Given the large number of gene products identified by SEREX, it would be useful to develop a classification system that could organize them into meaningful categories. The challenge is to decide on which characteristic or characteristics of the antigen should be chosen to form the basis for categorization. As many SEREX-defined genes are only partially defined structurally, and even fewer are defined functionally, it is clearly not plausible to attempt a comprehensive classification at this time. However, antigens with related characteristics can readily be identified at this point. For instance, antigens can be grouped according to their cellular location, and it is particularly striking that so many of the SEREX-defined antigens are nuclear proteins (e.g., enzymes and factors involved in DNA replication, transcriptional control, RNA elongation, DNA repair, including zinc finger proteins, RNA helicases, proteins related to the mitotic apparatus, and chromosome condensation proteins). Although it is tempting to conclude that the immune responses to these nuclear antigens are nonspecific and are related to the high cell turnover and necrosis associated with cancer, it would be premature to exclude the possibility that immunogenicity is a consequence of specific structural changes in these proteins or their expression patterns.

Another way to group the antigens would be by functional characteristics, and a broad array of functionally related proteins can be found among SEREX-defined antigens, including metabolic enzymes (e.g., lactic dehydrogenase, aldolases, pyridoxal kinase, adenylosuccinate lyase, glyceraldehyde-3-phosphate dehydrogenase), transcriptional and translational factors (e.g., zinc finger proteins, translation initiation factors), structural proteins (e.g., histone proteins, keratin, restin, and lamin), and stress proteins (e.g., heat shock proteins). In the context of tumor biology and tumor immunity, however, antigens of greatest interest would be those having a known relation to cancer or showing cancer-restricted expression or cancer-restricted immunogenicity. These tumor antigens would be more important for diagnostic or therapeutic applications (e.g., vaccine use). A number of SEREX antigens with these cancer-related characteristics have been identified and can be classified into one of the following six categories (Table 2).

Cancer-Testis Antigens

These antigens share the following characteristics (131):

1. predominant expression in gametogenic tissues and cancer,
2. coding genes frequently map to chromosome X and exist as multigene families,
3. immunogenic in cancer patients,
4. heterogeneous protein expression in cancer, and
5. in vitro activation by hypomethylation and/or histone deacetylase inhibition.

The count of CT antigens varies depending on the stringency of the defining criteria used, and up to 44 CT gene/gene families have been reported in the literature (132). However, some of these 44 genes showed substantial mRNA expression in several somatic tissues by RT-PCR, and probably should not be classified as CT genes in a strict sense. The number of immunotherapeutically relevant CT antigens is thus clearly lower. Table 3 summarizes the CT antigens that have been shown to elicit immune responses in cancer patients, and this group of gene products will likely be the focus of cancer vaccine trials in the foreseeable future. Of these, MAGE, BAGE, and GAGE were initially identified as cytotoxic T lymphocyte (CTL)-recognized antigens (47, 67), and SSX, NY-ESO-1, SCP1, CT7, CT8, CAGE1, cTAGE1, and XAGE1 were discovered by SEREX analysis. MAGE-C1, a gene identical to CT7, was independently cloned by representational difference analysis (RDA) using testicular cDNA after subtraction hybridization with other normal tissue cDNAs (68). Similar RDA approaches have also led to the identification of LAGE-1 (41), a second gene in the NY-ESO-1 family, and CT10 (27). Both LAGE-1 and CT10 have subsequently been isolated by SEREX using sera from cancer patients, confirming that they are immunogenic. OY-TES-1 was identified as the ortholog of a mouse CT antigen, and an antibody response to this gene was subsequently demonstrated in cancer patients (25/362) (133).

Of the MAGE family, MAGE-1, first defined by CTL epitope cloning (47, 67), was also isolated by SEREX from a case of melanoma (19). MAGE-4a, another member of the MAGE family, although not shown to elicit a CTL response yet, has been isolated from melanoma, ovarian cancer, and head/neck cancer by SEREX (25, 119). Other members of the MAGE family (e.g., MAGE-3 and MAGE-6), shown to be CTL targets (47, 67, 70), have also been identified by SEREX in breast cancer (36).

BAGE, another melanoma antigen recognized by CTL (67), was also identified by SEREX screening of a testis library with serum from glioma patients [clone HOM-TS-GLI-38 in the Cancer Immunome Database (150)].

GAGE genes have not been found by SEREX. However, a related gene family, XAGE, has been SEREX-defined (97, 104).

SSX2, also known as HOM-MEL-40, was isolated from melanoma by SEREX (19). SSX2 was originally recognized as a gene mapping to chromosome X that was involved in the t(X;18) translocation invariably associated with synovial sarcoma (71). Additional members of the SSX family have been cloned (22, 71, 134), and 6 of 9 (SSX1, 2, 3, 4, 5, and 7) members of the SSX family have been shown to be transcribed in testis. Of these, only SSX1, 2, and 4 show significant expression in cancer. Expression pattern analysis revealed expression of SSX1, 2, and 4 in a proportion of malignancies of various origins, whereas SSX5 is only rarely expressed (approx. 1% of the tumors examined), and SSX3 expression has only been shown in sarcomas (135). This pattern of discordant expression of the SSX genes and other CT genes mapped to chromosome X suggests the existence of gene-specific mechanisms for the activation of CT antigens, in addition to other general mechanisms of gene activation, such as global demethylation (72).

NY-ESO-1 was isolated from an esophageal squamous cell carcinoma by SEREX and has been shown to be expressed in 20% to 40% of several common tumor types, including breast cancer, lung cancer, prostate cancer, bladder cancer, head and neck cancer, and melanoma. A highly homologous gene (84% amino acid identity), LAGE1, has been found by RDA by Lethè et al. (41). A third gene member, ESO3, has been identified 300 kb centromeric to the NY-ESO-1 gene on chromosome Xq28. This gene, however, is ubiquitously expressed in all tissues and is not a CT antigen (136). Both NY-ESO-1 and LAGE-1 appear to be expressed in tumors at similar frequencies, and both are recognized by patients' sera (25). NY-ESO-1 has been identified by SEREX of esophageal cancer (28), melanoma (25), breast cancer, (35, 36) prostate cancer (104), and ovarian cancer (109), whereas LAGE-1 has been identified in melanoma (25), prostate cancer (104) and sarcoma (110).

SCP1 is a synaptonemal complex protein involved in chromosome reduction in meiosis (73). Originally detected in a subtractive testicular library with serum from a patient with renal cancer (23), SCP1 was also isolated by SEREX of melanoma (91), breast cancer (100) and cutaneous T cell lymphoma (124). SCP1 has the distinction of being the first CT antigen with a defined function. The finding of a meiotic protein aberrantly expressed in a somatic neoplastic cell raises the provocative question of its role in the chromosomal aneuploidy of cancer.

CT7 was isolated by allogeneic screening of the SK-MEL-37 melanoma cell line with serum from a melanoma patient (25). This gene encodes a protein of 1,142 amino acid residues, with a carboxyl terminus highly homologous to the MAGE-10 gene over a stretch of approx. 210 amino acid (57% identity, 75% homology, including conserved substitutions). Sequences N-terminal to this segment, however, show no homology to the MAGE family, having instead a striking repetitive pattern, with a core of ten almost exact repeats of 35 amino acids. This gene has also been isolated by Lucas et al. (68) using the RDA approach and has been designated MAGE-C1. Although Lucas et al. indicated three genes in the MAGE-C subfamily, namely MAGE-C1, C2, and C3, only CT7/MAGE-C1 contains the N-terminal repeats, and is thus immunologically and likely biologically distinctive.

CT8/HOM-TES-85 was isolated by Türeci et al. (24) by SEREX analysis of a subtracted testicular library with serum from a seminoma patient. It encodes a 36-kDa protein with a leucine zipper motif, characteristic of proteins involved in DNA binding and gene transcription.

CT10 was isolated from melanoma using RDA by Güre et al. (27). It is structurally closely related to CT7 but lacks the repetitive sequences of CT7. It maps to chromosome Xq27 in close proximity to CT7 and MAGE genes. ELISA analysis of 100 melanoma patient sera against recombinant CT10 protein revealed seroreactivity in two cases, demonstrating the immunogenicity of CT10. This finding was further confirmed by the discovery of CT10 as a SEREX-defined antigen in hepatocellular carcinoma, designated HCA587 by the authors of this study (106, 137).

CAGE, a DEAD box helicase protein, was identified by SEREX using a pool of sera from five gastric cancer patients (147). It is the product of an intronless gene on Xp22, sharing sequence homology to HAGE, another CT antigen (148).

CAGE-1 was discovered by the same group that identified CAGE, but these two genes are otherwise unrelated in structure. It was identified by screening a testis cDNA library against sera from lung cancer patients (96). It is a single-copy gene on chromosome 6, encoding a protein of 639 amino acids, 73 kDa in size.

cTAGE-1 was isolated by SEREX of T cell lymphoma (124). The predicted open reading frame (ORF) encodes a relatively short protein of 74 amino acids (GenBank Accession No. (AF177229) out of a 1.25 kb transcript. Whether this is indeed the gene product recognized by antibodies in cancer patients, however, has not been determined with certainty. cTAGE-1 is located on chromosome 18, and five members have been characterized of which two, cTAGE-1 and cTAGE-5, have been described as bearing a CT expression pattern (138).

XAGE-1 belongs to a family of at least three expressed genes (XAGE-1, -2, -3) (139, 140). Initially isolated as genes sharing homology in their carboxyl sequences to the GAGE and PAGE genes, XAGE genes were also located on Xp11 and were found to have similar cancer/testis expression patterns. SEREX analysis isolated XAGE-1 in non-small cell lung cancer (97) and prostate cancer (104), demonstrating its immunogenicity. Four transcripts were found in the lung cancer examined, with XAGE-1b being the dominant form.

ACRBP/OY-TES-1 was identified as the ortholog of a mouse CT antigen, and antibody response to this gene was subsequently demonstrated in cancer patients (133). OY-TES-1 is located on chromosome 12p13 and encodes the proacrosin binding protein sp32 precursor. It is thus one of the few CT gene products with a known function in germ cells.

Mutational Antigens

Several mutational antigens have been isolated by SEREX. A classic example is the tumor suppressor gene p53, which has been identified by SEREX of colon cancer (29), breast cancer (36), and ovarian cancer (109). In the case of colon cancer, a single base substitution (A to G) was identified, confirming this mutation as the basis for the observed immunogenicity. Other examples of mutational antigens in colon cancer were AD034, with a 32 bp frameshift mutation (92), and CDX2, with a single-base frameshift mutation (93). The CDX2 mutation is in a microsatellite sequence within the coding region, and is believed to result from microsatellite instability in this patient. A different type of mutation, namely translocation, was found in the E-cadherin gene, detected in the SEREX analysis of gastric cancer (34). These findings illustrate the capacity of SEREX to identify products of mutated genes. As discussed above, three genes coding for products identified by SEREX are clustered in chromosome 3p21, a region long known to be a hot spot of genetic aberrations in many cancer types and postulated to harbor tumor suppressor genes (53, 54). Two of the three genes, NY-REN-9 and NY-REN-10, were derived from renal carcinoma and correspond to LUCA-15 and gene 21, respectively (33). The third gene, NY-LU-12, isolated from lung cancer, was identical to gene 16, which maps to the telomeric breakpoint of a small cell lung cancer line, NCI-H740 (32). Although no mutation has been detected to date in these three genes, mutations may have been missed: wild-type rather than the mutated allele of the 3p genes might have been isolated in SEREX because the antibody elicited by the mutated 3p product cross-reacts with the wild-type product.

Differentiation Antigens

The classic example of a differentiation antigen recognized by SEREX is the melanocyte-specific protein tyrosinase (19). Other examples include NY-BR-1 in breast cancer (99), rab38 in melanoma (89), and NY-CO-27, a gene identical to galectin-4, in colon cancer (29). Normal tissue expression of galectin-4 is restricted to normal colon and small intestine. Because galectin-4 is localized to the leading edge of lamellipodia, it is thought to have a role in cell adhesion (44).

SOX, ZIC2, and other neuronal antigens were isolated from small cell lung cancer and could be classified as embryonal neuronal antigens (98). Four SOX group B genes (SOX1, SOX2, SOX3, and SOX21) were identified. SOX group B and ZIC2 genes encode DNA-binding proteins; these genes are expressed in early developmental stages in the embryonal nervous system and are downregulated in the adult. SOX2 mRNA can also be detected in some adult tissues, whereas ZIC2 is expressed in adult brain and testis.

CT antigens, because of their restricted expression in normal testis, are another special category of differentiation antigens. SCP1 and OY-TES-1, for instance, have specific functions in germ cells and represent true germ cell differentiation antigens. It is worthy to note, however, that many testis-specific transcripts and proteins are under such tight regulatory controls that they are almost never expressed in cancers other than germ cell tumors (69). This is a fully expected phenomenon, analogous to the exclusive expression of melanocytic markers in melanoma, CD20 in B cells, etc. The frequent expression of CT antigens in various types of tumors is an exception to this general rule. It suggests that the CT antigens, most of them with unknown function at present, are a distinct group of proteins in terms of their regulation and possibly their biological function.

Aside from the examples above, many other SEREX-defined genes, when analyzed for mRNA expression by RT-PCR, showed tissue-restricted expression with predominant expression in one or a few tissues. However, these genes often show low-level expression in other tissues, and are not strictly tissue-specific, particularly from the immunotherapeutic perspective. Some of these antigens are either amplified/overexpressed in cancer or show aberrant splice variants in cancer, and are more appropriately classified and discussed under these categories of tumor antigens.

Amplified or Overexpressed Antigens

Overexpression of normal gene products in cancer may be a major underlying mechanism for the immunogenicity of cancer antigens in cancer patients. Many SEREX-defined genes have been described as overexpressed in cancer based on various assays for mRNA quantitation, including Northern blot analysis, conventional RT-PCR, and real-time RT-PCR. Examples of amplified or overexpressed SEREX-defined antigens identified in earlier studies include carbonic anhydrase XII in renal cancer (52), galectin-9/HOM-HD-21 in Hodgkin's disease (21), eIF-4 gamma (30) in lung cancer, aldolase A in lung cancer (32) and breast cancer (35), KOC family genes in melanoma (25) and hepatoma (43), and AKT1 (37) and HER-2/neu (36) in breast cancer. Several mechanisms can account for amplified expression of gene products in cancer, including gene amplification (e.g., eIF-4 gamma), increased steady-state mRNA (e.g., KOC3), and increased protein stability (e.g., p53) (75). In our study of neuroblastoma, 4 genes identified were found to cluster on chromosome 17q21-23. It is known that unbalanced translocations often occur in neuroblastoma in this region, resulting in 17q21 gain in up to 83% of patients. These four gene products, among them topoisomerase II alpha (TOP2A), may thus be SEREX antigens due to gene amplification events. The frequency of gene amplification in SEREX-defined genes has been examined by Brass et al. (31), and 9 of 14 genes detected in a SEREX analysis of lung cancer were shown to be amplified by quantitative PCR, including three genes from chromosome 3, at least two of which were from a region known to be amplified in squamous cell carcinoma. This high frequency of gene amplification has not been reported in other SEREX studies, and most of the antigen overexpression events in cancer are likely due to epigenetic phenomena.

Splice Variant Antigens

Another category of SEREX-defined antigens is splice variants of genes that are differentially expressed in normal tissues. Since alternative splicing is a common event in many genes, it is not surprising that more than one transcript variants have been documented for many of the SEREX-defined genes. Many of the CT antigens, for examples, have different alternatively spliced variants, including LAGE-1, SSX, XAGE-1, etc. In the initial SEREX study by Sahin et al. (19), a splice variant of the intermediate filament protein restin was isolated from Hodgkin's lymphoma and was found to react with sera from both cancer patients and normal donors. In the SEREX analysis of colon cancer (29), two of the isolated antigens NY-CO-37 and NY-CO-38 represent differentially expressed isoforms of a previously unknown gene containing PDZ protein-protein interaction domains (45). In total, five splice variants of NY-CO-38 have been defined [PDZ-37, PDZ-45 (NY-CO-37), PDZ-54, PDZ-59, and PDZ-73 (NY-CO-38)]. One of these variants, PDZ-54, is normally expressed in normal kidney and brain but not in normal colon. However, PDZ-54 is expressed in all cases of colon cancer tested.

Viral Antigens

Human endogenous retrovirus (HERV)-related sequences have been shown to represent at least 1% of the human genome, and their gene products have been linked to the development of autoimmune and neoplastic diseases, including systemic lupus erythematosus (76), rheumatoid arthritis (77), and germ cell tumors (78). Of these sequences, HERV-K exists in 25 to 30 copies per haploid human genome, and HERV-K-encoded env and gag proteins are expressed consistently in germ cell tumors, leading to high-titer antibodies in 60% to 85% of patients with these tumors (78, 79). Although efforts to define the expression of HERV sequences in normal and neoplastic human tissues are only starting (151), endogenous HERV encoded env and gag proteins have been isolated by SEREX in renal cancer (19) and in prostate cancer (39). Serological screening showed antibodies against the HERV-K gag protein in a significant number of prostate cancer patients, suggesting that retroviral antigens may potentially be cancer vaccine targets.