Two nearly identical, gstD21(L) and gstD21(S) mRNAs, whose polyadenylation sites differ by 19 nucleotides, are transcribed from the intronless glutathione S-transferase D21 gene in Drosophila. Both mRNAs are intrinsically very labile, but exposure to pentobarbital renders them stabilized beyond what can be attributed to transcriptional activation. We have reconstituted this PB-mediated mRNA stabilization in a transgene (D21L) that contains the full-length gstD21(L) sequence. We have also constructed a similar transgene (D21L-UTR), which matches D21L but excluded the native 3'-UTR. D21L-UTR produces a relatively stable RNA, whose stability is unaffected by pentobarbital. Following pentobarbital treatment of wild-type flies, the levels of gstD21(L) and gstD21(S) mRNAs hold at a relatively constant ratio (L/S) of 1.4 +/- 0.2. In transgenic flies, heat shock induction of D21L mRNA changed the L/S ratio to 0.6 +/- 0.1, and it was further reduced to 0.3 +/- 0.1 as D21L mRNA accumulated in the presence of PB. The ratio returned nearly normal (1.1 +/- 0.1) as the D21L mRNA decayed over 12 h after terminating induction. In constrast, when D21L-UTR was present, the ratio remained constant (1.7 +/- 0.2) even under various induction conditions and during recovery. Thus, the 3'-UTR, which was the critical difference between these two transgenes, must have some role in determining the L/S ratio. Induced D21L mRNA alone is not sufficient to cause reversible changes in the ratio. Such changes require the presence of pentobarbital. Therefore, pentobarbital may regulate this L/S ratio by affecting the choice of polyadenylation sites for the gstD21 mRNAs through sensing the concentrations of the native 3'-UTR sequences.