gp100/pmel 17 Is a Murine Tumor Rejection Antigen: Induction of “Self”-reactive, Tumoricidal T Cells Using High-affinity, Altered Peptide Ligand

Female C57BL/6 (H-2^b) mice, 6–10 wk old, obtained from Frederick Cancer Research Center (Frederick, MD) and maintained in a barrier facility, were used for all experiments. EL-4 thymoma (H-2^b) and the derived β-galactosidase (β-gal)–transfected clone E22 have been described previously (12). B16 (H-2^b), hereafter named B16.WT, is a spontaneous murine melanoma expressing gp100, MART-1, tyrosinase, tyrosinase-related protein (TRP)-1, and TRP-2 by FACS^® and Western blot analysis (data not shown). B16.B7-1 is a hyperpigmented clone of B16.WT that was stably transfected using a Moloney mouse leukemia virus encoding the gene for B7-1 driven by a LTR promoter. JB/MS is a pigmented, chemically induced melanoma expressing gp100, provided by Dr. Vincent Hearing (National Cancer Institute, NIH, Bethesda, MD). 293K^b and 293K^bD^b are highly transfectable human renal 293 cells stably transfected with K^b and K^b plus D^b, respectively. RMA/S (H-2^b) is a cell line deficient in endogenous peptide loading (13). EL-4, B16.WT, RMA/S, MCA205, and MC38, a C57BL/6-derived colon carcinoma, were maintained in complete medium (CM; RPMI 1640 with 10% heat-inactivated fetal bovine serum [FBS; Biofluids, Rockville, MD], 0.03% l-glutamine, 100 μg/ml streptomycin, 100 μg/ml penicillin, and 50 μg/ml gentamycin sulfate [NIH Media Center, Bethesda, MD]). B16.B7-1 and E22 were maintained in CM with 400 μg/ml of bioactive G418. 293K^b and 293K^bD^b were maintained in DMEM with 10% heat-inactivated FBS, 0.03% l-glutamine, 100 μg/ml streptomycin, 100 μg/ml penicillin, 50 μg/ml gentamycin sulfate, and 400 μg/ml of bioactive G418.

All synthetic peptides were synthesized using regular F-MOC chemistry. The following synthetic peptides were all synthesized by Peptide Technologies (Washington, D.C.) to a purity >99% by HPLC and amino acid (AA) analysis: KVPRNQDWL, spanning AAs 25–33 of human (h)gp100; EGSRNQDWL, spanning AAs 25–33 of mouse (m)gp100; ASNENMETM, spanning AAs 366–374 of the nucleoprotein of influenza A (NP); DAPIYTNV, spanning AAs 96–103 of Escherichia coli β-gal (12); and SIINFEKL, spanning AAs 257–264 of chicken OVA. The NP, hgp100, and mgp100 peptides are restricted by H-2D^b, and the β-gal and OVA peptides by H-2K^b.

All recombinant vaccinia viruses (rVVs) used in this study were generated by insertion of the foreign genes by homologous recombination and subsequent purification of recombinant progeny as described by Earl et al. (14). rVVs encoding the human and mouse gp100 epitopes as minigenes were constructed using a recombination plasmid, pKT1401, containing an early promoter driving the signal sequence of the adenovirus E3/19k protein followed by an additional Ala using code GCA, resulting in a putative signal peptide cleavage site, and fused to oligonucleotides encoding the epitope sequences for mgp100, hgp100, or β-galactosidase as described above. pKT1401 was recombined into the VV locus encoding the small subunit of viral ribonucleotide reductase. Correct integration was checked by PCR-based viral genome analysis using primers flanking the viral locus encoding the small subunit of ribonucleotide reductase (12). rVVmgp100, rVVmMART-1, rVVmTyr, rVVmTRP-1, and rVVmTRP-2 were based on the plasmid pSC65, in which the completely synthetic early/late promoter pS_E/liter drives expression of the antigen and the early/late promoter p7.5_E/liter drives expression of the LacZ gene (15). The rVVS_E/Lβ-gal has been described previously (15, 16). Cloning of the genes for mgp100 and mMART-1/Melan-A has been described previously (11), and the cDNAs for hgp100, mTyr, mTRP-1, and mTRP-2 were gifts from Dr. Y. Kawakami (Surgery Branch, NIH; reference 9), Dr. H. Yamamoto (Tohoku University, Sendai, Japan; reference 17), Dr. S. Shibahara (Friedrich Miescher Institut, Basel, Switzerland; reference 18), and Dr. V. Hearing (Laboratory of Cell Biology, NCI, NIH; reference 19), respectively. The rVVhgp100, recombinant fowlpox virus (rFPV)hgp100, rFPVmgp100, and rFPVβ-gal were provided by Therion Biologics Corp. (Boston, MA). Plasmid DNA constructs (pDNA) were based on the pCDNA3 backbone, and encoded hgp100, mgp100, or β-galactosidase under the control of the CMV promoter. Expression of rVV, rFPV, and pDNA were confirmed by immunostaining (20) as well as by Western blot of transfected and infected cells using antisera (21, 22) provided by Dr. V. Hearing. Recombinant adenoviruses were provided by Dr. Bruce Roberts (Genzyme Corp., Framingham, MA), and encoded the genes for hgp100, mgp100, or β-galactosidase under the control of the CMV promoter (11).

Mice were vaccinated twice at 3-wk intervals by a hand-held helium-driven device (Geniva Inc., Middletown, WI) with 1 μg of hgp100 DNA (23), and 3 wk after the second vaccination splenocytes were cultured with rVVhgp100-infected dendritic cells (DCs), generated as previously described (16) in CM with 30 IU/ml rhIL-2 (a gift from Chiron Corp., Emeryville, CA) for 7 d, and were subsequently restimulated every 7–10 d with 2 × 10⁵ B16.B7-1 (60,000 rad) and 5 × 10⁶ irradiated splenocytes (1,500 rad) in CM with 30 IU/ml rhIL-2. T cell clones were derived by limiting dilution using 0.3 T cells per well cocultured with 4 × 10⁵ irradiated splenocytes and 2 × 10⁴ irradiated B16.B7-1 cells in U-bottomed 96-well plates. The generation of the β-gal reactive cell line has been described previously (12). All T cells were used between 5 and 10 d after restimulation. For peptide-induced T cells, mice were vaccinated as specified in figure legends. Spleens were harvested on day 21 after the last vaccination, separated into single cell suspension, and cultured in T-25 flasks in 15 ml CM at 5 × 10⁷ cells per flask. Peptide was added to a final concentration of 1 μM. After 6 d of culture, cells were washed in CM and used in a cytokine release assay.

To determine cytokine release, 5 × 10⁴ target cells per well were added to effector T cells in round-bottomed 96-well plates at an E/T ratio of 1:1 for established T cell lines and 5:1 for secondary cultures, and were incubated for 24 h in CM. Supernatants were collected and tested using the mIFN-γ ELISA kit (Endogen, Cambridge, MA) or the mGM-CSF ELISA kit (R&D Systems, Minneapolis, MN) according to manufacturer's protocol. rVV-, rFPV- and rAd-infected targets were infected at multiplicities of infection of >10, >10, and >100, respectively, incubated for 2 h, and then washed twice in CM. Peptide-pulsing was performed at 1 μM in CM for 2 h at 37°C and followed by two washes in CM. Transfected targets were prepared using Lipofectamine (GIBCO BRL, Gaithersburg, MD) following manufacturer's guidelines.

RMA/S cells were incubated at room temperature for 24 h, pulsed with peptide at the indicated concentrations for 45 h at room temperature, and then incubated at 37°C for 4 h to allow turnover of “empty” MHC class I molecules (13). Cells were then stained with FITC-conjugated D^b-specific mAb KH95 (PharMingen, San Diego, CA) for 1 h at 4°C and staining was asserted using a FACScan^® (Becton Dickinson, Sunnyvale, CA).

For adoptive transfer experiments, mice were injected with 2 × 10⁵ B16 intravenously on day 0 followed by intravenous injection of the indicated number of T cells on day 3. rhIL-2 (3 × 10⁵ IU) was given twice daily intraperitoneally for 5 d starting immediately after adoptive transfer.

Alternatively, mice were injected intravenously with the indicated number of T cells, followed 5 d later by 2 × 10⁵ B16 intravenously. Mice were killed on day 17 after tumor injection and pulmonary nodules were enumerated. All in vivo experiments were performed in a blinded, randomized fashion.

To generate gp100-specific T cells, mice were immunized with mgp100 encoded in plasmid DNA or rVV, two vectors with proven efficacy in the generation of specific T cells (3, 23, 24). Splenocytes from immunized mice were restimulated in vitro using syngeneic DCs infected with appropriate rVV, rFPV, or rAd, and subsequent restimulations were performed with irradiated splenocytes and irradiated B16.B7-1, a clone of B16 retrovirally transduced to express the costimulatory molecule CD80 (B7-1). None of the cultures exhibited gp100-specific reactivity when tested by IFN-γ release against gp100-positive targets (data not shown). Since xenogeneic antigen can, in some instances, induce immune reactivity where the autologous antigen failed (25–27), mice were immunized with pDNA encoding hgp100.

A CD4⁻CD8⁺ lytic T lymphocyte line was generated by gene-gun administration of pDNA encoding hgp100 followed by restimulation of splenocytes ex vivo with DCs that had been infected with rVVhgp100. After two subsequent in vitro restimulations with B16.B7-1 melanoma, the T cells recognized B16 melanoma, as well as the immortalized normal melanocyte line Melan-A, with a high degree of specificity (Fig. 1). The EL-4 thymoma and the MC38 sarcoma were not recognized. However, recognition could be conferred upon 293 K^bD^b cells after infection with rVV encoding either mgp100 or hgp100, but not after infection with rVV containing any of the other known MDAs, tyrosinase, TRP-1, or TRP-2.

We sought to evaluate the relevance of mgp100 as a tumor rejection antigen. To exclude the potential influence of T cells with a specificity other than gp100, the bulk T cell line was cloned by limiting dilution. 12 clones were evaluated and each had reactivities identical to the bulk T cell line shown in Fig. 1. The gp100-reactive T cell clone (clone no. 9) was tested in vitro (Fig. 2,A) and in vivo (Fig. 2,B). Mice bearing 3-d-old B16 pulmonary nodules were infused with clone no. 9 followed by rIL-2, resulting in a dramatic tumor destruction (P <0.0001, 4 × 10⁶ β-gal–specific T cells + rhIL-2 versus 4 × 10⁶ clone no. 9 T cells + rhIL-2), whereas treatment was ineffective using control β-gal–reactive T cells or rIL-2 alone (P >0.5 versus no treatment) (Fig. 2 B). Separate gp100-reactive T cells clones were comparable in their ability to reject B16 in vivo (data not shown). These results suggest that gp100 functions as a true tumor rejection antigen in established murine melanoma.

We sought to understand why T cells could only be induced by the xenogeneic, human form of gp100 and not the self sequence. To determine whether the difference in immunogenicity resided in the actual peptides from gp100 that were recognized by T cells, or in differences in sequences outside of those recognized, we characterized the epitopes recognized by the T cell clone. To identify the MHC class I molecule that restricted gp100 recognition, we used the human renal cell carcinoma line, 293, stably transfected with the mouse restriction elements, K^b and D^b. Both lines were transiently transfected with plasmids encoding either mouse or human gp100, or with a control plasmid encoding NP. Only 293 K^bD^b cells (and not 293 K^b) cells expressing mgp100 or hgp100 triggered the release of IFN-γ from gp100-reactive T cells, suggesting that D^b was the dominant and perhaps only restriction element for gp100 recognition (Fig. 3 A).

At 626 AAs in length, many possible epitopes in gp100 could be recognized by gp100-reactive T cells. To reduce the number of candidate peptides, a set of progressively shorter versions of the gene encoding hgp100 was used (28). These shortened forms were generated by 3′ exonuclease digestion or PCR amplification of the original hgp100 cDNA. Fragments were then transfected into 293 K^bD^b cells that were used as targets for T cell recognition (Fig. 3 B). Even the shortest fragment of the hgp100 cDNA, with a length of 300 bp by gel electrophoresis, conferred a high degree of recognition upon the 293 transfectants, implying that a major epitope was located in the first 100 AAs of the molecule.

To identify the sequence of the epitope, 9-AA-long peptides were evaluated for their potential to bind to D^b using a computer-generated epitope forecast, based on previously published peptide binding data, that is designed to predict binding affinity for a variety of human and mouse MHC class I alleles (29, 30; and http://bimas.dcrt.nih.gov/molbio/hla_bind). Based on these predictions, sets of synthetic peptides were made from both hgp100 and mgp100, which although similar are not identical. When the peptides were added to cultures of gp100-reactive T cells, one peptide pair, gp100_25–33, was recognized with a high degree of specificity in a GM-CSF release assay (Fig. 3 C).

With the identification of a defined peptide epitope, we explored the immunological mechanism underlying the disparate ability of mgp100 and hgp100 to induce gp100-specific T cells. Mice were vaccinated with rVVmgp100, encoding mgp100, or rVVhgp100, encoding hgp100, and splenocytes were isolated after 3 wk and stimulated for 6 d in vitro with synthetic peptides corresponding to hgp100_25–33, mgp100_25–33, or OVA_257–264, and then tested in a GM-CSF release assay (Table 1, first two columns). Splenocytes from mice vaccinated with rVVmgp100 and restimulated with mgp100_25–33 or hgp100_25–33 peptide failed to develop peptide reactivity. Conversely, splenocytes from mice vaccinated with hgp100 and restimulated with either mgp100_25–33 or hgp100_25–33 peptide developed reactivity to both peptides. Similar results were obtained using splenocytes from mice vaccinated with recombinant fowlpoxvirus or plasmid DNA (data not shown). These observations indicated that T cells recognizing mgp100 could be induced exclusively by xenoimmunization with the hgp100 molecule.

Several mechanisms could account for the apparent immunological unresponsiveness to mgp100 and the ability of hgp100 to break it. Nonhomologous regions of the full-length hgp100 (which is 76% identical to mgp100 at the AA level; reference 11) could result in intramolecular epitope-spreading (31, 32) or facilitate antibody-mediated antigen capture by APCs (26, 33– 35). Alternatively, sequence differences in the relevant epitopes or their flanking sequences could result in differential proteolytic cleavage or transporter associated with antigen processing (TAP)–mediated transport across the endoplasmic reticulum (ER) membrane (36). To explore these possibilities, we constructed a series of rVV-containing minigenes encoding the relevant 9-AA T cell epitopes preceded by the E3/19K adenoviral ER-insertion signal sequence (ES), previously shown to result in TAP-independent transport of antigenic peptides (37), and followed by a double stop codon. These constructs eliminated differences in flanking sequences and other nonhomologous regions of the molecule.

Mice were vaccinated with rVVESmgp100, rVVEShgp100_25–33, or rVVESβ-gal_96–103, encoding the 9-AA mgp100_25–33, hgp100_25–33, and β-gal_96–103 peptides, respectively (38). Splenocytes were cultured for 6 d with mgp100_25–33 or hgp100_25–33 peptides then tested for specificity in a cytokine release assay (Table 1). Reactivities were similar to those obtained with rVV encoding full-length gp100 molecules: immunization with the mouse minigene gp100 construct failed to induce T cells, whereas immunization with the hgp100 construct induced T cells that were cross-reactive with both mgp100_25–33 and hgp100_25–33 peptides. These results were not consistent with a major role for differential peptide processing and transport, antibody-facilitated antigen presentation, or intramolecular immunodominance in the differential in vivo immunogenicity of mgp100 and hgp100.

To evaluate the relative avidity of the gp100-specific T cells for the mgp100_25–33 and hgp100_25–33 epitopes, we pulsed the peptides onto EL-4 cells (H-2^b) (Fig. 4). A striking difference in the relative T cell activities to the mgp100_25–33 and hgp100_25–33 peptides was observed. The mgp100_25–33 peptide was recognized at concentrations as low as 10⁻⁹ M, with half-maximal recognition occurring at about 10⁻⁸ M, whereas the hgp100_25–33 peptide was recognized at concentrations as low as 10⁻¹¹ M. Alanine-substitutions at positions 1, 2, or 3 in either the mouse or human versions of the synthetic peptides did not substantially alter recognition by gp100-reactive T cells (Fig. 5,A). Indeed, a triple-substituted peptide, AAARNQDWL, fully retained the ability to sensitize EL-4 for recognition. (Fig. 5,B). The first three AAs could even be deleted and a substantial degree of recognition remained (Fig. 5,B). However, substitution of any of the AAs at positions 4 through 9 abrogated recognition of both the hgp100_25–33 and mgp100_25–33 peptide variants. To determine the involvement of MHC class I binding affinity, an MHC class I stabilization assay was done on RMA/S cells, which lack activity of the TAP transporters (13). FACS^® analysis revealed 50% stabilization of D^b by hgp100_25–33 peptide at a concentration ∼100-fold lower than for mgp100_25–33 (Fig. 6). This indicated that the apparent avidity of gp100-reactive T cells for the hgp100_25–33 peptide could be largely attributed to its greater ability to stabilize the restricting MHC class I molecule, H-2D^b.

One major potential use for self-reactive T cells is their application in cancer treatment. To evaluate the usefulness of altered antigen in the development of a synthetic cancer vaccine capable of eliciting therapeutic T cells, mice were vaccinated once with rVVhgp100 and restimulated for three rounds with splenocytes pulsed with 1 μM mgp100_25–33 peptide. CD8⁺ T cells were tested for gp100 recognition by IFN-γ release (Fig. 7 A). T cells strongly recognized both mgp100_25–33 and hgp100_25–33 peptides, but not control OVA_257–264 peptide, and recognized as well melanoma B16.WT, B16.B7-1, and JB/MS, a chemically induced C57BL/6 melanoma. Control H-2^b, gp100⁻ tumor lines were not recognized.

The function of T lymphocytes generated using purely recombinant and synthetic forms of gp100 were evaluated by adoptive transfer of T cells to mice bearing tumors established for 3 d in vivo. Mice receiving mgp100_25–33–specific T cells experienced a significant reduction in tumor burden compared with mice receiving control T cells (Fig. 7,B). The survival of these cells was evaluated in vivo by transfer 5 d before tumor challenge, with T cells retaining their antitumor activity for this period (Fig. 7 C). This indicated that tumor-specific T cells induced with an antigen-based cancer vaccine can survive in vivo and mediate efficient tumor destruction.

The nonhomologous sequences flanking a MHC-restricted epitope could influence the immunogenicity of the epitope through a variety of mechanisms. The full-length hgp100 and mgp100 molecules are 76% identical at the AA level (11). Xenoimmunization could induce antibodies to nonhomologous determinants on the xenoantigen. When expressed on the surface of B cells that produce them, these antibodies could capture the xenoantigen and make it available for B cell processing and presentation on MHC class II to activate CD4⁺ T cells (35). This mechanism has been postulated to play a role in the initiation of human autoimmune diseases such as SLE, which is largely mediated by CD4⁺ T cells and autoreactive antibodies (25, 34). Recent data suggest that B cells may cross-present antigen on MHC class I after capture (39). However, there is a vigorous debate over the ability of B cells to activate “virgin” T cells (i.e., T cells that have not previously been activated by antigen) (33, 40). DCs can also capture immune complexes containing xenoantigen through Fc receptors and present it through MHC class I and class II pathways, inducing de novo activation of autoreactive T cells (41). In this scenario, the completely autologous mgp100 would not induce such antibodies and thus would fail to be captured and presented. The involvement of extra-epitope sequences in the immunogenicity of the human gp100 molecules is not consistent with the results shown in Table 1 where the immunogenicity of human full-length and minimal determinant constructs are compared. In fact, the two constructs elicit comparable CD8⁺ T cell responses to the 25–33 epitope.

Another mechanism by which xenoimmunization could enhance immunogenicity is by the processing of a given epitope. For example, the hgp100_25–33 peptide might be processed more efficiently than the mgp100_25–33 peptide, since the two differ in the three NH₂-terminal AAs, as well as in the AA sequences surrounding the peptides in the full-length molecule (36, 42). The differences between the processing of the human and mouse sequences is minimized and possibly eliminated by the use of minigene constructs preceded by ER-insertion signal sequences that bypass proteasome-mediated peptide liberation as well as TAP-mediated peptide transport (Table 1).

Indeed, the increased immunogenicity of human gp100 appeared to reside completely within the 9-mer peptide. The hgp100_25–33 peptide differed from its mouse counterpart in three NH₂-terminal AAs. In fact, these three differences were rather dramatic, with a positively charged lysine (K) replacing a negatively charged glutamic acid (E), a medium-sized valine (V) replacing a small glycine (G), and proline (P), which is a cyclic residue and known to reduce the number of possible conformations due to impaired hydrogen bonding, replacing serine (S), which may not induce such a structural distortion. Despite the presence in both peptides of optimal anchor residues at both dominant anchor positions, 5 (N) and 9 (L), the difference in the three NH₂-terminal residues resulted in a dramatically increased affinity of the hgp100_25–33 peptide for the mouse MHC class I allele, H-2D^b, compared with the mgp100_25–33 peptide (Fig. 6). Clearly, the ability of the peptide to stabilize D^b molecules on the surface of RMA-S cells could reflect the amount of peptide presented on the surface of APCs, which in turn could determine the activation of T cell precursors. In retrospect, it was fortuitous that the interspecies differences in the gp100 sequences created such a high binder, whereas the similarities preserved sufficient T cell receptor contact residues to allow cross-recognition.

The efficacy of the mechanism described here through which xenoimmunization induced autoreactive T cells required that a number of criteria be satisfied. There must be an MHC-binding epitope in the autologous protein which is naturally processed and presented that can be recognized by the available T cell repertoire. Most importantly, a homologous epitope from the xenogeneic protein must also be naturally processed and presented, but must be presented in the context of MHC at a higher density on the surface of the APCs. The chance that such an epitope will be found in any give xenoantigen may be small. However, an understanding of this mechanism clearly points the way to the rational design of immunogens based on the enhancement of the stability of peptide–MHC complexes.

The data presented here suggest that the main difference between the mgp100_25–33 peptide and its hgp100_25–33 analogue resides in binding to MHC class I. In human in vitro studies, modification of the second AA of the hgp100 peptide epitope hgp100_209–217 to methionine has been shown to significantly increase its affinity to HLA-A2, leading to dramatically increased ability to raise gp100-specific T cells from patient PBLs in vitro (43). The higher affinity peptide is also more effective in vaccinating patients in vivo, increasing gp100-specific T cell precursor levels, and possibly resulting in higher treatment response rates (43a). Similarly, other groups have reported increased immunogenicity of peptide epitopes with enhanced MHC class I binding (26, 44, 45).

Taken together, these data suggest that there is indeed unresponsiveness to gp100 in mice. However, the unresponsiveness is relative and can be broken by using a peptide homologue with higher affinity for MHC class I. One mechanism through which a peptide with higher MHC class I binding can break tolerance is based on the assumption that T cell tolerance exists to a level of antigen rather than to the identity of the antigen (44, 46, 47). In the case of gp100, CD8⁺ T cell precursors with the ability to recognize a certain amount of peptide in the context of MHC class I are inactivated, through either thymic and/or peripheral deletion or anergization. The remaining T cells have TCRs with an affinity that is too low to be triggered by the levels of gp100 peptide present on melanocytes or APCs in lymph nodes draining the skin. Therefore, these T cells are never activated, deleted, or anergized. Instead, they remain “ignorant” of MHC class I with mgp100_25–33 peptide. A vaccination with autologous mgp100 will not trigger these T cells unless the vaccine is able to significantly raise the amount of peptide–MHC class I complex on professional APCs to a level high enough to surpass the TCR threshold. Only then do the T cells become activated. We are currently evaluating the clinical efficacy of antitumor effects of gp100-based cancer vaccines that contain epitopes with enhanced stability of peptide–MHC complexes.

The authors wish to thank Deborah R. Surman and Mr. Paul J. Spiess for expert technical assistance; Ms. Martha Blalock for assistance with graphics; Drs. Dennis Panicali and Linda Gritz (Therion Biologics) for providing rVVs; and Dr. Bruce Roberts for providing recombinant adenoviruses.

AA

amino acid

β-gal

β-galactosidase

CM

complete medium

DC

dendritic cell

ER

endoplasmic reticulum

FPV

fowlpox virus

h

human

m

mouse

MART

melanoma antigen recognized by T lymphocytes

MDA

melanocyte differentiation

NP

the nucleoprotein from influenza A

pDNA

plasmid DNA

rVV

recombinant vaccinia virus

TAP

the transporter associated with antigen processing

TRP

tyrosinase-related protein