLA-B heavy chains; the anti–HLA-A2,A28 mAbs CR11-351, HO-4, and HO-5; the anti–HLA-A2,B17 mAbs HO-2 and MA2.1; the anti–HLA-B7 cross-reacting group mAb KS4; and the mAb HC-A2, which recognizes a determinant restricted to β₂-μ–free HLA-A heavy chains, were developed and characterized as described elsewhere 22,23,24,25,26,27. The rabbit anti–β₂-μ–free HLA class I heavy chain serum R5996-4 28 and the rabbit anti-HLA class I heavy chain cytoplasmic tail serum were obtained from Dr. N. Tanigaki (Roswell Park Cancer Institute) and Dr. H.L. Ploegh (Harvard Medical School, Boston, MA), respectively. Purified rabbit anti–mouse Ig antibodies and FITC-conjugated goat anti–mouse Ig antibodies (FITC-GAM) were purchased from Jackson ImmunoResearch Laboratories.

The HLA class I–specific synthetic oligonucleotide primers and probes listed in Table were synthesized on a BioSearch Cyclone DNA synthesizer (MilliGen/BioSearch), with the exception of those obtained from Dr. S.Y. Yang (Memorial Sloan-Kettering Cancer Center, New York). Oligonucleotide probes were radiolabeled with [γ-³²P]ATP (>5,000 Ci/mmol; Nycomed Amersham plc) in the presence of T4 polynucleotide kinase 29.

The plasmid RSV.5neo-HLA-A2.1 was obtained from Dr. P. Cresswell (Yale University School of Medicine, New Haven, CT). The HLA-A2 cDNA for synthesis of RNA probe was constructed by cloning a 500-bp EagI/PstI fragment from plasmid RSV.5neo-HLA-A2.1 into pCR-Script™ SK(+) cloning vector in an antisense orientation. The EagI/PstI fragment contains most of the exon 2 and exon 3 of the HLA-A2 cDNA. The RNA probe was synthesized by in vitro transcription using MAXIscript™ T₇ kit (Ambion, Inc.) with [α-³²P]UTP (800 Ci/mmol; Amersham Pharmacia Biotech) following the manufacturer's instructions. The synthesized RNA probe was purified through a 5% polyacrylamide gel and eluted following the procedure recommended by the manufacturer. The HLA-A2 cDNA containing the whole coding region and the HLA-A2 cDNA lacking exon 2 to be used for in vitro translation were amplified by PCR from plasmid RSV5neo-HLA-A2.1 and melanoma 624MEL28 cells, respectively, using the primers 5P2 and 3P2. PCR products were cloned into SalI/HindIII sites of pCR-Script™ SK(+) cloning vector.

The plasmid pHLA-A2, which carries the entire HLA-A2 gene including its regulatory sequences, and the neomycin resistance gene 30 were obtained from Dr. J.L. Strominger (Harvard University, Cambridge, MA).

Indirect immunofluorescence (IIF) staining was performed as described elsewhere 4. After staining, cells were analyzed by cytofluorometry on a FACS^® analyzer (Becton Dickinson). Results are expressed as log fluorescence intensity.

Radiolabeling of cells, indirect immunoprecipitation, SDS-PAGE under reducing conditions, isoelectric focusing (IEF), and fluorography were performed as described elsewhere 10,31.

Western blotting was performed as described elsewhere 32, with the modification that antibodies bound to proteins transferred to filters were detected using the ECL™ Western blotting detection reagents kit (Amersham Pharmacia Biotech) following the manufacturer's instructions.

Plasmid pHLA-A2 was transfected into cells using the electroporation method as described 10. After transfection, cells were cultured for 2 wk in medium supplemented with G418-sulfate at the final concentration of 0.4 mg/ml. Cell colonies were picked up and expanded in medium supplemented with G418-sulfate at the final concentration of 0.2 mg/ml.

cDNA was reverse transcribed from total RNA as described 33. Genomic DNA was prepared from cells using the cell lysis method 34. Amplification of cDNA and genomic DNA was performed as described elsewhere 15 using a DNA Thermal Cycler (Perkin-Elmer Cetus).

PCR products were size fractionated, transferred to nylon membranes, and hybridized with probes as described elsewhere 15.

DNA sequencing was performed as described elsewhere 35, using the Sequenase version 2.0 kit (United States Biochemical Corp.). PCR products were directly sequenced using products recovered from an agarose gel using the Geneclean II kit (Bio 101 Inc.). After denaturation at 98°C for 10 min in the presence of 10% DMSO, PCR products were cooled in ethanol/dry ice and then sequenced.

The RNase protection assay was performed using the RPA II™ Ribonuclease protection assay kit (Ambion, Inc.) following the manufacturer's instructions, except for increasing the hybridization temperature to 45°C.

HLA-A2 mRNA was translated in vitro using the TNT^® T₇ coupled reticulocyte lysate system (Promega Corp.) following the manufacturer's instructions. The translation products were analyzed by SDS-PAGE under reducing conditions.

The clone 624MEL28 was isolated by limiting dilution from the melanoma cell line 624MEL (HLA-A2, -A3, -B7, -B14, -Cw7, -Cw8), which had been established from a metastatic lesion 17. This cell line was subcloned at the 35th passage, since the broad profile of the IIF staining with mAbs suggested marked heterogeneity in HLA-A2 antigen expression. The clone 624MEL28 was not stained in IIF by anti–HLA-A2,B17 mAb HO-2 and anti–HLA-A2,A28 mAbs HO-4 and HO-5, all of which stained the autologous HLA-A2 antigen–bearing clone 624MEL38 (Fig. 1). The lack of staining of the 624MEL28 cells by anti–HLA-A2 mAb is not due to loss of an antigenic determinant, since the three mAbs used recognize distinct determinants on HLA-A2 heavy chains (Ferrone, S., unpublished results). The HLA-A2 antigen loss is selective, since, like the autologous clone 624MEL38, 624MEL28 cells were stained strongly by anti–HLA-A,B,C mAb W6/32 and by anti–HLA-A mAb LGIII-220.6, and weakly by anti–HLA-B mAb H2-89.1 and anti–HLA-B7 cross-reacting group mAb KS4 (Fig. 1). It is noteworthy that the intensity of staining by the latter six mAbs of the clone 624MEL28 is comparable to that of the autologous clone 624MEL38. Furthermore, IFN-γ did not induce HLA-A2 antigen expression by 624MEL28 cells, since they were not stained by anti–HLA-A2 mAbs after a 72-h incubation at 37°C with IFN-γ (100 U/ml; Fig. 1). These findings indicate that 624MEL28 cells have selectively lost HLA-A2 antigens.

624MEL28 cells used as a source of HLA class I antigens for immunoprecipitation experiments were treated with IFN-γ in order to increase the level of HLA class I antigen expression. No component was detected by IEF analysis in the immunoprecipitates with anti–HLA-A2, A28 mAb CR11-351 and with anti–HLA-A2,B17 mAb MA2.1 from intrinsically radiolabeled 624MEL28 cells (Fig. 2). Furthermore, HLA-A2 heavy chains were not detected by IEF in the immunoprecipitates with mAb W6/32 and with rabbit antiserum R5996-4. However, the latter immunoprecipitates contain HLA-A3, -B7, and -B14 heavy chains. These immunochemical findings corroborate the results of binding assays and indicate that 624MEL28 cells do not synthesize wild-type HLA-A2 heavy chains.

To investigate whether HLA-A2 antigen expression by 624MEL28 cells could be reconstituted by transferring a wild-type HLA-A2 gene, cells were transfected with a plasmid containing both a wild-type HLA-A2 gene and a neomycin resistance gene. Two clones selected in medium supplemented with G418 were both stained in IIF by anti–HLA-A2,B17 mAb HO-2 (Fig. 3). The staining intensity was increased when cells were incubated for 72 h at 37°C with IFN-γ (100 U/ml; data not shown). These results indicate that the lack of HLA-A2 antigen expression by 624MEL28 cells is caused by structural abnormality(ies) in the HLA-A2 gene.

Reverse transcription (RT)-PCR amplification of the mRNA corresponding to a region spanning from the middle of exon 1 to exon 3 of the HLA-A2 gene yielded a 218- and a 729-bp product from 624MEL28 cells. The latter product is 241 bp larger than the one amplified from 624MEL38 cells with the expected size of 488 bp, whereas the former product is 270 bp smaller. Both RT-PCR products are HLA-A2 specific, since they hybridized with the HLA-A2,A69–specific probe 95V (Fig. 4 A). These results indicate that there are two forms of the HLA-A2 mRNA in 624MEL28 cells, one with a 270-base deletion and the other with a 241-base fragment inclusion. The size of deleted and inserted fragments is compatible with that of exon 2 and intron 2 of the HLA-A2 gene, respectively, suggesting the lack of the sequence corresponding to exon 2 or insertion of intron 2 in HLA-A2 mRNA. These possibilities were corroborated by the generation of a 571-bp product, but not of a 330-bp product when cDNA corresponding to exons 2 and 3 of the HLA-A2 gene was isolated from 624MEL28 cells and amplified by PCR (Fig. 4 B). Exon 2 skipping and insertion of intron 2 in the two aberrant forms of HLA-A2 mRNA identified in 624MEL28 cells were conclusively proven by DNA sequencing of the 729-, 571-, and 218-bp RT-PCR products. Only the sequences of exons 1 and 3 of the HLA-A2 gene were found in the 218-bp RT-PCR product (Fig. 5). The sequence of intron 2 of the HLA-A2 gene was found in both the 729- and the 571-bp RT-PCR products. It is noteworthy that the sequence of intron 1 was not detected in the 729-bp RT-PCR product, indicating that the latter was derived from the mature HLA-A2 mRNA with intron 2 retention. A T to A base substitution was found at position 2 in intron 2. This mutation results in a splicing defect, since the U at this position in pre-mRNA is required for spliceosome recognition for pre-mRNA splicing. As shown in Fig. 6, the intron 2 retention causes a reading frameshift, which introduces the stop codon TGA at position 6–8 of the unspliced intron 2. Thus, the synthesis of the HLA-A2 polypeptide encoded by the large HLA-A2 transcript is blocked. In contrast, the small HLA-A2 transcript is in frame in spite of the exon 2 skipping. Therefore, a truncated HLA-A2 heavy chain lacking the α₁ domain is expected to be synthesized by 624MEL28 cells. No additional mutation was found in the remaining sequence of HLA-A2 cDNA.

To define the molecular lesion responsible for exon 2 skipping in the HLA-A2 mRNA, the region from exon 1 to exon 3 of the HLA-A2 gene was amplified from genomic DNA of 624MEL28 cells using HLA-A2–specific primers. Only the PCR products with the expected size of 980 bp were amplified from 624MEL28 cells as well as from the autologous 624MEL38 cells, which express HLA-A2 antigens (data not shown). These results rule out the presence of additional HLA-A2 gene copies with exon 2 deletion in 624MEL28 cells. Furthermore, DNA sequencing of the four clones containing the 980-bp PCR product identified a T to A substitution at position 2 of the 5′ splice donor site in intron 2 (Fig. 7). This mutation is the same mutation found in the large mutated HLA-A2 cDNA. No mutation was found at the 3′ acceptor site in introns 1 and 2 of the HLA-A2 gene. Therefore, the mutation at position 2 of intron 2, which inactivates the 5′ splice donor site, is responsible for both exon 2 skipping and intron 2 retention in the HLA-A2 mRNA in 624MEL28 cells.

Although the HLA-A2 mRNA with exon 2 skipping in 624MEL28 cells was expected to synthesize a truncated polypeptide, such a polypeptide was not detected by SDS-PAGE analysis of antigens immunoprecipitated from intrinsically labeled 624MEL28 cells by the anti–β₂-μ–free HLA class I heavy chain xenoantiserum R5996-4 or by an anti-HLA class I heavy chain cytoplasmic tail xenoantiserum. Furthermore, the truncated HLA-A2 polypeptide was not detected by testing a 624MEL28 cell extract with the two xenoantisera in Western blotting. To determine whether these results reflect a dramatic decrease in the level of the HLA-A2 mRNA lacking the exon 2 sequence because of the splicing defect, its steady state mRNA level in 624MEL28 cells was measured using the RNase protection assay. A ³²P-labeled antisense HLA-A2 RNA complementary to 496 bp of the region spanning most of exons 2 and 3 of the HLA-A2 gene was used as a probe. As shown in Fig. 8, digestion with RNase of this probe protected by mRNA isolated from 624MEL28 cells yielded a 268- and a 496-base fragment. The 268-base fragment, which has the same size as exon 3 in the probe, is derived from the probe protected from RNase digestion by the HLA-A2 mRNA lacking the exon 2 sequence. The protection of the probe by mRNA from 624MEL38 cells did not yield this fragment. The 496-base fragment is likely to derive from the probe protected by the HLA-A2 mRNA with unspliced intron 2, since a fragment with the same size was generated from the probe protected by the RNA synthesized in vitro by an HLA-A2 genomic DNA fragment containing exon 2, intron 2, and exon 3. This fragment is unlikely to derive from protection by a wild-type HLA-A2 mRNA from 624MEL28 cells, since such an mRNA was not detected by RT-PCR in these cells. The level of the 268- and 496-base fragments protected by mRNA isolated from 624MEL28 cells is very low, at least 20-fold less than that of the 496-base fragment derived from the probe protected by the wild-type HLA-A2 mRNA isolated from control 624MEL38 cells. These results suggest that the steady state level of HLA-A2 mRNA with exon 2 skipping is too low to produce a detectable level of truncated HLA-A2 heavy chain.

Translation experiments using a transcription/translation-coupled system tested whether the HLA-A2 mRNA lacking the exon 2 sequence can synthesize proteins in vitro. To this end, the mutated HLA-A2 cDNA containing the whole coding sequence, except the exon 2–encoded region, was isolated from 624MEL28 cells and used for in vitro mRNA synthesis. A 35-kD polypeptide, ∼10 kD smaller than the wild-type, was synthesized by mutated HLA-A2 mRNA (Fig. 9 A). The loss of the α₁ domain, which is encoded by exon 2, accounts for the reduction in size of the synthesized polypeptide. This interpretation is consistent with the results of the SDS-PAGE analysis of the translation products immunoprecipitated by three xenoantibodies. The truncated polypeptide was not immunoprecipitated by mAb HC-A2, which recognizes a determinant expressed on the α₁ domain of β₂-μ–free HLA class I heavy chains 36, but was immunoprecipitated by the anti–β₂-μ–free HLA class I heavy chain xenoantiserum R5996-4 (Fig. 9 B) and by the anti-HLA class I heavy chain cytoplasmic tail xenoantiserum (data not shown). These results indicate that the HLA-A2 mRNA with exon 2 skipping synthesizes in vitro a truncated HLA class I heavy chain lacking the α₁ domain. It is noteworthy that the intensity of the 35-kD translation product is about twice as low as that of the 44.6-kD translation product (Fig. 9A and Fig. C), although the extent of RNA synthesis by the mutated and the wild-type HLA-A2 cDNA is similar, as measured by the level of [³²P]UTP incorporation (Fig. 9 D). Therefore, the efficiency of in vitro translation of the HLA-A2 mRNA with exon 2 skipping is lower than that of the wild-type HLA-A2 mRNA.

This study has characterized for the first time the molecular lesion underlying the spontaneous selective loss of an HLA class I allele by melanoma cells. HLA-A2 antigen has been selectively lost by melanoma cells 624MEL28 because of a defect in the splicing of HLA-A2 pre-mRNA. Pre-mRNA is spliced into two aberrant forms of mature mRNA, one with exon 2 skipping and the other with intron 2 retention. The latter mRNA form is distinct from pre-mRNA since it does not contain the sequences of introns 1, 3, 4, 5, 6, and 7. The mRNA with intron 2 retention cannot be translated into a wild-type HLA-A2 heavy chain, since the intron 2 retention introduces a premature stop codon at position 6–8 of the unspliced intron. The mRNA form with exon 2 skipping is expected to be translated into a truncated HLA-A2 heavy chain lacking the α₁ domain. Although detected in in vitro translation experiments, such a polypeptide was not detected in 624MEL28 cells by testing cell extracts with xenoantibodies to distinct domains of HLA class I antigens in indirect immunoprecipitation and Western blotting assays. The latter finding may reflect the very low steady state level of the mRNA form with exon 2 skipping, as measured by the RNase protection assay, and the low translation efficiency, as suggested by the results of in vitro translation experiments.

The abnormal splicing of the HLA-A2 mRNA in 624MEL28 cells is caused by a mutation at position 2 of the 5′ splice donor site in intron 2, which is highly conserved in mammalian cells 37. This mutation inactivates the 5′ splice donor site, since pre-mRNA splicing requires the U at this position for spliceosome recognition 38. The importance of the conserved nucleotide at this position has been proven with site-directed mutagenesis experiments in which replacement of the T with a purine prevented in vivo splicing of the 12S mRNA 39 and in vitro splicing of β-globin mRNA 40. The base substitution at the 5′ splice donor site in intron 2 of the HLA-A2 gene results either in exon 2 skipping or in intron 2 retention in the two aberrant transcripts of the HLA-A2 gene in 624MEL28 cells, since no HLA-A2 gene copy with exon 2 deletion or with normal sequence at the 5′ splice donor site of intron 2 was detected in 624MEL28 cells.

The occurrence of either exon skipping or intron retention in pre-mRNA splicing caused by the same mutation at the 5′ splice donor site of an intron is an uncommon phenomenon. To the best of our knowledge, this abnormality has been detected only in the type III procollagen gene in members of a family with aortic aneurysms and easy bruisability 41 and in the β-hexoaminidase α chain gene in an Ashkenazi Jewish patient with Tay-Sachs disease 42. In both cases, as in our own, exon skipping and intron retention were caused by a mutation in the splice site at the 5′ end of the mutated intron. Exon skipping is a predominant phenotype caused by mutations at the 5′ splice donor site of an intron in mammalian cells 43. The mutation of a 5′ splice donor site inhibits the interaction of spliceosome with the 3′ and 5′ splice sites across the exon so that it blocks exon definition 44. If no cryptic site is activated, splicing of the exon leads directly to exon skipping. In contrast, intron retention is a rare phenomenon, found in only ∼6% of the cases with alternative splicing in mammalian cells 43. In the splicing of a small intron, the spliceosome uses the intron as the initial mode for selection of splice sites. A mutation of a 5′ splice donor site in a small intron leads to intron retention 45. The mutated intron 2 of the HLA-A2 gene in 624MEL28 cells is a small intron, as it is ∼241 bp in size. Thus, it is likely that “exon and intron definition,” two recognition mechanisms, are used for selection of splice sites in the splicing of HLA-A2 pre-mRNA.

The aberrant splicing of the HLA-A2 mRNA in the melanoma cell line 624MEL28 is different in several respects from the alternative splicing that results in skipping of exon 5 in the mRNA for HLA class I heavy chains 46 and of exon 3 or of both exons 3 and 4 in the mRNA for HLA-G heavy chains 47. First, exon 5 skipping may occur in the mRNA for both HLA-A and -B locus gene products in the same cell line. Second, more than one exon may be skipped in the mRNA for HLA-G heavy chain. Finally, normally spliced mRNA encoding the various antigens is more abundant than the corresponding mRNA resulting from alternative splicing. In contrast, in 624MEL28 cells exon 2 is skipped only in the mRNA transcribed by a mutated HLA class I heavy chain gene. No normally spliced HLA-A2 mRNA was detected in 624MEL28 cells.

To the best of our knowledge, the molecular lesion of the HLA-A2 gene identified in 624MEL28 cells is the first to have been characterized in a melanoma cell line with a spontaneous selective HLA class I allele loss. This lesion is distinct from that found in the γ-irradiation–induced HLA-A2 loss mutant melanoma cell line SK-MEL-29.1.22 in which a partial deletion of the HLA-A2 gene results in its transcriptional blockade 15. Furthermore, the molecular defect in 624MEL28 cells is different from the lesions underlying HLA class I allelic losses in other malignant cells. Deletion and transcriptional downregulation of the HLA-A11 gene cause selective loss of this allele in a colon carcinoma 48 and in a Burkitt's lymphoma 49 cell line, respectively. Furthermore, mutations in the HLA class I gene itself or in the upstream promoter region are likely to underlie the selective loss of HLA class I alleles described in two colon and two cervical carcinoma cell lines 50,51. It is noteworthy that one single mutational event is sufficient to cause the selective loss of an HLA class I allospecificity. In contrast, at least two mutational events that inactivate the two B₂m genes present in a cell are required to cause total HLA class I antigen loss. This difference may account for the higher frequency of a selective HLA class I allele loss than of total HLA class I antigen loss in melanoma cells 52.

Preliminary results suggest that the HLA-A2 antigen loss 624MEL28 melanoma cells do not induce an MAA-specific CTL response restricted by the expressed HLA class I alleles. These findings, which parallel similar data in mouse model systems 53,54, could account for the escape of HLA-A2 antigen loss melanoma cells from CTL-mediated recognition and for the high frequency of selective HLA-A2 loss in metastatic melanoma lesions 1,4. The negative impact of these variants on the outcome of T cell–based immunotherapy emphasizes the need to design strategies to induce MAA-specific CTL responses restricted not only by HLA-A2 antigens, but also by the other HLA class I alleles present in HLA-A2–positive patients with melanoma. An alternative, although not exclusive, strategy is to combine T cell–based immunotherapy with immunotherapeutic modalities that are not negatively affected by HLA class I antigen loss by melanoma cells.

The authors acknowledge the secretarial assistance of Ms. Harriett V. Harrison, Ms. Donna James, Ms. Elba Osorio, and Ms. Derina McDonald. The authors also thank Dr. H.L. Ploegh, Harvard Medical School, Boston, MA; Dr. J.L. Strominger, Harvard University, Cambridge, MA; Dr. N. Tanigaki, Roswell Park Cancer Institute; Dr. P. Cresswell, Yale University School of Medicine, New Haven, CT; and Dr. S.Y. Yang, Memorial Sloan-Kettering Cancer Center, New York, NY, for providing antibodies, oligonucleotide primers, and probes.

This work was supported by U.S. Public Health Service grant CA67108 awarded by the National Cancer Institute, Department of Health and Human Services.