Optimization of protein kinase inhibitors – how structural information can help

At the Cresset European User Group Meeting 2015, Dr Daniel Kuhn presented a case study of the optimization of a c-Met protein protein kinase inhibitor from a high-throughput screening (HTS) hit through to a clinical compound. He focused on how structural information can inform the drug design process and also included retrospective Spark analysis that showed its effectiveness at identifying promising bioisosteres.

When taking a hit through to a lead, there are multiple parameters that need to be optimized, including potency, off-target activity, kinase selectivity, and the intellectual property freedom to operate. In all of these cases, structural information can help to guide you in the right direction.

c-Met is an established cancer target. The HCF/c-Met pathway is frequently deregulated in human cancer, and the overexpression of c-Met and or HGF correlates with aggressive tumor behaviour and poor clinical prognosis.

The starting point for the work was an HTS screen for novel c-Met inhibitors, which resulted in an attractive hit [1]. The X-ray structure of the hit revealed the hydrogen bonding pattern and a bended conformation which formed a basis for a structural based drug design.

The goal of the early hit optimization project was to improve the c-Met activity, synthetic access and removal of PDE activity.

Retrospective Spark analysis

At the time of the project we did not have a Spark license, therefore this analysis was carried out retrospectively. We were interested to know whether Spark would be able to discover additional known c-Met actives.

Spark was used to search for replacements for thiadiazinone activity. A Spark search has been performed and the top ranked hits have been inspected visually. It identified known bioisosteres in the top ranks. For example it found triazolopyridazine moieties from known clinical c-Met inhibitors. In addition, Spark also found thiadiazone alternatives such as pyridizanones which was found at that time in the project.

In general, when we use Spark to perform scaffold-hopping analysis we found it to be very fast and very intuitive to use. It identified good alternatives for our inhibitors.

What can we learn from the structure information?

We used a docking program to predict successful binding modes. This returned useful information, but the scoring needed to be improved.

Another thing we were able to learn from the structural information beyond visual analysis of X-ray structures was the estimation of water contributions to binding. Retrospectively, we used a Watermap of the HTS hit to rationalize important hinge and ribose pocket interactions.

Discovery and optimization of Tepotinib

Further optimization efforts comprised the screening for solvent-exposed residues resulting in the methylpiperidine group with good potency and pharmacokinetic properties. Pyrimidine was found to be the best hinge replacement. We were pleased to find that the binding mode was confirmed. We found that Tepotinib is an exquisitely selective c-Met kinase inhibitor with long lasting cellular c-Met inhibition.

Conclusions

The X-ray structures provide valuable information for the optimization of kinase hits.

We found that it can be challenging to use structure-based drug design (SBDD) for selectivity optimization, but that the binding mode prediction usually works well. Consideration of water contributions to binding can provide more insight than structural X-ray analysis alone.

Spark is a valuable tool to generate scaffold hopping ideas which can be rapidly communicated to the medicinal chemist

A combination of ligand-based and structure-based approaches have been successfully applied to various kinase projects.

Reference

The structure of Tepotinib and some of the development work is published in the paper: Dorsch D. et al, Bioorg Med Chem Lett 6094 (2015).