These early optimization efforts were continued by classical medicinal chemistry in collaboration initially with Miroslav STRNAD and later with Michel LEGRAVEREND (Institut Curie, Orsay). Among purines efficient in the sub-micromolar range we selected 2-(R)-(1-ethyl-2-hydroxyethylamino)-6-benzylamino-9-isopropylpurine for further investigation. As a language convenience and by reciprocity, the chemical name was abandoned for an easier name, « roscovitine » ! Roscovitine was potent at inhibiting CDK1/cyclin B (IC50: 0.450 µM) and still quite selective. (R)-roscovitine was slightly more efficient than (S)-roscovitine. Roscovitine generated a wide interest as a pharmacological tool but also for clinical applications. (R)-roscovitine, under the name of CYC202 or Seliciclib, has now reached phase 2 clinical trials against various cancers and phase 1 clinical trial against glomerulonephritis. Roscovitine was also the start-point of an important combinatorial chemistry effort to generate even more efficient purine analogues. These efforts were also guided by the co-crystal structures of CDK2 with olomoucine and roscovitine (see below). They ultimately led Nathanael GRAY, in Peter SCHULTZ’s laboratory in Berkeley, to the identification of purvalanols which were very potent in vitro (IC50 in the 0.004-0.04 µM range) and highly selective. Many other related and potent purines were synthesized in numerous laboratories and confirmed the efficacy of 2,6,9-trisubstituted purines at inhibiting CDKs.
Co-crystallization with CDK2
Stimulated by discovery of the (relative) potency and the selectivity of olomoucine, and the crystal structure of CDK2 that had been solved a few months earlier by Sung-Hou KIM, in Berkeley. His results were accompanied by two surprises. Firstly, although olomoucine occupied (as expected) the ATP-binding pocket of CDK2, its purine ring and that of ATP were not at all orientated in the same manner. Secondly, isopentenyladenine was positioned in yet a third orientation. Later, roscovitine, purvalanols, and other purines were co-crystallized with CDK2. All these purines were orientated like olomoucine in the ATP-binding pocket of the kinase.
In contrast, the apparently related O6-cyclohexylmethylguanine (NU2058), and its optimized derivative NU6102, were found to bind to CDK2 in a different way than olomoucine/roscovitine/purvalanol. The purine/CDK2 co-crystal structures generated a lot of interest and, complemented with extensive structure-activity relationship studies, greatly stimulated the search for and rational optimization of new CDK inhibitors. Since olomoucine, more than 40 pharmacological inhibitors have been co-crystallized with CDKs. Despite a surprising chemical diversity, they are all flat, hydrophobic heterocycles which bind in the ATP-binding pocket through 2-3 hydrogen bonds with the backbone atoms of Leu83 and Glu81 in the active site and hydrophobic and van der Waals interactions.
3D representation of CDK5/p25 co-crystallized with a specific inhibitor targetting the ATP binding pocket of the kinase.
Selectivity: investigation with panels of enzymes and by affinity chromatography
A frequently asked question about kinase inhibitors relates to their selectivity, especially in view of their molecular mechanism of action (competition with ATP) which, intuitively, does not seem to be very favorable for high selectivity. The selectivity issue is usually approached by testing the compounds on a panel of purified, usually recombinant kinases, a time-consuming and unsatisfying approach (considering that only a small fraction of the 518+ kinases present in the human genome can be evaluated, and that potential non-kinase targets are not tested !). Nevertheless this approach showed that 2,6,9-trisubstituted purines were rather selective, essentially inhibiting CDK1, CDK2, CDK5, CDK7 and CDK9, but not CDK4 and CDK6. In addition to CDKs, the MAP kinases Erk1 and Erk2 were sensitive to 2,6,9-trisubstituted purines, although at much higher concentrations. In the absence of potent MAP kinase inhibitors, olomoucine was co-crystallized with Erk2.
As an alternative approach to purify and identify the targets of our purine CDK inhibitors we developed an affinity chromatography method with Nathanael GRAY. Purines were first immobilized through a linker to sepharose beads. Extracts of various cell types and tissues were then incubated with this matrix and, after stringent washing of the beads, the bound proteins were resolved by SDS-polyacrylamide gel electrophoresis and identified by microsequencing of internal peptides. This global approach confirmed that CDK1, CDK2, CDK5 and CDK7 were targets of purvalanols. It revealed that Erk1 and Erk2 are also important targets. The success of this method was a stimulus to extend it to other CDK inhibitors such as paullones and indirubins. This technology was recently successfully applied to roscovitine. This extremely useful information is being used to optimize second generation derivatives of roscovitine.
