We have used exons 2 and 3 of the rat a-tropomyosin gene to analyze the basis of mutually exclusive exon selection. The basis of the strict mutually exclusive behavior of this exon pair is enforced by the proximity of the exon 3 branchpoint to the 5' splice site of exon 2. With the exception of smooth muscle cells, exon 3 rather than exon 2 is incorporated into mRNA in all cell types. We show here, using both in vivo and in vitro cell-free systems, that this alternative exon selection is a consequence of general principles that govern 3' splice site selection. In the absence of exon 3, exon 2 is utilized efficiently in all cells. Selection of exon 3 is therefore the default result of a competition between exons 2 and 3 for the flanking constitutive splice sites. The basis of this competition is the relative strength of the polypyrimidine tract/branchpoint elements of the two exons. The major determinant of this splice site strength is the pyrimidine content adjacent to the branchpoint, and this involves no other sequence specificity. The branchpoint elements play an important but secondary role. The functional strengths of the different polypyrimidine tract/branchpoint combinations, as determined in cis competition assays, showed a perfect correlation with their binding affinities to a spliceosome component that interacts with the pre-mRNA in an ATP-independent manner. Selection of exon 3 in most cell types therefore reflects the preferential interaction of these splice site elements with constitutive splicing factors early in spliceosome assembly. The aspects of splice site selection analyzed here are likely to be of general applicability to constitutive and alternative pre-mRNA splicing.[Key Words: Rat c~-tropomyosin gene; exon selection; branchpoint polypyrimidine tracts; splice site selection] Received December 10, 1990; revised version accepted February 1, 1991.One of the major unresolved issues conceming premRNA splicing is the problem of splice site selection. Most eukaryotic genes possess multiple intervening sequences, or introns, each bounded by a pair of functional 5' and 3' splice sites. The open reading frame of the mature mRNA is preserved by the accurate excision of these introns from the primary transcript. Potentially, splice sites could pair in many different combinations disruptive to message integrity. That this does not occur to any great extent is testimony to the high fidelity of the splicing process. The role of limited consensus sequences at the ends of introns in specifying splice sites has long been recognized [Mount 1982). These elements include the practically invariant GU and AG dinucleotides at the 5' and 3' splice sites, respectively, the 5' splice site consensus sequence that base-pairs with U1 snRNA [Zhuang and Weiner 1986), the branchpoint consensus sequence that base-pairs with U2 snRNA {Parker et al. 1987; Wu and Manley 1989;, and the polypyrimidine tract between the branchpoint and the 3' splice site AG. However, comparisons of primary sequences have yet to produce a set of ...