Tales from a broad kinome profiler

After early promise, drugs targeting kinases have lacked the pharmacautical punch it was once thought they would deliver. Yet – as Lee Tessler discusses here – with new approaches, things are looking up

As a pharmaceutical target class, kinases are not the young bucks they used to be. The early success and lofty promises of rationally designed, small molecule kinase inhibitors have been dampened by a decade of compounds moving through clinical phases at unspectacular rates and a lack of success stories to match the concurrent breakthroughs in our understanding of the genome.

However, Pharma pipelines continue to fill with hundreds of new kinase inhibitors moving into clinical phases each year. Several key technologies have fuelled the continued innovation in the area of kinases, which is still perceived to be one of the most druggable classes of signalling biomolecules. These advances include broad enzymatic profiling, improved bioinformatics, and cell-based pathway analysis technologies.

Newer enzymatic profiling technologies have allowed higher-throughput analysis of kinome-wide activities at earlier stages of hit-to-lead. Similarity analysis of activity fingerprints spanning the majority of the known kinome has enabled new ways to reduce off-target effects (negative selection strategies) and for rational design on the basis of multi-kinase inhibition (polypharmacology strategies). Additionally, cell-based pathway analysis technologies have allowed higher-throughput analysis of signalling responses at earlier stages of hit-to-lead. The application of these cell-based tools intermittently and in conjunction with enzymatic profiling enhances confidence in target identification and lead optimisation.

A recent study by Yinghong Gao and colleagues in the April 15, 2013 issue of Biochemical Journal highlights a variety of advantages that early, frequent, and broad enzymatic profiling can offer to drug discovery researchers1.

The authors profiled 158 structurally diverse small molecules for their inhibitory activity towards 234 human recombinant kinases. The compounds used were previously defined in the literature as potent and specific kinase inhibitors. The resulting 70,000+ data points make up the third-largest kinase data submission to the ChEMBL database and shine light on best practices for off-target effect determination and polypharmacology strategies.

The first question the authors address is: “What assay format and assay conditions are ideal for generating activity fingerprint data?”

Several assay formats are available that measure binding of inhibitor to purified kinases, including high-throughput, competitive binding assays and thermal shift stability assays. Although there is significant correlation between binding assays and activity assays, profiling kinase inhibitors based only on binding affinities can result in false positive and false negative predictions of activity inhibition. Gao and colleagues used the widely accepted in vitro phosphotransferase activity assay to minimise these errors, which could otherwise create systematic biases in activity fingerprint data.

Unlike another recently published broad kinase activity screen2, Gao and colleagues used ATP concentrations at the Km for each reaction (within 15 µM), ensuring valid comparisons of activity fingerprints across the kinome. Although screening with fixed levels of ATP can provide operational simplicity for a lab, it can result in assays being performed in large deviations away (10+ fold) from the measured Km of ATP for particular reactions, precluding activity comparisons between kinases with very different affinities for ATP.

The next question the authors ask is: “How good are the tool compounds out there?” The answer depended on the compound; some provided the potency and specificity that the literature had indicated, while others showed contrary behaviours. For example, the TGF?R I inhibitors were very selective for their originally published targets (ALK4 and ALK5), while Aurora B was hit by many kinase inhibitors originally characterised for unrelated targets. Interestingly, compounds originally defined as classic tyrosine kinase inhibitors showed strong effects on serine/threonine kinases and vice versa. Clearly, scrutiny is warranted when researching the literature on tool compounds. For peace of mind, a broad profile of kinome activity with every tool compound used is priceless. Fortunately, due to improvements in screening technologies, obtaining profiles from fee-for-service providers is more affordable than ever.

What does this study reveal about global similarities among kinases? Because all kinases have ATP binding sites and share other regions of homology, the promiscuity of inhibitor-kinase interactions is a reality. From this study’s broad profiling data set, the investigators drew some general conclusions regarding the nature and magnitude of kinase promiscuity. Kinases were hit with between 0 and 68 inhibitors, with a median of 12. The sensitivity of the kinases was independent of the Km of ATP, and tyrosine kinases were overrepresented among the most sensitive kinases. On the basis of these observations, the authors suggest that researchers should routinely profile for off-target effects focusing on the subset of the kinome that is the most promiscuous. In fact, the top 20 most promiscuous kinases can effectively represent the entire kinome for the purposes of ranking inhibitor selectivity, with a Spearman R of 0.71 (?= 0.88).

Perhaps the most valuable information obtained from broad kinase profiling data is the structure-activity similarity matrix for compounds. Gao and colleagues clustered the dataset based on kinase activity and then on 2D chemical structure. The resulting similarity matrix, containing all compound pair-wise similarity measurement in terms of kinase activity profile and 2D structures, was used to show that molecules that share structural similarity may not share activity profiles, and vice versa. Using this similarity matrix, researchers can find compounds with the closest activity profiles to their compounds of interest for follow-up screening or to help explain experimental observations.

This comparison of activity fingerprints with chemical fingerprints opens up a novel strategy for multi-kinase inhibition. Not only can compounds with dissimilar scaffolds share a single kinase target, but they may in fact share activity against a broad set of kinases. Using this similarity matrix may enable drug developers to design compounds with desired polypharmacology using divergent scaffolds.

In summary, this study by Gao and colleagues clarifies the importance of choosing the right assay format and assay conditions for obtaining interpretable kinase activity fingerprints. It highlights the need for further characterisation of previously reported tool compounds, which have either never been broadly profiled or were profiled under unreliable assay conditions. The study also provides a quantitative perspective on global promiscuity in the kinome and how off-target effects can be ascertained and avoided in a cost-effective manner by routine screening of small panels. Finally, it points to strategies for multi-kinase inhibition by analysing similarity networks of broad kinase activity fingerprints during the lead optimisation phase.

References

1. Gao Y, Davies SP, Augustin M, Woodward A, Patel UA, Kovelman R, Harvey KJ. A broad activity screen in support of a chemogenomic map for kinase signalling research and drug discovery. Biochem J. 2013 Apr 15;451(2):313-28.

2. Anastassiadis T, Deacon SW, Devarajan K, Ma H, Peterson JR. Comprehensive assay of kinase catalytic activity reveals features of kinase inhibitor selectivity. Nat Biotechnol. 2011 Oct 30;29(11):1039-45.

Author Lee Tessler is the commercial manager for drug discovery biochemical tools and services at EMD Millipore