EBAG9 was originally described as a novel tumor associated antigen with a functional role in tumor-immune interactions [3
]. Cell culture supernatant from SiSo cells inhibited the proliferation of activated T cells and induced apoptotic cell death in receptor bearing cells. Furthermore, recombinantly expressed EBAG9 was suggested to bind to a yet unidentified receptor on activated immune cells and on K562 cells. From these observations and from the detection of apoptotic tumor-infiltrating lymphocytes surrounding 22-1-1 mAb stained tumor lesions [3
] it was inferred that EBAG9 is a new death receptor ligand involved in tumor immune escape [23
]. In the present study, we have extended our previous cell biological characterization of EBAG9, resulting in conclusions that are not consistent with the currently held view on the functional and clinical role of EBAG9.
We first addressed the question whether EBAG9 is a tumor-specific marker. Recently, we and others have pointed out that EBAG9 is highly conserved in phylogeny, and the gene-product was found to be expressed in all murine tissues examined [24
]. Applying immunohistochemistry on a representative selection of tumor and normal human tissue specimen, reactivity with the EBAG9-specific monoclonal antibody Ab-1, generated against a recombinant EBAG9 fusion protein, was seen in essentially all tissues and cell types examined. Staining in benign or malignant cell types was confined to a cytoplasmic pattern. In tumor infiltrating plasma cells a Golgi-like distribution could be clearly seen. Secondly, referring to our previous report [10
] we asked whether the cognate 22-1-1 antigen and the EBAG9 antigen are distinguishable in situ. In a side-by-side comparison of both antibodies in immunohistochemistry, we confirmed significant differences between 22-1-1 and Ab-1 antibody reactivities. Most striking examples for differences in immunostaining included normal glandular tissues, among them gastric epithelia, colon and normal prostate tissue. Of note, 22-1-1 constantly stained secreted matter from glandular tissues, whereas Ab-1 was always negative for mucus. In agreement with previous reports [5
], 22-1-1 staining is not limited to malignant tissue, but is also seen in normal gastric, colon and prostate epithelia. In case of adenocarcinomas, strong staining for the 22-1-1 mAb was observed in signet ring cell carcinomas with an intense labeling of the intracellular mucin. In sharp contrast those neoplastic cells remained almost negative with Ab-1. In most cases, the expression of the 22-1-1 antigen was enhanced in adenocarcinomas as compared to their normal counterpart. This was not the case for the antigen recognized by Ab-1. Other marked differences were seen with mucous secretions and with the subcellular distribution of the epitope recognized by Ab-1. This conclusion is substantiated by only partial correlation between endogenous EBAG9 protein levels and the corresponding expression profiles for the 22-1-1 antigen in human tumor cell lines, as determined by immunoblotting or flow cytometry, respectively.
The staining pattern obtained for 22-1-1 was almost identical to that described for the tumor-associated O-linked glycan Tn [25
], thus confirming our epitope identification [10
]. In addition, Tn antigen is found as soluble antigen in serum of tumor patients, an observation that is also shared with the occurrence of 22-1-1 soluble antigen [4
We also refined the subcellular localization of the EBAG9-encoded antigen. EBAG9-GFP localized predominantly to the Golgi cisternae and to small vesicles surrounding the Golgi apparatus, as evidenced by immuno-electronmicroscopy. A more functional analysis revealed sensitivity to BFA and a corresponding colocalization with the cis/medial Golgi marker, mannosidase II. These findings identify EBAG9 as a predominantly Golgi localized protein which is unlikely to be secreted. Our results shed doubt on the hypothesis that soluble 22-1-1 reactive material induces apoptotic cell death in activated immune cells or other receptor bearing cells. This earlier conclusion rests on two different experimental approaches, either the incubation of activated T cells or K562 cells with culture supernatant, as obtained from SiSo or MCF-7 cells, or the exposure to recombinantly expressed EBAG9 protein [3
]. Functional readout for both systems was the detection of apoptotic cell death. However, none of the apoptotic pathways have been elucidated yet. We have previously pointed out that full-length recombinant EBAG-GST is not soluble in aqueous solutions and requires the presence of detergent. Therefore, it was reasonable to suggest that cell viability in assays using recombinant EBAG9 was most likely affected by residual detergent [10
]. In our hands, cell culture supernatant from 22-1-1 positive MCF-7 cells failed to induce apoptosis in K562 cells. Likewise, the O-linked glycan recognized by 22-1-1, Tn (αGalNAc), did not induce apoptosis in K562 cells. Presently, we cannot reconcile our data with those published earlier.
In conclusion, our data strongly suggest that the antigens recognized by 22-1-1 and the EBAG9 antibody, Ab-1, are different. It follows that studies on the correlation of EBAG9 expression and clinical prognosis were correct as long as their screens were based on RT-PCR or immunoblotting with a polyclonal anti EBAG9 antibody [8
]. In contrast, functional and clinical studies on the 22-1-1 defined antigen (Tn) need to be revisited, and should be compared to other studies obtained with anti-Tn antibodies. At present, a direct link between the occurrence of Tn and expression levels of EBAG9 is still elusive, since we observed a correlation in some cell lines, but not in others. Provided that the physiologically occurring molecule EBAG9 is indeed dysregulated or mutated in tumors, this should prompt further investigations on the role of EBAG9 in the modulation of O-linked glycan expression.