Do All Roads Really Lead to Rome? Learnings from Comparative Analysis using SPR, NMR, & X-Ray Crystallography to Optimize Fragment Screening in Drug Discovery.

There are several biophysical methods developed to rapidly identify weakly binding fragments to a target protein. X-ray crystallography provides structural information that is crucial for fragment optimization, however there are several criteria that must be met for a successful fragment screening...

Full description

Saved in:
Bibliographic Details
Main Author: Hassaan, Engi
Contributors: Klebe, Gerhard (Prof. Dr.) (Thesis advisor)
Format: Doctoral Thesis
Published: Philipps-Universität Marburg 2019
Online Access:PDF Full Text
Tags: Add Tag
No Tags, Be the first to tag this record!
Summary:There are several biophysical methods developed to rapidly identify weakly binding fragments to a target protein. X-ray crystallography provides structural information that is crucial for fragment optimization, however there are several criteria that must be met for a successful fragment screening including the production of soakable and well-diffracting crystals. Therefore, having a reliable cascade of screening methods to be used as pre-screens prior to labor-intensive X-ray crystallography would be extremely beneficial. This would allow the filtering of compounds as the screening progresses so that only the most promising hits remain. But which method should be the one to start the screening cascade? In this work, various sets of fragment libraries were screened against three different proteins; namely tRNA guanine transglycosylase (TGT) an important protein in Shigella, membrane associated protein peroxin 14 (PEX14) of T. Brucei, and endothiapepsin (EP), to investigate whether different screening methods will reveal similar collections of putative binders. The detailed comparative analysis of the findings obtained by the different methods is discussed in this thesis. Shigellosis, an acute bacterial infection of the intestine, is caused by the gram-negative Shigella bacterium whose pathogenicity is reliant on virulence factors (VirF) required to invade epithelial cells. The expression of these VirF is modulated by TGT. Strategies developed to inhibit TGT include potent active-site inhibitors to block the binding of tRNA, thereby preventing the transcription of the virulence factors. Our 96-fragment library was screened against TGT using SPR, NMR, and X-ray crystallography, as described in Chapter 2. A total of 81 fragments were screened in SPR using a direct binding assay approach, revealing a hit rate of 12%. A total of 77 fragments were screened in NMR revealing a hit rate of 29%. High-resolution crystal structures were also collected for the entire fragment library by soaking, revealing a hit rate of 8%. Upon comparison of all discovered fragment hits no overlaps from all three methods were found. Several factors are responsible for this finding such as exclusion of fragments from individual screens due to technical reasons. In detail, four X-ray hits were excluded from the SPR and NMR screens, two SPR hits were discarded from the NMR screen, and five NMR hits were never subjected to the SPR screen. SPR and NMR are currently the most commonly applied primary fragment screening techniques, however, our results suggest that if they would have been applied as incipient methods of a screening cascade, they would have missed three binders discovered by a subsequently applied, more elaborate crystallographic screen. X-ray crystallography allows the detection of specific binders that may be too weak binders to be detected by SPR and even by NMR but can still provide valid structural information to support the search for appropriate starting points in lead discovery. Additionally, MD simulations of the apo wild type TGT have predicted the opening of a transient sub-pocket located above the guanine/preQ1 pocket, which suggested a strategy to target this new binding site for the design of new inhibitors against TGT following a structure- based drug design concept which is also discussed in section 2.3. The human African trypanosomiasis (HAT), also known as the sleeping sickness, is a vector-borne parasitic disease caused by T. brucei and transmitted to humans by bites of the tsetse fly. T. brucei lacks feedback allosteric regulation of early steps in glycolysis but compartmentalizes the relevant enzymes within organelles called glycosomes. PEX14, a peroxin protein essential for biogenesis of glycosomes, forms an important protein-protein interaction with PEX5, an import receptor that transports cytoplasmic glycosomal enzymes into the organelle. Disrupting the PEX14/PEX5 interaction leads to the accumulation of glycosomal enzymes in the cytosol, depletion of ATP, glucose toxicity, metabolic collapse and death of T. brucei. This disruption can be achieved through small molecules that bind to and block PEX14, preventing PEX5 binding. A previous NMR screening of a fragment library resulted in fragment hits that bind to the N-terminal domain (NTD) of T. brucei PEX14. In this project, we attempted to validate these hits through X-ray crystallography by soaking, to allow visualization of the fragment interactions. The promising fragment hits would then be optimized into more potent lead compounds. Crystallization of the NTD PEX14 with a mutation in the first residue (E1W) revealed blocked binding pockets, as described in Chapter 3. The purpose of the added tryptophan was to render fluorescent properties to the short NTD construct which lacked fluorescent amino acids. However, this tryptophan was found to block the binding pockets of its neighboring crystal mates in the protein crystal, rendering a crystal form impossible to use for soaking. Attempts to find new crystal forms with free pockets were unsuccessful, as the small size of the protein and the hydrophobic nature of tryptophan rendered tightly packed protein crystals that block the binding pockets of neighboring crystal mates. Virtual Screening to discover novel ligands for co-crystallization revealed a ligand that aids the crystallization of the E1W PEX14 variant in the same space group but with a slightly different packing. This produced a crystal form that proved successful for fragment soaking as it enabled the binding of two additional fragment hits binding to further protein pockets. Additionally, the wild type form of PEX14 which lacks the tryptophan residue and thus has free binding pockets was crystallized. This enabled the soaking of a previously designed lead compound in different pockets of the PEX5 binding site. By obtaining a crystal structure of this complex at a resolution of 1.8 Å, the feasibility of using wild type PEX14 crystals for further fragment screening has been demonstrated. Endothiapepsin is a member of the pepsin-like aspartic proteases responsible for the hydrolytic cleavage of peptide substrates. Owing to its high degree of similarity to other pharmacologically relevant aspartic proteases, it has served as the model enzyme for studying their mechanism and to discover first lead structures. In previous work done by other members from our group to identify and characterize endothiapepsin binders, X-ray crystallography was consulted as a primary fragment screening method and its hit identification potential was compared to several biochemical and biophysical screening methods. The fragment library screened was designed for general purposes and contained 361 entries. Comparison of the overlap in the hit rates of the different methods to that of X-ray crystallography revealed a low overlap, with the RDA having the highest overlap at 7% and MS having the lowest overlap at 1% followed by STD NMR at 3%. To understand the reason behind the low overlap, two of these screening techniques were prioritized for closer analysis as described in Chapter 4. The 71 X-ray detected fragment hits were selected and rescreened again with STD NMR under slightly different buffer conditions, in addition to WaterLOGSY NMR experiments. The second STD NMR screen detected almost double the amount of hits as the initial one, and the Water LOGSY screen had the highest correlation from the NMR methods to the X-ray hits at 69%. This comparative analysis also revealed the phenomena of active site fragment displacement by use of so-called reporter ligands and that non-deuterated water in STD NMR may lead to false negatives. The entire 361 fragment library was also screened with SPR using an inhibition in solution assay, adding another biophysical method for our comparative analysis to give us further insight of which conditions are crucial to maintain while transferring across different techniques. The resulting hit rate from SPR was 34%, correlating to an overlap of 11% with the X-ray hits - the highest correlation between screening methods reported by us thus far. Finally, we also studied fragment detection and cocktailing in crystallography in comparison to fragment cocktailing in NMR. From this we concluded that cocktailing in crystallography can also lead to false negatives due to fragment competitive behavior and can reveal a different binding mode for a given fragment compared to the adopted geometry found when soaked individually. As for NMR, despite the ability to detect competitive binding of fragments due to the temporary binding and unbinding events, the parallel binding and thus detection of fragments is not always guaranteed as seen in 20% of the fragments we screened, in addition to our observation that the detection of fragments in cocktail NMR may also depend on the comparison of the cocktail set they are a part of.
Physical Description:266 Pages