Novel Methods for In Vitro Selections and Analysis of DNA-Encoded Libraries

Kyle Edward Denton, Purdue University


DNA-encoded libraries are becoming increasingly used to identify chemical probes and hit molecules of pharmaceutical interest. DNA-encoded libraries allow for vastly greater library sizes (i.e., number of unique molecules) over those possible with traditional high-throughput screening. This is achieved by combining the encoding and sensitive detection properties of DNA with the molecular diversity made possible by combinatorial synthesis. The selection of molecules from these libraries has traditionally relied on affinity-based approaches with immobilized target proteins. While this approach has had many successes, it is limited to soluble protein targets, assays ligand binding only, and provides no information regarding the binding site of a hit ligand. In this work, two novel selection methods were explored to specifically address these limitations. In addition, a critical evaluation of affinity-based selection assays was conducted. Statistical robustness metrics and parameters for optimal affinity selection assay design were established. ^ The first goal was to develop a crosslinking-based selection strategy that would allow for a DNA-encoded library to be assayed for binding to a target protein in solution, prior to immobilization. In this approach, reactive groups are displayed on the encoded library members via DNA hybridization and are added post-library preparation. Several different reactive groups were investigated including general electrophilic and photocrosslinking groups. Reactive groups were evaluated for crosslinking efficiency and specificity with three ligand receptor systems. In a model selection assay, crosslinking allowed for increased enrichment of both high and low affinity ligands in comparison to a traditional affinity selection, which was unable to detectably enrich a weak ligand. ^ Protein kinases are integral members in cell signaling pathways and continue to be high profile targets in drug discovery. Nearly all kinase inhibitors compete with ATP and face challenges in selectively targeting a kinase. Inhibiting the protein-substrate interaction is an attractive alternative but is challenging with traditional small molecules due to the large, flat surface area of a protein-protein interaction surface. To address this challenge, a substrate-based selection approach was developed for tyrosine protein kinases. Selection of Src kinase substrates from a 3-mer combinatorial DNA-encoded library revealed a small group of unnatural peptidomimetic substrates. These substrates present scaffolds for derivation for substrate-to-inhibitor conversion and optimization to protein substrate-competitive kinase probes. ^ CBX8 is a chromodomain-containing member of the polycomb repressive complex 1 (PRC1) and is responsible for targeting of the complex to trimethyllysine posttranslational modifications on histone peptides, resulting in transcriptional repression. The inhibition of the binding of CBX8 to target genetic loci is hypothesized to prevent transcriptional repression. Several cancers have been characterized as CBX8-dependent, such as glioblastoma multiforme and MLL-rearranged leukemias. Targeting CBX8 has been challenging due to its high homology with other PRC1 chromodomains, such as CBX7. Using three well characterized peptide ligands of varying affinities for CBX7 and CBX8, affinity selection assay parameters were optimized, and a mock assay was designed to evaluate robustness statistically using DNA-sequencing data. Using the Z? statistic in an analogous fashion to the evaluation of high-throughput screening assays, optimal assays gave Z? factors of 0.72 and 0.67 for CBX8 and CBX7, respectively. To develop a selective chemical probe for CBX8, a DNA-encoded positional scanning library was prepared based on an optimal ligand for CBX7, and optimized selection assays were implemented in parallel to both CBX8 and CBX7. Off-DNA validation of hits in orthogonal assays correlated well with enrichment levels and selectivities observed in the library selections. Optimization of a hit peptide yielded the most potent and selective CBX8 inhibitor to date. This ligand demonstrated selective engagement of CBX8 within PRC1 directly from cell lysates.^




Casey J. Krusemark, Purdue University.

Subject Area


Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server