Inactivation of the retinoblastoma susceptibility (RB) gene plays a role in the pathogenesis of a variety of human malignancies. Recently, it has become feasible to study RB expression in archival tissues, and it is expected that immunohistochemical studies on routinely processed tumors will further elucidate the biologic and clinical significance of RB mutations. Our study was designed to address two issues that are critical for the interpretation of such studies, i.e., whether mutant RB protein (pRB) can reliably be distinguished from normal pRB and whether there are significant differences in the performance characteristics of various anti-RB antibodies. We studied cell blocks of 26 mutant RB cell lines (11 lines without any RB expression, nine lines expressing truncated mRNA/pRB, six lines carrying missense mutations) with five different anti-RB monoclonal antibodies, using a recently described procedure that includes an antigen retrieval step. The specific staining pattern for pRB was nuclear. Cytoplasmic staining was found to be nonspecific and could be strong. Some truncated and all full-length mutant pRBs localized to the nucleus, creating positive nuclear staining that might be indistinguishable from the staining pattern of cells carrying wild-type RB. The five antibodies tested showed significant differences in sensitivity, specificity, and background reactivity. Our data suggest that a significant subset of mutant pRB has preserved nuclear translocation capacity, that not all anti-RB antibodies are equally suitable for immunohistochemical analysis of RB expression, and that any such analysis is bound to include a certain, albeit probably small, number of positive stains, despite the absence of functional pRB.