Each transgene produces properly processed albeit abnormally unstable mRNA as well as several smaller RNAs in erythroid cells. These smaller RNAs are detected only in the cytoplasm and, relative to mRNA, are longer-lived and are missing sequences from either exon I or exons I and H. In this communication, we show by using genetics and Si nuclease transcript mapping that the premature termination of 1-globin mRNA translation is mechanistically required for the abnormal RNA metabolism. We also provide evidence that generation of the smaller RNAs is a cytoplasmic process: the 5' ends of intron 1-containing pre-mRNAs were normal, the rates of removal of introns 1 and 2 were normal, and studies inhibiting RNA synthesis with actinomycin D demonstrated a precursor-product relationship between full-length mRNA and the smaller RNAs. In vivo, about 50% of the full-length species that undergo decay are degraded to the smaller RNAs and the rest are degraded to undetectable products. Exposure of erythroid cells that expressed a normal human 13-globin transgene to either cycloheximide or puromycin did not result in the generation of the smaller RNAs.Therefore, a drug-induced reduction in cellular protein synthesis does not reproduce this aspect of cytoplasmic mRNA metabolism. These data suggest that the premature termination of 13-globin mRNA translation in either exon I or exon II results in the cytoplasmic generation of discrete mRNA degradation products that are missing sequences from exon I or exons I and II. Since these degradation products appear to be the same for all nonsense codons tested, there is no correlation between the position of translation termination and the sites of nucleolytic cleavage. mRNA exists to be translated into protein. However, evidence suggests that in the process of being translated, the abundance of an mRNA can be altered. For some mRNAs, abundance is increased if the coding region is not translated in full. Examples of this type include mRNAs for the replication-dependent human histones (10,18,26); human c-fos (19,41,42); murine c-myc (50); yeast MATel (35); and hamster, chicken, murine, and sea urchin P-tubulin (16,17,34,51,52). In each case, translation to at least a certain point within the coding region of the mRNA appears to be required for mRNA degradation, and nonsense codons that reside upstream of that point are associated with an increase in mRNA abundance. While similar in this regard, however, each of these mRNAs is not degraded as a function of the same features of translation. Histone mRNA degradation at the end of S phase requires that ribosomes translate to within a specific distance of a hairpin structure that resides in the 3' untranslated region. In contrast, c-fos, c-myc, and MATod mRNA degradation seems to require ribosome translation across a specific region of the coding sequence. And, as another variation, ,B-tubulin mRNA decay depends upon an excess of free tubulin that presumably interacts with the first 4 amino acids of an actively elongating, nascent peptide. (2,...