The late polyadenylation signal of simian virus 40 functions with greater efficiency than the early polyadenylation signal, in turn affecting steady-state mRNA levels. Two chloramphenicol acetyltransferase (CAT) transient expression vectors, pL-EPA and pL-LPA, that differ only in their polyadenylation signals were constructed by using the early and late polyadenylation signals, respectively. In transfections of Cos, CV-1P, or HeLa cells and subsequent Northern blot analysis of CAT-specific RNA, approximately five times more steady-state CAT mRNA was produced in transfections with pL-LPA than with pL-EPA. The basis for this difference was not related to the specific promoter used or to RNA stability. Overall, the difference in steady-state mRNA levels derived from the two plasmids appeared to be attributable to intrinsic properties of the two polyadenylation signals, resulting in distinctly different cleavage and polyadenylation efficiencies. Additionally, we found that the utilization of the late polyadenylation site was dramatically reduced by deletion of sequences between 48 and 29 nucleotides 5' of the AAUAAA hexanucleotide. This reduction of mRNA levels was shown not to be caused by altered stability of mutant precursor RNAs or mRNAs, suggesting that these upstream sequences constitute an element of the late polyadenylation signal and may cause, at least to some extent, the greater efficiency of utilization of the late polyadenylation site.