Substrate recognition by the hairpin ribozyme has been proposed to involve two short intermolecular helices, termed helix ! and helix 2. We have used a combination of three methods (cleavage of mismatched substrates, in vitro selection, and site-specific mutational analysis) to systematically determine the substrate recognition rules for this RNA enzyme. Assays measuring substrate cleavage in trans under multiple turnover conditions were conducted using the wild-type ribozyme and substrates containing mismatches in all sites potentially recognized by the ribozyme. Molecules containing single-and multiple-base mismatches in helix 2 at sites distant from the cleavage site (g_4c, u_sa, g-4c : u_sa) were cleaved with reduced efficiency, whereas those with mismatches proximal to the cleavage site (c_2a, a_~c, c_2a : a_3c) were not cut. Analogous results were obtained for helix 1, where mismatches distal from the cleavage site (u+Ta , u+sa, u+9a, u+Ta : U+sa : u+9a) were used much more eiiicienfly than those proximal to the cleavage site (c+4a, u_sa, g+6 c, c+4a : u+sa : g+6c). In vitro selection experiments were carried out to identify active variants from populations of molecules in which either helix 1 or helix 2 was randomized. Results constitute an artificial phylogenetic data base that proves base-pairing of nucleotides at five positions within helix 1 and three positions within helix 2 and reveals a significant sequence bias at 3 bp (c+4" G6, c-2" Gll, and a_ 3 9 U12 ). This sequence bias was confirmed at two sites by measuring relative cleavage rates of all 16 possible dinucleotide combinations at base pairs c+4" G6 and c-2" Gll. The strong sequence preference suggests that tertiary structure (e.g., base triples) or alternative secondary structures at these sites may be important for ribozyme function. Together with results from previous work, we conclude that hairpin ribozymes can be engineered to cleave substrates containing the sequence 5'-