Identification of HLA-A3-restricted Cytotoxic T Lymphocyte Epitopes from Carcinoembryonic Antigen and HER-2/neu by Primary in Vitro Immunization with Peptide-pulsed Dendritic Cells¹

CTLs recognize antigen on target and APCs³ as epitopes composed of peptide fragments, 8–12 amino acids long, that are complexed to MHC molecules (1, 2, 3). Tumor cells expressing epitopes derived from TAAs can be recognized and destroyed by CTLs (4). Definition of CTL epitopes from specific TAAs is thus an important step in the development of specific CTL-based cancer immunotherapies.

After the discovery of the MAGE family of TAA genes in 1991 (5), several CTL epitopes derived from numerous TAAs have been described (reviewed in Refs. 6, 7). However, most of these epitopes are derived from melanoma TAA (melanocyte lineage-specific antigens) and only limited numbers of epitopes have been described for TAAs that are expressed on more common types of tumors such as breast, lung, colon, and gastric carcinomas. For these solid adenocarcinomas, CEAs and HER-2/neu represent attractive TAAs expressed in these tumors.

CEA is a 180-kDa cell surface and secreted glycoprotein overexpressed on most human adenocarcinomas including colon, rectum, pancreas, and gastric as well as on 50% of breast and 70% of non-small cell lung carcinomas (8). HER-2/neu is a transmembrane glycoprotein with tyrosine kinase activity whose structure is similar to the epidermal growth factor receptor (9, 10, 11). Amplification of HER-2/neu gene and/or overexpression of the associated protein have been reported in many human adenocarcinomas of the breast (12, 13, 14), ovary (13), uterus (15, 16), prostate (17), stomach (18, 19, 20), esophagus (20), pancreas (21), kidney (22), and lung (23, 24).

Recently, some CTL epitopes have been reported from CEA (25, 26) and HER-2/neu (27, 28, 29, 30); however, all of them are restricted by HLA-A2, the most frequent MHC HLA supertype (40–50%) in humans. To develop CTL-based therapies for tumors expressing CEA and HER-2/neu applicable to the general population, it is important to identify epitopes restricted by other common MHC class I alleles. Herein, we describe two new epitopes, derived from CEA and HER-2/neu, recognized by CTL in the context of the HLA-A3 allele, found in a large percentage of the general population (21% Caucasians, 16% Blacks, 15% Hispanics, and 7% Orientals). More importantly, the peptides representing these CTL epitopes also bind well to other class I HLA alleles belonging to the HLA-A3-supertype family. Our results show that in vitro lymphocyte immunization (31) with DCs pulsed with these HLA-A3 binding peptides elicits CTL capable of killing HLA-A3⁺ tumor cells expressing CEA, HER-2/neu. These results are relevant to the development of CTL-based immunotherapy for solid epithelial tumors.

Peptides were prepared according to standard solid phase methods using an Applied Biosystems synthesizer and purified by high-performance liquid chromatography as described previously (32). The purity (>95%) and identity of peptides were determined by mass spectrometry analysis.

The binding of peptides to HLA-A*0301 molecules was measured based on the inhibition of the binding of a radiolabeled standard peptide to purified MHC molecules as described previously (32). Briefly, various doses of the test peptides were incubated for 2 days with 5 nm standard peptide (KVFPYALINK) and detergent-solubilized (0.05% NP40) purified HLA-A*0301 molecules (10 nm) in the presence of a mixture of protease inhibitors (1 mm phenylmethylsolfunyl fluoride, 1.3 mm 1.10 phenanthroline, 73 μm pepstatin, 8 mm EDTA, and 200 μm N-α-p-tosyl-l-lysine chloromethyl ketone—all from Sigma) and 1 μm β₂-microglobulin (Scripps Laboratories, San Diego, CA). The binding of peptides to HLA-A*1101, HLA-A*3101, and HLA-A*6801 molecules was measured by the same protocol as HLA-A*0301 assay using KVFPYALINK as the standard peptide. In the case of HLA-A*3301 assay, STLPETYVVRR was used as the standard peptide. The percentage of MHC-bound radioactivity was determined by gel filtration and the concentration of the tested peptides to inhibit 50% of the binding of the labeled peptide (IC₅₀) was calculated.

The EBV-transformed human B cell line EHM (homozygous for A*0301, obtained from the American Society for Histocompatibility and Immunogenetics Repository Collection) was used as target for peptide-mediated cytotoxicity assays. EHM was cultured in RPMI (Life Technologies, Grand Island, NY) supplemented with 10% FCS and antibiotics. The HLA-typed colon tumor cell lines SW403 (A2.1/A3), WiDr (A1/A24), and SW480 (A2.1/A24) were purchased from American Type Culture Collection (Rockville, MD). SW403 and SW480 were cultured in DMEM (Life Technologies) supplemented with 10% FCS and antibiotics. WiDr was cultured in MEM (Life Technologies) supplemented with 10% FCS and antibiotics. The expression of HLA-A3 on the cell lines was determined by flow-cytometric analysis using monoclonal antibody 0170HA (One Lambda, Canoga Park, CA). The expression of CEA and HER-2/neu was measured by flow-cytometric analysis using monoclonal antibodies CEM010 (Takara Shuzo, Japan) and c-neu Ab-5 (Oncogene Science, Uniondale, NY), respectively.

PBMCs from normal volunteers were purified by centrifugation in Ficoll-Paque (Pharmacia, Piscataway, NJ) from leukopheresis products. The Institutional Review Board on Human Subjects (Epimmune) approved this research, and informed consent for blood donation was obtained from all of the volunteers. The monocyte-derived and cytokine-generated DC were prepared as described previously (31, 33, 34). Briefly, PBMC were plated in 6-well plates (10 × 10⁶ cells/well), and nonadherent cells were washed off after a 2-h incubation at 37°C. The monocyte-enriched adherent cells were cultured in the presence of 50 ng/ml recombinant (r)-granulocyte-macrophage colony-stimulating factor and 1000 units/ml rIL-4 (both from Endogen, Cambridge, MA) in complete medium (RPMI 1640 supplemented with 5% human AB serum (Gemini Bio-Products, Calabasas, CA), 0.1 mm MEM nonessential amino acids, 1 mm sodium pyruvate, 2 mm l-glutamine, and 50 μg/ml gentamicin). After 7 days, the cytokine-treated cells, which display typical cell surface markers of DC (CD3⁻, CD14⁻, CD86⁺, HLA-class I^high, HLA-DQ⁺; data not shown) were harvested and used as APCs as described below.

For CTL induction, the cytokine-generated DCs were pulsed with peptide and used as APCs to stimulate autologous CD8⁺ T-cells. For the screening of CEA peptides, DCs were pulsed with 40 μg/ml peptide in the presence of 3 μg/ml β₂-microglobulin for 4 h at 20°C in PBS with 1% BSA without any pretreatments. For the screening of HER-2/neu peptides, one of the following pretreatments was used to increase the density of relevant peptide-MHC complexes on the surface of DCs: (a) before pulsing with the peptide, the DCs were treated with a mild acidic buffer to release MHC-bound endogenous peptides (35, 36). The acid-treated DCs, which contain large amounts of “empty MHC molecules,” were then pulsed with peptides as described below in more detail; or (b) to increase specific MHC/peptide complexes on DCs, peptide was introduced into the DC cytoplasm by osmotic lysis of pinocytic vesicles (37, 38).

Specifically, for the first protocol, DCs (∼2 × 10⁶) were washed twice with 1% BSA-0.9% NaCl and resuspended in 1 ml of cold citrate-phosphate buffer [0.13 m citric acid and 0.06 m sodium phosphate monobasic (pH 3.0)] containing 1% BSA and 3 μg/ml β₂-microglobulin. After a 2-min incubation on ice, 5 volumes of cold 0.15 m sodium phosphate monobasic (pH 7.5), containing 1% BSA, 3 μg/ml β₂-microglobulin, and 10 μg/ml peptide were added, and cells were harvested by centrifugation. For the second procedure, DCs (∼2 × 10⁶) were resuspended in 50 μl of 1% BSA-PBS containing 0.4 mg/ml peptide; then 50 μl of the hypertonic solution (1m sucrose and 20% polyethylene glycol 1500) were added to the suspension. After a 10-min incubation at 37°C, 10 ml of warm 1% BSA-PBS were added, and DCs were harvested by centrifugation after a 3-min incubation at 37°C. After one of these pretreatments, the DCs were then pulsed with 40 μg/ml peptide in the presence of 3 μg/ml β₂-microglobulin for 4 h at 20°C in PBS with 1% BSA.

The peptide-loaded DCs were irradiated with 4200 rads and mixed with CD8⁺ cells, obtained by positive selection with Dynabeads M-450 CD8 and Detachabead (both from Dynal, Lake Success, NY), at a ratio of 1:20. The cultures were set up in 48-well plates; each well contained 0.25 × 10⁵ APCs, 5 × 10⁵ CD8⁺ cells and 10 ng/ml rIL-7 (Genzyme, Boston, MA) in 0.5 ml of complete medium. On day 1, a final concentration of 10 ng/ml rIL-10 (Endogen) was added to the culture. On days 7 and 14, the responder cells were restimulated with autologous peptide-pulsed adherent APCs. To prepare the APCs, 2× 10⁶ irradiated autologous PBMCs in 0.5 ml 1% BSA/d-PBS were added to each well of the 48-well plates. After incubation at 37°C for 2 h, the nonadherent cells were washed off, and the adherent cells were incubated for 2 h with 10 μg/ml peptide and 3 μg/ml β₂-microglobulin in a final volume of 0.25 ml of 1% BSA/d-PBS per well. The supernatant of the responder cultures were aspirated, and fresh complete medium was added to the total volume of 0.5 ml/well. The excess peptide was removed from the adherent APC plate, and responder cultures were transferred to the corresponding wells containing peptide-pulsed APCs. Each individual well was restimulated separately. The next day of each restimulation, the final concentration of 10 ng/ml rIL-10 was added to the culture. The cultures were fed every 2–3 days with fresh medium containing final concentration of 50 IU/ml rIL-2. Cytotoxicity was first tested after two cycles of restimulation with peptides (day 21).

After 2–3 cycles of antigen restimulation, CTL clones were isolated from the wells containing tumor-killing CTLs by limiting the dilution method by plating 1 to 0.3 CTL per well in 96-well plates. The positive CTL clones were expanded with anti-CD3 monoclonal antibody (OKT-3) and IL-2 in the presence of PBMC and EBV-transformed cell lines as feeder cells (31, 39). All of the CTL clones were CD8⁺, CD3⁺, and CD4⁻ as determined by cytofluorometric analysis using specific monoclonal antibodies (data not shown).

Cytolytic activity was determined in a standard 4–6 h ⁵¹Cr release assay as described previously (40). Peptide-pulsed targets were prepared by incubating the cells with 10 μg/ml peptide overnight at 37°C. Tumor cell lines were treated with 100 units/ml IFN-|gg (Genzyme) for 48 h before assays to increase the level of MHC class I expression. Adherent target cells were removed from culture flasks with trypsin-EDTA solution. Target cells were labeled with 300 μCi of ⁵¹Cr sodium chromate (Dupont, Wilmington, DE) for 1 h at 37°C. Labeled target cells (10⁴ cells/well) and various numbers of effector cells were plated in a final volume of 0.2 ml in 96-well plates. For the first assay, after two cycles of restimulation with peptides (day 21), unlabeled K562 cells (3 × 10⁵ cells/well) were added to eliminate the nonspecific lysis by natural killer cells. After 4–6 h at 37°C, 100 μl of supernatant was collected from each well, and percent-specific lysis was determined according to the formula:

Antigen specificity and tumor reactivity were confirmed by cold target inhibition assay by determining the capacity of peptide-pulsed (10 μg/ml for 16 h at 37°C) unlabeled (cold) EBV-transformed cell line to inhibit the lysis of ⁵¹Cr-labeled tumor cell lines.

To identify HLA-A3-restricted CTL epitopes from CEA and HER-2/neu, the protein sequences of CEA and HER-2/neu were first analyzed for the presence of peptide sequences containing the HLA-A3 binding motif: a hydroxyl-containing residue (S or T) or hydrophobic amino acid (L, V, I, or M) at position 2 and a positively charged residue (R, H, or K) at the COOH-terminal of peptides consisting of 9 or 10 residues (32). Peptides containing the HLA-A3 binding motif were synthesized and tested for their capacity to bind purified HLA-A*0301 molecules in vitro. From CEA, four peptides were identified that bound to HLA-A*0301 molecules with an IC₅₀ ≤ 500 nm (Table 1). In the case of HER-2/neu, we identified twenty peptides binding to HLA-A*0301 molecules with an IC₅₀ ≤ 500 nm. The seven highest binding peptides from HER-2/neu are shown in Table 2.

Of four peptides that bound to HLA-A*0301 molecules with an IC₅₀ ≤ 500 nm, two (CEA[9₆₁] and CEA[10_198/554]) were randomly selected for testing their capacity to induce CTL by a primary in vitro immunization protocol using peptide-pulsed DCs as APCs (31). CEA[9₆₂₈] could not be tested because sufficient quantities (∼5 mg) with a high degree of purity (>95%) of this peptide could not be prepared.

When CEA[9₆₁] was tested for its capacity to induce CTL using lymphocytes derived from a normal healthy donor (HLA-A1/-A3), a CTL line (CD8⁺) derived from one culture of the 48 tested (Table 1) killed peptide-pulsed target cells (% specific lysis: 34% for EHM pulsed with peptide and 1% for EHM without peptide). This culture was also reactive with the A3⁺/CEA⁺ colon tumor cell line SW403 (% specific lysis: 12%) but did not kill a CEA⁺ tumor line not expressing the HLA-A3 allele (1% specific lysis of WiDr cells). The CTL in this culture were cloned by limiting dilution and were further expanded to obtain sufficient cells to perform more detailed specificity analyses. The results in Fig. 1,A demonstrate that one of the resulting clones derived from this CTL line effectively recognized the SW403 tumor cell line as well as peptide-pulsed target cells even at low effector to target ratios. The antigen specificity of the killing of SW403 tumor cells by this CTL clone was confirmed by showing that cold (unlabeled) targets that were pulsed with peptide CEA[9₆₁], but not targets pulsed with an irrelevant HLA-A3-binding peptide, specifically inhibited the lysis of SW403 cells (Fig. 1 B). Peptide CEA[9₆₁] was also found to elicit tumor-reactive CTLs in another HLA-A3⁺ healthy donor where a CTL clone that was similarly isolated, was capable of killing 63% of peptide-pulsed EHM cells and 41% SW403 tumor cells at an effector:target ratio of 2:1, whereas less than 2% of either unpulsed EHM cells and WiDr cells were lysed by the CTL.

In the case of peptide CEA[10_198/554], peptide-specific CTLs could not be detected after 2 cycles of antigen restimulation in the two HLA-A3⁺ normal healthy donors that were tested (one example is shown in Table 1).

The seven highest HLA-A3-binding peptides were tested for their capacity to induce CTLs. Of the seven peptides, three (HER2[9₆₈₁], HER2[9₈₅₂], and HER2[9₇₅₄]) were found to induce CTLs that specifically reacted with peptide-pulsed targets (Table 2).

The CTLs induced with HER2[9₇₅₄] (Table 2) were found to have high lytic activity toward peptide-pulsed cells (% specific lysis, 54% for EHM pulsed with peptide and 3% for EHM without peptide) and for the A3⁺/HER-2/neu⁺ colon tumor cell line SW403 (% specific lysis, 30%) but not to a HLA-A3 negative CEA⁺ tumor cell line (SW480). After two cycles of antigen restimulation, CTL clones were isolated from this line by limiting dilution. As shown in Fig. 2, one of the resulting CTL clones from this line recognized SW403 quite effectively (Fig. 2,A). The cytotoxicity against the SW403 tumor cells was shown to be specific inasmuch as it was inhibited with cold targets pulsed with peptide HER2[9₇₅₄] but not by cold targets pulsed with a control peptide (Fig. 2 B). It is important to note that peptide HER2[9₇₅₄] was also effective in inducing tumor-reactive CTLs with lymphocytes from a separate individual. A CTL clone isolated from a second normal volunteer killed 70% of peptide-pulsed EHM cells and 40% of SW403 cells whereas less than 5% of either unpulsed EHM or SW480 (negative control) cells were lysed by these CTLs (data not shown).

CTL activity was also elicited with peptide HER2[9₆₈₁] (Table 2). However, after one more restimulation cycle with antigen and APCs, the killing activity of these cultures against SW403 was not observed, although these cells continued to lyse the peptide-pulsed EHM cells. Similarly, in the case of HER2[9₈₅₂], only 1 of 40 wells was positive for the peptide-pulsed targets, but recognition of SW403 tumor cells could not be detected (Table 2). Therefore, we are not certain whether peptides HER2[9₆₈₁] and HER2[9₈₅₂] represent true CTL epitopes that are processed and expressed on CEA⁺ tumor cells.

HLA-A*0301 is the prototype allele of the HLA-A3 supertype family, formed by several MHC class I molecules that share overlapping peptide binding specificities (32). To determine whether the newly identified HLA-A3 epitopes from CEA and HER-2/neu could be recognized in the context of other alleles of this family, the binding affinity of CEA[9₆₁] and HER2[9₇₅₄] to other members of the HLA-A3 supertype (HLA-A*1101, -A*3101, -A*3301, and -A*6801) was determined. As shown in Table 3, CEA[9₆₁] bound all five alleles of the HLA-A3 supertype family tested with very high binding affinity, and HER2[9₇₅₄] bound to four alleles (except -A*6801) with IC₅₀ ≤ 500 nm. Preliminary data indicate that peptide CEA[9₆₁] is capable of inducing HLA-A11-restricted tumor-reactive CTLs. A CTL clone isolated from an HLA-A11 individual killed 93% of peptide-pulsed BVR (HLA-A11⁺ target cells) and 24% of TT cells (CEA⁺HLA-A11⁺ thyroid cancer line, ATCC: CRL-1803), whereash2% of either the unpulsed BVR cells or the SK-BR-3 tumor cells (A11⁺CEA⁻, ATCC: HTB-30) were lysed by these CTLs (data not shown). These results suggest that epitopes represented by peptide CEA[9₆₁] may be used to elicit CTLs in at least two allelic members (A3 and A11) of the HLA-A3 supertype.

In this study, we identified two HLA-A3-restricted epitopes, CEA[9₆₁] (HLFGYSWYK) and HER2[9₇₅₄] (VLRENTSPK), using peptide-pulsed DCs in an in vitro primary CTL induction protocol with PBMCs from normal healthy volunteers. The CTL induction protocol has been very effective for the identification of CTL epitopes. We have recently described several new HLA-A2-restricted CTL epitopes from gp100 (40), MAGE antigens, CEA and HER-2/neu (25), and most recently HLA-A3-and HLA-A11-restricted CTL epitopes from gp100 (41). It is interesting to note that although the induction of CTLs with peptide CEA[9₆₁] could be achieved by our conventional protocol using peptide-pulsed DCs (31), this method was unsuccessful with peptides from HER-2/neu (data not shown), all of which bind to HLA-A3 significantly less that peptide CEA[9₆₁] (Tables 1and 2). To determine whether increasing the amount of MHC/peptide complexes on the DCs would result in the induction of CTLs, we proceeded to enhance peptide loading by mild acid elution of endogenous bound peptides and/or osmotic loading of peptides in hypertonic solution. Our results indicate that somehow these modifications to the peptide-loading protocol of DCs were effective in triggering CTL responses to at least three of the HER-2/neu peptides (Table 2). Interestingly, using a computer-based prediction algorithm (42), the estimated half-time disassociation of peptide CEA[9₆₁] to HLA-A3 is 1350 s whereas that for peptide HER2[9₇₅₄] is 30 s, suggesting that more of the latter peptide would be required to produce sufficient MHC/peptide complexes for CTL stimulation. In summary, it is evident that MHC binding affinity (association and disassociation constants) of peptide to MHC is critical for the production of stable complexes that will interact with T-cell receptor molecules on CTL precursors.

To our knowledge, this is the first report demonstrating specific HLA-A3-restricted CTL epitopes for CEA and HER-2/neu. Bremers et al. (43) identified six peptides with high-binding affinity to HLA-A3 molecule by testing 34 CEA-derived peptides that contained the HLA-A3 binding motif using a MHC stabilization binding assay with EBV-transformed B cell lines. Interestingly, peptide CEA[9₆₁] had the second highest binding affinity in their assay system (43). However, this group did not determine whether any of the HLA-A3-binding peptides could induce tumor- (or peptide)-reactive CTLs.

The CEA[9₆₁] epitope binds with high binding affinity to all of the HLA molecules belonging to the HLA-A3 superfamily (HLA-A*0301, -A*1101, -A*3101, -A*3301, and -A*6801) suggesting that this peptide may be capable of eliciting CTLs restricted by all of these alleles. As mentioned above (in “Results”), in preliminary studies, peptide CEA[9₆₁] has been able to induce tumor-reactive CTLs in an HLA-A11⁺ individual. The possibility that this peptide may also trigger CTL responses for the other members of the HLA-A3 supertype family remains to be examined.

In summary, we have described two HLA-A3-restricted CTL epitopes that have the potential of inducing antitumor responses in individuals expressing MHC class I alleles belonging to the HLA-A3 superfamily. These results allow investigators to broaden the population coverage for the design and development of CTL-based antigen-specific immunotherapies for commonly found types of malignancies such as breast, lung, colon, and gastric carcinomas.