We have defined a DNA sequence that behaves as an RNA polymerase II termination signal by using the human HeLa cell transient expression system. Surprisingly, this sequence is tripartite, including part of the coding region of the sea urchin H2A histone gene together with two separate sequences in the 3' flanking region of the gene. We demonstrate that this signal functions both in its normal gene environment and also when placed within the human a-globin gene. However, we have failed to detect a discrete 3' terminus. Rather, our data indicate the, presence of an extremely heterogeneous series of nonpolyadenylated RNAs. These heterogeneous nonpolyadenylated RNAs are stable when transcribed from the intact histone gene but are highly unstable within the human a-globin gene. This provides evidence for the role of poly(A) in the stability of mnRNA.Initiation of transcription in eucaryotic RNA polymerase II (polII) genes is a highly regulated and in some instances well-defined process (see reference 28 for a review). In contrast, termination of transcription in polII genes is ill defined and may not be a regulated process. The principal cause of this uncertainty is the rapid posttranscriptional processing of primary RNA transcripts to mature mRNA (see reference 40 for a review). The fact that introns are spliced out of the polII gene transcript is well documented. Similarly, the generation of mRNA 3' termini for both polyadenylated [poly(A)+] mRNA and nonpolyadenylated [poly(A)-] histone mRNA by a processing event has been recently established (see references 5 and 44 for reviews).The best evidence for processing of mRNA 3' ends comes from studies on histone mRNA. Thus, the existence of a histone mRNA 3' processing activity in Xenopus laevis oocytes and Drosophila melanogaster tissue culture cell extracts has been clearly demonstrated (4,25,43). Furthermore, it has been shown that this activity is associated with a small nuclear RNA U7 (12, 52) in a similar way to the involvement of the small nuclear RNA Ul in intronic splicing (24,27,42,47). These results indicate that transcriptional termination is likely to occur separately from 3'-end processing in histone genes. Evidence of a similar nature in poly(A)+ mRNA genes is now becoming available. Moore and Sharp (37) have recently shown that synthetic RNAs extending beyond an adenovirus poly(A) site can be efficiently and accurately cleaved and polyadenylated by using in vitro cell extracts.Further evidence for transcriptional termination separate from nmRNA 3'-end formation comes from hybridization analysis of pulse-labeled RNA transcripts. Using these techniques, Nevins and Darnell (38) and Fraser et al. (11) deduced that the major late gene transcripts of adenovirus extend beyond the various late mRNA poly(A) sites to near the end of the viral genome. Similarly, transcription extends beyond the adenovirus early region 2 and 4 poly(A) sites and the simian virus 40 (SV40) late poly(A) site (10, 39). These viral transcription experiments were performed on pulse-* Corr...