Covalent Fragment Screening of a Challenging Protein Target Using X-ray Crystallography

In this article Jijin Ayanath describes how we at Sygnature were able to successfully identify covalent binders to a structurally challenging target. This was achieved through the development of a robust crystal soaking system, that allowed the screening of our focused covalent fragment library.

Fragment-based drug discovery (FBDD) has gained prominence as an effective strategy for initiating drug discovery, in recent years. FBDD uses smaller molecules (fragments) that explore a diverse chemical space, unlike the traditional High-Throughput screening (HTS) methods, where larger more complex molecules are screened. Fragments typically exhibit weak binding affinities in the micromolar to millimolar range and have low molecular weights (100 – 300 Da). With technological advancements, methods such as Surface Plasmon Resonance (SPR), Thermal shift assays, NMR and X-ray crystallography are being used to detect these weak interactions. Although crystallographic screening lacks direct evaluation of Kd, what sets it apart from the rest is the ability to visualize fragment binding to a target at atomic level. The details of the specificity and interactions between the fragment and residues in the target protein obtained through crystallography provides a compelling foundation for iterative optimization of drug candidates.

Covalent Fragments at Sygnature Discovery

Covalent fragments offer unique advantages by binding irreversibly through their electrophilic war heads to nucleophilic residues such as cysteine, serine, threonine, lysine etc. on protein targets. Covalent fragment screening using crystallography can yield specific hits to proteins deemed ‘undruggable’. The hits identified may be further optimized for enhancing selectivity and binding affinity, thereby serving as promising lead candidates for drug discovery.

Our curated covalent fragment screen library at Sygnature Discovery has approximately 400 covalent fragments which were hand-picked from multiple commercial libraries and filtered using Rule-of-3 with focus on Cysteine-specific war heads. Many of the fragments have acrylamide warheads, frequently found in approved covalent drugs. Solubility analysis of the 405 fragments was performed using Dynamic Light Scattering (DLS) instrument. For each sample, Normalized intensity of Static Light Scattering (SLS) was measured at the desired concentration in the crystal-soaking buffer, allowing us to choose the best soluble 300 fragments for soaking crystals.

Figure 1: A) A data table showing the normalized intensity readings collected for a few samples. A visual image rating: green, yellow and red corresponding to if there is no precipitate, light precipitate or heavy precipitate in the sample is also given. B) The Normalized intensity values are plotted, and soluble fragments are selected for fragment screening based on their values and visual image rating.

Screening covalent fragments

Our client’s protein target posed significant challenges as it lacked deep pockets for inhibitor binding and had multiple binding partners with overlapping protein-protein interaction sites. However, the presence of numerous cysteine residues and the need for highly specific inhibition made it ideal target for covalent fragment screening. The screening process involves soaking protein crystals in high concentrations of covalent fragments and diffracting them to generate electron density maps, which reveal the presence of bound fragments. Robust, reproducible and good diffraction-quality crystals are required in large numbers to successfully carry out a fragment screening experiment. The client provided pure protein as well as the initial crystallization conditions. The crystals were reproduced at 8 ^oC in multiple conditions in 96-well plates. The crystallization set-up was optimized to generate enough crystals for conducting a fragment screen. Roughly ~ 1000 crystals are required for completing a 300-fragment screen. A soaking strategy was also developed by testing DMSO tolerance of the crystals by both visual inspection and diffraction experiments. Additional factors such as, the pH of the soaking buffer and its salt concentration were optimized. Once the strategy was in place, soaking was done, and 300 crystals were flash-frozen (one crystal/fragment).

Figure 2: A) Single crystals that were reproduced from the client protein B) Crystals soaked in high concentration of fragments after optimizing soaking conditions.

Results

The data collection was done at the Canadian Light Source (CLS). The diffraction obtained was in the range of 1.8 Å-2.5 Å. Data processing was done using PanDDA software (Pearce et al., 2017) which revealed 20 hits of covalent fragments binding to different cysteines on the target. These hits represent promising starting points for further optimization into potent and selective inhibitors.

Figure 3: A) A pie chart showing the resolutions of the data collected from Canadian Light Source (CLS), data collected was in the range of 1.8 – 2.5 Å. B) A close-up view of the client’s protein, highlighting a fragment hit in two Cysteines, and the corresponding electron density C) A close-up view of the client’s protein, highlighting a fragment hit in three Cysteines, and the corresponding electron density.

Conclusion

The combination of a focused covalent fragment library, robust crystal soaking process, and high-resolution X-ray screening enabled the successful identification of covalent binders to a structurally challenging target. This project demonstrates Sygnature Discovery’s capability in integrating covalent chemistry and structural biology for clients seeking to capitalize on opportunities for drug discovery.

If you are interested in running a fragment screen with us at Sygnature, please don’t hesitate to get in touch via info@peakproteins.com

Reference

Pearce, N., Krojer, T., Bradley, A. et al. A multi-crystal method for extracting obscured crystallographic states from conventionally uninterpretable electron density. Nat Commun 8, 15123 (2017)