Control of complexity constraints on combinatorial screening for preferred oligonucleotide hybridization sites on...

We have explored the use of short (10-mer), fully sequence-randomized oligonucleotide libraries for affinity-based screening in solution for energetically preferred sites of hybridization of a model 47-nucleotide (nt) mutant Ha-ras mRNA stem-loop fragment. In characterizing the model, binding studies using either a gel mobility-shift assay or an RNase ONE footprinting assay indicated the presence of a greatly preferred hybridization site for individual antisense RNA oligonucleotides on the 5'-most side of the ras RNA 19-nt loop. However, initial attempts to affinity-titrate combinatorial uniform 2'-O-methyl-substituted oligonucleotide libraries for selective binding to this 5'-loop site using an RNase ONE footprinting assay that can discriminate between binding to different sites on ras RNA were unsuccessful. By reducing the complexity of the library to a mix of seven RNA oligonucleotides complementary to a range of sites on ras RNA and with no self-complements, footprinting evidence for binding was obtained but was characterized by ras RNA site-specific binding constants differing dramatically from binding constants for individual oligonucleotides. The library complexity was reduced further to three different cases of two RNA oligonucleotides, one of which for all cases was the highest affinity 5'-loop complement. Detailed kinetic and thermodynamic binding analyses revealed a good fit of the data to independent (5'-loop and ascending stem sites), competitive (overlapping 5'-loop sites), or mutually allosteric (5'-loop and 3'-loop sites) formalisms and an energetics description showed that ras 5'-loop site-specific binding could be achieved by affinity titration only for the independent case. Reconstruction of events with the full complexity library suggested that there was the emergence of multiple, linked binding interactions and implied that successful hybridization affinity screening would be achieved only if all possible bimolecular binding interactions of individual library oligonucleotides with target RNA could be made mutually independent. Accordingly, by holding the calculated concentration of unique oligonucleotide sequences of a full complexity DNA library well below the value for the dissociation constant for binding of individual complement to the 5'-loop site and then titrating the concentration of ras RNA through this value, hybridization specific to the 5'-side of the ras loop was demonstrated as assayed either by sequential gel mobility-shift resolution of bimolecular complexes and RNase ONE footprinting in situ in gel slices or by RNase H cleavage of complexes in solution. Because this strategy uses an unbiased oligonucleotide library it should combinatorially identify energetically preferred hybridization sites on folded RNA targets of any sequence and of undetermined structure. This should enable a focused in vitro optimization of antisense oligonucleotide length, sequence, and chemical composition for preferred site binding affinity and specificity which, in turn, may be expected to provide for enhanced biological potency and specificity (Lima et al., 1996). Finally, the complexity constraints encountered and the fundamental requirement to control them presented here also should be applicable to interactions with any biomolecule target of any chemical class of combinatorial library when screened in solution in pooled mixes.