In the case of the great majority of SEREX-defined antigens, it is unclear why they elicit a humoral immune response. The important point would be to distinguish antigens that have no direct relevance to cancer (e.g., antigens detected by preexisting autoantibodies or antibodies elicited by antigens related to necrotic tumor products) from antigens that have some casual relation to cancer etiology or cancer phenotype. Past studies have demonstrated a high frequency of autoantibodies to known normal tissue autoantigens in cancer patients (80), and it would not be surprising that a significant proportion of the currently defined SEREX antigens are autoantigens of this kind. However, the importance of these antigens in the context of cancer antigens should not be dismissed without understanding the reason for their immunogenicity in cancer patients. For instance, mutational events in the tumor may elicit antibodies that cross-react with the corresponding nonmutated counterparts in normal cells. Under these circumstances, the immunogenic stimulus for the antibody response in cancer patients is in fact a mutated or altered gene product, but, because the resulting antibody cross-reacts with the wild-type protein, the wild-type gene (from admixed normal cells or nonmutated alleles) might be isolated in SEREX. Unless the mutated gene is identified, such antibodies would be classified as conventional autoantibodies. This scenario has been recognized as a major problem in interpreting the fact that mutations have been detected so rarely in SEREX-defined genes (29). Sequencing of repeated isolates of the gene from the cancer, microdissected cancer cells, or cell lines derived from the cancer would be one way to address this critical issue. Given the large number of SEREX-defined antigens, however, this clearly represents a daunting task.

With regard to SEREX antigens with obvious or suspected cancer relatedness, immunogenicity can be ascribed to one of several mechanisms: gene activation or repression, mutation, amplification, mRNA overexpression, or expression of abnormal splice variants. The immune response to CT antigens is clearly related to the anomalous expression of gene products in cancer that are normally only expressed in primitive germ cells. CT antigen expression in cancer has been ascribed to abnormal demethylation (72, 134), although other mechanisms may well be involved. Anomalous antigen expression also appears to be the basis for certain paraneoplastic syndromes affecting the central nervous system. These syndromes are believed to result from autoimmune recognition of neural antigens aberrantly expressed by nonneural cancers, and specific autoantibodies are often found to be associated with specific tumor types (81). For example, in paraneoplastic cerebellar degeneration, a syndrome seen in patients with breast and ovarian cancers, autoantibodies have been found to react with neuronal antigens, including CDR34, an antigen strongly expressed in Purkinje cells of the cerebellum (82). The experimental precedent for immunogenicity due to anomalous activation of a gene in cancer comes from the study of the TL system of antigens in the mouse (83). In some mouse strains (TL+ strains), TL is a normal alloantigen whose expression is limited to thymocytes. In other strains (TL- strains), no normal cell types express TL. However, leukemias arising in TL- as well as TL+ strains can express TL and a strong humoral immunity against TL can be elicited in TL- mice.

With regard to mutations as a basis for immunogenicity, p53 and CDX2 are good examples among SEREX-defined antigens. In the case of p53, however, it is unclear whether mutation or accumulation of high levels of p53 in cells harboring p53 mutations represents the initial antigenic stimulus leading to the development of p53 antibodies. Nevertheless, what is clear is that the resulting p53 antibodies recognize wild-type p53 sequences rather than showing specificity for mutated sequences (84).

Amplified expression appears to be one of the most frequent reasons for the immunogenicity of antigens isolated by SEREX, and many examples of antibodies to overexpressed gene products in cancer have been detected in SEREX analysis. Thus, the immune system appears poised to respond to quantitative as well as qualitative changes in antigen expression in cancer cells. Until the basis for amplification or overexpression has been understood and the specificity established, however, the relation between antigen overexpression and antibody response can only be regarded as a strong correlation rather than a causal relationship.

Finally, another stimulus for an immune response has been postulated to be splice variants of genes, which are differentially expressed in normal tissues but aberrantly expressed in cancer. For example, PDZ-54, one of five splice variants of NY-CO-38 normally expressed in kidney, brain, but not colon, was found to be expressed in colon cancer. Thus, tolerance may not extend to normal splice variants that are aberrantly expressed in a cancer, but this idea remains to be formally demonstrated.

Seroepidemiology of SEREX-defined Antigens

Serology plays a central role in three phases of SEREX analysis.

1. In the initial identification of reactive clones
2. In screening small panels of sera from normal individuals and cancer patients for antibody (petit serology)
3. In large-scale surveys of human sera (grand serology)

Petit serology, using the isolate as target antigen, provides some indication of cancer-specific recognition of the antigen, whereas grand serology, using recombinant protein as target antigen, establishes the seroreactivity pattern of humans with or without cancer on a larger scale. Although only a small percentage of SEREX-defined antigens has been subjected to petit or grand serology, there is a growing list of antigens that show a promising degree of cancer-specific recognition in petit serology. For instance, in the SEREX analysis of four colon cancers by Scanlan et al. (29), 6 of 48 antigens isolated in the study showed a cancer-restricted recognition pattern in tests with 16 normal sera and 29 colon cancer sera. In a SEREX analysis of four renal cancers by Scanlan et al. (33), 12 of the 65 antigens isolated in the study showed a cancer-restricted recognition pattern in tests with 19 normal serum and 32 renal cancer patients. As a rule, the highest reactivity frequency with these antigens is 20% to 25% of patients, and there is a distinctive seroreactivity pattern with each of the antigens. As a consequence, the combined use of the six restricted antigens in the colon cancer panel detected 69% of sera from colon cancer patients, and the 12 restricted antigens in the renal cancer panel detected 72% of sera from renal cancer patients. Reactivity is not restricted to patients with the corresponding cancer type; sera from patients with other forms of cancer (e.g., lung or breast cancers) recognize a proportion of the antigens derived from colon and renal cancer. Table 4 shows an updated seroepidemiologic survey by Scanlan et al. (29, 33, 36, 141) of seven SEREX-defined antigens showing a cancer-restricted recognition pattern identified in our analysis of colon and renal cancer. This idea of testing multiple antigens with the hope of reaching a panel that would offer adequate sensitivity and specificity for serological diagnosis of cancer has also been tested in other more recent SEREX studies (104, 107, 115, 128). However, this goal is yet to be reached for any tumor type as of today.

Petit serology, although useful to identify antigens worthy of future study, is laborious and has several limitations:

1. It requires large amounts of sera.
2. Sera must be preabsorbed to remove Escherichia coli or phage reactivity.
3. Only small numbers of sera can be tested at one time.

For this reason, ELISA tests with recombinant protein (grand serology) offers a number of advantages (e.g., does not require preabsorbed sera, requires small amounts of sera, a large number of sera can be tested, and the analysis is quantitative). However, some degree of sensitivity (approx. 1 log) is sacrificed in grand serology as compared to petit serology. NY-ESO-1 is the first SEREX-defined antigen to be analyzed in grand serology (85). No reactivity was found with 70 sera from normal individuals. Antibody to NY-ESO-1 was found in approx. 10% of sera from unselected patients with melanoma and ovarian cancer. To investigate the relation between NY-ESO-1 expression in the tumor and antibody response, a series of 62 melanoma patients were tested, 15 with NY-ESO-1+ tumors and 47 with NY-ESO-1- tumors. The conclusions were clear - NY-ESO-1 antibody was only found in patients with NY-ESO-1+ tumors, and up to 50% of patients with advanced NY-ESO-1+ tumors had NY-ESO-1 antibody.

SEREX-defined antigens showing cancer-restricted seroreactivity offer a range of opportunities for cancer diagnosis and disease monitoring. To explore these applications, the current approach using ELISA technology and recombinant SEREX-defined antigens provides a satisfactory methodology. The modified "dot blot"-based revision of the petit serology technique, SADA (141) and SMARTA (142), are also useful alternatives when the goal is to screen a larger number of SEREX-defined antigens against a limited number of serum samples. In the future, however, protein chip technology holds great promise for miniaturized, rapid, and large-scale screening of human sera for antibodies against SEREX-defined antigens.

T-cell Recognition of SEREX-defined Antigens

The detection of tyrosinase and MAGE-1 by SEREX, two tumor antigens initially recognized by epitope cloning as targets for CD8 T cells, established the critical principle that the analysis of humoral immunity to tumor antigens has the potential for identifying CD8 T-cell recognized antigens. In addition, because production of IgG antibodies is known to require CD4 T-cell help, SEREX analysis can be viewed as a way to define the CD4 T-cell repertoire against human tumor antigens. A number of laboratories are developing approaches for defining the peptide targets for CD8+ and CD4+ T-cell recognition of SEREX-defined antigens. NY-ESO-1, one of the first antigens isolated by SEREX, provides a model for defining the T-cell recognized peptides of a tumor protein initially identified by antibody (86). In the case of CD8 T cells, an HLA-A2+ melanoma patient with high-titered NY-ESO-1 antibody was also found to have strong CTL reactivity against the autologous NY-ESO-1+ melanoma. To investigate the possibility that NY-ESO-1 was the target for the CD8 recognition in this patient, COS cells were cotransfected with HLA-A2 and the NY-ESO-1 coding gene, and these transfectants were found to be lysed by CTLs from the patient with high-titered NY-ESO-1 antibody. Additionally, the reactivity of these CTLs cotyped with NY-ESO-1 expression in a panel of HLA-A2+ melanoma. To identify the NY-ESO-1 peptide epitopes recognized by the CTLs, a series of overlapping peptides were synthesized on the basis of known HLA-A2 peptide-binding motifs, and three of these peptides were found to be specifically recognized. Subsequent studies with CTLs from other HLA-A2+ patients with NY-ESO-1+ tumors and NY-ESO-1 antibody showed recognition of these HLA-A2 restricted peptides. For the identification of CD4-recognized NY-ESO-1 peptides, a similar general strategy was followed (87). CD4 T cells from two patients with NY-ESO-1+ melanoma and NY-ESO-1 antibody recognized NY-ESO-1 target cells pulsed with NY-ESO-1 protein in an HLA-DRB4 0101-0103-restricted fashion in enzyme-linked immunospot (ELISPOT) analysis. Overlapping NY-ESO-1 peptides were synthesized, and three of these were recognized by CD4 T cells in ELISPOT and proliferation assays using peptide-pulsed target cells.

Protocols used to detect T cell reactivity to a molecularly defined protein and to identify the epitopes within it, although somewhat variable from laboratory to laboratory, are usually based on similar approaches, as exemplified above for NY-ESO-1. Using these methods, T cell reactivity to several SEREX-defined antigens have been studied. NY-ESO-1 protein has been extensively analyzed, leading to the identification of multiple CD8+ and CD4+ HLA-restricted peptide epitopes. Two open reading frames (ORF) are known to exist in NY-ESO-1, encoding proteins of 180 and 58 residues, and T cell epitopes were found in both ORFs (74). A similar analysis has now been extended to SSX2, another CT antigen, and an HLA-A2 CD8+ T cell epitope has been defined (145, 146), as well as CD4+ T cell epitopes (154). Aside from CT antigens, T cell reactivity toward other SEREX-defined antigens has also been described and CD8+ epitopes have been defined, e.g. in NY-BR-1 (155) and coactosin-like protein (121).

Because the definition of targets for T-cell recognition is a far more complex and laborious task than defining antibody targets, current technologies place a limit on the number of antigens that can be analyzed from the T-cell perspective. In my opinion, SEREX-defined antigens eliciting high-titered antibodies with a cancer-restricted pattern in a substantial number of patients constitute the most promising targets for T-cell analysis. Newer techniques involving efficient transfection of coding genes with viral or nonviral vectors, better methods for long-term propagation and stabilization of specifically reactive CD8 and CD4 T cells, and new approaches to identify and expand low frequency, specific T-cell populations should facilitate T-cell analysis of SEREX-defined antigens.

Cancer Immunome

The past decade has seen enormous strides in our understanding of the immune response to human cancer. In major part, this has been due to the development of methodologies capable of defining the antigenic targets on cancer cells that elicit an immune response (19, 67, 88). The cloning of T-cell recognized epitopes by Boon et al. (67) and by Kawakami and Rosenberg (88) has provided a growing list of tumor peptides that allows detailed monitoring of CD8 T-cell responses to these antigens in cancer patients and offers promising targets for cancer vaccine development. SEREX technology, because it is generally applicable to all tumor types and is less technically demanding than T-cell epitope cloning, holds promise for greatly extending the understanding of the immune response to cancer. In fact, identifying the complete repertoire of immunogenic gene products in human cancer - what is becoming known as the cancer immunome - is now an achievable goal for tumor immunology. Since the establishment of the SEREX database in 1997, later incorporated into the Cancer Immunome Database (150), 2593 sequences derived from 2169 clones have been deposited (February 2004), most of them contributed by the LICR SEREX Collaborative Group. Many of the genes have been isolated repeatedly by SEREX, from the same and/or from different tumor types, indicating that these gene products are highly immunogenic in the human host. On the other hand, even in a very recent study of lung cancer (144), only about one third of the isolated genes were already in the database, suggesting that the pool of immunogenic cancer antigens, although apparently finite in size, is still far from completely defined.