It is well established that MHC class I molecules present peptides from endogenous proteins, such as virus or tumour antigens, to CD8+ T lymphocytes. This implies that expression of MHC class I molecules on tumours is also mandatory for an effective T cell response against neoplasias. Indeed, numerous murine models exist (such as IC9, 3LL, P815) in which the existence of MHC presented tumour antigens and a protective CTL response have been well documented (reviewed in Hämmerling et al, 1987). However, the key question of whether cytolytic T cell attack has a role in human cancer remains unsolved. Similarly, the role of NK cells is unclear; these seem to lyse cells with low levels of MHC class I expression more efficiently. So far, human tumour specific antigens that can be presented by HLA molecules have not been identified on the molecular level. For a subpopulation of patients with malignant melanoma, the existence of tumour antigen can be deduced from the existence of tumour specific cytotoxic T cells isolated from TIL. However, for other epidemiologically more important tumours such as carcinomas of the colon, breast and lung, even indirect evidence is still missing. It is thus unknown how many tumours express tumour specific antigen at all and whether these putative antigens are of unique specificity or are shared by certain groups of (histologically related) neoplasias. Most reports agree that malignant cells in general have a more or less pronounced propensity to express class I molecules abnormally and often in the sense of hypoexpression or (sublocus selective) loss. Mechanisms inducing aberrant expression are numerous. Immunohistochemical studies revealed that an abnormal content in surface class I/beta 2m may be associated with other aspects of dedifferentiation of the tumour and hence may eventually correlate with biological signs of an increased grade of malignancy. Consequences of defective MHC class I expression for the survival of a malignant clone may theoretically consist of an escape from cytotoxic T cell attack or in an increased susceptibility for NK mediated lysis. In view of the fact that a particular antigen in a cell will be presented only by one or few HLA alleles on the cell, one would not expect that immunoselection would lead to a loss of all HLA alleles. It can be deduced from the few existing clinical studies on the prognostic impact of aberrant MHC class I expression that immunoselection by one way or the other is not a relevant mechanism in terms of tumour biology.(ABSTRACT TRUNCATED AT 400 WORDS)