Complex gel permeation assays for screening combinatorial libraries.

Gel permeation methods have been commonly used to screen combinatorial libraries synthesized on a solid support. We report here three screens of combinatorial libraries using gel permeation assays. These include a simple enzymatic assay to identify inhibitors of the influenza enzyme neuraminidase, and two more complex assays designed to screen for inhibitors of the interleukin-8 (IL-8)-IL-8 receptor and the urokinase-urokinase receptor interactions, respectively. The IL-8 ligand-receptor assay makes use of IL-8 receptor-expressing cells attached to a membrane, thus enabling washing steps as part of the assay. The urokinase ligand-receptor assay employs an enzyme-linked immunosorbent assay-type format, previously thought to be amenable only to well-based assays. The results of these three screens are reported here, including the discovery of a novel series of acyclic inhibitors of neuraminidase. The development of complex assays in a gel permeation format allows for the routine screening of combinatorially as well as noncombinatorially made compound collections against virtually any kind of target, and is being widely used in our high throughput screening operations.

Structural and functional characterization of potent antithrombotic oligonucleotides possessing both quadruplex and duplex motifs.

We report the results of a selection for single-stranded DNA oligonucleotide ligands to the serine protease thrombin using recently developed methods. This selection yielded a family of DNA sequences that conform to a consensus structure comprised of a unimolecular quadruplex motif and complementary flanking sequences capable of forming an additional Watson-Crick duplex motif. This novel quadruplex/duplex structure was not reported in a previous selection for DNA molecules which bind to thrombin [Bock et al. (1992) Nature 355, 564-566]. All quadruplex/duplex molecules tested bound to thrombin with higher affinity than quadruplex structures lacking the duplex structure. However, binding affinities did not always correlate with inhibitory potency since some molecules with high affinity were not potent inhibitors in vitro. 1H NMR spectroscopy studies demonstrated that the complementarity of bases in the duplex portion of a selected sequence allows it to form multimolecular structures. Constraining these molecules to the unimolecular quadruplex/duplex structure by bridging the 5' and 3' ends of the duplex motif with either triethylene glycol or disulfide bonds improved their thrombin inhibitory activity. All bridged quadruplex/duplex molecules were more potent inhibitors than molecules with only a quadruplex motif. Bridging the ends of these structures not only increased thrombin inhibition but also improved resistance to nucleases in serum more than 40-fold over the unbridged quadruplex. In addition, we have found that both the length and sequence of the duplex motif are important for inhibition.

In vitro evolution of intrinsically bent DNA.

DNA fragments which are intrinsically bent or curved migrate anomalously during electrophoresis through polyacrylamide gels. Starting with an initial population of approximately 10(12) unique DNA sequences, DNA which exhibited the kind of anomalous mobility associated with DNA bending was selected and enriched using a variation of the SELEX procedure. After seven rounds of selection and amplification, the vast majority of the remaining population of DNA fragments migrated as bent DNA. Cloning and sequencing of 30 individual sequences from this population has yielded information regarding the relationship between DNA sequence and bending. Some of the previous conclusions on DNA bending have been confirmed while others have been modified, by the results presented here. In addition, the dinucleotide base step CA/TG, which had not been thought to be a major factor in DNA bending, appears to be important.

E. coli promoter spacer regions contain nonrandom sequences which correlate to spacer length.

The -10 and -35 regions of E. coli promoter sequences are separated by a spacer region which has a consensus length of 17 base-pairs. This region is thought to contribute to promoter function by correctly positioning the two conserved regions. We have performed a statistical evaluation of 224 spacer sequences and found that spacers which deviate from the 17 base-pair consensus length have nonrandom sequences in their upstream ends. Spacer regions which are shorter than 17 base-pairs in length have a significantly higher than expected frequency of purine-purine and pyrimidine-pyrimidine homo-dinucleotides at the six upstream positions. Spacer regions which are longer than 17 base-pairs in length have a significantly higher than expected frequency of purine-pyrimidine and pyrimidine-purine hetero-dinucleotides at these positions. This suggests that the nature of the purine-pyrimidine sequence at the upstream end of spacer regions affect promoter function in a manner which is related to the spacer length. We examine the spacer sequences as a function of spacer length and discuss some possible explanations for the observed relationship between sequence and length.