Generation of Human T-cell Responses to an HLA-A2.1-restricted Peptide Epitope Derived from α-Fetoprotein¹

HCC³ is one of the most common fatal tumors (1, 2), with an annual global incidence of 1.2 million (3). In the United States, approximately 13,000 new cases are diagnosed each year, and the median survival is generally less than 6 months (1, 4). Resection, transplantation chemoembolization, alcohol injection, and cryoablation are potentially curative, but only in small, localized tumors (5, 6, 7). Unfortunately, most patients have advanced disease at diagnosis, and current systemic therapies are largely ineffective. The development of novel treatment strategies is greatly needed. Most gene therapy efforts use suicide gene (8, 9, 10, 11, 12), tumor suppressor gene (13, 14, 15), or cytokine-based strategies (16, 17, 18).

AFP is expressed during fetal development but transcriptionally repressed shortly after birth (19). Certain tumors, principally HCC and germ cell tumors, express AFP, and its measurement in serum plays an important role in diagnosis and in monitoring responses to treatment (20). Human AFP is translated as a 609-aa protein that is cleaved to yield a 590-aa secreted protein (21). The regulation of human and murine AFP genes has been extensively studied and is largely at the level of transcription (22, 23). The normal function of AFP is unknown. It has been hypothesized to play a role in serum component transport because AFP has been shown to bind fatty acids, steroids, and heavy metals (24, 25, 26). In addition, there have been reports that AFP may have an immunosuppressive role in fetal development (27, 28, 29).

The idea that AFP can serve as a target for immunotherapy is not new. Efforts were reported in earlier tumor immunology literature that involved attempts to generate humoral responses (30, 31, 32). These were predictably unsuccessful due to high circulating levels of AFP that neutralized antibody. However, AFP-producing tumors would be expected to process and present AFP-derived peptide fragments on their cell surface in the context of major histocompatibility molecules. These MHC-restricted AFP peptides could potentially be recognized by the immune system, provided that these T cells were not clonally deleted during the ontogeny of the immune system. Both murine and human T-cell repertoires appear to contain self-reactive T-cell clones for such proven and putative tumor rejection antigens as MART-1 (33, 34), MAGE (35, 36), gp100 (33, 37), carcinoembyronic antigen (38, 39, 40, 41), and others. It would be surprising if potential AFP-reactive clones could not be marshalled with an appropriate set of activation signals in an immunostimulatory environment.

Our strategy in examining human T-cell responses to AFP was guided by our parallel studies of human T-cell responses to the well-characterized melanoma antigen MART-1 (42). Robust responses could be generated in vitro by DCs genetically engineered to express MART-1. DCs transduced with a recombinant MART-1 AdV expressed this melanoma antigen at high levels and correctly processed and presented the immunodominant HLA-A2.1-restricted MART-1_27–35 peptide. MART-1 engineered human DCs could be used to generate specific human T-cell responses in vitro. In addition, we have reported a murine MART-1 model in which potent CTLs, cytokine-producing T cells, and protective immunity are generated after immunization with MART-1-engineered DCs (43, 44). We have exploited these potent antigen-presenting cells to investigate human T-cell responses to AFP. We report, for the first time, that the human T-cell repertoire can recognize AFP, and we characterize the response to an HLA-A2.1-restricted epitope.

The University of Wisconsin Genetics Computer Group Program “find patterns” was used to screen the hAFP sequence (GenBank accession numbers J00077, J00076, and V01514) and identify 9- and 10-mer peptides that contained (a) two strong binding anchor residues at positions 2 and 9 (or 10); (b) only one anchor residue; or (c) no anchor residue but other positive binding residues. Of the three groups of peptide sequences, those that contained more than one residue thought to abolish binding were eliminated.

HLA-A2.1 donors and cell lines were screened with the BB7.2 (HLA-A2)-specific antibody and a goat-antimouse-FITC secondary antibody (Caltag, South San Francisco, CA) and subtyped by PCR and direct sequence analysis by the UCLA Tissue Typing Laboratory. The K562, HepG2, Hep3B, B95-8, BB7.2, and W6/32 cell lines were obtained from American Type Culture Collection (Rockville, MD). The M202 human melanoma cell line was generated in our laboratory from a surgical specimen and has been described previously (45). T2 cells were provided by Peter Cresswell (Yale University School of Medicine). JY cells (HLA-A2.1 homozygous) were provided by Martin Kast (Loyola University Cancer Center). Jurkat/A2K^b cells were provided by Linda Sherman (Scripps Research Institute). EBV-transformed LCLs were generated by incubating PBMCs from HLA-A2.1+ donors with supernatant from B95-8 cells. Cell lines were cultured in RPMI 1640 (Life Technologies, Inc., Gaithersburg, MD) or IMDM (JY cells; Life Technologies, Inc.) with 10% FBS (Omega Scientific, Tarzana, CA) and pennicillin/streptomycin/fungizone (Life Technologies, Inc.). Anti-β₂-microglobulin and the anti-CD4, -CD14, -CD19, and -CD56 NA/LE depletion antibodies were obtained from PharMingen (San Diego, CA), anti-pan class I antibody was prepared from concentrated supernatant of the W6/32 hybridoma, anti-HLA-A2 antibody was prepared from concentrated supernatant of the BB7.2 hybridoma, and CD4-FITC, CD8-PE, and CD16-PE antibodies were obtained from Caltag.

The stable transfectant cell lines LCL<AFP, M202<AFP, Jurkat A2/K^b<AFP (Jurkat/AFP), and Jurkat A2/K^b/MART1 (Jurkat/MART) were prepared by lipofection of parental cells with an expression plasmid (VR1012hAFP and VR1012 from Vical) expressing hAFP and cotransfection with the hygromycin-expressing plasmid pCEP4 (Invitrogen, Carlsbad, CA) or pRcCMVMART1neo (44), followed by selection in hygromycin (Boehringer Mannheim) at 50 μg/ml or in G418 at 500 μg/ml. Expression of AFP was confirmed by RT-PCR (primers 5′-GCAACCATGAAGTGGGT and 3′-CTCTCTCTCTCTAGAAACTCCCAAAGCAGCACGAGT) and AFP radioimmune assay (UCLA Medical Center Clinical Labs), and expression of MART-1 was confirmed by RT-PCR as described previously (45). IL-2 was provided by Hoffman-LaRoche (Nutley, NJ), IL-7 was obtained from Biosource (Camarillo, CA), IFN-γ was provided by Dr. Steven Dubinett (UCLA), and keyhole limpet hemocyanin and β₂-microglobulin were obtained from Sigma (St. Louis, MO).

AdVhAFP contains the 1.9-kb hAFP cDNA originally cloned by RT-PCR using the primers listed above and driven by the CMV promoter/enhancer in a pAC-CMVpLpA AdV type 5 backbone. The virus was prepared by recombination of this plasmid with pJM17, which contains the 35-kb AdV genome, deleted in the E1 region, in 293 cells that provide the E1 genes in trans. Recombinant viruses were released into the medium, purified by limiting dilution, and amplified on 293 cells. The empty AdV vector, AdVRR5, has been described previously (46) and served as a control. All viruses used have been purified on CsC1 gradients as described previously (46).

Peptides were synthesized by Chiron (Victoria, Australia) and at the UCLA Peptide Synthesis Facility using standard f-moc technology.

Each peptide was tested for concentration-dependent binding to T2 cells in a HLA-A2.1 stabilization assay (47, 48). T2 (TAP-deficient) cells that had been incubated at RMT the previous night to increase cell surface MHC class I molecule expression were then incubated overnight with each peptide over a range of peptide concentrations from 0.1–100 μm. In the T2 cell line, only MHC molecules that are filled with 8-10-mer peptides are stable on the cell surface. Stability of HLA-A2.1 was assayed by flow cytometry after staining the cells with an anti-HLA-A2 antibody (BB7.2) and goat antimouse-FITC. The HLA-A2.1 strongly binding Flu matrix peptide (aa 58-66; GILGFVFTL; Flu) was used as a positive control.

Peptide binding to HLA-A2.1 was determined using HLA-A2.1+ JY cells. Cells were washed twice in PBS before stripping off peptides with 2 ml of citrate-phosphate buffer [0.131 m citric acid and 0.066 m Na₂HPO₄ (pH 3.2)] for 90 s (49). After washing once in serum-free medium (IMDM), cells were resuspended and seeded on a 96-well plate (Costar) at 1 × 10⁵ cells/well. Different concentrations of each peptide (50 μm-50 nm), along with 100 nm human β ₂-microglobulin (Sigma), were added to each well and incubated for 4 h at RMT. Negative controls were either no peptides or the HLA-A1 binding MAGE-3 peptide. Flu was used as a positive control. After incubation, cells were washed twice before adding 5 μg/ml W6/32 antibody and incubated on ice for 30 min. Surface HLA-A2.1 expression was detected by using a microtiter plate reader quantifying the hydrolysis of the β-galactosidase substrate chlorophenol red-β-d-galactopyranoside monosodium salt (Boehringer Mannheim) after staining the cells with goat antimouse IgG F(ab′)₂ antibody conjugated with β-galactosidase (Southern Biotechnology Associates, Birmingham, AL). The relative binding strength of each peptide is expressed in absorbance of wells containing the peptide over absorbance in wells without peptides.

HLA-A2.1 LCLs were stripped of surface class I peptides and β ₂-microglobulin with a mild pH 3.2 citrate-phosphate acid buffer that makes MHC molecules unstable (50). Each peptide was immediately pulsed onto stripped cells at 200 μm for 1 h in the presence of β₂-microglobulin at 3 μg/ml at RMT. Excess peptide was washed off, and the cells were incubated at 37°C for 0, 2, 4, and 6 h. Cells were washed at the end of each time point and stained for cell surface HLA-A2 expression and then analyzed by flow cytometry. The peptide-MHC class I complex was considered stable if the mean fluorescence intensity increased at least 1.5-fold from cells that were stripped but not pulsed with peptide.

Peptide-specific CTLs were generated as described previously (42) to various peptides. Briefly, normal donor HLA-A2.1 PBMCs were pulsed with 50 μg/ml peptide for 90 min at RMT in serum-free IMDM (Life Technologies, Inc.), washed, and cultured on day 0 with IL-7 (10 ng/ml) and keyhole limpet hemocyanin (5 μg/ml) in RPMI 1640 10% autologous serum at 3 × 10⁶ cells/1.5 ml per well. Cells were restimulated weekly by removing the nonadherent cells from the culture and adding them to fresh, autologous, peptide-pulsed, washed, and irradiated PBMCs at a 1:1 ratio. IL-2 was added twice weekly at 10 units/ml. After 3 weeks of culture, the cultures were tested for cytotoxicity and/or cytokine release.

DCs [prepared as described from PBMCs incubated with GM-CSF and IL-4 (51, 52)] were transduced with AdVhAFP or AdVMART1 at a MOI of 1000 for 2 h. Transduced DCs were washed, irradiated, and plated at 2–5 × 10⁵ cells/well in a 24-well plate to serve as stimulators for CTL generation. Autologous nonadherent cells were depleted of CD4, CD14, CD19, and CD56+ cells by magnetic bead depletion (Dynal, Lake Success, NY) to prepare CD8+ enriched responder cells (population generally at least 80% CD8+; data not shown). The CD8+ cells were plated with the transduced DCs at 2 × 10⁶ cells/well in 5% autologous medium plus IL-7 at 10 ng/ml to generate CTLs. Cultures were supplemented with IL-2 at 10 units/ml every 3–4 days. The CD8+ CTLs were restimulated weekly with fresh, autologous, AdV-transduced DCs at a ratio of 1 DCs: 5-10 CD8+ CTLs. Most cultures were phenotyped for CD4+ and CD8+ cells on a weekly basis.

Target cells were harvested, washed, counted (T2 cells were peptide-pulsed at 50 μg/ml) and chromated with 100 μCi of Na₂Cr⁵¹O₄ (Amersham, Arlington Heights, IL), with shaking incubation at 37°C for 1.5–2 h. CTLs were washed, counted, and diluted to the desired concentrations in RPMI 1640/10% AB (or RPMI 1640/10% FBS for murine T cells), and plated in triplicate wells in a V-bottomed 96-well plate (Costar). Target cells were washed three times, diluted to 5 × 10⁴ (or 1 × 10⁵ for murine assay) cells/ml, and plated with CTLs. To control for nonspecific lysis, a 10–50-fold excess of unchromated (cold) K562 cells was added to most target populations before adding to CTLs (in human assay). The plates were spun briefly at 1000 rpm and incubated for 4–5 h. Supernatants were harvested and counted in a gamma counter. Triplicate wells were averaged, and the percentage of specific lysis was calculated as follows: [sample − spontaneous release] /[maximum release − spontaneous release].

To determine the frequency of antigen-induced cytokine-producing T cells, the ELISPOT technique was used (53, 54). T-cell restimulation was performed with 3–5 × 10⁶ CTLs incubated with 1 × 10⁵ autologous LCLs pulsed with specific or nonspecific peptides or tumor cell lines in 1 ml of RPMI 1640/10% autologous serum medium/well of a 24-well plate. LCLs were peptide-pulsed in 1 ml of serum-free IMDM (Life Technologies, Inc.) at RMT for 1–2 h with 50 μg/ml peptide. The cells were rinsed before plating with CTLs. An additional well was prepared with restimulator cells without CTLs as a negative control. The 24-well plates were incubated in a humidified incubator at 37°C for 48 h. The ELISPOT microtiter plates (Millipore, Bedford, MA) were coated with purified IFN-γ-, GM-CSF, or tumor necrosis factor α (PharMingen) antibody in coating buffer [0.1 m NaHCO₃ (pH 8.2)] at 4 μg/ml and stored at 4°C overnight. The next day, the restimulated CTLs were rinsed and set to identical cell concentrations in serum-free X-Vivo-10 (Life Technologies, Inc.). The blocked (PBS/10% FBS) plates were then washed with PBS, followed by one wash with X-Vivo-10. The restimulated CTLs were plated in duplicate wells, in each of three dilutions, and then incubated at 37°C for 24 h. The plates were washed 10× with PBS/Tween, and secondary biotinylated antibody was added at 3 μ g/ml and incubated at 4°C overnight.

Finally, the plates were washed 10× with RMT PBS/Tween, and a 1:2000 dilution of avidin-peroxidase (Vector Laboratories) was added, and the plates were incubated in the dark at RMT for 1–2 h. The color substrate 3-amino-9-ethylcarbazole (Sigma) was prepared in formamide/0.05 m NaOAc buffer (pH 5.0) and filtered. The plates were then washed again in PBS/Tween. H₂O₂ solution was added to the color substrate solution, and the substrate solution was added to the washed plate. The reaction was stopped by rinsing in tap water. The spots were counted under a dissecting microscope.

HLA-A2.1/K^b tg female mice (created by Dr. Linda Sherman, Scripps Research Institute, La Jolla, CA) were originally purchased from Harlan-Sprague Dawley (Indianapolis, IN) and are currently bred by the animal facility of the Department of Radiation Oncology at UCLA and handled in accordance with the animal care policy of the UCLA. For peptide immunizations, mice received 100 μg of AFP or control peptide emulsified 1:1 in CFA (Difco, Detroit, MI) s.c.

DCs were differentiated from murine bone marrow progenitor cells following the Inaba method (55), with modifications (44). Day +8 nonadherent and loosely adherent cells contained DC aggregates with a high level of MHC class I and II, B7.1 (CD80), B7.2 (CD86), CD1d, CD18, and CD44-positive cells that were superior stimulators of a mixed lymphocyte reaction (data not shown). In vitro cultured DCs were transduced in 15-ml conical tubes (Costar) in a final volume of 1 ml of RPMI 1640/2% FCS to which the virus stock was added at a moi of 100 viral plaque-forming units/DC. Transduction was carried out for 2 h at 37°C, and the DCs were then washed extensively and resuspended at 5 × 10⁵ DCs/0.2 ml PBS/animal for injection into mice. Cell counts were determined using a hemocytometer, with viability assessed by trypan blue exclusion. In all cases, viability exceeded 95%.

Two weeks after priming (by AdV/DC or peptides), splenocytes (3 × 10⁶/cells well) were activated ex vivo with irradiated, mitomycin C-treated Jurkat<AFP or Jurkat/MART (5 × 10⁵/cells well) in 2 ml of RPMI 1640/10% FBS and 50 units/ml IL-2 in 24-well plates for 6 days.

Groups of HLA-A2.1/K^b tg mice were primed by AdV-transduced DC immunizations, and 2 weeks later, splenocytes (4 × 10⁶/well) were activated ex vivo with an optimal concentration of peptide in 1 ml of complete medium (RPMI 1640/10% FBS) in 24-well plates. For tg mice primed through peptide immunizations, the draining popliteal and inguinal lymph nodes were removed 10 days after immunization, and a single cell suspension was prepared. Lymph node cells (5 × 10⁶ cells/ml) were cultured in T-25 flasks in an equal volume with irradiated Jurkat-A2/K^b cells (stably transfected with either MART-1 or hAFP at 1 × 10⁵ cells/ml) plus 50 units/ml IL-2.

After culture for 48 h for both splenocyte and lymph node cell cultures, cells were washed and transferred by serial dilution (from 2 × 10⁵ to 2 × 10⁴ cells/well) to 96-well microtiter plates precoated with capture antibodies (IFN-γ or IL-4; PharMingen) at 2 μg/ml in serum-free medium (X-Vivo). After 24 h, cells were removed, and spots were visualized using biotinylated secondary antibodies and avidin-D-peroxidase in conjunction with 3-amino-9-ethylcarbazole substrate, as described above. The frequency of antigen-specific cells was determined from the difference between the number of spots seen with and without antigen during restimulation.

We demonstrate that the human T-cell repertoire can recognize a HLA-A2.1-restricted peptide epitope derived from AFP using the following lines of evidence.

DCs genetically engineered to express tumor antigens by adenoviral transduction are potent inducers of T-cell responses in vitro and in vivo. We first used DCs transduced with an AdV-expressing human AFP (AdVhAFP) to generate AFP-specific, HLA-A2.1-restricted T-cell responses using CD8+ enriched in vitro T lymphocyte cultures.

Human DCs differentiated from loosely adherent peripheral blood progenitors in GM-CSF and IL-4 express hAFP mRNA in a viral dose-dependent manner when transduced by AdVhAFP (Fig. 1,A). T-cell cultures stimulated weekly with AdVhAFP/DC generated AFP-specific T cells that recognized AFP-expressing cells by both cytotoxicity assay (Fig. 1,B) and IFN-γ ELISPOT analysis (Fig. 1,C). These T cells lysed M202<AFP transfectants but not parental M202 cells (Fig. 1,B), and ELISPOT analysis shows a greater frequency of IFN-γ-secreting cells responding to AFP-transfected cells compared to parental cells (Fig. 1 C).

We next sought to identify potential HLA-A2.1-restricted hAFP peptide epitopes recognized by these AFP-specific T cells. A computer-based analysis of the published human AFP coding sequence was performed to identify 9- and 10-mer peptides whose sequences conformed to the well-characterized binding motif for HLA-A2.1 (56, 57, 58, 59). Here, we report the complete analysis of hAFP_542–550 (GVALQTMKQ). This 9-aa peptide has one HLA-A2.1 anchor residue in position 2 and binds with low affinity but forms a stable complex. Table 1 presents the HLA-A2.1 binding properties of hAFP_542–550 and those of two well-characterized HLA-A2.1 immunodominant peptides, Flu M1_58–66 and MART-1_27–35.

Because AdVhAFP/DC-stimulated T cells from human lymphocyte cultures specifically recognized hAFP-transfected targets in both CTL and ELISPOT assays, we wished to determine whether these T cells would also recognize hAFP_542–550 in the context of HLA-A2.1. Therefore, after 7–21 days of culture, AdVhAFP/DC T cells were tested for both cytotoxicity and the frequency of hAFP_542–550-specific IFN-γ cytokine-producing cells by ELISPOT (Fig. 2,A and C). As an internal control, we prepared AdVMART1/DC-stimulated cell cultures, which, as we have reported previously, will generate T cells that recognize the A2.1-restricted MART-1_27–35 immunodominant peptide (Ref. 42); (Fig. 2 B and D). AdVhAFP/DC T cell cultures were specifically cytotoxic for T2 cells pulsed with AFP_542–550 and, when restimulated with autologous peptide-pulsed LCLs, contained over four times as many IFN-γ -secreting cells specific for AFP_542–550 compared to MART-1_27–35.

The hAFP_542–550 synthetic peptide was then used to stimulate human T cells in vitro. Bulk T-cell cultures were generated from PBMCs pulsed with hAFP_542–550 and tested between weeks 3 and 9 of expansion for the ability to kill both peptide-pulsed (Fig. 3) and AFP-expressing targets (Fig. 4). These cultures generated hAFP_542–550 pep-tide-specific CTLs (Fig. 3,A) and an increased frequency of hAFP_542–550-specific T cells (Fig. 3,C) compared to controls for both IFN-γ (Fig. 3,C) and GM-CSF cytokines (data not shown). Results for MART-1_27–35 peptide internal control cultures are also shown (Fig. 3, B and D).

hAFP_542–550 T-cell cultures were tested for cytotoxicity and cytokine release against both hAFP-expressing HCC lines (HepG2 and Hep3B) and stably transfected targets M202 (M202< AFP) and LCL (LCL<AFP). The hAFP_542–550 peptide-specific lymphocytes were specifically cytotoxic for the HLA-A2.1-positive cell line HepG2 and not for the A2.1-negative cell line Hep3B (Fig. 4,A). The killing of HepG2 was increased with IFN-γ treatment (which up-regulates class I) and decreased with β₂-microglobulin blocking antibody. The low level of Hep3B killing was eliminated by the addition of unchromated K562 cells (a natural killer cell target), whereas this did not alter the killing of HepG2 cells (data not shown). Similarly, hAFP transfectants were recognized by hAFP_542–550-specific lymphocytes to a much greater extent than parental cells in both cytotoxicity (Fig. 4,B) and IFN-γ synthesis (Fig. 4 C), indicating that this peptide is naturally processed and presented in these HLA-A2.1-positive cell lines.

HLA-A2.1/K^b tg mice were used to determine whether the murine T-cell repertoire could recognize hAFP_542–550 in the context of HLA-A2.1. Splenocytes from mice immunized with DCs transduced with AdVhAFP were restimulated in vitro with hAFP_542–550 and assayed for both cytotoxicity and IFN-γ production by ELISPOT. AdVhAFP/DC immunization induced a strong AFP_542–550-specific response (Fig. 5,A and C), whereas naive splenocytes showed no cytotoxicity and had few IFN-γ-producing cells (data not shown). Conversely, AdVMART1/DC-immunized mice specifically recognized the immunodominant MART-1_27–35 peptide (Fig. 5 B and D).

To confirm that in vivo immunization with hAFP_542–550 peptide results in AFP-specific responses, tg mice were immunized s.c. in the foot pads with 100 μg of hAFP_542–550 or MART-1_27–35 peptides in CFA. Cytotoxicity and IFN-γ-specific ELISPOT assays were performed with recovered splenocytes and lymph node cells restimulated in vitro with Jurkat/AFP- or Jurkat/MART-transfected cell lines. Immunization with hAFP_542–550 and subsequent restimulation with Jurkat/AFP induced both peptide-specific (Fig. 6,A) and AFP-specific (Fig. 6,C) cytotoxicity as well as large numbers of IFN-γ-producing cells (Fig. 6 E). Lymphocytes from PBS-injected mice showed neither cytotoxicity nor IFN-γ production, regardless of restimulation (data not shown). Mice immunized with MART-1_27–35 peptide produced MART-1-specific responses.

This is the first report that the human T-cell repertoire can recognize AFP in the context of MHC. One of the HLA-A2.1-restricted peptide epitopes recognized by these T cells is the 9-mer hAFP_542–550 (GVALQTMKQ). This result was obtained using two basic strategies. In the first strategy, DCs genetically engineered to express hAFP were shown to generate AFP-specific T-cell responses in CD8-enriched human lymphocyte cultures. These AFP-specific T cells also recognized AFP_542–550-pulsed cells in both cytotoxicity and cytokine release assays. In addition, we demonstrated that AFP_542–550 peptide-generated T cells recognized both hAFP-expressing HCC lines and hAFP-transfected targets in human lymphocyte cultures and in HLA-A2.1/K^b tg mice. Thus, we provide compelling evidence that: (a) the human T-cell repertoire recognizes AFP in the context of HLA-A2.1; and (b) AFP_542–550 is naturally processed and presented by AFP-engineered DCs and AFP-transfected targets serving as a class I-restricted immunogen and antigen, respectively.

The findings from this human AFP model support the important concept that certain “self” proteins overexpressed by tumor cells can be recognized by CD8 T cells in a MHC class I-restricted fashion. For properly selected antigens, this autoimmune response can be an effective antitumor immune response (42, 60, 61).

It is clear that the T-cell repertoire is capable of recognizing these “self” peptides. For MART-1 and gp100 immunodominant peptides, the HLA-A2 binding affinities are intermediate due to the absence of optimal residues at either the second or ninth anchor positions (59, 62, 63, 64, 65, 66). One hypothesis that has been forwarded is that high affinity peptides expressed at high levels on the cell surface may generate tolerance due to strong negative selection in the thymus during initial T-cell education (50, 67). Lower binding or “subdominant” determinants may be capable of stimulating peptide-responsive T cells not deleted from the T-cell repertoire. Work by the Melief group has provided evidence that the off-kinetics, or dissociation rate, of a peptide bound to class I is highly predictive of the immunogenicity of that peptide (50). This group analyzed known self, immunogenic epitopes from melanoma antigens (gp100 and MART-1) and found that their stable binding affinity in a soluble class I reconstitution assay was low, but they had very slow off-kinetics. Like many tumor antigen peptide epitopes (68, 69, 70, 71, 72, 73), hAFP_542–550 has one anchor residue and intermediate binding affinity to HLA-A2.1. Molecular modeling of GVALQTMKQ in the HLA-A2.1 binding groove showed that the valine side chain at P2 was buried in the B pocket of the MHC molecule, whereas the side chains in the central part of the peptide (L4-T6) were mostly exposed. The lack of a hydrophobic side chain (V or L) at P9 may explain the weak affinity of this peptide for HLA-A2.1. We are currently investigating single and double amino acid substitutions in this peptide to potentially increase binding and immunogenicity while retaining AFP specificity.

The HLA-A2.1/K^b tg mice allow an in vivo analysis of AFP epitope processing and presentation in the context of HLA-A2.1. Several reports have already shown the comparable range of epitopes presented by these mice and humans (74, 75). In these mice, AFP genetic immunization generated hAFP_542–550-specific responses, and peptide hAFP_542–550 immunization generated AFP-specific responses to cells endogenously expressing the gene. Together, these data confirm that AFP can serve as a tumor-specific antigen in HCC and is a suitable target for T-cell-mediated immunotherapy strategies in this disease.

We thank Dr. Eli Sercarz and Dr. Mitchell Kronenberg for helpful discussions, Dr. C. J. Melief and Dr. S. Van der Burg for suggestions on the use of the off-kinetics assay, and Dr. Linda Sherman for suggestions on the use of the HLA-A2.1/K^b tg mice.