For most of the mammalian mRNAs that have been shown to be reduced in abundance by a nonsense or a frameshift mutation that generates a nonsense codon, reduction takes place while the mRNA is nucleus-associated rather than after the mRNA has been exported to the cytoplasm (reviewed in Maquat LE, 1995, RNA 1:453-465). A variety of mechanisms have been put forth to explain how a nonsense codon could affect the abundance of nuclear mRNA. Some mechanisms have implicated nonsense codon recognition in the nucleus prior to splicing. Among the best-studied nonsense transcripts that manifest nonsense-mediated alterations in nucleus-associated metabolism are those that derive from human alleles for the glycolytic enzyme triosephosphate isomerase (TPI). Nonsense codons within TPI transcripts have been shown to reduce the half-life of completely spliced TPI (mRNA that co-purifies with nuclei (Belgrader P et al., 1994, Mol Cell Biol 14:8219-8228). However, whether or not nonsense codon recognition within TPI transcripts takes place prior to or after splicing remained unresolved. To address this issue, codons that span two exons, i.e., are disrupted by an intron prior to pre-mRNA splicing, were converted to nonsense. If nonsense codon recognition were to precede splicing, then the disrupting intron would be expected to preclude nonsense codon recognition by preventing the physical juxtapositioning of the codon nucleotides. In the absence of nonsense codon recognition, there would be no nonsense-mediated reduction in TPI mRNA abundance. The results of northern (RNA) blot hybridization demonstrated that the two nonsense codons of this type that were studied reduced the level of total, nuclear and cytoplasmic TPI mRNA to an average of 12% of normal, consistent with each nonsense codon being competent to mediate nuclear mRNA degradation. The possibility that the nonsense codons reduced TPI mRNA abundance by altering TPI mRNA abundance or splicing was eliminated by using RT-PCR to demonstrate that the level of each intron within pre-mRNA was essentially unaffected and cDNA sequencing to demonstrate that splice site choice was unaltered. Furthermore, missense codons that harbored some of the nonsense codon changes were found to have little effect on mRNA abundance. These findings, plus the previous finding that a suppressor tRNA abrogates the decay of TPI mRNA brought about by a nonsense codon residing within a single exon (Belgrader P, Cheng J, Maquat LE, 1993, Proc Natl Acad Sci USA 90:482-486), argue strongly that nonsense codon recognition in the nonsense-mediated decay of TPI mRNA takes place after splicing.