The phosphoinositide 3-kinases are structurally closely related lipid kinases, which catalyze the ATP-dependent phosphorylation of phosphoinositide substrates1,2
. Together with the serine/threonine protein kinase B (PKB), PI3Ks constitute a central signalling hub that mediates many diverse and crucial cell functions like cell growth, proliferation, metabolism and survival1,3
. The observation that PI3Ks acting downstream of receptor tyrosine kinases (RTKs) are the most commonly mutated kinases in human cancers has spurred an immense interest in understanding the structural mechanisms how these mutations upregulate PI3K activity and in developing selective and drug-like PI3K inhibitors4,5
PI3Ks can be grouped into three classes based on their domain organisation6
. Class I PI3Ks are heterodimers consisting of a p110 catalytic subunit and a regulatory subunit of either the ‘p85’-type (associated with class IA PI3Ks with the isoforms p110α/β/δ) or the ‘p101/p84/p87’-type (associated with class IB PI3K p110γ). The p110 catalytic subunit consists of an adaptor-binding domain (ABD), a Ras-binding domain (RBD), a C2 domain, a helical domain and the kinase domain7-10
Mutant mice and inhibitor studies have shown less functional redundancy for the various class I PI3K isoforms than previously anticipated. While p110α and p110β are ubiquitously expressed, p110γ and p110δ are predominantly found in haematopoietic cells11-13
. Genetic deregulation of PI3K activity (oncogenic gain-of-function mutations, overexpression) has been implicated in cancer (all class I PI3K isoforms)14-17
, diabetes (p110α)18
, thrombosis (p110β)19
, rheumatoid arthritis (p110γ and p110δ)20
and asthma (p110γ and p110δ)21,22
. Consequently, the selective inhibition of individual PI3K isoforms using small molecule and ATP-competitive inhibitors is a promising therapeutic strategy23
. However, since all active-site side chains in contact with ATP are completely conserved throughout all class I PI3K family members (Supplementary Fig. 1
), this is a challenging objective. Furthermore, in order to minimize undesired and often poorly understood toxic side effects, such inhibitors ideally would have to show no cross-reactivity towards off-pathway targets24
The earliest generation of small molecule and ATP-competitive PI3K inhibitors including the pan-selective LY29400425
were important tools to investigate PI3K-mediated cellular responses in the laboratory but their low affinities (LY294002), instability (wortmannin) as well as non-selectivity and toxicity limited their clinical use. However, further chemical modifications of some of these early inhibitors significantly helped to improve their drug-like properties. For example, PWT-458 (Wyeth) and PX-866 (Oncothyreon) are modified wortmannin-based PI3K inhibitors with improved pharmacological properties that are currently in phase I clinical trials27,28
The first crystal structures of p110γ in complexes with pan-selective PI3K inhibitors29
made it possible to begin to rationalize PI3K isoform-selective inhibitors like AS604850 (Merck-Serono) for p110γ30
. However, many of these inhibitors retained off-target activities and, partially due to the lack of crystal structures of other PI3K isoforms and PI3K related protein kinases (PIKKS), these unwanted side effects were difficult to rationalize.
Noteworthy, the development of multi- and pan-selective PI3K inhibitors as well as dual PI3K/mTOR or PI3K/tyrosine kinase31
rather than isoform-selective PI3K inhibitors remains a valid therapeutic strategy. XL-147 (Exelixes), which is currently evaluated in combination with other cancer therapeutics is in phase I/II clinical trials for the treatment of non-small lung cancer and GDC-0941 (Roche)32
, also in phase I trials for the treatment of breast cancer33
, are examples of pan class I selective PI3K inhibitors. NVP-BEZ235 (Novartis), currently in phase I/II trials for breast cancer34
and SF1126 (Semaphore), a RGDS peptide conjugated prodrug of LY294002 in phase I trials35
are examples of dual-selectivity PI3K/mTOR inhibitors.
Recently, several new class I PI3K isoform-selective inhibitors showing improved selectivities and potencies have been reported and some of them have entered clinical trials: CAL-101 (Calistoga), a derivative of the highly p110δ-selective inhibitor IC8711436
with increased potency, entered stage I clinical trials for the treatment of acute myeloid leukaemia (AML) and B-cell chronic lymphoid leukaemia (CLL). The p110β-selective AZD6482 (AstraZeneca) is in clinical phase I for the treatment of thrombosis. Strikingly however, despite a growing list of such isoform-selective compounds, little is known about what determines isoform-selectivity on a structural level.
Impaired PI3Kδ signalling results in severe defects of innate and adaptive immune responses and suggested that targeting of this isoform would be a beneficial therapeutic strategy20,24
. To elucidate the molecular mechanisms of isoform-selectivity of PI3Kδ inhibitors, we report the crystal structure of the catalytic core of p110δ, both free and in complexes with a broad panel of novel and mostly p110δ-selective PI3K inhibitors. Our study provides the first detailed structural insights into the active site of a class IA PI3K occupied by non-covalently bound inhibitors. Furthermore, our structures suggest mechanisms to achieve p110δ selectivity and to increase potency of inhibitors without sacrificing isoform-selectivity. To obtain these structures, we developed a unique expression and purification scheme that has now been extended to all class IA PI3K isoforms.
With our new set of p110δ crystal structures and better models of flexibility resulting from molecular dynamics simulations we are now starting to understand why p110δ can be more easily deformed to open an allosteric pocket in which p110δ-selective inhibitors can be accommodated.