The search for new therapeutic strategies is one of the main research fields in translational cancer research. The serine/threonine kinase polo-like kinase 1 (Plk1)[1] attracts great attention in the field of cancer therapy because it exhibits generally elevated activity in cancer cells [2, 3] and is a negative prognostic factor for cancer patients.[4] The importance of Plk1 activity as a measure for the aggressiveness of a tumor results from its important role in mitotic checkpoints.[5–8] Plk1 inhibition by antisense oligonucleotides, small interfering RNAs, antibodies, or dominant-negative mutants has resulted in reduced Plk1 expression and activity in vitro and in vivo.[8–15] A first generation of Plk1 inhibitors targeting the active conformation has entered clinical trials.[16–18] Here, we present the structure-based identification and biochemical validation of a novel potent (IC50= 200 pM) inhibitor of inactive Plk1 as a potential starting point for lead structure optimization. The high degree of conservation of kinase structure due to the same catalytic mechanism, the same cosubstrate (ATP) and similar protein folding poses the problem of inhibitor selectivity.[19] Kinases undergo conformational changes between the active and the inactive conformation by switching crucial structural elements: the aC helix, the activation loop with the conserved DFG motif as anchor, and the glycine-rich Ploop (Figure 1, figure S1 in the Supporting Information). An additional hydrophobic pocket (allosteric site), in which amino acid residues are less conserved, is accessible in the inactive conformation.[19] As a consequence, inhibitors of kinases in the inactive conformation (type II inhibitors) are more selective over other kinases than inhibitors of the active conformation (type I).[19]

We performed structure-based virtual screening for potential type II Plk1 inhibitors using a comparative protein model (homology model) in the absence of known reference ligands—a strategy that has been successful in other hit and lead finding projects already.[20–22] To cope with the structural ambiguities of the homology model, we combined pharmacophore screening and automated ligand docking methods,[23, 24] and transferred this concept to a model of the inactive conformation of Plk1.