About half of Caenorhabditis elegans genes have a 1-2 bp mismatch to the canonical AAUAAA hexamer that signals 3' end formation. One rare variant, AGUAAA, is found at the 3' end of the mai-1 gene, the first gene in an operon also containing gpd-2 and gpd-3. When we expressed this operon under heat shock control, 3' end formation dependent on the AGUAAA was very inefficient, but could be rescued by a single bp change to create a perfect AAUAAA. When AGUAAA was present, most 3' ends formed at a different site, 100 bp farther downstream, right at the gpd-2 trans-splice site. Surprisingly, 3' end formation at this site did not require any observable match to the AAUAAA consensus. It is possible that 3' end formation at this site occurs by a novel mechanism--trans-splicing-dependent cleavage--as deletion of the trans-splice site prevented 3' end formation here. Changing the AGUAAA to AAUAAA also influenced the trans-splicing process: with AGUAAA, most of the gpd-2 product was trans-spliced to SL1, rather than SL2, which is normally used at downstream operon trans-splice sites. However, with AAUAAA, SL2 trans-splicing of gpd-2 was increased. Our results imply that (1) the AAUAAA consensus controls 3' end formation frequency in C. elegans; (2) the AAUAAA is important in determining SL2 trans-splicing events more than 100 bp downstream; and (3) in some circumstances, 3' end formation may occur by a trans-splicing-dependent mechanism.