Previous studies in this laboratory suggested that in adult liver, either the gene for the tumor-type poly(A) polymerase is poorly transcribed or the mRNA for this enzyme is largely not expressed. To test these possibilities, total RNA from rat liver and Morris hepatoma 3924A RNA were isolated by using a guanidine thiocyanate method; poly(A+) RNA and poly(A-) RNA were separated by oligo(dT)-cellulose chromatography and used for translation in a rabbit reticulocyte lysate system. After in vitro translation, the products were immunoprecipitated with either purified anti-tumor poly(A) polymerase antibodies or control immunoglobulins. When the polypeptides translated from poly(A+) or poly(A-) hepatoma RNA were precipitated with immune sera, a unique [35S]methionine-labeled 35-kilodalton (kDa) protein was observed. This band was not apparent when control serum was used for the immunoprecipitation. The radiolabeled 35-kDa polypeptide was not evident when the products were incubated with highly purified tumor nuclear poly(A) polymerase prior to immunoprecipitation. Prior incubation of the translation products with bovine serum albumin instead of poly(A) polymerase had no effect on the immunoprecipitation. This 35-kDa protein was not apparent when liver poly(A+) RNA was used to direct translation. These data demonstrate that (a) the tumor enzyme is not synthesized as a precursor, (b) tumor mRNA, but not normal liver mRNA, contains detectable sequences coding for tumor-type poly(A) polymerase, and (c) poly(A) polymerase mRNA also exists as a poly(A-) population.