Identification of Peptide Vaccine Candidates for Prostate Cancer Patients with HLA-A3 Supertype Alleles

Prostate cancer is one of the most common cancers among elderly men (1). Despite the fact that androgen withdrawal therapy is transiently effective for prostate cancer, there is no effective therapy against recurrent hormone-refractory and metastatic prostate cancer. For such patients, specific immunotherapy may be a promising option because prostate cancer–reactive T cells could detect multiple metastases with fine specificity. Thus far, many cancer-related antigens and their peptides that are recognized by CTLs have been identified (2). In addition, several antigenic peptides derived from prostate-specific antigen (PSA; refs. 3–5), prostate-specific membrane antigen (PSMA; refs. 6, 7), prostatic acid phosphatase (PAP; refs. 8, 9), or prostate stem cell antigen (10–12) have been identified. As is the case with melanocyte differentiation antigens for melanoma, these prostate-related antigens have been considered promising targets in specific immunotherapy for prostate cancer patients (13). However, the peptide vaccine candidates identified to date have been focused on the HLA-A2 and HLA-A24 alleles because of the higher worldwide frequency of these alleles.

Based on the structural similarities of the group of HLA alleles and the peptide binding motif analysis, the following supertypes have been proposed: HLA-A2, HLA-A3, HLA-B7, and HLA-B44 supertype alleles (14). Among them, the A3 supertype allele is found in 38% of Caucasians, 53% of Chinese, 46% of Japanese, and 43% of North American African Americans and Hispanics (14). It has been shown that the same epitope peptides, derived from viral protein and melanoma antigen, were recognized by different HLA-A3 supertype alleles (15–17). In this study, we tried to identify prostate-related antigen-derived peptides that are applicable for HLA-A3 supertype⁺ prostate cancer patients, to facilitate the development of a peptide-based anticancer vaccine for prostate cancer patients with alleles other than HLA-A2 and HLA-A24.

Patients. Peripheral blood mononuclear cells (PBMC) were obtained from prostate cancer patients and healthy donors who had provided written informed consent. These included HLA-A11⁺, HLA-A31⁺, and HLA-A33⁺ patients, but PBMCs from HLA-A3⁺ or HLA-A68.1⁺ patients were not available because of their extremely low frequency (1.6% and 0.5%) in the Japanese population (18). None of the participants was infected with HIV. Twenty milliliters of peripheral blood were obtained, and PBMCs were prepared by Ficoll-Conray density gradient centrifugation. All of the samples were cryopreserved until they were used for the experiments. The expression of HLA-A11, HLA-A31, and HLA-A33 molecules on PBMCs of cancer patients was determined by flow cytometry using the following antibodies: anti-HLA-A11 monoclonal antibody (mAb; One Lambda, Inc., Canoga, CA), anti-HLA-A31 mAb (One Lambda), and anti-HLA-A33 mAb (One Lambda).

Cell lines. LNCaP is an HLA-A*0201⁺ prostate carcinoma cell line. C1R-A11, HLA-A31, and HLA-A33 are sublines that were stably transfected with the HLA-A1101, HLA-A3101, and HLA-A3303 genes, respectively. The expressions of HLA-A11, HLA-A31, and HLA-A33 molecules on these sublines were previously reported (17). To generate LNCaP sublines expressing each of the HLA-A11, HLA-A31, and HLA-A33 molecules, an HLA-A1101, HLA-A3101, or HLA-A3303 plasmid cDNA was inserted into the eukaryotic expression vector pCR3.1 (Invitrogen, Carlsbad, CA) by a method reported previously (19). Electroporation was done using a Gene Pulser (Bio-Rad, Richmond, CA). LNCaP-A11, HLA-A31, and HLA-A33 are sublines that were stably transfected with the HLA-A1101, HLA-A3101, and HLA-A3303 genes, respectively. All of the cell lines were maintained in RPMI 1640 (Invitrogen) with 10% FCS.

Induction of peptide-specific CTLs from peripheral blood mononuclear cells. Assays for the detection of peptide-specific CTLs were done according to a previously reported method with several modifications (20). PBMCs (1 × 10⁵ cells per well) were incubated with 10 μL/mL of each peptide in quadruplicate in a U-bottomed-type 96-well microculture plate (Nunc, Roskilde, Denmark) in 200 μL of culture medium. The culture medium consisted of 45% RPMI 1640, 45% AIM-V medium (Life Technologies, Gaithersburg, MD), 10% FCS, 100 units/mL of interleukin-2, and 0.1 mmol/L MEM nonessential amino acid solution (Life Technologies). Every 3 or 4 days, half the culture medium was removed and replaced with new medium containing the corresponding peptide (20 μg/mL) and 100 units/mL interleukin-2. On the 15th day of culture, the cultured cells were separated into four wells. Two wells were used for the culture with the corresponding peptide-pulsed C1R-A11, C1R-A31, or C1R-A33 cells, and the other two were used for the culture with HIV peptide–pulsed C1R-A11, C1R-A31, or C1R-A33 cells. The successful induction of peptide-specific CTLs was judged to be positive when a significant value of P < 0.05 was reached by a two-tailed Student's t test and when the difference in IFN-γ production compared with the HIV peptide was >50 pg/mL. After an 18-hour incubation, the supernatant was collected and the level of IFN-γ was determined by ELISA.

Peptides. The 10 PSA-derived peptides, 12 PAP-derived peptides, and 20 PSMA-derived peptides that are listed in Table 1 were prepared based on the binding motifs to the HLA-A11, HLA-A31, and HLA-A33 molecules (21). All peptides were of >90% purity and were purchased from the Biologica Co.(Nagoya, Japan). Influenza (Flu) virus–derived, EBV-derived, tyrosinase-related protein 2–derived, and HIV-derived peptides were used as controls binding to HLA-A3 supertype alleles. All peptides were dissolved with DMSO at a dose of 10 mg/mL.

Cytotoxicity assay. Peptide-stimulated PBMCs were tested for their cytotoxicity against LNCaP, LNCaP-A11, LNCaP-A31, or LNCaP-A33 by a standard 6-hour ⁵¹Cr release assay. Phytohemagglutinin-activated T cells were used as a negative control. Two thousand ⁵¹Cr-labeled cells per well were cultured with effector cells in 96-round-well plates at the indicated effector/target ratio. Immediately before the cytotoxicity assay, CD8⁺ T cells were positively isolated using a CD8-positive Isolation Kit (Dynal, Oslo, Norway). The specific ⁵¹Cr release was calculated according to the following formula: (test cpm − spontaneous cpm). Spontaneous release was determined by the supernatant of the sample incubated with no effector cells, and the total release was then determined by the supernatant of the sample incubated with 1% Triton 100-X (Wako Pure Chemical Industries, Osaka, Japan). In some experiments, 10 μg/mL of either anti-HLA-class I (W6/32: mouse IgG2a), anti-HLA-DR (L243: mouse IgG2a), anti-CD4 (NU-TH/I: mouse IgG1), anti-CD8 (NU-TS/C: mouse IgG2a), or anti-CD14 (H14: mouse IgG2a) mAb were added into wells at the initiation of the culture.

Cold inhibition assay. The specificity of peptide-stimulated CTLs was confirmed by a cold inhibition assay. In brief, ⁵¹Cr-labeled target cells (2 × 10³ cells per well) were cultured with the effector cells (2 × 10⁴cells per well) in 96-round-well plates with 2 × 10⁴ cold target cells. C1R-A11, C1R-A31, and C1R-A33, which were prepulsed with either the HIV peptide or a corresponding peptide, were used as cold target cells.

Detection of peptide-specific immunoglobulin G. Peptide-specific immunoglobulin G (IgG) levels in plasma were measured by ELISA as previously reported (22). Briefly, a peptide (20 μg per well)–immobilized plate was blocked with Block Ace (Yukijirushi, Tokyo, Japan), and 100 μL per well of plasma sample diluted with 0.05% Tween 20-Block Ace was added to the plate. After a 2-hour incubation at 37°C, the plate were washed and further incubated for 2 hours with a 1:1,000-diruted rabbit anti-human IgG (γ-chain specific; DAKO, Glostrup, Denmark). The plate was washed, and then 100 μL of 1:100-diluted goat anti-rabbit IgG-conjugated horseradish peroxidase (En Vision, DAKO) were added to each well, and the plate was incubated at room temperature for 40 minutes. After the plate was washed again, 100 μL per well of tetramethyl benzidine substrate solution (KPL, Guildford, United Kingdom) were added, and the reaction was stopped by the addition of 1 mol/L phosphoric acid. To estimate peptide-specific IgG levels, we compared the absorbance values of each sample with those of serially diluted samples, and the values are shown as absorbance units/mL. To confirm the specificity of IgG-reactive to relevant peptides, samples were cultured in peptide-coated plates, and the levels of peptide-specific IgG in the supernatant were determined by ELISA. IgG reactive to a corresponding peptide was judged significant when the absorbance in 1:100-diluted plasma exceeded 1.5 times the absorbance of no peptide control samples. Significance was evaluated for each plasma sample.

Statistics. The statistical significance of the data was determined using a two-tailed Student's t test. P < 0.05 was considered statistically significant.

Detection of immunoglobulin G reactive to the prostate-specific antigen, prostatic acid phosphatase, and prostate-specific membrane antigen peptides. First, we prepared 42 peptides derived from PSA, PAP, and PSMA, based on the binding motifs to the HLA-A3, HLA-A11, HLA-A31, HLA-A33, and HLA-A68.1 molecules (Table 1). Although these five HLA-A alleles share binding motifs, we preferentially considered the binding capacity to HLA-A11, HLA-A31, and HLA-A33 molecules, because both HLA-A3⁺ and HLA-A68.1⁺ are found very rarely in Japanese. We next examined the ability of these peptides to be recognized by IgGs of prostate cancer patients, because we have previously observed that IgGs reactive to CTL-directed peptides are frequently detected in the plasma of patients with several types of cancers (22, 23), and that IgGs reactive to PSA-derived or PSMA-derived CTL-directed peptides are detectable in the plasma of prostate cancer patients and healthy donors (24, 25). IgG reactive to a corresponding PSA, PAP, or PSMA peptide was judged significant when the absorbance in 1:100-diluted plasma exceeded 1.5 times the absorbance of no peptide control samples. Significance was evaluated for each plasma sample. The results of HLA-A3 supertype⁺ cancer patients are shown in Table 2. IgG reactive to the PSA_16-24 was detected in the plasma of prostate cancer patients most frequently in 10 PSA peptides. IgGs reactive to either the PAP_155-163 or PAP_248-257 peptide were detected in the plasma of prostate cancer patients more frequently than IgGs reactive to the other 12 PAP peptides. IgGs reactive to either the PSMA_207-215 or PSMA_431-440 peptide were detected in the plasma of prostate cancer patients more frequently than IgGs reactive to the other 20 PSMA peptides. In contrast, the PSA_36-45, PAP_19-28, or PSMA_199-207 peptide was rarely recognized by plasma IgG of HLA-A3 supertype⁺ prostate cancer patients. All results for the HLA-A3 supertype⁺ or supertype⁻ prostate cancer patients and HLA-A3 supertype⁺ or supertype⁻ healthy donors are summarized in Table 3. Although the positive percentages differed, IgGs reactive to peptides were detected in healthy donors. Healthy donors included both males and females, whereas there was no difference between the sexes in positive percentages (data not shown).

We further confirmed the validity of the assay of peptide-specific IgGs. Representative results are shown in Fig. 1. The specificity of IgGs against the PSA_16-24, PAP_155-163, PAP_248-257, PSMA_207-215, and PSMA_431-440 peptides was confirmed by an antibody absorption test. Thus, the levels of IgGs reactive to each peptide in the plasma of patients were significantly diminished by incubating the samples in wells coated with the respective peptides.

Induction of peptide-specific CTLs from the peripheral blood mononuclear cells of prostate cancer patients. We next determined whether or not five peptide candidates that were frequently recognized by IgGs in cancer patients could induce peptide-specific CTLs from the PBMCs of HLA-A11⁺, HLA-A31⁺, and HLA-A33⁺ prostate cancer patients. The PSA_36-45, PAP_19-28, and PSMA_199-207 peptides were employed as controls that were recognized by IgG less frequently. The PBMCs were stimulated in vitro with each of PSA-, PAP-, and PSMA-derived peptides or control peptides and examined for their IFN-γ production in response to corresponding peptide-pulsed C1R-A11, C1R-A31, or C1R-A33 cells. Representative results of 15 A3 supertype⁺ prostate cancer patients are shown in Table 4. Only the positive results are shown. Successful induction of peptide-specific CTLs was judged to be positive when P < 0.05 and when the difference in IFN-γ production compared with the control HIV peptide was >50 pg/mL. Thus, the PSA_16-24, PAP_155-163, PAP_248-257, PSMA_207-215, and PSMA_431-440 peptides induced corresponding peptide-reactive CTLs from the PBMCs of three, one, one, zero, and zero of five HLA-A11⁺ cancer patients; three, three, one, two, and one of five HLA-A31⁺ cancer patients; and two, three, four, three, and two of five HLA-A33⁺ cancer patients, respectively. The results from HLA-A3 supertype⁻ cancer patients and healthy donors are summarized in Table 5. No peptide-specific CTLs were induced from the PBMCs of HLA-A3 supertype⁻ patients or healthy donors. Although healthy donors consisted of males and females, peptide-specific CTLs were induced from the PBMCs of HLA-A3 supertype⁺ in both sexes (data not shown). These findings indicate that the PSA_16-24, PAP_155-163, PAP_248-257, PSMA_207-215, and PSMA_431-440 peptides have the potential to generate peptide-specific CTLs in the PBMCs of prostate cancer patients and healthy donors with HLA-A3 supertype alleles.

Induction of prostate cancer–reactive CTLs from the peripheral blood mononuclear cells of prostate cancer patients with HLA-A3 supertype alleles. We further determined whether or not CTLs induced by in vitro stimulation with each of the PSA_16-24, PAP_155-163, PAP_248-257, PSMA_207-215, and PSMA_431-440 peptides could show cytotoxicity against prostate cancer cells. To confirm HLA-A3 supertype allele–restricted cytotoxicity, LNCaP sublines expressing each of the HLA-A11, HLA-A31, and HLA-A33 molecules were prepared (data not shown). The PBMCs from HLA-A11⁺ patient 9 and healthy donor 14, HLA-A31⁺ patient 22, and HLA-A33⁺ patient 19 were stimulated with each of the PSA_16-24, PAP_155-163, and PSMA_207-215 peptides, and it was determined whether or not peptide-reactive CTLs could show cytotoxicity against prostate LNCaP cells expressing the HLA-A11, HLA-A31, and HLA-A33 molecules, respectively (Fig. 2). The PBMCs from patient 9 and healthy donor 14, which were stimulated in vitro with each of the PSA_16-24, PAP_155-163, and PSMA_207-215 peptides, exhibited a higher level of cytotoxicity against LNCaP-A11 than against LNCaP and HLA-A11⁺ T-cell blasts. Similarly, these peptides possessed the ability to induce prostate cancer–reactive CTLs from PBMCs of HLA-A31⁺ patient 22 and HLA-A33⁺ patient 19. That is, these peptide-specific CTLs showed a higher level of cytotoxicity against LNCaP-A31 or LNCaP-A33 than against LNCaP or T-cell blasts. Similar results were observed in the cases with the PAP_248-257 and PSMA_431-440 peptides (data not shown). Together, these results indicate that the PBMCs stimulated in vitro with the PSA_16-24, PAP_155-163, PAP_248-257, PSMA_207-215, or PSMA_431-440 peptide could show cytotoxicity against prostate cancer cells in an HLA-A11-, HLA-A31-, or HLA-A33-restricted manner.

Peptide-specific and CD8⁺ T cell–dependent cytotoxicity against prostate cancer cells. We further tried to identify the cells responsible for the cytotoxicity of peptide-stimulated PBMCs. Purified CD8⁺ T cells were used in the following two experiments. As shown in Fig. 3, the cytotoxicity of the PSA_16-24, PAP_155-163, or PSMA_207-215 peptide-stimulated PBMCs from HLA-A11⁺ patient 9 and HLA-A11⁺ healthy donor 10, HLA-A31⁺ patient 22, and HLA-A33⁺ patient 19 was significantly inhibited by the addition of anti-HLA class I mAb but not by the addition of anti-HLA class II (HLA-DR) or anti-CD14 mAb. Similar results were observed in the cases with the PAP_248-257 and PSMA_431-440 peptides (data not shown). These results indicated that the cytotoxicity of peptide-stimulated PBMCs against prostate cancer cells was dependent on HLA class I–restricted CD8⁺ T cells. In addition, their cytotoxicity against LNCaP-A11, LNCap-A31, and LNCap-A33 was significantly suppressed by the addition of corresponding peptide-pulsed unlabeled C1R-A11, C1R-A31, and C1R-A33 cells but not by HIV peptide-pulsed unlabeled C1R-A11, C1R-A31, or C1R-A33 cells (Fig. 4). Similar results were observed in the cases with the PAP_248-257 and PSMA_431-440 peptides (data not shown). The results of the cold competition assay indicate that the cytotoxicity of peptide-stimulated PBMCs against prostate cancer cells could be ascribed to the corresponding peptide-specific CD8⁺ T cells.

Cytotoxicity of peptide-stimulated peripheral blood mononuclear cells against prostate cancer cells sharing HLA-A3 supertype alleles. Finally, we determined whether or not peptide-stimulated PBMCs positive for one of the HLA-A3 supertype alleles could show cytotoxicity against prostate cancer cells expressing other alleles of the HLA-A3 supertype. As shown in Fig. 5, the PBMCs from patients 9, 2, and 14, which were stimulated in vitro with each of the PSA_16-24, PAP_155-163, and PAP_248-257 peptides, showed higher levels of cytotoxicity against LNCaP-A11, LNCaP-A31, and LNCaP-A33 cells compared with LNCaP cells. The cytotoxicity against phytohemagglutinin-stimulated T-cell blasts was significantly low compared with that against LNCaP-A11, LNCaP-A31, and LNCaP-A33 cells. These results indicate that peptide-stimulated PBMCs could show cytotoxicity against prostate cancer cells sharing HLA-A3 supertype alleles.

In this study, we first screened 42 peptide candidates based on their ability to be recognized by IgGs of prostate cancer patients followed by testing their ability to induce HLA class I–restricted CTL activity. Thus, we identified five peptides that were frequently recognized by IgG and that were able to induce CTLs. Interestingly, IgGs reactive to some peptides were also detected in the plasma of healthy donors, irrespective of the expression of HLA-A3 supertype alleles and sex. We previously detected IgGs reactive to PSA-derived or PSMA-derived CTL-directed peptides with HLA-A24-binding motifs in the plasma of healthy donors (24, 25). Why were these peptide-reactive IgGs detected in healthy male and female donors, and what is their biological significance? In healthy males, prostate-related antigens are nonmutated self-antigens, whereas immunologic tolerance to them might be incomplete. In this study, IgGs reactive to prostate-related antigen-derived peptides were detected in healthy females. This finding might indicate that the reactivity of peptide-reactive IgGs is a result of cross-reactivity to exogenous non–self-antigens. Another possibility is that prostate-related antigens might be expressed in some normal tissues. Indeed, some reports suggest that PSA and PAP are expressed in normal tissues, including the normal anal glands of male and urethral glands of both sexes (26, 27). In addition, we incidentally found that normal pyloric glands in gastric mucosa are positive for the PAP expression (28). Although we do not yet have a clear understanding of the biological significance of peptide-specific IgGs in normal individuals, we now have a clue about their role in antitumor immune responses. In our clinical trials, we observed that peptide vaccination frequently resulted in the induction of IgGs reactive to given CTL-directed peptides, and the induction of IgGs reactive to vaccinated peptides was positively correlated with clinical responses and the survival of vaccinated patients (29, 30). Furthermore, we recently analyzed 113 vaccinated patients with various types of cancers and found that augmentation of peptide-specific IgGs after peptide vaccination can predict the prolonged survival in advanced cancer patients who have been vaccinated (31). These lines of evidence suggest that the measurement of peptide-specific IgG could be a good laboratory marker to predict patient prognosis after peptide vaccination. This possibility is under investigation in our laboratory.

We previously reported that humoral responses were not restricted to HLA class I molecules (23, 32). In this study, IgG reactive to the PSMA_431-440 peptide was detected in 16 of 20 HLA-A3 supertype⁺ prostate cancer patients but not in six HLA-A3 supertype⁻ patients. IgG reactive to this peptide was rarely detected in healthy donors, irrespective of the expression of HLA-A3 supertype alleles. Although we have no clear explanation for this observation, the preferential detection of IgG may suggest the usefulness of this peptide in peptide-based immunotherapy for HLA-A3 supertype⁺ prostate cancer patients.

Peptide-specific CTLs induced by stimulation with the PSA_16-24 peptide from the PBMCs of an HLA-A11⁺ patient showed cytotoxicity not only against HLA-A11⁺ LNCaP cells but also against HLA-A31⁺ and HLA-A33⁺ LNCaP cells. Similar results were observed when the PAP_155-163 and PAP_248-257 peptides were used. These findings suggest that these peptide candidates are applicable to prostate cancer patients positive for HLA-A11, HLA-A31, and HLA-A33 alleles among the HLA-A3 supertype alleles. There was a similar report that HLA-A31-restricted tyrosinase-related protein 2 peptide-specific T cells showed reactivity to the peptide presented by HLA-A33 molecules (15). As to why peptide-stimulated CTLs were promiscuous toward other alleles, we suppose the following possibility. The optimal COOH-terminal amino acid of A31- or A33-binding peptides is arginine, whereas that of A11-binding peptides is lysine (33, 34). In addition, the binding affinity of a tyrosinase-related protein 2 peptide, which was identified with HLA-A31-restricted T cells, to HLA-A11 can be improved significantly by the substitution of arginine with lysine at the COOH terminus of the peptide (15). These findings suggest that arginine at the COOH terminus of a peptide could be optimal to all of A11, A31, and A33 molecules. Indeed, three peptides that showed promiscuous reactivity in Fig. 5 carry arginine at the COOH terminus. Although we suppose that arginine at the COOH terminus may be a key factor for promiscuous reactivity, further studies are needed to elucidate this observation.

The other issue requiring elucidation is why the positive frequencies of IgG reactive to the peptides were relatively higher, whereas those of peptide-specific CTL precursors were lower, in prostate cancer patients compared with those in healthy donors. These differences may be due in part to the fact that the magnitude of prostate-related antigen expression in prostate cancer is superior to that in the normal prostate, resulting in the induction of strong IgG responses. On the other hand, T-cell responses might be suppressed in prostate cancer patients because of tumor progression and also in part by preceding therapies. This assumption is partly supported by the fact that CTL responses to EVB, Flu, and tyrosinase-related protein 2 peptides, taken as positive controls, were also suppressed in prostate cancer patients compared with healthy donors.

Because tumor-reactive CTLs recognize peptides in the context of HLA class I molecules on tumor cells and because the frequencies of HLA alleles vary among ethnic populations, ethnicity may influence the reactivity of tumor-reactive T cells. It is well known that CTL reactivity to a MART-1 peptide is stringently limited to HLA-A2-positive patients (35). The identification of peptide vaccine candidates to date has focused on the HLA-A2 and HLA-A24 alleles because of the higher worldwide frequency of these alleles (14). The main purpose of the present study was to extend the possibility of peptide-based immunotherapy for prostate cancer patients by focusing on the A3 supertype alleles; we focused on these alleles because of their wide expression in many ethnic populations. Because a number of peptide candidates for HLA-A2⁺ or HLA-A24⁺ prostate cancer patients have already been identified, the present study should facilitate the development of a peptide-based anticancer vaccine for the majority of prostate cancer patients throughout the world.