Complete loss of HLA class I antigen expression on melanoma cells: a result of successive mutational events

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-gamma 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.

Analysis of HLA-E expression in human tumors

Downregulation of MHC class Ia molecule expression is a widespread mechanism used by tumor cells to escape antitumor T-cell-mediated immune responses. However, it is not known why NK cells cannot lyse these MHC class-Ia-deficient tumor targets. Tumors must select additional routes of escape from NK cells. An attractive hypothesis is that the aberrant expression of nonclassical HLA class Ia molecules in tumors provides the required inhibitory signal to NK cells, rendering tumor cells resistant to NK lysis. To analyze the possible role of HLA-E molecules in providing tumor cells with an NK escape mechanism, we studied the cell surface expression of this HLA class Ib molecule in a variety of tumor cell lines with well-defined HLA class Ia alterations. Tests were done with the monoclonal antibody 3D12 recognizing cell surface HLA-E molecules. Our results indicate that HLA-E was mainly detected in leukemia-derived cell lines. In addition, HLA-E was detected in tumor cell lines of different origin. This expression was related with the availability of free beta(2)-microglobulin (beta(2)m) in the cytoplasm of tumor cells. Expression was detected in tumor cell lines showing an imbalance in heavy chain/beta(2)m expression, particularly in tumor cell lines with alterations in the expression of heavy-chain genes. Several lines of evidence favor these conclusions: (1) In the FM55 and NW145 melanoma tumor systems, the reduction in HLA class Ia expression paralleled the increased cell surface detection of HLA-E. (2) A cervical tumor (808) and a melanoma cell line (R22.2) expressing a single HLA-A1 allele also expressed HLA-E. (3) The addition of human beta(2)m to tumor cell lines that expressed the HLA-E(G) allele increased HLA-E cell surface expression. (4) There was no HLA-E cell surface expression in tumor cell lines with total loss of HLA class Ia expression, including cell lines with low transcription of HLA class I heavy chains or with beta(2)m mutations. Our findings suggest that the biological consequences of these cumulative genetic and molecular changes in tumor cells lead to the appearance of HLA-E in a limited number of tumor cell lines with peculiar phenotypic and genotypic characteristics, namely: HLA-class Ia downregulation, free beta(2)m and HLA-E(G) genotype. The aberrant HLA-E expression might be of particular biological relevance in those HLA tumor phenotypes that express a single HLA-A allele when NK inhibition is markedly reduced due to the downregulation of HLA-B and -C alleles.

Analysis of HLA expression in human tumor tissues

Cancer cells can be detected and destroyed by cytotoxic T lymphocytes in many experimental tumor systems, and--as has been well-documented--in some human tumors. In humans however, most diagnosed tumors are not eliminated by T cells but grow steadily, invading and metastasizing until the host is destroyed. Evidence is accumulating that progressive tumor growth occurs not because the immune system is defective or deteriorated, but because the cancer cell is capable of developing a variety of strategies to escape immune recognition. In addition, cancer cells acquire new biological properties to generate invasive capacity in order to migrate and colonize new tissues. Major histocompatibility complex (MHC) antigens are molecules that are specialized in communicating with the T cell receptor and natural killer (NK) cell ligands. With the former, they use the interaction with peptides derived from processed cellular and exogenous proteins to monitor self and non-self status. With the latter, they determine the degree of activation and killing capacity of NK cells by interacting with NK receptors. Any change in the MHC profile of tumor cells (including classical and nonclassical MHC molecules) may therefore have a profound influence on the immune recognition and immune rejection of cancer cells. We have reviewed the data from our laboratory and other groups, and have presented a standardized procedure for analyzing the MHC profile of human tumors with special emphasis on the quality and laboratory use of the material obtained from microdissected tumor samples. Appropriate tissue processing is of particular relevance, since it is not possible to obtain tumor cell lines from most patients. Oncologists require rapid information on the MHC profile of the tumor if gene therapy is envisaged to restore normal MHC class I gene expression.

Impaired surface antigen presentation in tumors: implications for T cell-based immunotherapy

The identification of tumor-associated antigens has suggested new possibilities for cancer therapy. However, multiple mechanisms may contribute to the ability of tumor to escape antitumor immune responses. Tumor antigen heterogeneity, modulation of HLA expression and immune suppressive mechanisms may occur at any time during tumor cell progression, and can affect the outcome of therapeutic immune intervention. In particular, the appearance of altered HLA class I phenotypes during tumor development may have important biological and medical implications due to the role of these molecules in T and NK cell functions. Exhaustive tumor tissue studies are necessary before deciding whether a particular patient is suitable for inclusion in T cell-based immunotherapy protocols.

Rexpression of HLA class I antigens and restoration of antigen-specific CTL response in melanoma cells following 5-aza-2'-deoxycytidine treatment

Cell surface expression of HLA class I/peptide complexes on tumor cells is a key step in the generation of T-cell-based immune responses. Several genetic defects underlying the lack of HLA class I expression have been characterized. Here we describe another molecular mechanism that accounts for the complete absence of HLA class I molecule expression in a tumor line (MSR3-mel) derived from a melanoma patient. Hypermethylation of the MSR3-mel DNA, specifically of HLA-A and -B genes, was identified, which resulted in loss of HLA class I heavy chain transcription. Treatment of MSR3-mel cells with the demethylating agent 5'-aza-2'-deoxycytidine (DAC) allowed HLA-A and -B transcription, restoring cell surface expression of HLA class I antigens and tumor cell recognition by MAGE-specific cytotoxic T lymphocytes. The MSR3-mel line was obtained from a metastatic lesion of a nonresponding patient undergoing MAGE-3.A1 T-cell-based peptide immunotherapy. It is tempting to speculate that the hypermethylation-induced lack of HLA class I expression is the cause of the impaired response to vaccination. This study provides the first evidence that DNA hypermethylation is used by human neoplastic cells to switch off HLA class I genes, thus providing a new route of escape from immune recognition.

Analysis of HLA class I expression in different metastases from two melanoma patients undergoing peptide immunotherapy

We characterized the HLA class I alterations in five metastases obtained from two patients with melanoma immunized with Melan A/MART-1, tyrosinase and gp100 tumor peptides. All three metastases analyzed in the first patient (NW145) showed a similar HLA class I alteration with a dual population of melanoma cells. One population was HLA class I antigen positive and the other had loss of heterozygosity (LOH) in the short arm of chromosome 6 leading to an HLA haplotype loss (A02011, B4007, Cw1). The absence of HLA-A2 antigen may explain why this patient did not develop HLA-A2 restricted, Melan A/MART-1 specificity immunization, since this HLA molecule is the restriction element for the tumor peptides used. However, this HLA-deficient population was not selected after peptide immunotherapy. The primary tumor in this patient presented LOH in region 6q, but only in the vertical growth phase of the lesion, whereas LOH at 6p was observed only in DNA from metastatic material. The second patient (NW16) also presented two metastatic lesions with an identical HLA molecular defect, i.e. HLA B locus downregulation (HLA B51011: serological B51; B1503: serological B70). One lesion expressed the tumor antigen (Melan A/ MART-1), but the other did not. Interestingly, the antigen-positive metastasis regressed after peptide immunotherapy, whereas the other progressed rapidly. These findings provide the first indication that multiple metastases generated in the same host can have identically altered HLA class I phenotypes.

Beta2-microglobulin gene mutation is not a common mechanism of HLA class I total loss in human tumors

One hundred and sixty-two tumor samples were analyzed for HLA class I expression using immunohistological techniques. HLA class I total loss (phenotype no. I) was detected in 31 cases (19%), comprising 20 colorectal, 3 laryngeal, and 2 bladder carcinomas and 6 melanomas. Twenty-one cases were selected for molecular analysis due to a higher proportion of tumor cells versus stroma cells (75%). We investigated whether beta2-microglobulin mutation was responsible for HLA downregulation. Single-strand conformation polymorphism and sequencing analysis of DNA samples was performed. Alterations were detected only in melanomas M78 (a point mutation in the initiation ATG sequence), M79 (a mutation in codon 31 producing a stop codon), and M34 (a TTCT deletion introducing a termination codon signal). We found no beta2-microglobulin gene mutation in the other 18 samples. Loss of heterozygosity in 15q close to the beta2-microglobulin gene was found in 5 cases. We conclude that HLA class I total loss can frequently occur without beta2-microglobulin gene mutations.

Multiple mechanisms of immune evasion can coexist in melanoma tumor cell lines derived from the same patient

Progressive tumor growth may be associated with suppression of the immune response. Many different mechanisms may contribute to immune evasion. We investigated some of these mechanisms in melanoma cells lines generated from two patients. These cell lines show a complex pattern of altered HLA expression; however, the resulting phenotype did not satisfactorily explain the simultaneous evasion of T and NK cell cytotoxicity. Two additional alterations have now been detected in these melanoma cell lines: (1) resistance to FAS-induced apoptosis caused by defective FAS gene expression, and (2) constitutive expression of immunosuppressive cytokines. Our results show that several of the major mechanisms for immune evasion may coexist in a single tumor. This suggests that tumor progression may give rise to an extremely resistant phenotype, which may be an impediment to some immunotherapeutic strategies. We hypothesize that the simultaneous presence of several mechanisms involved in tumor immune evasion must be the result of progressive selection of characteristics that are advantageous for tumor survival in a competent host. Our findings do not support the possibility that FASL expression is a common mechanism of evasion of immune response in melanoma cells.

Immunoselection by T lymphocytes generates repeated MHC class I-deficient metastatic tumor variants

Alteration of MHC class I molecule expression is a widespread mechanism used by tumor cells to evade T cell responses. It has long been proposed that the origin of these MHC class I-negative or -deficient tumor variants is T cell immune selection. However, there are no experimental or clinical data to substantiate this hypothesis, and this issue is currently the subject of debate. Here we report that an H-2 class I-negative fibrosarcoma tumor clone generated MHC class I-negative spontaneous lung metastases in immunocompetent syngeneic BALB/c mice. Interestingly, the same B9 clone generated MHC class I-positive metastatic nodes, under basal conditions, in athymic nu/nu BALB/c mice. This phenomenon was observed in the metastatic nodules generated after a period of in vivo growth but not in the primary tumors growing locally in the footpad. These findings support the hypothesis that the H-2 phenotype of metastatic nodes is influenced by the T cell repertoire of the host, since in the absence of this T cell pressure (i.e., in nude mice) the metastatic nodes 'recovered' H-2 class I expression. In addition, 2 different phenotypes were found when the metastatic nodules obtained from immunocompetent mice were treated with IFN-gamma. One phenotype, present in 83% of the colonies, was characterized by resistance of the Ld molecule to IFN-gamma induction, due to a deletion involving the Ld gene. The second phenotype (17% of the colonies) was similar to the original B9 clone and was characterized by the response of K, D and L class I genes to IFN-gamma. These data provide evidence that the changes in MHC class I expression during tumor development might not be random but could be predictable.

Multiple mechanisms underlie HLA dysregulation in cervical cancer

The consistent dysregulation of HLA expression in cervical neoplasia is likely to influence the natural history of the disease and prospects for cell-mediated vaccine therapies. We have studied the underlying mechanisms in eight new cervical cancer cell lines derived from primary tumour biopsies. At least five independent mechanisms leading to changes in HLA expression were seen: HLA class I allelic transcription but no protein; abnormal HLA class I allelic transcription; no HLA-B locus transcription; loss of heterozygosity (LOH); no gammaIFN-mediated upregulation of HLA class I expression, and/or no interferon-gamma (gammaIFN)-mediated HLA class II induction. These were evident in different combinations in 7/8 cell lines showing that multiple, mostly irreversible mechanisms not overridden by gammaIFN, are responsible for HLA dysregulation in cervical neoplasia. Point mutations were responsible for lack of HLA-A2 expression in two cases. In cell line 808, the mutation encodes a stop codon in exon 3; in cell line 778, mutation of the first intron acceptor site leads to use of an alternative AG site in exon 2, resulting in a frameshift and a stop codon after the translation of only 38 amino acids. Tumour cells showing specific HLA class I loss may have selective advantage in the face of tumour-specific cytotoxic T cells (CTL). Such immune escape mechanisms present a major obstacle for the success of CTL-mediated therapies in cervical cancer.

Microsatellite instability analysis in tumors with different mechanisms for total loss of HLA expression

Down-regulation of the expression of major histocompatibility complex molecules is a frequent event that is associated with the poor immunogenicity of tumor cells. Acquired resistance to T-cell-based immunotherapy has been associated with loss of functional beta2-microglobulin expression. This anomaly appears to be particularly relevant in tumors exhibiting a defect in DNA-mismatch repair, and induces structural abnormalities in HLA cell-surface expression that are not reversible by cytokine treatment. We examined HLA expression in 118 melanoma, colon or larynx tumors to identify total loss of HLA class I expression with or without somatic beta2-microglobulin gene mutation. Microsatellite instability was investigated in these tumors to determine whether a replication error phenotype (RER+) implied a particular alteration in HLA phenotype. A total of 7.6% of the tumors showed the RER+ phenotype, and 12.7% were HLA-ABC-negative. In the RER+ group, only one tumor was HLA-ABC-negative and no beta2-microglobulin mutation was identified. In contrast, in the HLA-ABC-negative group, only one tumor showed microsatellite instability. None of the three melanomas that contained beta2-microglobulin mutation exhibited the mutator phenotype. These findings suggest that beta2-microglobulin mutation in human melanoma tumors may arise through a mechanism that does not necessarily involve microsatellite instability. Our results also indicate that somatic mutations of the beta2-microglobulin gene are not the main mechanism of total loss of HLA expression.

Chromosome loss is the most frequent mechanism contributing to HLA haplotype loss in human tumors

Loss of heterozygosity (LOH) in the short arm of chromosome 6 (6p) was detected in samples obtained from colon (13.8%), larynx (17.6%) and melanoma (15.3%) tumors. The parallel study of HLA-antigen expression in tumor tissues using locus- and polymorphic-specific antibodies in combination with LOH microsatellite analysis on 6p allowed us to establish that LOH in chromosome 6 is a representative phenomenon in most tumor cells present in a given tumor tissue. In most cases, specific HLA alleles had been lost in a predominant population of tumor cells, indicating that LOH is a non-irrelevant mutation that probably confers a selective advantage for survival of the mutant cell. We also demonstrate that LOH frequently occurred through chromosome loss rather than somatic recombination. LOH at all loci studied on the p and q arms of chromosome 6 was observed in at least 56.2% (9/17) cases. This HLA-associated microsatellite analysis was a useful tool for classifying tumors as LOH-positive or -negative, and therefore to consider a patient as a potential non-responder or responder in a vaccination trial.

Looking for HLA-G expression in human tumours

Tumour and virus infected cells escape CTLs responses by losing some or all HLA class I molecules. However the NK escape mechanism that uses the HLA-A, -B, and -C tumour deficient variants is unknown. To determine whether HLA-G is expressed on tumour cells and thus favours tumour escape by abolishing NK lysis, we studied HLA-G in a large panel of human tumour tissues and human tumour cell lines of different origin that were previously characterized for HLA-A, -B, and -C expression. We studied HLA-G mRNA transcripts using RT-PCR, and HLA-G1 expression by FACS and immunohistochemical techniques. We found several mRNA transcripts of HLA-G isoforms in most of the samples studied. However, we detected no cell surface expression of HLA-G1 using two specific monoclonal antibodies (mAbs) (87G and 01G). We cannot, however, exclude the possibility that some isoforms other than HLA-G1 may be expressed in some tumours.

Expression of HLA G in human tumors is not a frequent event

Expression of HLA G may be a way for tumor cells to escape immuno-surveillance. HLA G is selectively expressed by extravillous trophoblast in the human placenta, a tissue that does not express HLA A or B molecules. It is tempting to propose that tumor cells resemble this unique HLA class I phenotype as they frequently lose classical HLA A, B and C class I expression. Such peculiar HLA class I distribution would in theory allow tumor cells to escape from T- and NK-cell cytotoxicity. To determine whether HLA G is expressed on tumor cells, we studied HLA G mRNA levels using RT-PCR and HLA G cell-surface expression by immunohistological techniques in a panel of 50 human solid tumor tissues, 31 tumor cell lines of different origin, 4 autologous mucosa samples and 3 peripheral white cell samples. We found mRNA transcripts of different HLA G isoforms in most of the samples studied. However, we did not detect cell-surface expression of HLA G using 3 specific monoclonal antibodies (MAbs; 87G, 01G and G223). HLA G was detected only in the U937 myelomonocytic cell line after stimulation with IFN-gamma. We favor the hypothesis that HLA G plays a minor role, if any, in providing an inhibitory signal to NK cells to escape immunosurveillance. We cannot, however, exclude the possibility that some other HLA G isoforms may be expressed in some tumors.

Differential effect on U937 cell differentiation by targeting transcriptional factors implicated in tissue- or stage-specific induced integrin expression

The inhibition of transcription factor functions was used to define their role in phorbol ester-induced cellular differentiation of a monocytic cell line, U937. We demonstrate a differential effect on cell adhesion and differentiation: antisense or competitive binding with double-stranded oligonucleotides antagonized the functions of AP-1, NF-kappaB, and PU.1 transcriptional factors. In the presence of phorbol 12-myristate 13-acetate (PMA), U937 cells attached to the plastic surface and cells were characterized by marked expression of beta2-integrin molecules on the cell surface. We show that the in vivo differentiation of U937 cells appears to occur normally in the absence of AP-1 activity. In contrast, the addition to the cell culture of phosphorothioate oligonucleotides that contained the NF-kappaB or PU.1 binding sites significantly inhibited U937 differentiation. The absence of NF-kappaB led to pleiotropic effects with a clear reduction in the expression of integrin and other lineage-specific myeloid antigens on the cell surface. In contrast, the absence of PU.1 had a more restricted effect on integrin expresion on the cell surface, probably as a result of blockage of CD18 gene expression.

Mutations of the beta2-microglobulin gene result in a lack of HLA class I molecules on melanoma cells of two patients immunized with MAGE peptides

Mutations have been identified in the beta2-microglobulin gene of tumor cells of two metastatic melanoma patients who received immunizations with MAGE peptides. One mutation abolishes the start codon whereas the other introduces a premature stop codon. The second beta2-microglobulin allele of both tumors appears to be lost on the basis of sequence data and loss of microsatellite heterozygosity. The lack of beta2-microglobulin gene product results in the absence of HLA class I antigens on the surface of the tumor cells. This may explain why the tumors of both patients progressed despite the immunization treatment and shows the necessity of analyzing in depth the antigen presentation capability of the tumor cells for the interpretation of clinical trials involving anti-tumor vaccination.

Differential expression of MHC class II genes in lung tumour cell lines

Molecular characterization of HLA class II expression was investigated in five lung tumour cell lines at the protein and mRNA levels. The cell lines exhibited a differential expression of HLA-DR, HLA-DP and HLA-DQ products and also showed differences in the inducibility of HLA class II genes by gamma-IFN. Gamma-IFN stimulation induced only HLA-DR expression to varying degrees in three cell lines, while only one cell line showed stimulation for HLA-DP and none for HLA-DQ antigens. These results suggest locus-specific regulation for the three loci. The presence of DR protein on the cell-surface membrane was always positively correlated with the presence of HLA-DR mRNA in the cells. After treatment with 5-azacytidine in the A549 cell line, which expressed the lowest values, there was no effect on HLA class II levels. This suggested that methylation does not play an important role in the lack of MHC class II antigen expression. In addition to studying mRNA levels of HLA class II antigens, we analysed mRNA of the proto-oncogene c-myc and observed a positive correlation of two mRNA: the increments in HLA-DR expression were associated with increments in c-myc expression. This suggests a relationship between the regulatory and HLA-DR antigens control the expression of c-myc and HLA-DR antigens in lung tumour cell lines.

Unresponsiveness to interferon associated with STAT1 protein deficiency in a gastric adenocarcinoma cell line

HC class I expression can be up-regulated by interferons (IFN) and other cytokines. Both IFNalpha and IFNgamma have been shown to exert their effects via a recently discovered signalling pathway by inducing tyrosine phosphorylation of their receptors. Receptors for interferons and other cytokines signal through the action of associated protein tyrosine kinases of the JAK family (Janus kinase) and latent cytoplasmic transcriptional activators from the STAT family (signal transducers and activators of transcription). Here we report a gastric adenocarcinoma cell line, AGS, that is defective in its response to either IFNalpha or IFNgamma. AGS cells display selective alterations only in MHC class I inducibility and not in constitutive MHC class I expression. In nuclear extracts of AGS cells, no binding activity to interferon-responsive elements (GAS/ISRE) was observed. We found that AGS cells showed an extremely low level of STAT1 expression, which may be responsible for the absence of biological response to IFN. Because STAT1-deficient cells are highly sensitive to infection by virus, the absence of these proteins may also contribute to the tumor phenotype, giving the tumor a selective advantage, by inhibiting cell growth suppression mediated by IFN and abetting escape from the T cell antitumor response.

Microsatellite instability in cervical intraepithelial neoplasia

We studied the incidence of microsatellite instability (MI) in lesions defined as cervical intraepithelial neoplasia (CIN). The human papillomavirus (HPV) status of the tissues was also determined. DNA from tissue samples and autologous lymphocytes were studied for five loci located within or adjacent to the DNA mismatch repair genes. Replicate errors were detected in 7 out of 47 (14.8%) samples of cervical tissue from 24 women. Our results indicate that the defect in DNA repair-associated genes does not appear to be necessary for the selection of clones which progress from dysplasia to carcinoma. Although HPV DNA of highly oncogenic types (16/18) was detected in most cervical lesions and may be an important factor for MI, we also detected MI in two loci in HPV-negative normal tissue, indicating that further events can also be involved in mismatch repair defects.