There is an open reading frame (ORF) between ATG2089 and TGA2656 in the early E3 transcription unit of subgroup B adenovirus 3 (Ad3) that could encode a protein of 20,500 MW (20.5K). This ORF also exists in Ad7, Ad11, and Ad35 (subgroup B). An antipeptide antiserum made in rabbits against the predicted Ad3/7 20.5K protein immunoprecipitated two diffuse bands with apparent molecular weights of about 22K and 36K from Ad3- or Ad7-infected cells. These bands were also detected in immunoblots. These bands were not seen in mock-infected cells or cells infected with Ad3 or Ad7 mutants that delete the gene for 20.5K. In vitro transcription and translation of the 20.5K gene yielded a protein of about 18K, suggesting that this may be the primary translation product and that the 22K and 36K forms of 20.5K arise by post-translation processing. Pulse/chase experiments suggest that the half-life of the 22K form is short, and that this form is further modified to the 36K species. In accord with these results and as judged by its predicted sequence, 20.5K appears to be a membrane glycoprotein with a potential N-terminal signal sequence, a second hydrophobic putative transmembrane domain, and two potential Asn-linked glycosylation sites. The 20.5K protein was synthesized throughout the course of the infection. Ad3 and Ad7 mutants lacking 20.5K grew as well as wild-type Ad3 and Ad7, indicating that, in common with subgroup C E3 proteins, the 20.5K protein is dispensable for virus replication in cultured cells. The 20.5K gene is totally absent from the E3 region of Ad2 and Ad5 (subgroup C). The gene is also absent or highly diverged in the E3 region of Ad12 (subgroup A) and Ad40 (subgroup F). Given that E3 genes may counteract host defenses, the 20.5K protein may contribute to the unique pathogenic properties of subgroup B human adenoviruses.