We identified a novel 8.1-kb human melanoma gene, MG50,derived from subtractive hybridization with a squamous lung carcinoma cell line, LU-1. 6.8 kb containing an open reading frame were sequenced, and the structure of the encoded 1496 amino acid protein was deduced. With HLA-A2.1-transduced Drosophila cells as antigen-presenting cells, we identified six epitopes restricted by HLA-A2.1 that elicited CTLs in vitro. Reactivity of the CTLs to melanoma cells containing MG50 indicated that the epitopes were displayed naturally. Significant cross-reactivity of CTLs immunized against a melanoma cell line that lacked HLA-A2.1 indicated that at least four of the epitopes were also recognized in a different HLA class I context, most likely HLA-A*6802. By quantitative reverse transcription, MG50 message was found in one of two skin melanoma cell lines, an ocular melanoma cell line, two of four metastatic skin melanomas, two of three mammary carcinomas, one of two colon carcinomas, and an ovarian carcinoma. Of six normal tissues, MG50 was found only in a specimen of normal skin and was absent from a congenital nevus. It is likely that MG50is the gene for the interleukin 1 receptor antagonist because a reported sequence of cDNA from the latter had a sequence of 528 bases in the 3′ region, a long contiguous base sequence, and 176 encoded amino acids identical with those of MG50. MG50 is one of the few melanoma-associated antigens that is not a differentiation antigen or a mutated protein. Because of its nature, it may prove to be important in the pathogenesis of the tumors in which it is found, as well as an immunogen and target for immunotherapy.

Advances in the understanding of immunological reactivity to melanoma have been at the forefront of the field of tumor immunology. After the empirical demonstration that melanoma responded to a variety of biological treatments, such as vaccines,IL3-2, and IFN-α2b, considerable impetus was given to the determination of which antigens were recognized by the immune response. Van der Bruggen et al.(1), drawing upon their work in the P815 mastocytoma in mice, demonstrated in 1991 an epitope called MAGE-1, the first such epitope recognized by human CTLs. Since then,other investigators have found a variety of melanoma antigens,including MART-1/Melan A, gp 100, and tyrosinase. Boon et al. (2) showed that the screening of melanoma expression libraries with CTLs derived from autologous vaccine-immunized patients was an effective method of detecting neoantigens. Other groups have also used this procedure in their successful quest for new melanoma antigens (3).

In the late 1980s, we began our attempts to isolate melanoma genes with a method known as subtractive hybridization, which had been used successfully by Hedrick et al.(4) to isolate the T-cell receptor. This procedure depends upon the hybridization of cDNA from one type of cell with an excess of mRNA from another cell that is closely related but is likely to differ in the critical gene of interest. Single-stranded cDNAs, which presumably represent gene fragments unique to the cell type of interest, can be separated from double-stranded hybrids. Our strategy was to identify novel rare genes with a frequency <1:20,000 that might encode tumor neoantigens, which appeared to be among the less frequent cellular antigens. By subtractive hybridization with mRNA from a squamous lung carcinoma cell line (LU-1) as the partner, we identified 12 cDNAs that were found only in a melanoma cell line MSM-M-1 (5), a component of our allogeneic lysate vaccine (6). We chose the squamous lung carcinoma because it was another tumor but was unrelated histologically to the melanoma. We purposely avoided using normal melanocytes as a partner, because we did not want to rule out rare genes that encoded melanocyte differentiation antigens. Of the 12 cDNAs, 6 had sequences that were at that time unreported in any of the gene sequence databases such as GenBank and were, therefore, considered “novel.”

We decided to explore two clones in more detail, because of their distribution among tumor cells and normal cells. One of the cDNAs[“clone 159” in our original paper (5)] was found to be identical with PMP-22, a gene found in Schwann cells and involved in the pathogenesis of Charcot-Marie-Tooth disease (Uchiyama,C., Ph.D. thesis, University of Southern California, 1994). A second cDNA, “clone 50,” has been studied in more detail, requiring almost a decade because of its large size and complex structure. The gene from which clone 50 was derived, which we have called MG50 (registered in GenBank as D2S448) has been localized to chromosome 2p25.3 by fluorescence in situhybridization (7).

In our laboratory, Weiler (8) was able to determine the sequence of 1.3 kb at the 3′ end of the cDNA molecule but could not go beyond that point. Only recently have we been successful in sequencing nearly all of MG50, through a variety of approaches. Here we present that sequence, as well as the protein encoded by the large open reading frame. Their identical sequence in the 3′ region indicates that MG50 may be closely related to or identical with the IL-1 receptor intracellular binding protein (IL-1 receptor antagonist). We further present evidence that MG50 encodes a new melanoma antigen containing at least six naturally expressed melanoma epitopes that are recognized by CTLs in the context of HLA-A2.1.

Sequencing MG50.

The original MG50 cDNA sequence was 500 bp in length. This was used to screen several libraries in an attempt to isolate larger cDNA clones corresponding to MG50(8). Thus, a 1.3-kb cDNA clone was obtained at the 3′ end of the molecule and from its transcript was determined to be 8.1 kb. Although this clone represented only an untranslated sequence, it was valuable in deducing the remainder of the gene. Homology searches on the Blast-NIH and WashU-Merck EST Project databases with 50–100-bp regions as probes revealed one sequence that partially extended beyond and partially overlapped the extreme 5′ end of our clone. With this we constructed a hypothetical upstream extension of our 1.3-kb clone, sequentially searching and matching database clones containing at least 95%homology. The various cDNA clones were pieced together to generate additional sequences 5′ to our 1.3-kb clone.

The first long extension to the 1.3-kb clone was 1200 bp toward the 5′end. We verified that the 5′ sequences hypothetically constructed were linked to the 1.3-kb cDNA clone, and in the proper order, by constructing sense primers from the GenBank-derived sequence, and antisense primers from within our 1.3-kb cDNA sequence. The first sense and antisense primers were 22 bases in length. cDNA from melanoma cell line MSM-M3 (hereafter abbreviated M3), which contains MG50,was then used as a template for these primers in a PCR. The PCR was also performed with the same cDNA using sense and antisense primers from sequences contained entirely within the 1.3-kb cDNA clone. cDNA from the M3 melanoma cell line was amplified in a PCR with the sense and antisense primers located within the 1.3-kb cDNA, and more importantly, cDNA from the M3 melanoma cell line was also amplified with a sense primer located upstream from the 1.3-kb cDNA and an antisense primer located within our 1.3-kb cDNA clone (Fig. 1). These results indicate that the sequence generated from GenBank DNA clones is an additional 5′ sequence contiguous with the original 1.3-kb cDNA clone.

With additional primers spanning most of the 2.5-kb sequence, we demonstrated that the entire 2.5-kb sequence was contiguous in our melanoma cell lines. We subcloned these PCR products into the BlueScript plasmid vector and verified the correct insert size by restriction endonuclease digestion. These plasmid/PCR product recombinants were then sequenced with the dideoxy chain termination procedure. We found that the sequence we generated from our own melanoma cDNA clones was virtually identical with the 2.5-kb hypothetical sequence generated from overlapping GenBank clones.

This approach of constructing a hypothetical sequence based upon fragment homology and then testing the validity of the extension with PCR was used as a means of extending the sequence well into the 5′region upstream from the 22-kb sequence we first added to the 1.3-kb cDNA.

When Nomura and colleagues reported in the gene database a number of large-sized cDNAs obtained from the human genome by random priming PCR,4we used the sequence of bases that overlapped our known MG50sequence to deduce a long extension toward the 5′ end. These sequences were from a human myeloblast line (KG-1) derived from bone marrow and were located on chromosome 2, where MG50 is found. We confirmed by PCR that we had the full sequence from the 3′ end to 6.8 kb, or 1.3 kb from the 5′ end of the molecule. Unfortunately, none of the methods that we used subsequently, such as anchored PCR and marathon PCR, could extend the sequence further. However, a long open reading frame was obtained from which the amino acid sequence of a large protein was deduced.

Quantitative PCR for Determining the Presence of MG50 in Archived Specimens and Cell Lines.

The method was as described by Kan-Mitchell et al.(9). In brief, cDNA libraries were constructed from each specimen and cell line, with avian myeloblastosis virus reverse transcriptase (Seikagaku American, Inc., St. Petersburg, FL), from acidified guanidinium thiocyanate-extracted total RNA. Each was then incubated with 23-base sense and antisense primers from within the 1.3-kb sequence, subjected to the PCR for 30 cycles, and the reaction was analyzed in 6% polyacrylamide gels with 7 murea, 89 mm Tris-borate, and 2 mm EDTA. The T7 polymerase promoter sequence was attached to the 5′ primer to convert amplified cDNA to cRNA, for an additional 500-fold amplification. In parallel, reverse transcription-PCR for β-actin was also performed, with bases 2104–2127 of exon 3 (sense) and bases 2409–2432 of exon 4 (antisense)genomic sequence as primers. Radioactive RNA products (associated with[α-32]CTP) were visualized by autoradiography, and the amount of radioactivity was determined by liquid scintillation. By the calculations of Danenberg et al.(10), we estimated the initial amount of cDNA template. A comparison of cpm/μl cDNA of MG50 RNA with that of β-actin provided a measurement of the relative expression of MG50 DNA among the specimens. A ratio of MG50 cDNA toβ-actin cDNA ≥0.06, 2 SD, was considered a significant amount of MG50.

Deducing Putative Epitopes Binding to HLA-A2.1.

The principal algorithm used was that of Falk and Rotzschke(11), where the motif most suggestive of an HLA-A2.1-binding peptide was *[LM]***[CFILMVWY]**[VL]. A computer program selected regions within the deduced protein product of MG50 with this motif. Motifs for HLA-A1 binding,**[TSM][DEAS]****Y (12), and HLA-B7 binding,*P******Y, were also used to scan the molecule for possible binding to other common HLA class I molecules.

Binding Inhibition (Competitive Binding) Assay for Putative Epitopes.

The method used for measuring the binding of putative epitopes to HLA class I molecules was that of del Guercio et al.(13). Briefly, 2 × 106 HLA-A*0201-transfected Drosophilacells were incubated for 24 h with copper sulfate to induce HLA expression. HBVc peptide 18–27 (FLPSDFFPSV) was radiolabeled with 125I as a standard. Unlabeled putative epitopes at various concentrations were incubated with the Drosophilacells at 26°C for 4 h, together with radiolabeled HBVc. The cells were then layered on fetal bovine serum and centrifuged to separate free and cell-bound peptide. Binding inhibition was calculated as 100 × (1 − binding of HBVc in the absence of unlabeled peptide/binding of HBVc in the presence of unlabeled peptide). Optimal inhibition occurred at a concentration of 100 μg/ml of peptide. More than 20% inhibition of binding of the HBVc standard at 100 μg/ml of peptide was considered sufficient for the peptide to be examined further for immunogenicity.

Generation of CTLs with Peptides (Method 1).

A modification of the method of Rivoltini et al.(14) was used. PBMCs were plated at a concentration of 1.5 × 106/ml in 2-ml capacity wells, on 24-well plates, in Iskove’s modified DMEM + 10%human AB serum (CM). One μg/ml of peptide was added to this preparation. Two days later, 12 IU/ml of IL-2 were added. Weekly restimulation of the PBMCs was performed by harvesting, washing, and replating responders in new 24-well plates at a concentration of 2.5 × 105 cells/ml in CM(5 × 105 cells/well). Autologous PBMCs were thawed, washed twice, resuspended at 5–8 × 106 cells/ml, and pulsed with 10 μg/ml of peptide in 15-ml tubes for 3 h at 37°C. These PBMCs were irradiated at 3000 cGy, washed once, and added to the responders in 24-well plates at stimulator:responder ratios ranging from 3:1 to 10:1. The next day, 12 IU of IL-2 were added. This weekly stimulation was performed two to three times, judging by the activity of the CTLs in a 4-h 51Cr release assay and by IFN-γ release. After the final stimulation, the CTLs were stored frozen in 10% DMSO in CM for further studies.

Generation of CTLs with Peptides (Method 2).

Drosophila cells were also used as APCs in an alternative and more sensitive method (15). The Schneider S2 Drosophila cell line (deposited with the American Type Culture Collection as CRL 10974) was transduced with HLA-A2.1, CD80(B7-1), and CD54 (intercellular adhesion molecule-1) with a pRmHa-3 plasmid vector. The Drosophila cells were grown in Schneider’s medium (106 cells/ml) with 10%fetal bovine serum and CuSO4 at 27°C, the optimal temperature for these insect cells. They were harvested,washed, and resuspended in X-press medium (BioWhittaker) containing 100μg/ml of the putative MG50 epitope. CD8+ T cells were obtained from the PBMCs of a normal donor by positive selection with anti-CD8 immunobeads(Dynabeads; Dynal, Lake Success, NY) and a second antibody(16). At least 92% of the resultant population were CD8+, with the ∼4% CD16+and 4% CD4+. After incubation with the MG50 peptide at 27°C for 3 h, the S2 Drosophila cells were incubated with the CD8+ T cells at 37°C at a ratio of 1:10 in RPMI 1640 containing 10% autologous serum. Two days later, 20 IU of IL-2 and 30 IU of IL-7 were added to the growth medium. Incubation was continued for 1 week, after which the Drosophila cells were replaced with autologous, irradiated PBMCs (3000 cGy) and peptide. This was repeated for one further round of stimulation, after which the CD8+ T cells were tested for cytotoxicity by a 4-h 51Cr release assay.

Sequence of MG50 and Its Encoded Protein.

The variety of methods we used to sequence the long gene, described in“Materials and Methods,” may be summarized as follows:(a) screening libraries with the original 500-bp cDNA clone to obtain longer cDNA clones; (b) sequencing a 1.3-kb cDNA clone at the 3′ end of the molecule obtained by this process;(c) performing repeated homology searches of public gene databases with 50–100-bp regions as probes to find matching regions;(d) constructing hypothetical upstream extensions,sequentially searching and matching database regions with 95% or greater homology; and (e) using PCR with sense and antisense primers from within a region of verified sequence and the hypothetical extension, to test whether that hypothetical extension was correct.

Through these methods, we were able to obtain the sequence of the MG50 gene, which is shown in Fig. 2 A. A number of internal poly(A) sequences were found, which may have accounted for regions of compression in the molecule and caused many of the difficulties that we encountered in sequencing. 6.8 of the 8.1-kb were identified, continuous from the 3′ end to 1.3 kb from the 5′ end. The remaining 1.3 kb at the 5′ end were unidentifiable by any of the methods we used. Similarly, the investigators who reported a cDNA sequence overlapping with that of MG50 in the 5′ region did not obtain sequence at the very end of the molecule.5

Nonetheless, a large open reading frame was found, from which a long protein sequence was deduced. The sequence of the protein is shown in Fig. 2 B. Many leucine-serine regions were found within the molecule, which suggested that it might bind to chondroitin sulfate. There were regions within the protein that had similarities to several different molecules. Among these is a considerable homology in regions 965-1057 and 1084–1102 with human peroxidase precursors,including eosinophil peroxidase precursor, myeloperoxidase precursor,and anti-thyroid peroxidase precursor. Perhaps the most striking similarity was the 75% homology with a molecule related to human peroxidases: peroxidasin. Peroxidasin is a newly described protein combining peroxidase and extracellular matrix motifs, which is involved in the formation of extracellular matrix during the embryogenesis of hemocytes in Drosophila. It is also involved in phagocytosis and defense in that organism (17). MG50 had other regions with homology to human thrombospondin, human contactin,human heparin sulfate core protein, and human axonin-1.

Probable Identity of MG50 as the Gene Encoding the IL-1 Receptor Antagonist Protein.

Most recently, in the course of review for a patent for MG50, we were made aware of information from the United States Patent Office, suggesting that it might be closely homologous with another gene. A cDNA clone encoding a portion of IL-1 receptor intracellular binding protein (18) was isolated by Drs. L.-L Lin and J. Graham. The entire clone was 1571 bp, comprising 528 bp in the coding region. These were identical with the 3′ terminal portion of the MG50 gene. The amino acid sequence of the encoded protein fragment of 176 amino acids was also identical with a portion of the MG50 protein. The doubly underlined region in Fig. 2,Ashows the sequence of bases 3961 to 4488 in MG50 and in Fig. 2 B, the corresponding amino acid sequence of peptides 1321 to 1496 in our gene, which match exactly with the cDNA and encoded peptide of Lin and Graham (18). The contiguous downstream noncoding region of the Lin-Graham cDNA contained a sequence identical with that of MG50, completing its 3′ end. Thus, we think that it is likely that MG50 is the full-length IL-1 receptor binding protein (IL-1 receptor antagonist; IL-1Ra) gene.

Distribution of MG50 by Reverse Transcription-PCR.

“Clone 50” was originally chosen for further study because it was expressed in four of five melanoma cell lines (MSM-M1, MSM-M3, MSM-M4,and UCLA-M-21), a breast cancer cell line (734B), and a glioblastoma(U138 MG) but not in myeloid (K562) or lymphoid (Daudi) tumor cell lines, concanavalin A-induced lymphoblasts, or normal spleen. This distribution was determined by in situ colony hybridization with radiolabeled cDNA probes from human cells and tissues(5). By PCR, we also found MG50 in an embryonal rhabdomyosarcoma and in normal placenta (data not shown).

We repeated studies of the distribution of the gene with a quantitative PCR (9) on archived tissues and cell lines. By a quantitative reverse transcription-PCR, comparing cpm/μl cDNA library, of MG50 and β-actin as an internal control, MG50 was demonstrable mainly in melanoma and breast cancer cells (Table 1). The skin melanoma cell line M3, ocular melanoma cell line OCM-3, and two of four metastatic skin melanoma specimens were “positive” by this assay. In addition, specimens of two of three mammary carcinomas,an ovarian carcinoma, and one of two specimens of colon carcinoma (a sigmoid carcinoma) were “positive.” Specimens of soft tissue sarcoma, pancreatic carcinoma, squamous cell carcinoma of the lung, and pseudomyxoma peritonei were negative. Of the six normal tissues tested,a specimen of skin was “positive,” although a congenital nevus specimen was negative. In addition, EBV-transformed lymphoblastoid cells from patient M14 were negative.

MG50 appeared to be tumor associated, i.e.,relatively restricted to several disparate types of tumor, but absent from many normal tissues. However, our limited analysis thus far does not allow us to state its distribution among tumors with certainty. Further screening of a wider variety of tissues and tumors, as we are now beginning with cDNA microarrays, should clarify these important points. We do know that, as with many tumor-associated genes and antigens, MG50 is also present in normal fetal tissue (fetal liver and spleen). A DNA base sequence that we used to deduce the upstream base structure of MG50 was reported by Hillier et al. (IMAGE Consortium, Washington University, St. Louis, MO), from liver and spleen material derived from a 20-week postconception male fetus.

Examination of the Secondary Structure of the Protein.

The protein sequence encoded by the large open reading frame was 1496 amino acids. Its secondary structure was kindly examined by Drs. A. F. Kirkin and J. Zeuthen, (Danish Cancer Society Institute of Cancer Biology, Copenhagen, Denmark) with the GOR algorithm for secondary structure prediction of Garnier et al.(19). From their analysis, the protein was predicted to contain a paucity(23%) of long α-helical structures and, therefore, was thought to be likely to unfold without difficulty. The other percentages of the residues were: sheet, 28.6%; turns, 27.9%; and coil, 21.6%. (Full data of the analysis are available upon request.) Consequently, Kirkin and Zeuthen further predicted that the protein was likely to be degraded by antigen-processing cells, exposing its epitopes to CTL.6

We also analyzed the protein for “PEST” sequences, i.e.,those rich in proline, glutamic acid, serine, threonine, and aspartic acid residues (20). Such sequences are characteristic of regulatory proteins with a high turnover rate. They may also improve the potential for expression of epitopes if found outside the region of the epitopes themselves (21). A 15-amino acid peptide in the untranslated region, 1793–1807,(RPEQEPLPDGSSQGR) had a very high PEST score of 8.05, with a PEST mole fraction of 38.68% and a hydrophobicity index of 26.44, but no PEST-rich sequence within the 1496 amino acid translated protein was identified. In fact, there were 18 regions of PEST-poor amino acids, two of which (68–87 and 614–635)included sequences we later found to be immunogenic.

Binding of Putative Epitopes Encoded by MG50 to HLA Class I Molecules.

We examined the sequence of the MG50 protein with several algorithms based upon motifs known to bind to various HLA class I molecules. A decamer (1801–1810), which we now know is from the untranslated region, was studied early in the course of this investigation. Its sequence (RPRPEQEPLP) suggested that it might bind to HLA-B7 histocompatibility antigen; therefore, it was tested for binding to that molecule (13) through the courtesy of Dr. Esteban Celis, then at Cytel Corporation (San Diego, CA). The decamer was found to have an extremely high binding affinity for HLA-B7, 1.96 nm. Despite, or perhaps because of, the exceedingly strong binding affinity, the peptide was not found to be immunogenic for CTLs,as tested by in vitro immunization with matched autologous PBMCs. Whether alternative in vitro immunization procedures, such as with dendritic or Drosophila cells as APCs, will give different results has not been tested. One potential HLA-A1 binding motif, DVTSGNTVY, was also identified (peptide 272–280), but the peptide has not yet been studied further.

However, there were 12 peptides whose structure suggested binding to HLA-A2.1, an HLA class I molecule found in 40–50% of Caucasians and Asians. Each of the peptides had a leucine or methionine in position 2 and a leucine or valine in position 9. Table 2 illustrates the amino acid sequence of the 12 peptides. Their binding to HLA-A2.1 was assayed by competition against the strongly binding HBVc peptide 18–27 (HBc18–27; FLPSDFFPSV) as the standard (13). No binding was found with five of the peptides in a preliminary assay. Fig. 3 shows the results of the binding assay on the remaining seven peptides. One of the seven (844–852) bound only weakly and was eliminated from further consideration, whereas the other six had a moderate to strong binding affinity that provided an impetus for their testing as immunogens.

Immunogenicity of MG50 Peptides.

The six remaining peptides with significant binding affinity to HLA-A2.1 were tested for immunogenicity in vitro in two ways. We attempted immunization with irradiated autologous HLA-A2+ PBMCs as APCs (14). In addition, we immunized CD8+ T cells with APCs consisting of Drosophila cells transduced with HLA-A2.1,CD80 (B7-1), and CD54 (intercellular adhesion molecule-1). In both instances, we attempted to immunize naïve CD8+ T cells derived from normal volunteers. We did not test CD8+ T cells from melanoma patients to avoid possible artifacts from in vivo tumor-induced anergy.

All six peptides were capable of immunizing CTLs de novo to recognize APCs presenting those peptides when presented on HLA-A2.1+ T2 or HLA-A2.1-transduced Jurkat cells(Figs. 4 and 5). Two of these peptides, 65–73 and 624–632, were identified by the method of Rivoltini et al.(14) and confirmed with Drosophila as APCs; the other four were identified by the more sensitive procedure with Drosophila APC. Fig. 4shows the results of immunization with peptides 65–73, 68–76,624–632, and 1050–1058 with autologous PBMCs as APCs. Testing was performed with T2 cells after 3 weeks of stimulation in vitro. Peptides 65–73 and 624–632 were able to generate immunity against themselves, whereas by this method peptides 68–73 and 1050–1058 were not.

Fig. 5 shows the results of immunization by each of the six peptides with Drosophila cells as APCs. After 3 weeks of immunization, CD8+ CTLs were tested against the corresponding peptide as presented on HLA-A2.1+T2 or HLA-A2.1-transduced Jurkat (JA2) cells over a range of E:T ratios. At a 50:1 ratio, the background reactivity against T2 or JA2 cells alone was generally high, but at ratios of 10:1 and 2:1 a clear distinction of reactivity against peptide-bearing cells was obtained. All six putative epitopes were found to be immunogenic by this method.

Reactivity of Peptide-immunized CTLs against Melanoma Cell Lines.

Most importantly, all six peptides, 65–73, 68–76, 209–217, 624–632,1050–1058 and 1243–1251, immunized CD8+ T cells to react against the long-term HLA-A2.1+ melanoma cell lines MSM-M7, MSM-M14, and MALME-3M and against the short-term(<6 months) HLA-A2.1+ IW melanoma cell line(Fig. 6). Immunization against the melanoma cell lines was accomplished equally well when the peptides were introduced singly or together in Drosophila APCs (data not shown). There was no reactivity of CTLs that lysed MALME-3M melanoma cells against control MALME-3 fibroblasts. These data indicated that the peptides were epitopes that were naturally expressed on melanoma cells. We did not have a melanoma cell line that was HLA-A2.1+ but did not express MG50, which would have been a final control for specificity of the reaction of CTLs.

Test of CTLs against Melanoma Cells Bearing “Inappropriate” HLA Class I Molecules.

As shown in Fig. 6, we also tested the epitopes against a “control”melanoma cell line, M3, which was a high expresser of MG50mRNA but was HLA-A2.1 negative. By serotyping, M3 bears the following alleles: HLA-A68, HLA-A30, HLA-B13, HLA-B27, HLA-C6, and HLA-C8. None of these alleles was found in the HLA-A2.1+ melanoma cell lines to which the CTLs otherwise reacted. An epitope that was “classically” restricted solely by HLA-A2.1 (HLA-A*0201) was 624–632, against which CTLs were elicited that did not react against M3 cells. Reactivity of CTLs directed against peptide 65–73 to M3 cells was low but only slightly less than to the other melanoma cell lines. However, the other four putative epitopes, 68–76, 209–217, 1050–1058, and 1243–1251,although leading to a similar degree of lysis of the HLA-A2.1 melanoma cell lines as 65–73 and 624–632, also elicited CTL reactivity against the M3 melanoma cell line (Fig. 6).

We assume that the reactivity of CTLs against epitopes 68–76, 209–27,and 1243–1251 as expressed on M3 cells was restricted by HLA-A*6802, which is closely related in amino acid sequence to HLA-A*0201. Although we have not formally tested whether cross-reactivity involved HLA-B and HLA-Calleles, no HLA-B or HLA-C alleles of M3 were common to the other melanoma cell lines

Also, because we did not clone the CTLs, we cannot be certain whether cross-reactivity was characteristic of the majority of the CTLs or was attributable to a subpopulation of “nonfastidious” (degenerate)CTLs.

Sequence of the Six Likely Epitopes.

To summarize, the sequences of the HLA-A2.1-restricted epitopes are:65–73, CMHLLLEAV; 624–632, VLSVNVPDV; 68–76, LLLEAVPAV; 209–217,TLHCDCEIL; 1243–1251, RLGPTLMCL; and 1050–1058, WLPKILGEV. The last four sequences may also be recognized in the context of other HLA class I alleles.

The identification of new melanoma genes and their corresponding antigens has been a rapidly paced field during this decade since the initial publication of Van der Bruggen et al.(1). There are no obvious similarities in structure between MG50 and any of the melanoma antigens described previously. Thus, MG50 differs from published sequences of MAGE, BAGE, GAGE, PRAME, and NY-ESO-1 (the cancer/testis-specific antigens), gp100, tyrosinase, MART-1/Melan-a, TRP-1 and TRP-2 (the melanocyte differentiation antigens), and CDK4, β-catenin, and MUM-1(mutated or aberrantly expressed antigens). MG50 protein is one of the few melanoma-associated antigens that is not a melanocyte differentiation antigen or a mutated protein. The MG50 gene may be important not only because of its novelty, but because it is a large gene encoding at least six HLA class I-restricted epitopes recognized by human CTLs. Even large viral proteins usually express only one or two CTL-defined epitopes besides their more numerous helper T cell-defined epitopes (22).

Although we have not yet investigated the distribution of MG50 in a large number of normal tissues and tumor cells,evidence from this report and our previous studies (5)suggests that the gene is relatively restricted to tumors such as melanoma, breast cancer, ovarian cancer, and glioblastoma. MG50 was absent from hematopoietic (lymphocytic and myelocytic) tumor cell lines and archived specimens of normal tissues,with the exception of skin, but a cDNA sequence from MG50was identified by others in fetal liver and spleen. Because MG50 appeared in our early work to be relatively tumor associated and potentially a useful immunogen and target for immunotherapy, we persisted in our attempts to sequence MG50over nearly a decade.

We are hopeful that the MG50 protein will serve as a useful source of immunogenic peptides for vaccination against several types of cancer,overcoming some of the problems with immunization with the cancer/testis (e.g., MAGE) group of antigens. The disappointing results thus far with those peptides may have several causes. Kirkin et al.(21) suggested that antigens recognized by CTLs have a structure that permits efficient unfolding of the molecule, denaturation, and ubiquitination. The cancer/testis group are highly organized stable structures, with enrichment of long stretches of α helices, and may not easily be subject to degradation and proteasomal cleavage. In contrast, the melanoma differentiation antigens, such as tyrosinase, are less highly organized, which may explain their improved immunogenicity over the MAGE group. Examination of the structure of the MG50 protein with the GOR algorithm for stable structural proteins (19) revealed that it had a paucity of long α-helical structures. It was concluded that MG50 was likely to unfold easily and be degraded, presumably into peptides recognizable by CTLs.6 The presence of PEST-rich sequences outside the region of the epitopes would have further strengthened the possibility of efficient and rapid degradation of the protein. Unfortunately, the MG50 protein contains only one PEST-rich sequences at its 3′ end, and no others within the translated region. Nevertheless, we have some confidence from its structure alone that the whole MG50 protein, or the gene encoding it, will offer a feasible method of immunization against a large variety of epitopes of various HLA restrictions. There are undoubtedly additional epitopes binding to HLA molecules other than HLA-A2.1, beyond the six we have identified for that common haplotype.

Its CTL-defined epitopes, which were naturally represented on melanoma cells, indicated that the novel gene MG50 encoded a previously undescribed melanoma antigen. We found six epitopes recognized by CTLs within the MG50 protein, which is very different from the paucity of CTL-defined epitopes even on large-sized viruses(22). Two epitopes, 65–73 (CMHLLLEAV) and 624–632(VLSVNVPDV), were demonstrable on melanoma cells and were classically HLA-A2.1 restricted. Four other epitopes, 68–76 (LLLEAVPAV), 209–217(TLHCDCEIL), 1050–1058 (WLPKILGEV), and 1243–1251 (RLGPTLMCL), were recognized not only on HLA-A2.1+ melanoma cell lines but also on the M3 melanoma cell line (HLA-A68, HLA-30). This is not entirely surprising because HLA-A*0201 and HLA-A*6802 are members of the same HLA supertype (23). We have also described“nonfastidious” CTLs in patients with skin or ocular melanoma that cross-reacted against melanoma cell lines with HLA class I haplotypes outside the HLA-A2 supertype (24, 25). However, we have not determined whether the four novel epitopes we have described here can be recognized outside the HLA-A2 supertype family.

For active immunotherapy, our bias has been toward the use of polyvalent or multiepitopic immunogens rather than a single peptide(26, 27). Whether the gene, the whole protein, or selected epitopes will prove to be most useful in creating a synthetic melanoma or breast cancer vaccine is uncertain. The methods we used to demonstrate the presence of epitopes, especially the use of purified APCs such as Drosophila or dendritic cells, can be adapted for use in in vivo or in vitro vaccination. Several methods of incorporating epitopes into dendritic cells have been described (28), including our own procedure for adding signal sequences that conduct nonamers into the endoplasmic reticulum (29). Likewise, several viral vectors conducting minigenes or entire genes into dendritic cells have been developed. We favor the use of lentiviral vectors, some of which have the capacity to transduce genes as large as 10 kb into the DNA of nonreplicating cells. The characteristics of translated MG50 protein should permit it to unfold and be degraded within APCs and then be recognized by CD4+ and CD8+ T cells.

Besides cancer vaccines, melanoma epitopes could be used for in vitro immunization of CTLs, to be administered as adoptive immunotherapy. In a Phase I trial, 5 × 106 CTLs immunized ex vivo against a tyrosinase epitope, YMNGTMSQV, with Drosophila cells as APCs elicited a clinical partial remission lasting over 1 year in 1 of 10 patients.7A comparison of these strategies will enable us to determine which construct and therapeutic approach with MG50 are most useful for specific immunotherapy.

Sequence data at the 3′ end of the gene strongly suggest that MG50 is identical with the IL-1 receptor antagonist protein(IL-1Ra) gene, cDNA of which was identified independently by Lin and Graham of Genetics Institute (Cambridge, MA; PCT/US96/06363; Ref.18). Like the panoply of other genes for tumor-associated antigens, such as tyrosinase and gp100 in melanoma, and c-erbB2 and the mucins in adenocarcinomas, MG50 is not unique to tumors. However, like the others, it may be overexpressed in tumors and thus identifiable by CTLs as a potential target for cellular immunotherapy,regardless of its expression in normal cells.

The function of the IL-1Ra in a melanoma cell is uncertain but may represent another example of the antagonism of the tumor cell to cytokines that aid the host in rejecting it. IL-1 is a proinflammatory cytokine produced by bone marrow and bone marrow-derived cells such as macrophages. A component of what was originally called“lymphocyte-activating factor” (30, 31, 32), IL-1 production is stimulated from macrophages by immunological adjuvants such as Bacillus Calmette Guérin,in vivoand in vitro(33, 34, 35). IL-1 is important in the inflammatory cascade for up-regulating the expression of adhesion molecules, recruitment of neutrophils, and inducing other cytokines(36). These activities and others are blocked efficiently by the IL-1 receptor antagonist protein. After administration of lipopolysaccharide (endotoxin), IL-1Ra levels increase within 2 h in the serum of normal human subjects (37). The secretory IL-1Ra gene is controlled through three lipopolysaccharide-responsive elements, one of which was found to be an nuclear factor-κB binding site (38). In contrast, IL-18,which induces IL-1, does not induce anti-inflammatory cytokines such as IL1-Ra or IL-10 (39).

Among its positive activities, IL-1Ra was useful in overcoming experimental “postmenopausal osteoporosis,” decreasing bone loss and bone resorption in ovariectomized rats (40). In the tumor setting, however, it is plausible that, similar to IL-10,transforming growth factor-β, and prostaglandin E2, IL-1Ra functions to subvert the activity of antitumor lymphocytes. IL-1 released in the vicinity of the tumor by macrophages may be blocked in its activity if IL-1Ra binds to IL-1R in antitumor lymphocytes. Burger et al.(41) found high levels of IL-1Ra in the ascites of 18 patients with advanced ovarian cancer. Ascites tumor cells spontaneously released IL-1Ra and TNF-α/β binding protein BP55, and tumor-associated macrophages released IL-1Ra and the TNF-α/β binding protein BP-75. Lymphocytes from the ascites of these patients released no TNF-α or TNF-β. Although our major interest has been in T cell-defined epitopes, it is possible that targeting of MG50 with monoclonal antibodies or antisense oligonucleotides will be useful for selective inhibition of tumor-induced immunosuppression mediated by IL-1Ra.

Fig. 1.

A 2% agarose gel of PCR products stained with ethidium bromide. Lane 1, DNA molecular weight markers. Lane 2, product from a PCR using sense and antisense primers both located within the 1.3-kb cDNA region but without cDNA. Lane 3, product from a PCR of melanoma M3 cDNA using sense and antisense primers, both located within the 1.3-kb cDNA. Lane 4, product from a PCR using a sense primer located within a GenBank cDNA sequence that was 5′ to the 1.3-kb cDNA fragment but without cDNA. Lane 5, product from a PCR of melanoma M3 cDNA using a sense primer located within a GenBank cDNA sequence that was 5′ to the 1.3-kb cDNA fragment and an antisense primer located within the 1.3-kb cDNA region. Note that: (a) M3 cDNA was amplified with primers from within the 1.3-kb region; and, more importantly (b) a primer from within the 1.3-kb region and one within the GenBank sequence region also amplified M3 cDNA. This showed that the two regions were contiguous. The full region was later sequenced to prove this point conclusively (see text).

Fig. 1.

A 2% agarose gel of PCR products stained with ethidium bromide. Lane 1, DNA molecular weight markers. Lane 2, product from a PCR using sense and antisense primers both located within the 1.3-kb cDNA region but without cDNA. Lane 3, product from a PCR of melanoma M3 cDNA using sense and antisense primers, both located within the 1.3-kb cDNA. Lane 4, product from a PCR using a sense primer located within a GenBank cDNA sequence that was 5′ to the 1.3-kb cDNA fragment but without cDNA. Lane 5, product from a PCR of melanoma M3 cDNA using a sense primer located within a GenBank cDNA sequence that was 5′ to the 1.3-kb cDNA fragment and an antisense primer located within the 1.3-kb cDNA region. Note that: (a) M3 cDNA was amplified with primers from within the 1.3-kb region; and, more importantly (b) a primer from within the 1.3-kb region and one within the GenBank sequence region also amplified M3 cDNA. This showed that the two regions were contiguous. The full region was later sequenced to prove this point conclusively (see text).

Close modal
Fig. 2.

A, DNA sequence of 6.8-kb of gene MG50. B, sequence of a 1496 amino acid protein encoded by gene MG50. The doubly underlined portion of sequences in A and B are the regions of identity with 528 bases and the corresponding 176 amino acids encoded by the translated region of the cDNA reported by Lin and Graham (Ref. 14; MG50 bases 3961–4488 and amino acids 1321 to 1496.) The remaining 1044 bases they reported, for a total of 1571, were also identical, completing the sequence at the 3′end of MG50.

Fig. 2.

A, DNA sequence of 6.8-kb of gene MG50. B, sequence of a 1496 amino acid protein encoded by gene MG50. The doubly underlined portion of sequences in A and B are the regions of identity with 528 bases and the corresponding 176 amino acids encoded by the translated region of the cDNA reported by Lin and Graham (Ref. 14; MG50 bases 3961–4488 and amino acids 1321 to 1496.) The remaining 1044 bases they reported, for a total of 1571, were also identical, completing the sequence at the 3′end of MG50.

Close modal
Fig. 3.

Results of a binding assay with seven MG50peptides, all putatively HLA-A2.1-restricted epitopes. The top line in the graph is the standard HBVc peptide 18–27, against which the others were compared.

Fig. 3.

Results of a binding assay with seven MG50peptides, all putatively HLA-A2.1-restricted epitopes. The top line in the graph is the standard HBVc peptide 18–27, against which the others were compared.

Close modal
Fig. 4.

Immunogenicity of four MG50 peptides. Cytolytic activity of PBMCs against four MG50 peptides immunized with autologous PBMCs as antigen-presenting cells. Target cells were T2 cells pulsed with the specific peptide. Only two of four peptides were determined to be immunogenic by this means.

Fig. 4.

Immunogenicity of four MG50 peptides. Cytolytic activity of PBMCs against four MG50 peptides immunized with autologous PBMCs as antigen-presenting cells. Target cells were T2 cells pulsed with the specific peptide. Only two of four peptides were determined to be immunogenic by this means.

Close modal
Fig. 5.

CTL activity of CD8+ effectors generated against six MG50 peptides with Drosophila cells as APCs. HLA-A2.1 transduced Jurkat lymphoma cells pulsed with specific peptide as targets. All peptides were found to be immunogenic by this strategy.

Fig. 5.

CTL activity of CD8+ effectors generated against six MG50 peptides with Drosophila cells as APCs. HLA-A2.1 transduced Jurkat lymphoma cells pulsed with specific peptide as targets. All peptides were found to be immunogenic by this strategy.

Close modal
Fig. 6.

Reactivity against melanoma cell lines of CD8+ CTL generated against various epitopes of MG50. MSM-M-7,MSM-M-14, and MALME-3M were HLA-A2.1+ long-term melanoma cell lines; IW was an HLA-A2.1+ short-term cell line that had been passaged in culture fewer than 6 months. MALME-3 was an HLA-A2.1+ fibroblast line, which served as a control for the MALME-3M melanoma. MSM-M-3 was an HLA-A2.1 negative long-term cell line that expressed MG50, which was used as a control for HLA class I specificity of the CTLs.

Fig. 6.

Reactivity against melanoma cell lines of CD8+ CTL generated against various epitopes of MG50. MSM-M-7,MSM-M-14, and MALME-3M were HLA-A2.1+ long-term melanoma cell lines; IW was an HLA-A2.1+ short-term cell line that had been passaged in culture fewer than 6 months. MALME-3 was an HLA-A2.1+ fibroblast line, which served as a control for the MALME-3M melanoma. MSM-M-3 was an HLA-A2.1 negative long-term cell line that expressed MG50, which was used as a control for HLA class I specificity of the CTLs.

Close modal

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1

Supported by USPHS Grant RO1-CA57846 and a grant from the A-H Foundation.

3

The abbreviations used are: IL, interleukin;MAGE, melanoma antigen; MG, melanoma gene; HBVc, hepatitis B virus core; PBMC, peripheral blood mononuclear cell; CM, complete medium;APC, antigen-presenting cell; TNF, tumor necrosis factor.

4

T. Nagase, N. Seki, K. Ishikawa, O. Ohara, and N. Nomura. Prediction of the coding sequences of unidentified human genes. VI. The coding sequences of 80 new genes (KIAA 0201–KIAA 0280) deduced by analysis of cDNA clones from human cell line KG-1 and brain, August 2, 1996.

5

N. Nomura, personal communication, October 4,1996.

6

A. F. Kirkin and J. Zeuthen, personal communication, 1998.

7

M. S. Mitchell, D. Darrah, D. Yeung, S. Halpern, A. Wallace, V. Jones, and J. Kan-Mitchell. Adoptive immunotherapy of melanoma with cytolytic T lymphocytes immunized in vitro against a tyrosinase epitope: Phase I trial,submitted for publication.

Table 1

Quantitative PCR for MG50 DNA performed on tumor cell lines and archived tissues

cpm/1 μl cDNA libraryRatio
MG50β-actin
Cell Line    
MSM-M1 1.8 × 104 5 × 107 0.0004 
MSM-M3 4.9 × 105 4.8 × 106 0.10              a 
OCM-3 1.7 × 104 1.6 × 105 0.12 
Lymphoblastoid cell line from melanoma patient 0.0 2.9 × 106 0.0 
Archival tumor tissues    
Mammary CA 1 × 105 1.7 × 107 0.0059 
Mammary CA 4.9 × 105 4.3 × 106 0.11 
Mammary CA 7.4 × 105 1.1 × 107 0.067 
Soft tissue sarcoma 9.6 × 104 2.7 × 107 0.0036 
Squamous cell CA 2.4 × 104 2.1 × 106 0.011 
Pancreatic CA 8.7 × 103 8.2 × 105 0.011 
Ovarian CA 2.2 × 105 1 × 106 0.22 
CA of sigmoid 2.5 × 104 4.4 × 105 0.057 
Colorectal CA 5.8 × 104 5.2 × 106 0.011 
Pseudomyxoma peritonei 1.7 × 105 1.1 × 107 0.015 
Melanoma metastatic, WW 140 2.2 × 105 6.6 × 105 0.33 
Melanoma metastatic, WW 135 3.7 × 104 1 × 107 0.0019 
Melanoma, metastatic, WW 129 5.6 × 104 2.7 × 106 0.0207 
Melanoma, metastatic, WW 141 5.3 × 104 3 × 104 1.77 
Archival normal tissues    
Stomach 1.7 × 103 
Skin 2.3 × 103 1.8 × 104 0.13 
Congenital nevus, WW 145 1.7 × 104 2 × 106 0.0085 
Kidney 2.1 × 103 2.3 × 105 0.0091 
Lung 5.8 × 104 9 × 106 0.0006 
Breast 1.2 × 104 1.5 × 106 0.0077 
cpm/1 μl cDNA libraryRatio
MG50β-actin
Cell Line    
MSM-M1 1.8 × 104 5 × 107 0.0004 
MSM-M3 4.9 × 105 4.8 × 106 0.10              a 
OCM-3 1.7 × 104 1.6 × 105 0.12 
Lymphoblastoid cell line from melanoma patient 0.0 2.9 × 106 0.0 
Archival tumor tissues    
Mammary CA 1 × 105 1.7 × 107 0.0059 
Mammary CA 4.9 × 105 4.3 × 106 0.11 
Mammary CA 7.4 × 105 1.1 × 107 0.067 
Soft tissue sarcoma 9.6 × 104 2.7 × 107 0.0036 
Squamous cell CA 2.4 × 104 2.1 × 106 0.011 
Pancreatic CA 8.7 × 103 8.2 × 105 0.011 
Ovarian CA 2.2 × 105 1 × 106 0.22 
CA of sigmoid 2.5 × 104 4.4 × 105 0.057 
Colorectal CA 5.8 × 104 5.2 × 106 0.011 
Pseudomyxoma peritonei 1.7 × 105 1.1 × 107 0.015 
Melanoma metastatic, WW 140 2.2 × 105 6.6 × 105 0.33 
Melanoma metastatic, WW 135 3.7 × 104 1 × 107 0.0019 
Melanoma, metastatic, WW 129 5.6 × 104 2.7 × 106 0.0207 
Melanoma, metastatic, WW 141 5.3 × 104 3 × 104 1.77 
Archival normal tissues    
Stomach 1.7 × 103 
Skin 2.3 × 103 1.8 × 104 0.13 
Congenital nevus, WW 145 1.7 × 104 2 × 106 0.0085 
Kidney 2.1 × 103 2.3 × 105 0.0091 
Lung 5.8 × 104 9 × 106 0.0006 
Breast 1.2 × 104 1.5 × 106 0.0077 
a

Ratios of MG50/β-actin≥0.06, >2 SD from mean, are shown in italics. Two of three mammary carcinomas, an ovarian carcinoma, a sigmoid carcinoma, skin melanoma M3, ocular melanoma OCM-3, and two of four metastatic skin melanomas were “positive” by this assay. Of the six normal tissues tested,“skin” was “positive,” but significantly, a congenital nevus specimen was negative.

Table 2

MG50 putative HLA-A2-restricted peptides, as predicted by algorithms for motif

PeptidesAmino acid position
12345678910
33–41  
65–73  
68–76  
209–217  
624–632  
844–852  
1050–1058  
1132–1140  
1243–1251  
1251–1259  
1407–1415  
1414–1422  
PeptidesAmino acid position
12345678910
33–41  
65–73  
68–76  
209–217  
624–632  
844–852  
1050–1058  
1132–1140  
1243–1251  
1251–1259  
1407–1415  
1414–1422  

We gratefully acknowledge the expert collaboration of Drs. Ann Moriarty and Didier LeTurcq of R. W. Johnson Pharmaceutical Research Institute, San Diego, California, in identifying the epitopes of MG50 with their Drosophila cell technology. We are very grateful to Dr. Alexei Kirkin and Prof. Jesper Zeuthen, Danish Cancer Institute, Copenhagen, Denmark, for their analysis of the secondary structure of MG50 with the GOR algorithm. The diligent efforts of Dr. Sarah Weiler in sequencing approximately one-third of the gene at the 3′ end of the molecule, in partial fulfillment of the Ph.D. degree, and of Dr. Thomas Kaido, University of California, San Diego, California, in further sequencing of the molecule during a postdoctoral fellowship are also acknowledged with gratitude. We thank Dr. Peter Cresswell, Yale University School of Medicine, New Haven, CT, for his generous provision of T2 cells. Dr. James Feramisco, University of California, San Diego, California,contributed very helpful guidance during the sequencing of the major portion of the gene.

1
Van der Bruggen P., Traversari C., Chomez P., Lurquin C., De Plaen E., Van den Eynde B., Boon T. A gene encoding an antigen recognized by cytolytic T lymphocytes on a human melanoma.
Science (Washington DC)
,
254
:
1643
-1647,  
1991
.
2
Boon, T., Coulie, P., Marcand, M., Weynants, P., Wolfel, T., and Brichard, V. Genes coding for tumor rejection antigens: perspectives for specific immunotherapy. Important Adv. Oncol., 53–69, 1994.
3
Kawakami Y., Eliyahu S., Delgado C. H., Robbins P. F., Rivoltini L., Topalian S. L., Miki T., Rosenberg S. A. Cloning of the gene coding for a shared human melanoma antigen recognized by autologous T cells infiltrating into tumor.
Proc. Natl. Acad. Sci. USA
,
91
:
3515
-3519,  
1994
.
4
Hedrick S. M., Cohen D. I., Nielsen E. A., Davis M. M. Isolation of cDNA clones encoding T cell-specific membrane-associated protein.
Nature (Lond.)
,
308
:
149
-153,  
1984
.
5
Hutchins J. T., Deans R. J., Mitchell M. S., Uchiyama C., Kan-Mitchell J. Novel gene sequences expressed by human melanoma cells identified by molecular subtraction.
Cancer Res.
,
51
:
1418
-1425,  
1991
.
6
Mitchell M. S., Kan-Mitchell J., Kempf R. A., Harel W., Shau H., Lind S. Active specific immunotherapy of melanoma. Phase I trial of allogeneic melanoma lysates and a novel adjuvant.
Cancer Res.
,
48
:
5883
-5893,  
1988
.
7
Weiler S. R., Taylor S. M., Deans R. J., Kan-Mitchell J., Mitchell M. S., Trent J. M. Assignment of a human melanoma associated gene MG50 (D2S448) to chromosome 2p25.3 by fluorescence in situ hybridization.
Genomics
,
22
:
243
-244,  
1994
.
8
Weiler, S. R. Molecular characterization of a novel human melanoma associated gene (MG50), Thesis/Dissertation, University of Southern California, 1993.
9
Kan-Mitchell J., Liggett P. E., Taylor C. R., Rao N., Granada E. S., Danenberg K. D., White W. L., Van Eldik L. J., Horikoshi T., Danenberg P. V. Differential S100 β expression in choroidal and skin melanomas: quantitation by the polymerase chain reaction.
Investig. Ophthalmol. Vis. Sci.
,
34
:
3366
-3375,  
1993
.
10
Danenberg P. V., Horikoshi T., Danenberg K. D. Quantitation of thymidylate synthase, dihydrofolate reductase, and DT-diaphorase gene expression in human tumors using the polymerase chain reaction.
Cancer Res.
,
52
:
108
-116,  
1992
.
11
Falk K., Rotzschke O. Consensus motifs and peptide ligands of MHC class I molecules.
Semin. Immunol.
,
5
:
81
-94,  
1993
.
12
Grey H. M., Ruppert J., Vitiello A., Sidney J., Kast W. M., Kubo R. T., Sette A. Class I MHC-peptide interactions: structural requirements and functional implications.
Cancer Surv. Mol. Mechanisms Immune Response
,
22
:
37
-49,  
1995
.
13
del Guercio M. F., Sidney J., Hermanson G., Perez C., Grey H. M., Kubo R. T., Sette A. Binding of a peptide antigen to multiple HLA alleles allows definition of an A2-like supertype.
J. Immunol.
,
154
:
685
-693,  
1995
.
14
Rivoltini L., Kawakami Y., Sakaguchi K., Southwood S., Sette A., Robbins P. F., Marincola F. M., Salgaller M. L., Yannelli J. R., Appella E. Induction of tumor-reactive CTL from peripheral blood and tumor- infiltrating lymphocytes of melanoma patients by in vitro stimulation with an immunodominant peptide of the human melanoma antigen MART-1.
J. Immunol.
,
154
:
2257
-2265,  
1995
.
15
Cai Z., Brunmark A., Jackson M. R., Loh D., Peterson P. A., Sprent J. Transfected Drosophila cells as a probe for defining the minimal requirements for stimulating unprimed CD8+ T cells.
Proc. Natl. Acad. Sci. USA
,
93
:
14736
-14741,  
1996
.
16
Luxembourg A. T., Borrow B., Teyton L., Brunmark A. B., Peterson P. A., Jackson M. R. Biomagnetic isolation of antigen-specific CD8+ T cells usable in immunotherapy.
Nature (Lond.) Biotechnology
,
16
:
281
-285,  
1998
.
17
Nelson R. E., Fessler L. I., Takagi Y., Blumberg B., Keene D. R., Olson P. F., Parker C. G., Fessler J. H. Peroxidasin: a novel enzyme matrix protein of Drosophila development.
EMBO J.
,
13
:
3438
-3447,  
1994
.
18
Lin, L-L., and Graham, J. Interleukin-1 receptor intracellular ligand proteins and inhibitors of ligand binding. Patent no. PCT/US96/06363. Washington, DC: United States Government Printing Office, 1995.
19
Garnier J., Gibrat J. F., Robson B. GOR method for predicting protein secondary structure from amino acid sequence.
Methods Enzymol.
,
266
:
540
-553,  
1996
.
20
Rogers S., Wells R., Rechsteiner M. Amino acid sequences common to rapidly degraded proteins: the PEST hypothesis.
Science (Washington DC)
,
234
:
364
-368,  
1986
.
21
Kirkin A. F., Dzhandzhugazyan K., Zeuthen J. Melanoma-associated antigens recognized by cytotoxic T lymphocytes.
Acta Pathol. Microbiol. Immunol. Scand.
,
106
:
665
-679,  
1998
.
22
Yewdell J. W., Bennink J. R. Cell biology of antigen processing and presentation to MHC class I molecule-restricted T lymphocytes.
Annu. Rev. Immunol.
,
52
:
1
-123,  
1992
.
23
Sette A., Sidney J., del Guercio M. F., Southwood S., Ruppert J., Dahlberg C., Grey H. M., Kubo R. T. Peptide binding to the most frequent HLA-A class I alleles measured by quantitative molecular binding assays.
Mol. Immunol.
,
31
:
813
-822,  
1994
.
24
Kan-Mitchell J., Huang X.-Q., Steinman L., Oksenberg J. R., Harel W., Parker J. W., Goedegebuure P., Darrow T. L., Mitchell M. S. Clonal analysis of in vitro-activated CD8+ cytotoxic T lymphocytes from a melanoma patient responsive to active specific immunotherapy.
Cancer Immunol. Immunother.
,
37
:
15
-25,  
1993
.
25
Huang X.-Q, Mitchell M. S., Liggett P. E., Murphree A. L., Kan-Mitchell J. Non-fastidious, melanoma-specific CD8+ cytotoxic T lymphocytes from choroidal melanoma patients.
Cancer Immunol. Immunother.
,
38
:
399
-405,  
1994
.
26
Mitchell M. S. A personal (biased) perspective on cancer “vaccines.”.
Oncol. Res.
,
9
:
459
-465,  
1997
.
27
Mitchell M. S. Perspective on allogeneic melanoma lysates in active specific immunotherapy.
Semin. Oncol.
,
25
:
623
-635,  
1998
.
28
Hsu F. J., Benike C., Fagnoni F., Liles T. M., Czerwinski D., Taidi B., Engleman E. G., Levy R. Vaccination of patients with B-cell lymphoma using autologous antigen-pulsed dendritic cells.
Nat. Med.
,
2
:
52
-57,  
1996
.
29
Minev B., Chavez F., Dudouet B. M., Mitchell M. S. Synthetic insertion signal sequences enhance class I presentation of a peptide from the melanoma antigen MART-1.
Eur. J. Immunol.
,
30
:
2115
-2124,  
2000
.
30
Gery I., Waksman B. H. Potentiation of the T-lymphocyte response to mitogens.
II. The cellular source of potentiating mediator(s). J. Exp. Med.
,
136
:
143
-155,  
1972
.
31
Gery I., Gershon R. K., Waksman B. H. Potentiation of the T-lymphocyte response to mitogens.
I. The responding cell. J. Exp. Med.
,
136
:
128
-142,  
1972
.
32
Gery I., Gershon R. K., Waksman B. H. Potentiation of cultured mouse thymocyte responses by factors released by peripheral leucocytes.
J. Immunol.
,
107
:
1778
-1780,  
1971
.
33
Mitchell M. S., Kirkpatrick D., Mokyr M. B., Gery I. On the mode of action of BCG.
Nat. New Biol.
,
243
:
216
-218,  
1973
.
34
Mitchell M. S., Mokyr M. B., Kahane I. Stimulation of lymphoid cells by components of BCG.
J. Natl. Cancer Inst.
,
55
:
1337
-1343,  
1975
.
35
Mokyr M. B., Mitchell M. S. Activation of lymphoid cells by BCG in vitro.
Cell Immunol.
,
15
:
264
-273,  
1975
.
36
Colagiovanni D. B., Shopp G. M. Evaluation of interleukin-1 receptor antagonist (IL-1RA) and tumor necrosis factor binding protein (TNF-BP) in a rodent abscess model of host resistance.
Immunopharmacol. Immunotoxicol.
,
18
:
397
-419,  
1996
.
37
Kuhns D., Alvord W., Gallin J. Increased circulating cytokines, cytokine antagonists, and E-selectin after intravenous administration of endotoxin in humans.
J. Infect. Dis.
,
171
:
145
-152,  
1995
.
38
Smith M., Jr., Carl V., Lodie T., Fenton M. Secretory interleukin-1 receptor antagonist gene expression requires both a PU.1 and a novel composite NF-κB/PU.1/GA-binding protein binding site.
J. Biol. Chem.
,
273
:
24272
-24279,  
1998
.
39
Puren A. J., Fantuzzi G., Gu Y., Su M. S., Dinarello C. A. Interleukin-18 (IFN-γ-inducing factor) induces IL-8 and IL-1β via TNFα production from non-CD14+ human blood mononuclear cells.
J. Clin. Investig.
,
101
:
711
-721,  
1998
.
40
Kimble R. B., Vannice J. L., Bloedow D. C., Thompson R. C., Hopfer W., Kung V. T., Brownfield C., Pacifici R. Interleukin-1 receptor antagonist decreases bone loss and bone resorption in ovariectomized rats.
J. Clin. Investig.
,
93
:
1959
-1967,  
1994
.
41
Burger R., Grosen E., Ioli G., Van Eden M., Brightbill H., Gatanaga M., DiSaia P., Granger G., Gatanaga T. Host-tumor interaction in ovarian cancer. Spontaneous release of tumor necrosis factor and interleukin-1 inhibitors by purified cell populations from human ovarian carcinoma in vitro.
Gynecol. Oncol.
,
55
:
294
-303,  
1994
.