Evidence for a linear search in bimolecular 3′ splice site AG selection

In most eukaryotic introns the 3' splice site is defined by a surprisingly short AG consensus found a variable distance downstream of the branch site. Exactly how the spliceosome determines which AG to use, however, is not well understood. Previously we showed that when the branch site and 3' splice site AG are supplied by separate RNA molecules, there is a strong preference for use of the 5'-most AG in the 3' splice site-containing RNA. Here we show that this apparent 5'-->3' directionality holds even when this RNA contains four tandem repeats of a 6-nt sequence containing AG. Exactly the same pattern of 3' splice site choice was observed when the same tandem repeats were incorporated into a full-length splicing substrate. When the 3' splice site AG is supplied by a separate RNA, that RNA must be linear with an unobstructed 5' end. Similarly, the branch-containing RNA must be truncated immediately 3' to the polypyrimidine tract. A model is presented that incorporates these observations and reconciles previously proposed mechanisms for 3' splice site selection.