Identification of a New Endoplasmic Reticulum-resident Protein Recognized by HLA-A24-restricted Tumor-infiltrating Lymphocytes of Lung Cancer¹ Free

Many genes encoding tumor antigens recognized by CTLs have been identified from melanoma cDNA (1, 2, 3, 4, 5, 6, 7, 8). CTL-directed tumor antigens have also been identified from tumors other than melanomas,including HER2/neu (9, 10), prostate-specific antigen(11, 12), SART-1 (13), SART-3(14), and cyclophilin B (15). Immunotherapy with some of the peptides capable of inducing HLA-class I-restricted CTLs has been shown to result in tumor regression in HLA-A0201⁺ melanoma patients (16, 17). These results indicate that identification of the peptides capable of inducing CTLs may provide a new modality of cancer therapy. However, little information is available on peptide-based immunotherapy for cancer patients other than melanoma patients. This may be attributable in part to the relative dearth of studies on target molecules recognized by CTLs infiltrating into cancers other than melanomas. Lung cancer is among the most commonly occurring malignancies in the world and is one of the few that continues to show an increasing incidence. The HLA-A24 allele is found in ∼60% of the Japanese population, 20% of Caucasians, and 12% of Africans(18). Therefore, we have studied antigens recognized by HLA-A24-restricted CTLs established from T cells infiltrating into lung adenocarcinoma and have reported a new CTL-directed tumor antigen residing in the ER.³

A bladder carcinoma cell line, HT1376, was used for preparation of the cDNA library. COS7 and C1R-A2402 (an HLA-A2402 transfectant) cells were used for the transfection and peptide-pulse experiments, respectively. The other cell lines used in this study were as follows: lung adenocarcinomas (1-87, PC-9, and A549), lung squamous cell carcinomas(Sq-1 and QG56), small cell lung carcinomas (LK79 and LC-65A), large cell lung carcinoma (86-2), esophageal squamous cell carcinomas (KE3,KE4, TE8, and TE9), head and neck cancers (HSC-4 and Kuma-1),hepatocellular carcinomas (HAK-3 and KYN-1), colon adenocarcinomas(KM12LM, colo201, and colo205), gastric adenocarcinoma (MKN28), and Epstein-Barr virus-transformed B-cell line (Bec-1). The origins and HLA genotypes of these cell lines have been described previously (13, 19, 20).

The expression-gene cloning method was used to identify a gene-encoding tumor rejection antigens recognized by the CTL line,GK-CTL, as reported elsewhere (15). Briefly, cDNA derived from the HT1376 bladder carcinoma cells was ligated to SalI/NotI site of the expression vector pSV-SPORT-1 (Life Technologies, Inc., Gaithersburg, MD). A total of 1 × 10⁵ clones from the cDNA library were screened, and two positive pools (6A1 and 4E6, each containing 100 clones) obtained from the first screening were subdivided for further screening. This study described the results for the 4E6-derived genes, whereas those for the 6A1-derived genes have been reported elsewhere (15). A total of 400 clones from the 4E6 were analyzed for their bioactivity to stimulate IFN-γ production by the GK-CTLs at the second and third screening,and one positive (4E6-2B9) and one negative (4E6-6F2) clone were provided for this study. DNA sequencing was performed by the dideoxynucleotide sequencing method using an AutoRead Sequencing kit(Pharmacia Biotech, Uppsala, Sweden) and analyzed using an ALF express DNA Sequencer (Pharmacia Biotech). Full-length cDNA clones were obtained from an HT1376 and a PBMC cDNA library (SuperScript Human Leukocyte cDNA library in pCMV-SPORT; Life Technologies) by the standard colony hybridization method with ³²P-labeled cDNA probe (clone 4E6-2B9) as reported previously (13). We tentatively designated this gene an ART-4 gene. The difference of the nt sequence at position 758 of ART-4 cDNA between PBMCs and HT1376 cells was confirmed by both repeated nt sequencing and susceptibility of PCR products against AciI restriction enzyme digestion. Amplification was performed for 30 cycles using the primers 5′-ATCCAAGCAGATCCAG CAGG-3′ (sense) and 5′-AGTGTGAGCAGAACACTCGG-3′(antisense).

Preparation of RNA, transfer to nylon membranes, and Northern hybridization have been described elsewhere (13). A ³²P-labeled 600-bp fragment of NcoI cut ART-4 cDNA was used as a probe. The membranes were washed and then autoradiographed. The relative expression of the ART-4 mRNA was calculated with the following formula:

A genomic DNA panel of hybrids with 1000-kb resolution made from hybrid cells of irradiated human HFL cells and hamster A23 cells (Gene bridge 4 Radiation Hybrid Panel; Research Genetics, Huntsville, AL; Ref.21) was used as a template for PCR. A 768-bp fragment of the ART-4 gene was amplified by PCR from the panel. PCR amplifications were performed with a sense primer, F2 (111–130;5′-ATCCAAGTGCTTGCACTCACA-3′), and an antisense primer, 4R878 (859–878;5′-AGTGTGAGCAGAACACTCGG-3′), under the following conditions: denaturing at 95°C for 1 min, annealing at 58°C for 1 min, and extension at 72°C for 1 min for 35 cycles. The PCR products were subsequently dot-blotted on a nitrocellulose filter and hybridized with the ³²P-end-labeled sequence-specific oligonucleotide probe R3 (452–471; 5′-AGGTTCTCAGGCTCACAAGC-3′), which specifically hybridized with the human ART-4 gene but not with any hamster-derived genes. After washing, the filter was subjected to autoradiography as described previously (22). The results obtained were analyzed using the radiation hybrid map software of the Whitehead Institute/Massachusetts Institute of Technology Center for Genome Research.

For preparation of the ART-4/GST protein, the BamHI site was introduced to the 5′-terminal of the ART-4 gene by a linker PCR. The primer pair used for the PCR was as follows: sense BamHI primer 5′-CTCACGGATCCAACATGGCTCC-3′ and antisense primer 5′-AATCGATGAGCTCACCTTAACCTT-3′. The amplified fragment (nt 3–364) was digested with BamHI and SacI and ligated to a SacI/NotI fragment of pSVSPORT/ART-4. Subsequently, the ligated fragment of the ART-4 gene containing nt positions 3–1733 was cloned into the BamHI/NotI site of the pGEX-4T-2vector (Pharmacia Biotech). In this construct, the GST gene flanked the 5′ terminal of the ART-4 gene. The nt sequence of the construct was confirmed by sequencing. Purification of the ART-4/GST fusion protein from the bacterial lysate was performed according to the manufacturer’s procedure. Polyclonal anti-ART-4/GST Ab was prepared by immunization of rabbits with purified ART-4/GST protein by the methods reported previously (13, 23).

For preparation of the ART-4/myc, the XbaI site was introduced to the 5′ flanking region of the stop codon by PCR. The primer pair used for the PCR was as follows: sense primer 5′-CGCTGCCATGGCTGTTT-3′ and antisense XbaI primer 5′-TGCGGGAACTCGCTCTAGAC-3′. The amplified fragment (nt 821-1256) was digested with XbaI and NcoI and ligated to a KpnI/NcoI fragment of pSVSPORT-ART-4. Subsequently, the ligated fragment of the ART-4 gene containing nt positions 1–1251 was cloned into the KpnI/XbaI site of the pcDNA3.1/Myc-Hisvector (Invitrogen, Carlsbad, CA). A myc gene flanked the 3′terminal of the ART-4 gene as a tag (ART-4/myc). To prepare the ART-4/GFP protein, the SalI site was introduced to the 5′ flanking region of the stop codon by PCR. The primer pair used for the PCR was as follows: sense primer 5′-TCGAATTCCACGCAGCCAA-3′ and antisense SalI primer 5′-AACTCGGTCGACACCTTTTCTTCAC-3′. The amplified product (nt 1–1267) was digested with EcoRI and SalI and cloned into an EcoRI/SalI site of pEGFP-N2 vector(Clontech, Palo Alto, CA). A GFP gene flanked the 3′terminal of ART-4 gene as a tag (ART-4/GFP).

Preparation of cytosol and nuclear fractions of tissues or cells have been described (23). Obtained fractions were separated by SDS-PAGE, blotted to a polyvinylidene difluoride membrane, and subjected for the Western blot. Precise methods of the Western blot analysis have been described previously (23).

Intracellular localization of the ART-4/GFP fusion protein in the transfectants was analyzed as follows. COS7 cells were transfected with the ART-4/GFP, followed by serial observation under a Zeiss confocal Ar-Kr laser scanning microscope with both fluorescence and visible rays or fluorescence only. Localization of the ART-4/GFP protein was recorded under an FITC filter (520 nm). The exposure sequences and imaging were controlled by LSM version 3.70 imaging software. The visible-ray image was studied by differential interference microscopy. COS7 cells transfected with Living Colors subcellular localization vectors pEYFP-ER,pEYFP-Golgi, or pEYFP-Mito (Clontech) were also analyzed to obtain representative fluorescence patterns of ER, Golgi,or mitochondria-resident proteins, respectively. The pEYFP-ER vector encodes a fusion protein consisting of EYFP,the ER targeting sequence of calreticulin, and the ER retrieval sequence, KDEL. pEYFP-Golgi encodes a fusion protein consisting of EYFP and a sequence of the NH₂-terminal 81 amino acids of humanβ1,4-galactosyltransferase to sort the fusion protein to the Golgi apparatus. pEYFP-Mito encodes a fusion protein consisting of EYFP and the mitochondrial targeting sequence from subunit VIII of human cytochrome c oxidase.

Subcellular organelles were fractionated by velocity-controlled sucrose gradient fractionation (24, 25). In brief, 5 × 10⁷ COS7 cells transfected with the ART-4/myc gene were washed with ice-cold STE buffer [0.25 m sucrose, 1 mmphenylmethylsulfonyl fluoride, 10 μg/ml aprotinin, 20 mm Tris-HCl (pH 8.0), and 1 mm EDTA]. Cells were suspended in a hypotonic buffer (0.025 m sucrose-STE), incubated for 30 min on ice, and homogenized with a tight fitting Dounce homogenizer. The homogenate was centrifuged at 500 × gfor 5 min to remove the nuclei and undisrupted cells. The supernatant was layered on a discontinuous sucrose gradient consisting of 1 ml of 2 m sucrose, 3 ml of 1.3 msucrose, 3 ml of 1 m sucrose, and 2.5 ml of 0.6 m sucrose, and then centrifuged at 40,000 rpm(284,000 × g) for 4 h at 4°C in a P40ST rotor (Hitachi, Tokyo, Japan). Fractions were collected from the top of the tube. Sucrose concentration in each fraction was as follows:the fraction numbers 1 and 2, 0.6 m; number 3,interface between 0.6 and 1 m; numbers 4 and 5, 1 m; number 6, interface between 1 and 1.3 m; numbers 7 an 8, 1.3 m;number 9, interface between 1.3 and 2 m; and number 10, 2 m. Each fraction was mixed with SDS-sample buffer and subjected to Western blot analysis. For detection of ER and Golgi marker proteins, rabbit anti-NADPH cytochrome P450 reductase (StressGen, Victoria, British Columbia, Canada) and anti-protein kinase C μ (Santa Cruz Biotechnology, Santa Cruz, CA)Abs were used, respectively (26, 27). Normal rabbit serum was used as a negative control.

Eighteen different synthetic peptides derived from the deduced amino acid sequence of ART-4 (purity was >70%) with binding motifs for HLA-A2402 molecules in the literature (28), including motifs of tyrosine or phenylalanine at position 2, and of isoleucine,leucine, phenylalanine, or tryptophan at position 9, were obtained from Biologicala (Nagoya, Japan). An HIV-derived peptide (RYPLTFGWCF)capable of binding to HLA-A24 molecules was used as a negative control(15). Peptides of >95% purity were used for experiments of dose dependency and CTL induction. The estimated score of half time of dissociation of each ART-4 peptide for HLA-A24 molecules was calculated based on a computer search for HLA peptide motifs(29) as follows: ART-4_50–59, 30;ART-4_61–69, 198;ART-4_75–84, 75;ART-4_156–164, 36;ART-4_238–246, 20;ART-4_265–274, 150;ART-4_309–317, 20; or ART-4_365–373, 150. The value of ART-4_13–20 was not obtained by the analysis because it is an 8-mer peptide. For detection of antigenic peptides recognized by the GK-CTLs, COS7 cells (2 × 10⁴) were transfected with HLA-A2402or control HLA-A2601 cDNA. Two days after the transfection,peptides at a concentration of 10 μm, unless stated otherwise, were added to the culture. Two h later, the supernatant was removed, and the GK-CTLs (1 × 10⁵) were added to the culture and incubated for an additional 18 h, and the concentration of IFN-γ in the culture supernatants was measured by ELISA (limit of sensitivity, 10 pg/ml) in a duplicate assay. The surface phenotype of the CTL line and sublines was investigated by an immunofluorescence assay with FITC-conjugated anti-CD3, anti-CD4, or anti-CD8 mAb (15). For inhibition of CTL activity, 10 μg/ml of anti-class I (W6/32,IgG2a), anti-HLA-A24 (A11.1 M, IgG3), anti-CD8 (Nu-Ts/c, IgG2a),anti-class II (H-DR-1, IgG2a), and anti-CD4 (Nu-Th/i, IgG1) mAbs were used as reported previously (15). Anti-CD14 (JML-H14,IgG2a) or anti-CD13 mAb (MCS-2, IgG1) was used as an isotype-matched control mAb. Two-tailed Student’s t test was used for the statistical analysis in this study.

PBMCs (1 × 10⁶ per well) from HLA-A2402⁺ healthy donors or cancer patients were incubated with 10 μm of the peptide in a 24-well plate in the presence of 100 units/ml interleukin 2 as reported previously (15). At days 7 and 14 of culture, the cells were restimulated at a responder to stimulator ratio of 4:1 with the irradiated (30 Gy) autologous PBMCs as APCs that had been incubated with the same peptide (10 μm) for 2 h. Responder cells were harvested at day 21 of the culture and were further cultured in a 96-well U-bottomed microculture plate in the presence of irradiated autologous PBMCs (2 × 10⁶ cells/well) as APCs that had been pulsed with a corresponding peptide. Seven to 10 days later, the expanded cells were transferred to a 24-well plate and cultured for additional 14–25 days with interleukin 2 alone. The cytotoxic activity was measured by a standard 6-h ⁵¹Cr-release assay at different E:T ratios, as reported previously (15).

The CTL line used for identification of the gene was the HLA-A24-restricted and tumor-specific effector CTL (GK-CTL) line with CD3⁺CD4⁻CD8⁺phenotype that was established from T cells infiltrating into the lung adenocarcinoma, and its characteristics have been reported elsewhere(15). A total of 10⁵ cDNA clones from the cDNA library of the HT1376 tumor cells were tested for their ability to stimulate IFN-γ production by the GK-CTLs after cotransfection with HLA-A2402 into the COS-7 cells. After repeated cycles of the screening, one clone (4E6-2B9) was confirmed to encode a tumor antigen recognized by the GK-CTLs when cotransfected with HLA-A2402 but not when cotransfected with control HLA-A2601 (Fig. 1). The 4E6-6F2 clone was served as a negative control in this experiment.

The clone 4E6-2B9 contained a 1584-bp-long DNA insert. Expression of this gene at the mRNA level was investigated by Northern blot analysis. A band of ∼1.8 Kbp in size was observed in each of the 21 different tumor cell lines tested, including 3 lung adenocarcinomas and 2 lung squamous cell carcinomas, and in all of the normal tissues tested(heart, brain, placenta, lung, liver, skeletal muscle, kidney,pancreas, spleen, thymus, prostate, testis, ovary, small intestine, and colon). Representative results are shown in Fig. 2,A. The relative level of mRNA expression ranged from 0.90 to 1.24 in tumor cell lines and from 0.78 to 4.40 in normal tissues, when the levels of the expression in the KE4 esophageal cancer cells were taken as 1.0 (Fig. 2 A). In the normal tissues, relatively high levels of expression were observed in the pancreas (4.4), kidney(2.8), and ovary (2.4), and low level expression was seen in skeletal muscle (0.6) and PBMCs (0.7). These results suggest that this gene is expressed ubiquitously at the mRNA level, and that the clone 4E6-2B9 with a length of 1584-bp might be a truncated form.

A cDNA library of HT1376 cells was further screened to obtain the full-length of cDNA clones containing the sequence of 4E6-2B9 by a standard colony hybridization technique, and a 1733-bp cDNA clone was obtained. A cDNA clone of the same length was also cloned from the cDNA library of PBMCs of a healthy donor using the 4E6-2B9 as a probe. The nt sequences of these clones are shown in Fig. 2 B. This gene was tentatively designated ART-4. The sequence of ART-4 cloned from HT1376 cDNA is available from EMBL/GenBank/DDBJ (accession number AB026125). The nt sequences of these clones were identical with the exception of the position 708(guanine in HT1376 versus adenine in the PBMCs). This difference is attributable to a genetic polymorphism rather than to a point mutation, because the position 708 was a guanine in the HT1376,RMG I, HSC-4, and LC-1 cells but an adenine in the PBMCs, MKN28, and 86-2 cells. PBMCs from another healthy donor and a tissue sample of the testis contained both guanine and adenine at position 708. The amino acids translated from these codons were different (arginine or glutamine). This gene showed 99% similarity at the nucleotide level with the previously reported gene HP-10 (30)isolated from placental cDNA and having an unknown function(30). Despite the high homology between the ART-4 and HP-10 gene at the nt levels, the homology between the two genes at the deduced amino acids level was only ∼50%. This discrepancy was mainly attributable to the difference of start positions for the translation and the subsequent frame shift of the codons. We next attempted to clone the HP-10 from a human placental cDNA library using PCR methods;however, the nt sequences of all 20 clones obtained from PCR products were identical with that of ART-4. Furthermore, the susceptibility of these clones to several restriction enzymes that distinguish the differences between ART-4 and HP-10 verified the sequencing results (data not shown). Therefore, we could not confirm the presence of HP-10 in human samples under the used conditions.

The deduced amino acid sequence of the ART-4 gene is shown in Fig. 2,C. Hydrophobicity analysis showed that a protein encoded by the ART-4 gene had a hydrophobic region at the NH₂ terminus (positions 1–21), potentially allowing it to act as a signal peptide (Fig. 2,D). There are nuclear localization signals at positions 396–412 and 401–407 and a di-lysine (KKXX) motif-like ER membrane retention signal (FVKK) at the COOH terminus positions 408–411 (Fig. 2,C). The ability of the ART-4 gene to stimulate IFN-γ production by the GK-CTLs was confirmed (Fig. 2 E), i.e., the ART-4 gene stimulated IFN-γ production by the CTLs in a dose-dependent manner when COS7 cells transfected with both the ART-4 and HLA-A2402 gene were used as stimulator cells.

Chromosome mapping of the ART-4 gene was performed by the radiation hybrid mapping method as reported elsewhere(21). A genomic DNA panel of hybrids with 1000-kb resolution made from hybrid cells of irradiated human HFL cells and hamster A23 cells was used as a template for PCR. The PCR with the primer pair F2 and 4R878 amplified the 768-bp fragment from genomic DNA isolated from PBMCs (data not shown). The PCR product did not contain an intron. A faint band corresponding to a PCR product of the same size was also amplified from hamster A23-derived genomic DNA under the same conditions. Therefore, the PCR products from the radiation hybrid panel were dot-blotted and subsequently hybridized with the human ART-4 sequence-specific oligonucleotide probe R3. This probe specifically hybridized with the PCR product from the human HFL cell-derived genomic DNA but not with the hamster A23-derived genomic DNA (data not shown). Database analysis for the radiation hybrid mapping indicated that the ART-4 gene was located on the chromosome 4, 2.12 centirays distal from the AFM350VH9marker gene, compatible with the classical 4q31.22.

Expression of ART-4 at the protein level in various cells and tissues was examined by Western blot analysis using rabbit anti-ART-4/GST polyclonal Ab. This method visualized an M_r 46,000 band of a recombinant ART-4 protein after cleavage with thrombin (data not shown). When the ART-4 gene was transfected to COS7 cells, an intensive M_r 46,000 band was observed in both the nuclear and cytosol fractions (Fig. 3,A). Furthermore, both the polyclonal anti-ART-4/GST Ab and anti-myc mAb recognized an M_r 51,000 band in the nuclear and cytosol fractions of COS7 cells transfected with the ART-4/myc gene. The difference in migration between these bands (M_r 46,000 and 51,000) was probably attributable to the myc-tag peptide(M_r ∼5,000). The M_r 46,000 band was detected in both the cytosol and nucleus of all of the tested tumor cell lines established from various organs, including the lung(n = 10), esophagus (n = 3), stomach (n = 3), head and neck(n = 4), uterus (n = 8), ovary (n = 8), and breast(n = 6). It was also detectable in leukemia cell lines, with the exception of the three cell lines (Jurkat, MOLT-4,and RPMI8402) from T-cell leukemia. This band was also detectable in both the cytosol and nucleus of the majority of cancer tissues from various organs tested, including the lung (n = 18), head and neck (n = 9), stomach(n = 8), uterus (n = 16), and breast (n = 8). In contrast,an M_r 46,000 protein of ART-4 was not detectable in either the cytosol or nuclear fraction of any of the normal tissues tested, except for those of the testis, placenta, and fetal liver. Some of these results are shown in Fig. 3,B, and a summary is shown in Table 1.

The ART-4/GFP protein was found in both the cytosol and nucleus of COS7 cells transfected with the ART-4/GFP gene (Fig. 3,C). On the basis of a comparison of the fluorescence distribution pattern of ART-4/GFP to that of control EYFP-fusion proteins (EYFP-ER, EYFP-Golgi, and EYFP-Mito), the ART-4/GFP protein in the cytosol might have been expressed at the ERs, but it would not have been expressed on the Golgi apparatuses or the mitochondrias. Furthermore, this ART-4 protein was detected in the fractions 1, 2, 7,and 9 when the subcellular organelles from the ART-4/myctransfectants were fractionated by the velocity-controlled sucrose gradient method (Fig. 3 D). The distribution pattern of this protein was very similar to that of NADPH cytochrome P450 reductase, a marker protein for ER-resident protein, but differed largely from that of protein kinase Cμ, a Golgi marker protein. These results together with the presence of an ER membrane retention signal at the COOH terminus at positions 408–411 strongly suggest that the ART-4 protein is an ER-resident protein.

Each of the 18 different ART-4-derived synthetic peptides with motifs of binding to HLA-A2402 molecules was loaded onto the HLA-A2402-transfected COS7 cells at a concentration of 10μ m and then tested for its ability to induce IFN-γ production by the GK-CTLs. Two of these peptides,ART-4_13–20 and ART-4_75–84, stimulated significant levels of IFN-γ production in a dose-dependent manner (Fig. 4, A and B). Similar dose-dependent effects of the two peptides were also observed when the peptides were loaded onto irradiated autologous PBMCs or C1R-A2402 cells (data not shown). The other 16 peptides failed to stimulate IFN-γ production by the CTLs. To confirm the presence of peptide-specific CTLs, the sublines were established from the parental GK-CTL line by a limiting dilution culture at 1 or 10 cells/well, and their peptide specificities were tested. Among the 68 different sublines tested, 7 sublines recognized C1R-A2402 cells pulsed with ART-4_13–20 but not with either ART-4_75–84 or HIV-derived peptide as a negative control. The CTL activity of those 7 sublines was inhibited by 10 μg/ml of anti-A24, anti-class I HLA (W6/32) or anti-CD8 mAb but not by anti-class II (H-DR-1), anti-CD4, or isotype-matched irrelevant control mAbs (anti-CD13 and anti-CD14). Representative results for one subline, #1-2, are shown in Fig. 4,C. The other 6 sublines recognized C1R-A2402 cells pulsed with ART-4_75–84 but not those pulsed with the other peptides, and their CTL activity was inhibited by anti-HLA-A24,anti-class I HLA (W6/32), or anti-CD8 mAb but not by the other mAbs. Representative results for one subline, #10-1, are shown in Fig. 4 D. The remaining 55 sublines failed to respond to any peptides tested (data not shown).

ART-4_13–20 and ART-4_75–84peptides were tested for their ability to induce CTLs from the PBMCs of six HLA-A24⁺ patients with lung cancer (four with adenocarcinomas and two with squamous cell carcinomas) and five HLA-A24⁺ healthy donors. The PBMCs stimulated with either peptide in the four patients (three with adenocarcinoma and one with squamous cell carcinoma) showed significant levels of cytotoxicity to HLA-A24⁺ 11-18 lung cancer cells but not to HLA-A24⁻ QG-56 lung cancer cells. Low but significant lytic activity of the CTL against 11-18 cells was probably attributable to the low expression of HLA-A24 molecules but not to that of the ART-4 peptides on the 11-18 cells, because a significant augmenting effect of exogenously loaded ART-4 peptide onto the 11-18 cells on the recognition by the CTLs was not observed (data not shown). Results of the three cases are shown in Fig. 5,A. In all three cases, the PBMCs failed to lyse HLA-A24⁺ PHA-blasts, VA13 fibroblast cells, or K562 tumor cells. These PBMCs possessed 20–60% of CD3⁺CD4⁻CD8⁺cells, and the remaining cells were mostly CD3⁺CD4⁺CD8⁻cells at the time of assays (data not shown). To confirm the peptide specificity, sublines derived from the patient 1 shown in Fig. 5,A were further established from the peptide-induced PBMCs,followed by a test of their reactivity to ART-4_13–20, ART-4_75–84,or HIV peptide. Three of 69 sublines tested from the PBMCs stimulated with the ART-4_13–20 responded to the C1R-HLA-A2402 cells pulsed with ART-4_13–20. IFN-γ production by these sublines was inhibited by 10 μg/ml of anti-HLA-A24, anti-class I HLA (W6/32), or anti-CD8 mAb but not by anti-class II (H-DR-1), anti-CD4, or irrelevant control mAbs (anti-CD13 and anti-CD14). The results for one subline,#1-10⁻⁹, are shown in Fig. 5,B. On the other hand, 2 of 17 sublines tested from the PBMCs stimulated with the ART-4_75–84 responded to C1R-HLA-A2402 cells pulsed with ART-4_75–84peptide, and the CTL activity of these sublines was inhibited by 10μg/ml of anti-HLA-A24, anti-class I HLA (W6/32), or anti-CD8 mAb but not by the other mAbs. The results for one subline,#6-10⁻⁷, are shown in Fig. 5 C. In the remaining two cancer patients, HLA-A24-restricted CTL activity was not induced in the PBMCs stimulated with either peptide. In these PBMCs, the percentages of CD3⁺CD4⁻CD8⁺cells were very low (<8%). HLA-A24-restricted CTL activity was also undetectable in the PBMCs of all healthy donors (n = 5) after stimulation with either peptide (data not shown).

This study reports that the ART-4 gene encodes antigenic epitopes recognized by the HLA-A24-restricted and tumor-specific CTLs. The ART-4 had 99 and 50% homology with the HP-10 at the nt and the deduced amino acid levels,respectively. The HP-10 was originally cloned from a human placenta cDNA library using human hexokinase mAb (30). However, the HP-10 did not show such hexokinase activity,and the function of the HP-10 protein has been reported as regulation of protein phosphorylation only in Escherichia coli, not in eukaryotes (30). When we screened the HT1376 and PBMC cDNA libraries to obtain the full-length ART-4 gene by colony hybridization, none of the obtained clones had the HP-10sequence. To confirm the existence of the HP-10 gene in human samples, we then tried to clone the HP-10 gene from a human placenta cDNA library by means of PCR. We analyzed 20 clones isolated from the PCR products by nt sequencing and also by digestion with restriction enzymes, and the results indicated that all of them were the ART-4 gene (data not shown). With the exception of HP-10, none of the genes reported previously have significant homology to the ART-4 gene, based on a search of the literature. Therefore, we were unable to confirm the presence of the HP-10 gene in human samples under the used conditions. The HP-10 might be a family gene of the ART-4gene, the message of which was rarely expressed in the placenta,HT-1376 tumor cells, and PBMCs.

There are several motifs in the sequence of the ART-4 protein: a nuclear localization signal at positions 396–412, a cAMP- or cGMP-dependent protein kinase phosphorylation site at positions 294–297, and a di-lysine (KKXX)-like ER membrane retention signal at the COOH terminus. The KKXX motif has been found previously at the COOH terminus of several ER-resident types I and III membrane proteins and shown to be responsible for the localization to the ERs(31). Membrane-topology prediction suggests that the ART-4 is a type Ib membrane protein. Results of the present laser-confocal microscopic analysis of the GFP-tagged ART-4 transfectants suggest that the ART-4 protein was localized at the ER and in the nucleus but not at the Golgi or on the plasma membrane. The localization of ART-4 at the ER was further confirmed by the subcellular organelle fractionation of a tagged ART-4 protein in the transfectants. All of the results suggest that the ART-4 is an ER-resident protein encoding tumor-rejection antigen recognized by the HLA-A24-restricted CTLs. We have identified recently cyclophilin B as the tumor-rejection antigen recognized by the GK-CTLs(15), and this molecule is also known as an ER-resident protein involved in the signal transduction of cellular proliferation(32). ER is well known as a major site of peptide loading to HLA molecules (33, 34). Therefore, it might be interesting to study the relationship between CTL-directed epitopes and ER-resident proteins. However, currently identified functional domains involved in the signal transduction of cellular proliferation, such as kinase domains or src homologous domains, are not found in the deduced ART-4 sequence. Therefore, we were unable to hypothesize any biological functions of this newly identified ER-resident protein,and further studies on the function of the ART-4 will be needed.

Under the present conditions, the ART-4 protein was not detectable by Western blot analysis with polyclonal anti-ART-4/GST Ab in any of the normal tissues, except for the testis, fetal liver, and placenta, despite the ubiquitous expression of the ART-4gene at the mRNA level. We have reported similar expression patterns for the SART-1 (13) and SART-3 (14) proteins,but the mechanisms involved in the discrepancy between the mRNA and protein expression are presently unclear.

Among the 18 peptides with HLA-A24 antigen-binding motifs tested, the two ART-4-derived peptides, ART-4_13–20 and ART-4_75–84, were consistently recognized by the HLA-A24-restricted CTLs in repeated experiments. Because of the presence of each of the peptide-specific CTL sublines, the parental GK-CTL line would consist of a mixture of these peptide-specific CTL clones. A dose dependency was observed in the two peptides, and >0.01μ m of each the peptides were required for the recognition of target cells by the CTLs. Both peptides were able to induce HLA-A24-restricted CTLs from PBMCs of the HLA-A24⁺ lung cancer patients but not from those of the HLA-A24⁺ healthy donors. These results suggest that there is a preferential presence of the CTL precursors reacting to the ART-4 epitopes in PBMCs of these cancer patients. In contrast, the CTL precursors may not exist in PBMCs of healthy donors.

Regardless of the numerous unsolved issues, including identification of its biological roles, the ART-4 antigen might be an ideal target molecule for use in specific immunotherapy of relatively large numbers of cancer patients, because it could be a cancer-specific protein expressed in the cytosol (ER) of the vast majority of cancers from various tissues with different histological types. The other tumor-rejection antigens, including MAGE family antigens (1, 7), SART-1 (13), and SART-3 (14), are also preferentially expressed in cancer cells with different histological types. Although the mutated antigens, including mutant CDK4 (35) and mutant CASP 8 (36), are recognized by CTLs, these types of antigens are expressed only in certain cancer cells, and the mutated products are usually detected in only a small part of tumor samples. Thus, these molecules would not be applicable as targets in specific immunotherapy of a large number of cancer patients.

The HLA-A24 allele is found in 60% of Japanese (with 95% of these cases being genotypically A2402), in 20% of Caucasians, and in 12% of Africans (18). The two ART-4-derived peptides were able to induce HLA-A24-restricted and tumor-specific CTLs in PBMCs of lung cancer patients. These ART-4 peptides might be appropriate molecules for use in the specific immunotherapy of HLA-A24⁺patients with lung and other cancers.