Haplotype Loss of HLA Class I Antigen as an Escape Mechanism from Immune Attack in Lung Cancer

One of tumor escape mechanisms from the host's immunosurveillance system (i.e., a haplotype loss of HLA class I antigens) has been detected in various tumor cells. We hypothesize that the majority of tumor cells with normal HLA class I expression were attacked and eradicated by CTLs, and only a minority with an abnormal expression of HLA class I antigens could escape the host's immunosurveillance system. Using HLA class I–transfected tumor variants as stimulators in A904L lung cancer cell line, which has a haplotype loss of HLA class I antigens, both the transfected HLA-A26 and HLA-B39–restricted CTL lines were induced from autologous lymphocytes. However, only one HLA-B39–restricted CTL clone (CTL G3b) was established, and it was then used to identify the antigen. SGT1B [suppressor of G2 allele of SKP1 (SGT1), suppressor of kinetochore protein (SKP1)] was identified as the antigen recognized by CTL G3b. Further experiments using 13 subclones from a primary culture of A904L were found to confirm our above-mentioned hypothesis. Tumor cells with a normal HLA class I expression may thus be killed by CTL at an early stage of carcinogenesis, and only tumor cells with a haplotype loss of HLA class I antigens can escape an immune attack and develop into clinical cancer.

Marked progress in the field of tumor immunology using newly provided cellular and molecular technology has now enabled us to identify the tumor-associated antigens recognized by CTLs and antibodies (1–5). Evidence has shown that both cellular and humoral immunities against various targets exist even in advanced cancer patients. It is well known that CTLs exhibit HLA class I–restricted cytolytic activity against tumor cells; however, tumor cells have several mechanisms that allow them to escape from a host's immune surveillance system. However, the exact mechanisms of such tumor cells have yet to be elucidated.

It is widely known that tumor cells often cause a loss or down-regulation of HLA class I molecules on their surface, and this phenomenon is thought to be one of tumor escape mechanisms (6). An abnormal HLA class I expression of tumor cells has been reported in melanomas (7), squamous cell carcinomas of the head and neck (8), breast cancer (9), lung cancer (10), and carcinomas of the uterine cervix (11), and such an abnormal expression has become a major hurdle in performing CTL-based cancer immunotherapy. A haplotype loss of HLA class I antigens is also one of the most common causes of an abnormal HLA expression in various types of tumors (12). We have reported previously that DNA genotyping showed a haplotype loss of HLA class I antigens in 6 of 15 (40%) lung cancer cell lines established in our laboratory (13). We hypothesize that the majority of tumor cells with normal HLA class I expression bearing a strong antigen are attacked and eradicated by CTLs; thereafter, only the tumor cells with an abnormal expression of HLA class I antigens can avoid the host's immunosurveillance system and survive. The novel strategies of our study were as follows: (a) transfection of the deleted HLA class I cDNAs to a lung cancer cell line with haplotype loss of HLA class I antigens, (b) induction and establishment of CTL clones against the HLA-transfected tumor cells from autologous regional lymph node lymphocytes (RLNL), and (c) identification of the tumor antigen that is presented in the context of the transfected HLA class I antigens. The purpose of this study is to clarify the precise roles of HLA class I loss in tumor development.

Culture medium. The culture medium consisted of RPMI 1640 (Life Technologies, Grand Island, NY) supplemented with 10% heat-inactivated FCS (Equitech-Bio, Ingram, TX), 10 mmol/L HEPES, 100 units/mL penicillin G, and 100 mg/mL streptomycin sulfate (14).

Cell lines. A904L (HLA-A*2402, B*0702, and Cw*0702) is a lung large cell carcinoma cell line derived from patient A904 (15, 16). A904-EBV-B (HLA-A*2402/2603, B*0702/3901, and Cw*0702), an autologous EBV-transformed B-cell line, was produced from peripheral blood mononuclear cells (PBMC) by infection with the supernatant from the EBV producer line B95.8. Comparisons of the HLA genotype between these cell lines revealed that A904L had lost the HLA class I haplotype of HLA-A*2603, B*3901, and Cw*0702. 293-EBNA1 (HLA-A*0301, B*0702, and Cw*0702) is an adenovirus-transformed human kidney cell line that was purchased from Invitrogen (Carlsbad, CA). K562 is an erythroleukemia cell line, which is sensitive to natural killer cell cytotoxicity. All of these cell lines were maintained in 10% FCS-containing culture medium. HLA genotypes of the cell lines were tested by the Shionogi Co. (Osaka, Japan), and HLA peripheral blood serotypes were examined by SRL (Tokyo, Japan).

Preparation of dendritic cells. Autologous dendritic cells were generated from PBMCs. The PBMCs were suspended in culture medium (1 × 10⁶/mL) and were allowed to adhere in a six-well flat-bottomed culture plate (Iwaki glass, Tokyo, Japan). After 2 hours at 37°C, the adherent cells were cultured to induce immature dendritic cells in culture medium containing 500 units/mL recombinant human interleukin (IL)-4 (Serotec, Oxford, United Kingdom) and 1,000 units/mL recombinant granulocyte macrophage colony-stimulating factor (Sigma Chemical, St. Louis, MO) for 7 days. On day 5, the culture medium with 50 units/mL recombinant tumor necrosis factor-α was added to induce mature dendritic cells from the immature dendritic cells for 2 days. The culture medium was replenished every 3 days. On day 7, both A904-mature and A904-immature dendritic cells were harvested and then used for assays.

Induction of autologous lymphoblastoid cells. The nonadherent cells were collected from autologous PBMC and then were used as sources of lymphoblastoid cells. A904-phytohemagglutinin (PHA)-blast cells (HLA-A*2402/2603, B*0702/3901, and Cw*0702) were induced by stimulating the nonadherent cells with 0.001 mL PHA-P (Difco Laboratories, Detroit, MI) and 200 units/mL recombinant IL (rIL)-2 (kindly donated by Takeda Chemical Industries, Osaka, Japan). A904-pokeweed mitogen (PWM)-blast cells were induced by stimulating with 10 μg/mL PWM (Sigma Chemical).

Detection of loss of heterozygosity at 6p21.3 as a cause of HLA haplotype loss. According to information from an available database,¹

HLA class I genes are located in 6p21.3. For microsatellite analysis of 6p, genomic DNA was extracted from A904L and A904-EBV-B cells using the Qiagen genomic DNA purification kit (Valencia, CA). The PCR amplification was done on 20 ng DNA using pairs of primers specific for 6p microsatellite markers. In this study, three 6p microsatellite regions (detected by D6S497, D6S2414, and D6S1019) of both A904L and A904-EBV-B cells were compared to detect the loss of heterozygosity (LOH). The forward primer was fluorolabeled (PE Applied Biosystems, Foster City, CA), and conditions of PCR amplification were 94°C for 5 minutes, 94°C for 30 seconds, 55°C for 30 seconds, and 72°C for 30 seconds for 35 cycles and 72°C for 10 minutes. The PCR products were applied to ABI PRISM 310 automated sequencers (PE Applied Biosystems) and the results were analyzed by the GeneScan program (PE Applied Biosystems).

Cloning of HLA-A*2603 and B*3901 cDNAs from A904-EBV-B cells. Both HLA-A*2603 and HLA-B*3901 cDNAs were cloned from A904-EBV-B. Briefly, the total RNA of A904-EBV-B was extracted using the RNeasy Mini kit (Qiagen) and then was converted to cDNA using an oligo(dT) primer (Amersham Biosciences, Piscataway, NJ). The PCR was done using 1 μL plaque-forming units DNA polymerase (Stratagene, Heidelberg, Germany), 20 pmol forward primers (HLA-A; OKY-5, 5′-ACTGGGCGGATCCGGACTCAGAATCTCCCCAGACGCCGAG-3′, HLA-B; OKY-13, 5′-CGGGATCCGCCGAGATGCTGGTCAT-3′), and reverse primer (OKY-2; as described above). The HLA-A amplification protocol consisted of denaturation for 1 minute at 94°C, annealing for 1 minute at 57°C, and extension for 1 minute at 72°C for a total of 35 cycles. The HLA-B amplification protocol consisted of denaturation for 1 minute at 94°C, annealing for 1 minute at 65°C, and extension for 1 minute at 72°C for a total of 36 cycles. The PCR product was purified and cloned using the pcDNA3.1/V5-His TOPO TA Expression kit (Invitrogen). The cloned HLA-A*2603 and HLA-B*3901 cDNAs in the vector pcDNA3.1 were digested and thereafter ligated into another expression vector, pEAK10 (Edge BioSystems, Gaithersburg, MD), and then were used to transform bacteria TOP10 P3 (Invitrogen) according to the manufacturer's instructions. The transformed bacteria colonies were selected in a medium containing 50 μg/mL ampicillin and expanded for plasmid DNA preparation. The plasmid DNA was extracted and purified using QIAprep plasmid spin columns (Qiagen). The sequence of the cloned HLA cDNAs was confirmed using ABI PRISM 310 automated sequencers.

Transfection of autologous HLA-A*2603 or B*3901 into CD80-transfected A904L cells. The tumor cell line A904L-CD80 was established previously from a stable transfection of human CD80 cDNA to augment the efficacy of CTL induction (17, 18). In the present study, each HLA-A*2603 or HLA-B*3901 cDNA was transfected into A904L-CD80 independently; then, HLA class I–transfected autologous tumor cells were newly established. Briefly, A904L-CD80 (1 × 10⁶/mL) was cultured in six-well plates (Iwaki glass) at 37°C in a 5% CO₂ atmosphere and then was transfected with 1 μg of the pEAK10 plasmid containing autologous HLA-A*2603 or HLA-B*3901 cDNA using a LipofectAMINE reagent (Life Technologies) according to the user's manual. These cell lines were selected by 0.9 μg/mL puromycin (Life Technologies), and viable cells were then expanded to continuous cell lines (designated as A904L-A26 and A904L-B39, respectively). A stable transfection of the deleted HLA class I allele was confirmed by reverse transcription-PCR (RT-PCR) and sequencing as described above.

Induction of CTL lines and establishment of CTL clone. Autologous RLNLs obtained at the time of surgery were prepared and frozen in a deep freezer at −130°C until use as described previously (13, 19). After 1.4 × 10⁷ RLNLs were rapidly thawed, each of 2 × 10⁶ RLNLs was seeded in seven wells from 24-well plates (Iwaki glass) and then was weekly stimulated with 2 × 10⁵ of irradiated (100 Gy) A904L-A26 or A904L-B39 cells, respectively. The RLNLs were incubated with culture medium containing 25 units/mL rIL-2, 5 ng/mL IL-4, and 5 ng/mL rIL-7 (Genzyme, Cambridge, MA) for 3 weeks at 37°C in a 5% CO₂ atmosphere. One week after the third stimulation, the RLNLs were harvested, washed, and tested for their CTL activity. To establish CTL clones from CTL lines, a limiting dilution method was done as reported previously (20).

Monoclonal antibodies and flow cytometry. The hybridomas (HB-95 and HB-145) were purchased from the American Type Culture Collection (Rockville, MD). The culture supernatants of ATCC HB-95 (W6/32; anti-HLA class I) and HB-145 (IVA12; anti-HLA class II) were used to analyze the HLA restriction of T-cell clones. FITC-conjugated Nu-TH/I (anti-CD4, Nichirei Corp., Tokyo, Japan) and PE-conjugated Leu-2a (anti-CD8, Becton Dickinson, Mountain View, CA) were prepared for phenotypic analysis. The cells (2 × 10⁵/mL) were incubated with either monoclonal antibodies (mAb) or control IgG for 30 minutes at 4°C. Thereafter, they were washed and tested using a flow cytometer (EPICS XL, Coulter International, Fullerton, CA; ref. 19).

CTL assay. The lytic activity of CTLs was assessed by a standard 4-hour ⁵¹Cr release assay (13, 14). The CTL activity was also assayed in duplicate using a Human IFN-γ ELISA Test kit (Biosource, Camarillo, CA; ref. 16). In a blocking assay, one fourth of diluted culture supernatant of hybridomas was added to the coculture of CTLs and tumor cells (20).

T-cell receptor usage analysis. A T-cell receptor (TCR) Vβ usage of CTL G3b was assessed by RT-PCR and sequencing (16). Total RNA was extracted from CTL G3b and converted to cDNA, which served as a template for PCR amplification using a panel of Vβ-specific forward primer (OKY-43, 5′-TGCTCCCCTATCTCTGGGCA-3′) and the TCR-D-J joining region-specific reverse primer (OKY-400, 5′-CCATTCTGAGCTTCTTGAC-3′). The second PCR was done using the diluted first PCR products (1:100) as a template with the same forward and the newly designed TCR-D-J joining region-specific reverse primer (OKY-401, 5′-GTATCCGAATGCTGTCCGAGG-3′).

Construction and screening of the plasmid cDNA library. mRNA was extracted from the A904L cells and converted to cDNA with the Superscript Choice System (Life Technologies) using an oligo(dT) primer [5′-ATAAGAATGCGGCCGCTAAACTA(T)18VZ-3′; V = G, A, or C; Z = G, A, T, or C] containing a NotI site at its 5′ end. The cDNA was ligated to HindIII-EcoRI adaptors (Stratagene), phosphorylated, digested with NotI, and then inserted at the HindIII and NotI sites of the expression vector pCEP4 (Invitrogen). The bacteria TOP10 (Invitrogen) were transformed by electroporation with the plasmids and selected by 50 μg/mL ampicillin. The library containing 1.1 × 10⁵ cDNA clones was divided into 1,110 pools. Each pool containing ∼100 cDNA clones was amplified for 4 hours after extraction of plasmid DNAs using the QIAprep 8 plasmid kit (Qiagen). For the screening of this cDNA library, a transient transfection was done using a LipofectAMINE reagent. Briefly, 293-EBNA1 (3 × 10⁴ cells/mL) in a flat-bottomed 96-well dish were cotransfected with 100 ng plasmid DNA from each pool of the library and 50 ng pcDNA3.1 containing HLA-B*3901 cDNA. After 24 hours, CTLs (3 × 10³/mL) were incubated with these transfectants in culture medium containing 25 units/mL rIL-2 for 18 hours. The supernatants were then collected and production of IFN-γ was assessed. The positive pool of cDNA library was subcloned; finally, single cDNA clone was isolated as the antigen-coding gene recognized by the CTL.

Determination of the antigenic region recognized by CTL G3b. Four mini genes of SGT1B (SGT1B_1-202 corresponding to nucleotide position 1-202, SGT1B_1-342, SGT1B_1-535, and SGT1B_1-697) were obtained using PCR and a pcDNA3.1/V5-His TOPO TA Expression kit. To identify the location of the antigenic peptides, they were transiently cotransfected with HLA-B*3901 cDNA into 293-EBNA1 cells and then were underwent the screening as described above.

Peptides and CTL assay. To determine the presence of epitope peptide, 17 kinds of nonapeptides (>80% purity, listed in Fig. 4C) between codons 48 and 95 of SGT1 were synthesized and their sequences were verified by mass spectrometry. The peptides were obtained in a lyophilized form and dissolved in 100% DMSO at 10 μg/mL, and a stock solution of 100 ng/mL was then prepared by further dilution in the culture medium. A904-PHA-blast cells (3 × 10⁴/mL) were loaded with the peptides for 1 hour at room temperature and then were washed with culture medium twice. The CTLs (3 × 10³/mL) were then added in 100 μL culture medium containing 25 units/mL rIL-2 and incubated for 18 hours. The supernatants (70 μL) were collected to measure IFN-γ concentration.

Expression of SGT1 by reverse transcription-PCR. Total RNAs extracted from various normal tissues and tumor cell lines were converted to cDNAs as described above. The cDNAs served as templates for PCR amplification using SGT1-specific forward (OKY-365, 5′-CAGCAACAGCGACTACG-3′) and reverse primer (OKY-362, 5′-CCATTCTGAGCTTCTTGAC-3′). The amplification protocol consisted of denaturation for 1 minute at 94°C, annealing for 1 minute at 61°C, and extension for 2 minutes at 72°C for a total of 35 cycles.

Cloning of the parental A904 cells and HLA-B39-specific reverse transcription-PCR. The frozen stocked A904L cells at the primary culture were rapidly thawed, seeded at 0.3 cells per well in 96-well flat-bottomed plates, and then coincubated with 1 × 10⁵ irradiated (100 Gy) A904L cells at 37°C in a 5% CO₂ atmosphere because of the augmentation of cell-to-cell contact. Four weeks later, a total of 13 subclones of A904L were established. Total RNAs of A904L subclones were independently extracted and converted to cDNAs, which served as templates for PCR amplification using the HLA-B39-specific forward primer (OKY-204, 5′-GAACACACAGATCTGCAAGACCAACA-3′) and the reverse primer (OKY-2, 5′-ACTGCCCGAATTCTCTCAGTCCCTCACAAGGCAGCTGTC-3′).

Loss of heterozygosity on a region of chromosome 6 in A904L cells. To confirm whether A904L cells had a regional deletion of one allele on 6p, we analyzed microsatellite regions surrounding HLA class I genes in DNA extracted from A904L and A904-EBV-B cells. For all of the three regions, a retention of heterozygosity (ROH) was found in DNA from A904-EBV-B cells; however, LOH was found in DNA from A904L cells for two microsatellite regions (detected by D6S2414 and D6S1019; Fig. 1). The results indicate that, in A904L cells, the normal copy of HLA gene is lost because of a regional deletion of 6p21.

Establishment of HLA class I–transfected lung cancer cell lines. We reported previously that 15 lung cancer cell lines were established from 570 lung cancer specimens (2.6%), and a DNA genotype analysis showed haplotype loss of HLA class I antigens in 6 of the 15 (40%) lung cancer cell lines (13). One of them, A904L, possessed HLA-A*2402, B*0702, and Cw*0702 and lost the haplotype of HLA-A*2603, B*3901, and Cw*0702 compared with A904-EBV-B. The HLA-transfected lung cancer cell lines, A904L-A26 and A904L-B39, showed a similar level of HLA class I expression as A904L by flow cytometry (data not shown). Tumor doubling times of A904L, A904L-A26, and A904L-B39 cells are almost same (∼23 hours).

Induction of CTL lines. As described in Materials and Methods, 2 × 10⁶ RLNLs were stimulated thrice with 2 × 10⁵ irradiated A904L-A26 or A904L-B39 in the presence of IL-2, IL-4, and IL-7. The cytolytic activities of the thus-induced CTL lines were shown in Fig. 2. Figure 2A shows that the CTL line induced by stimulation with A904L-A26 showed a greater response to A904L-A26 than to parental A904L. The CTL line induced by stimulation with A904L-B39 also exerted higher cytolytic activity against A904L-B39 than parental A904L (Fig. 2B).

Establishment of CTL clones. We established three CTL clones from the CTL line specific for A904L-B39. One of them, CTL G3b, lysed A904L-B39 and A904-EBV-B but not A904L, A904-PHA-blast, A904-PWM-blast, A904-mature dendritic cells, A904-immature dendritic cells, or K562 (Fig. 3A). CTL G3b was also assessed for the cytokine production in response to tumor cells. High amount of IFN-γ production by CTL G3b was detected in response to A904L-B39 but not in response to A904L or A904L-A26 (Fig. 3B). IFN-γ production in response to A904L-B39 was completely inhibited by the addition of anti-HLA class I mAb but not by anti-HLA class II mAb (Fig. 3B). These results indicate that CTL activity of CTL G3b is certainly restricted by transfected HLA-B*3901 molecules. On the other hand, we also established three CTL clones from the CTL line specific for A904L-A26. However, all of them lysed A904L, A904L-A26, and A904L-B39 equally, thus suggesting that they might be restricted by the remaining HLA class I molecules (data not shown).

T-cell receptor usage of the CTL clone G3b. By a TCR analysis, the frequency of CTL G3b, which had Vβ5 and the same TCR-D-J joining region, was found to be 1/4.6 × 10⁶ of RLNLs.

Identification of gene encoding antigen recognized by CTL G3b. Using the cDNA expression cloning method described in Materials and Methods, cDNA 301.9 was isolated as the antigen-coding gene recognized by CTL G3b. CTL G3b only responded against 293-EBNA1 cotransfected with both HLA-B*3901 cDNA and cDNA 301.9 (Fig. 3C). cDNA 301.9 was 938 bp long and contained an open reading frame. According to DNA homology search using the available database (Genbank),²

the nucleotide sequence of cDNA 301.9 was found to be identical to the suppressor of G2 allele of SKP1B (SGT1B, accession no. AY321358), except for a deletion from nucleotide position 938 to the end of SGT1B (21). SGT1B was described to be a splicing variant of SGT1A (the original: SGT1, accession no. NM006704; ref. 22).

Identification of antigenic peptide recognized by CTL G3b. To identify the antigenic peptides of SGT1B, four types of plasmids containing SGT1B_1-697, SGT1B_1-535, SGT1B_1-342, and SGT1B_1-202 were manufactured (Fig. 4A). When both the plasmids and HLA-B*3901 cDNA were cotransfected into 293-EBNA1 cells, IFN-γ production by CTL G3b significantly decreased by the transfection of SGT1B_1-202-coding plasmid (Fig. 4B). This implies that the antigenic epitope recognized by CTL G3b is located between 202 and 342 bp of SGT1B. Seventeen nonapeptides located in this region were synthesized as listed in Fig. 4C. IFN-γ production by CTL G3b was evaluated against the A904-PHA-blast cells loaded with each of the 17 different peptides of SGT1B at an optimal 0.1 μmol/L concentration (Fig. 4C). Two peptides (CHILLGNYC and YCQFRAYCHI) were picked up as candidates of the antigenic peptide. To confirm the antigenic peptide, the A904-PHA-blast cells were pulsed with them (YCQRAYCHI and CHILLGNYC) and ADAKKSLEL (as a negative control) at various concentrations (Fig. 4D). The IFN-γ production by CTL G3b was detected in response to CHILLGNYC-loaded A904-PHA-blast cells in a dose-dependent manner, and CTL G3b thus recognized the peptide CHILLGNYC. CTL G3b actually responded against the peptide-pulsed A904L-B39 more than A904L-B39 (data not shown).

Expression of SGT1B in allogeneic tumor cell lines and normal tissues. To examine the expression of SGT1B in tumor cell lines and normal tissues, RT-PCR was done. As far as we tested, both SGT1A and SGT1B were broadly expressed in 10 lung cancer cell lines, 1 melanoma cell line, and 1 esophageal cancer cell line, and no significant difference in the expression levels among them was detected (Fig. 5A). Both of them were also ubiquitously expressed in all 20 normal tissues, A904-EBV-B, and A904-PHA-blast (Fig. 5B).

Absence of HLA-B39 in the parental A904L cells. To confirm whether the haplotype loss of HLA class I antigens occurred by in vivo immunoselection or in vitro generation, we attempted to obtain various subclones of A904L cells from the primary culture. A total of 13 subclones of A904L were established; then, their cDNAs were tested by RT-PCR for HLA-B39 expression. None of the 13 A904L subclones expressed HLA-B39. In addition, CTL G3b did not produce IFN-γ in response to these A904L subclones (Fig. 6). These results indicated that the haplotype loss of HLA class I antigens did not occur by in vitro selection but occurred by in vivo selection.

Tumor cells usually develop as a result of multistep processes, characterized by the accumulation of genetic mutational events (23). Garrido et al. reported that tumor cells often cause a loss or down-regulation of HLA class I antigens during tumor development, which is considered to be one of the mechanisms of immune escape (24). They also described that the multiplicity of molecular mechanisms leading to altered HLA expression within a tumor was suggested to increase number of immunoselected variants during its multistep evolution (25). In the early stages of cancer, it is likely that tumor cells with a normal HLA class I expression are killed by CTLs, when strong antigenic peptides are expressed in the context of HLA class I molecules.

Previous studies have shown that various molecular systems regulate expression of HLA class I molecules. For example, mutations of the transporter associated with antigen processing and β₂-microglobulin cause total loss of HLA class I antigens (7, 26). On the other hand, HLA class I gene mutations induce incomplete HLA class I alleles, which are unable to present the antigen. In a melanoma cell line, a point mutation of HLA-A2 gene of the tumor cells has been reported to cause an incomplete translation of HLA-A2 and induce HLA-A2 heavy chains without the α1 domain (27).

A haplotype loss of HLA class I antigens was found in 6 of 15 (40%) lung cancer cell lines and total loss of expression of HLA class I antigens was found in 3 of 15 (20%) cell lines established in our laboratory (13). Hiraki et al. also reported haplotype loss of HLA class I in 3 of the 7 (42.7%) lung cancer cell lines (10). A haplotype loss of HLA class I antigens has been reported to be caused by LOH on 6p21 (28–30). Furthermore, this HLA haplotype loss might result in a simultaneous deletion of tumor suppressor genes, such as p21 located in 6p21, which may be directly involved in either tumor development or progression in vivo (31, 32). We analyzed the microsatellite regions on 6p surrounding the HLA class I genes in DNA extracted from A904L and A904-EBV-B cells (Fig. 1). The microsatellite regions near HLA class I genes on 6p21 (D6S2414 and D6S1019) revealed LOH in A904L cells. We therefore conclude that, in A904L cells, the normal copy of HLA gene is lost because of a regional deletion of 6p21.

In the present study, CTL G3b recognized A904L-B39 but not the parental A904L. All of 13 subclones of A904L derived from the frozen primary culture exhibited loss of HLA-B39 (Fig. 6). This suggests that the haplotype loss of HLA class I antigens of A904L did not occur by in vitro selection but instead by in vivo immunoselection in which (a) the haplotype loss of HLA class I antigens might occur in a small population of A904L cells at early stage of carcinogenesis, (b) A904L cells with a normal expression of HLA class I antigens might be attacked and eradicated by the CTLs, and (c) A904L cells with a haplotype loss of HLA class I antigens might progress to clinical cancer.

As an antigen recognized by CTL G3b restricted by the transfected HLA-B*3901, cDNA 301.9 was isolated and found to be identical to SGT1B, which is a splicing variant of SGT1A (21). Kitagawa et al. reported that SGT1A, which is distributed widely from yeast to humans, was found to be associated with SKP1 in SCF (SKP1/Cullin-1/F-box protein) ubiquitin ligases and was an essential requirement for the ligase-dependent cell cycle regulation (22). The SCF complexes, which recruited specific substrates and catalyzed their ubiquitylation for proteasome degradation, played a broad role in regulating the stability and activity of many proteins in diverse physiologic processes (23). Both SGT1A and SGT1B were located on chromosome band 13q14.13 and contained a tetratricopeptide repeat domain (TPR domain), two variable regions (VR1 and VR2), a CS (Chord and Sgt1) domain, and a SGS (SGT1-specific) domain. SGT1B, however, contained an additional 33 amino acids encoded by a region between exons 5 and 6 of SGT1A in the TPR domain and a deletion of one amino acid (serine) at position 110 of SGT1A (21). An cDNA 301.9 was almost identical to SGT1B, except for loss of end of COOH-terminal region. The TPR domain is considered to consist of three or more TPR motifs, which are highly degenerated, and 34–amino acid repeats. The TPR domain plays an important role in controlling of the cell cycle, transcription, and protein transport (33). The antigenic peptide recognized by CTL G3b was located in the TPR domain (Fig. 4).

As shown in Fig. 3A, CTL G3b lysed A904L-B39 and A904-EBV-B but not A904L, A904-PHA-blast, A904-PWM-blast, A904-mature dendritic cells, A904-immature dendritic cells, or K562. Karanikas et al. described that the explanation for the lysis of autologous tumor cells but not normal cells by self-antigen–specific CTL clone is probably due to (a) the high level of the surface expression of HLA class I molecules on immortalized and rapid growing cells or (b) the different pathway for the antigenic peptide presentation between immortalized cells and normal cells (34). For example, some antigenic peptides in virally induced tumor-like cells (such as EBV-transformed B cells) have been reported to be processed by immunoproteasomes but not by standard proteasomes; therefore, the immortalized EBV-B cells may possibly be recognized by the immune system (34–36). In fact, CTL G3b recognized and lysed A904-EBV-B; therefore, A904-EBV-B may present the peptide CHILLGNYC of SGT1B by immunoproteasomes in the context of HLA-B39 and thus be recognized by CTL G3b.

In the present study, we revealed that the tumor cells with a haplotype loss of HLA class I antigens (A904L) maintained their ability to present tumor-associated antigens in the context of the lost HLA class I antigens because the HLA-transfected tumor cell (A904L-B39) was recognized by CTL G3b in a HLA-B*3901-restricted manner. CTL G3b, which recognized the transfected HLA class I/antigenic peptide complex, was found to infiltrate the draining lymph nodes. Furthermore, due to an attack of CTL, an immune response can be induced the SGT1B gene, which plays a key role for cell cycle regulation. Both SGT1A and SGT1B are ubiquitously expressed in cancer cells and normal cells (Fig. 5); however, only immortalized cells, including cancer cells, processed SGT1B protein to the antigenic peptide CHILLGNYC. The identification of such antigens restricted by deleted HLA class I antigens may therefore help us to obtain a better understanding of the above-described tumor escape mechanisms and may therefore eventually lead to the development of more effective cancer immunotherapy.

Grant support: Grants-in-aid for Scientific Research from the Ministry of Education, Science and Culture of Japan.

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.

We thank Dr. Katsumi Kitagawa (St. Jude Children's Research Hospital) and Dr. Hidetaka Uramoto and Dr. Manabu Yasuda of our laboratory for critical comments and Takashi Fukuyama, Yoshika Nagata, Kahoru Noda, Ayako Yamasaki, and Miki Shimada for their technical help.