Getting to the Surface: A Link Between Tumor Antigen Discovery and Natural Presentation of Peptide-MHC Complexes Free

Despite their origin from self-tissue, tumor cells can be immunogenic, triggering immune responses that can profoundly influence tumor biology. The clinical implications are obvious: It may be possible to amplify or induce antitumor immune responses to achieve tumor rejection in patients. Cytotoxic T lymphocytes (CTLs) are a primary mediator of antitumor immunity. The same rules of engagement that dictate conventional T-cell responses to intracellular pathogens also govern CTL recognition of tumor cells, namely antigen specificity and restriction to self-MHC. Rearranged T-cell receptors on CTL recognize the three-dimensional complex of cognate peptide antigen bound in the groove formed between the α1 and α2 domains of surface MHC class I molecules. Like other antigens, tumor-associated antigens (TAA) are degraded by the proteasome into short peptides, transported into the endoplasmic reticulum, packaged in the groove of newly synthesized MHC molecules, and delivered as peptide-MHC (pMHC) complexes to the cell membrane. Engagement of the appropriate T-cell receptor by these pMHC complexes conveys signals to the CTL, which, in the context of other costimulatory signals, activate the CTL to proliferate, produce cytokines, and ultimately seek out and lyse target cells presenting the same antigen (Fig. 1).

An increased understanding of TAA has been instrumental to the development of tumor immunotherapies. Although most TAA characterized thus far are derived from wild-type products of the human genome, both preclinical studies and clinical trials of prototype vaccines and adoptive cellular therapies have shown that T-cell tolerance against self-TAA is incomplete and that immune responses against cancer can be generated in patients with occasionally dramatic tumor regression.

To date, several hundred human TAA epitopes have been described (1), raising the conundrum, “How can we discern which of these epitopes are the best candidates for tumor-rejection antigens?” The method of discovery must be taken into consideration. TAA were first characterized based on the molecular dissection of tumor-reactive T cells isolated from patients (2, 3). Other investigators eluted peptides from the groove of tumor MHC molecules, sequenced them, and tested them for T-cell reactivity (4, 5). Eventually, bioinformatic programs were developed to simplify the approach by predicting which epitopes from a particular protein would be most likely to bind to common MHC alleles based on their biochemical properties (6, 7). Predicted peptides can be synthesized, confirmed experimentally to bind to MHC, and used to generate peptide-specific CTL that kill tumor cells in an antigen-specific, MHC-restricted manner. Prostate-specific antigen, carcinoembryonic antigen (CEA), proteinase 3, her2/neu, Wilms' tumor antigen-1, survivin, telomerase, and many more TAA have been characterized in this fashion (1, 8).

A chief advantage of this approach, often termed “reverse immunology,” is to focus experimental resources only on those candidate TAA with desirable expression profiles or functions, e.g., proteins broadly expressed in cancer but not in normal cells, or proteins critical to tumor growth and development such that deletion or mutation as a means of immune escape would itself lead to death of the tumor cell (8). A major disadvantage of reverse immunology, however, is that experimental verification that the predicted peptide is actually found on the surface of antigen-presenting cells and tumor cells has generally been a late step in the process. Neither processing nor presentation of pMHC is guaranteed a priori for any given 9 or 10 amino acid peptide of interest, even if mathematically predicted or experimentally shown to bind tightly to a particular MHC allele (Fig. 2). In our experience with tumor antigens, <50% of predicted peptides for which a specific T-cell receptor repertoire exists can actually be used to generate CTL that kill tumors in vitro. Without efficient presentation of pMHC on the surface of professional antigen-presenting cells, antigen-specific CTL priming is minimal. Without presentation of pMHC on the surface of tumor cells, antigen-specific CTL killing is impossible.

In this issue of Clinical Cancer Research, Miyahara and colleagues describe a refinement in the search for TAA in which the natural processing and presentation of candidate peptide epitopes is required at multiple experimental steps for the epitope to prove its immunologic worth (9). The investigators describe two novel epitopes—one each from the TAA MAGE-A4 and SAGE and both restricted to the HLA-A*2402 allele of MHC class I—that are, by virtue of the discovery process, highly likely to be found on the surface of HLA-A*2402+ tumor cells and, therefore, accessible to peptide-specific, HLA-A*2402–restricted CTL.

For the priming phase, the investigators took advantage of a useful immunologic tool, the HLA-transgenic mouse. This mouse expresses a chimeric MHC molecule consisting of human β2 microglobulin linked to the α1 and α2 domains from a particular human MHC class I allele, and the murine H-2D^b α3, transmembrane, and cytoplasmic domains. Use of human α1 and α2 domains restricts epitopes identified by this method to the human HLA allele in question. The authors first generated transgenic mice expressing HLA-A*2402, an allele found in >60% of Japanese patients and with moderate prevalence in other patient populations. These mice were then crossed onto a β2 microglobulin-deficient background. Because β2 microglobulin is necessary to stabilize MHC class I molecules at the cell surface, its absence minimizes expression of endogenous mouse MHC class I molecules that would produce confounding results. The investigators immunized these mice with a cDNA plasmid encoding full-length MAGE-A4 or SAGE, harvested CD8⁺ splenocytes, and tested for memory responses to several peptides that they had previously predicted to bind HLA-A*2402.

These experiments identified two promising epitopes, one each from MAGE-A4 and SAGE. To confirm that these epitopes would also be processed and presented by human cells, parallel experiments were done in vitro by stimulating human CD8⁺ T cells with autologous CD4⁺ phytohemagglutinin blasts electroporated with mRNA encoding MAGE-A4 or SAGE. Phytohemagglutinin T-cell blasts are used as surrogate antigen-presenting cells in these studies, given their ease of preparation from a small number of precursor cells. Previous studies with professional antigen-presenting cells, such as dendritic cells and activated B cells (10, 11), have shown that RNA transduction is a highly efficient means of introducing TAA into nondividing antigen-presenting cells, giving the translated protein direct access to the cellular processing and presentation machinery. Stimulation with mRNA-transfected T-cell blasts led to the expansion of MAGE-A4 and SAGE-specific CTL that lysed an HLA-A*2402+ cell line pulsed with the peptide epitopes identified in the murine study. Specific CTL clones were established and, importantly, shown to kill HLA-A*2402+ tumor cells expressing MAGE-A4 or SAGE, but not tumor cells lacking either the TAA or the appropriate HLA allele. This series of experiments shows an approach to identify MHC-restricted TAA epitopes that are endogenously processed and presented by tumor cells and, therefore, may represent clinically relevant targets for immunotherapy. Knowing the dominant epitope is not only critical for designing peptide-based immunotherapies but also facilitates immune monitoring of patients receiving more complex treatment formulations, such as those incorporating full-length copies of TAA genes or whole tumor cells expressing the TAA.

Other approaches for evaluating pMHC complexes on the cell surface have been explored, including mass spectroscopy of peptides eluted from MHC molecules (4, 12, 13), anti-pMHC antibodies generated using phage technology (14), and soluble recombinant T-cell receptor molecules that bind pMHC with high avidity (15). Each of these approaches is clearly useful in revealing pMHC on the surface of tumors and antigen-presenting cells. In some cases, fewer than 10 specific copies of pMHC per cell can be detected biochemically (12). Nevertheless, specific CTL responses are also exquisitely sensitive; even trace amounts of pMHC can activate CTL (16–19). These various approaches can be complementary. For example, HLA-A*0201–transgenic mice vaccinated with full-length cDNA for the cytochrome P450 isoform 1B1 mounted strong CTL responses specific for the peptide CYP190, which had been predicted to bind to HLA-A*0201 (13). Notably, CYP190 peptide was also the only epitope from cytochrome P450 1B1 reported in the groove of HLA-A*0201 from human tumor cells by mass spectroscopy (13).

As tempting as it is to make the biochemical detection—rather than just the immunologic demonstration—of pMHC complexes a requirement for the validation of candidate TAA epitopes, one caveat is that epitopes with the most abundant pMHC on the surface of cells may not necessarily be optimal for inducing antitumor CTL responses. Low abundant or cryptic epitopes from TAA can also drive CTL reactivity, circumventing tolerance induction and drawing upon a higher avidity T-cell receptor repertoire than typically exists for abundant or dominant epitopes (20). Cryptic epitopes may be undetectable biochemically, but they can induce prominent CTL responses following heteroclitic modification of MHC-binding anchor residues or residues that directly interact with the T-cell receptor. One example is the CEA605 epitope, which in one study was not observed in the groove of HLA-A*0201 on tumor cells by mass spectroscopy unlike other HLA-A*0201–binding epitopes from CEA (12). Substitution with aspartate for asparagine at position 610 increases potency for CTL induction in vitro (21). When used as part of a vaccine in patients with CEA-expressing cancers, the CEA-derived peptide generated CTL specific for both the wild-type and modified epitope and caused significant tumor regression in a few patients (22). Similarly encouraging results have been reported for recombinant vaccines that use full-length CEA modified at this residue (23).

Collectively, these studies show that a complex array of bioinformatics, biochemistry, and immunology can be used to whittle down the expanse of the human proteome to those peptide epitopes most likely to mediate T-cell rejection of tumors. Immunologic demonstration of natural processing and presentation of peptide epitopes is a key criterion and discovery methods that help guarantee this property will be important as efforts continue in the development of novel immunotherapies for cancer.