Phosphoinositide 3-kinases (PI3Ks) are major regulators of many cellular functions, and hyperactivation of PI3K cell signalling pathways is a major target for anticancer drug discovery. PI3Kα is the isoform most implicated in cancer, and our aim is to selectively inhibit this isoform, which may be more beneficial than concurrent inhibition of all Class I PI3Ks. We have used structure-guided design to merge high-selectivity and high-affinity characteristics found in existing compounds. Molecular docking, including the prediction of water-mediated interactions, was used to model interactions between the ligands and the PI3Kα affinity pocket. Inhibition was tested using lipid kinase assays, and active compounds were tested for effects on PI3K cell signalling. The first-generation compounds synthesized had IC50 (half maximal inhibitory concentration) values >4 μM for PI3Kα yet were selective for PI3Kα over the other Class I isoforms (β, δ and γ). The second-generation compounds explored were predicted to better engage the affinity pocket through direct and water-mediated interactions with the enzyme, and the IC50 values decreased by ∼30-fold. Cell signalling analysis showed that some of the new PI3Kα inhibitors were more active in the H1047R mutant bearing cell lines SK-OV-3 and T47D, compared with the E545K mutant harbouring MCF-7 cell line. In conclusion, we have used a structure-based design approach to combine features from two different compound classes to create new PI3Kα-selective inhibitors. This provides new insights into the contribution of different chemical units and interactions with different parts of the active site to the selectivity and potency of PI3Kα inhibitors.

Introduction

Phosphoinositide 3-kinases (PI3Ks) are a family of lipid kinases that catalyze the phosphorylation of a 3′-hydroxyl group on the inositol ring of phosphatidylinositol containing embedded membrane phospholipids. The PI3K enzymes are divided into three different classes based on structural differences and substrate specificities, and have been the subject of many reviews and commentaries [18]. Class I PI3Ks catalyze the conversion of phosphatidylinositol (4,5)-bisphosphate (PIP2) into phosphatidylinositol (3,4,5)-trisphosphate (PIP3), and these enzymes can be further divided into Class IA and Class IB subfamilies. PIP3 then acts as a second messenger activating many downstream signalling proteins including PDK-1 and AKT to control cell outcomes including proliferation, survival, metabolism, motility and transformation. Class IA PI3Ks are heterodimers consisting of one catalytic subunit p110α, p110β or p110δ, tightly bound to one regulatory subunit p85α (and its splice variants p55α and p50α).

The PIK3CA gene that encodes the p110α protein is frequently mutated in cancer [9,10], with most mutations occurring in two hotspot regions. One hotspot is located in the helical domain and involves the amino acid substitutions E542K, E545K and Q546K, whereas the other hotspot is found in the kinase domain and involves substitution of H1047 with R, L or Y (Catalogue of Somatic Mutations in Cancer, http://cancer.sanger.ac.uk/cosmic). Mutations at both hotspots activate the enzyme, increasing enzymatic activity, AKT activation and the transforming potential of cells in vitro, as well as contributing to tumorigenesis in vivo [8].

As hyperactive PI3Kα signalling is associated with cancer, selectively inhibiting this enzyme may be expected to have increased therapeutic benefit over the concurrent inhibition of all Class I PI3Ks [11]. PI3Kα-selective inhibitors, such as the tool compound A66 or clinical agent BYL719 (Alpelisib) [12,13], have enabled a clearer understanding of PI3Kα function in cell signalling and survival, as well as the potential for patient selection in clinical trials [14]. Jamieson et al. used the PI3Kα-selective inhibitor A66 [15,16] to show that AKT phosphorylation in cell lines harbouring the PI3Kα H1047R mutation was more responsive to PI3Kα blockade compared with some harbouring the E545K mutation, and that this responsiveness was associated with growth delay of an SK-OV-3 xenograft in vivo [17]. Others have also shown that proliferation and viability of breast cancer cell lines harbouring PIK3CA mutations are more sensitive to BYL719 treatment than cell lines with the wild-type PIK3CA gene [18,19].

PI3K inhibitors bind to the ATP substrate-binding site located in a cleft between the N- and C-terminal lobes of the kinase domain. Four regions have been defined that serve as the main ligand interaction sites. These include the hinge region in the adenine pocket, the affinity pocket, the specificity pocket [2022] and a non-conserved region of the C-terminal lobe [23]. The latter two are exploited by inhibitors to achieve selectivity both between the PI3K isoforms and over other kinases.

Well-studied PI3Kα-selective inhibitors occupy chemically distinct compound classes (Figure 1). One can be represented by PIK-75 and its derivatives [2426], another with superior selectivity is represented by BYL719 [13], whereas CNX-1351 is a covalent modifier of the p110α subunit, targeting the p110α-specific amino acid Cys862 [27]. Based on predicted or observed active-site interactions, these inhibitors engage different areas of the ATP-binding site. PIK-75 and related compounds are predicted to interact with the backbone amide of Val851 in the linker region through a nitrogen atom in the central bicyclic imidazo[1,2-a]pyridine core and engage a region in the N-terminal wall of the kinase domain close to the specificity pocket termed the Trp-shelf [28]. Unlike PIK-75, the highly selective PI3Kα blockers represented by A66 and BYL719 do not have a central bicyclic scaffold (Figure 1) [12,13]; instead, they are biaryl-based, with a central aminothiazole unit connected to a second aromatic feature. Both the crystal structure of BYL719 [protein data bank (PDB) code 4JPS] [13] and the predicted binding mode of A66 and related analogues [17,29,30] show that the aminothiazole ring of the scaffold makes a hydrogen bond contact with Val851, and the PI3Kα-specific amino acid Gln859 via an (S)-pyrrolidine carboxamide group [13,17]. Binding to Gln859 confers both PI3Kα selectivity and potency [12,17,31]. Whereas more potent analogues of PIK-75, such as 1 (Figure 1), are predicted to engage the affinity pocket with a cyano group [28], BYL719 and its analogues use a larger substituted aromatic unit in this region.

PI3Kα-selective inhibitors.

Figure 1.
PI3Kα-selective inhibitors.

A66 and BYL719 are biaryl based inhibitors with a central aminothiazole unit. PIK-75 and compound 1 have a bicyclic imidazo[1,2-a]pyridine core. CNX-1351 covalently modifies the p110α subunit of the PI3Kα enzyme.

Figure 1.
PI3Kα-selective inhibitors.

A66 and BYL719 are biaryl based inhibitors with a central aminothiazole unit. PIK-75 and compound 1 have a bicyclic imidazo[1,2-a]pyridine core. CNX-1351 covalently modifies the p110α subunit of the PI3Kα enzyme.

Based on the different binding modes for PIK-75 and BYL719, and the effect of the carboxamide unit on potency in both A66 [17] and BYL719 [13], we considered that fusing the bicyclic heteroaromatic core of PIK-75 with the carboxamide of the latter series might be sufficient to maintain potency with respect to PI3Kα and also improve selectivity for this enzyme over other related kinases. Within this is also a better understanding about the contribution of the nitroaromatic-sulphonamide unit to the potency of the broader PIK-75 class of compounds. Furthermore, there are examples of PI3Kγ inhibitors with a core structure similar to the new series of PI3Kα inhibitors in the present study. This also prompted us to question whether the addition of a chemical unit able to contact Gln859 was sufficient to switch the selectivity of a PI3Kγ inhibitor.

Materials and methods

PI3K expression and purification

Recombinant baculovirus containing coding sequences for both the full-length p85α regulatory subunit and either of the full-length p110α, β or δ catalytic subunits, or recombinant baculovirus containing a coding sequence for the p110γ catalytic subunit alone, were used to infect Sf9 cells (Life Technologies, Carlsbad, CA, U.S.A.) at a ratio of virus stock to cell culture volume empirically determined to produce maximal protein expression after 72 h. Cell cultures were then centrifuged at 666×g for 5 min at 4°C, and the pellets were resuspended in an equal volume of 20 mM Tris and 137 mM NaCl (pH 8.0), flash-frozen in liquid nitrogen and stored frozen at −20°C. The frozen cell pellets were resuspended with gentle agitation in 25 mM Tris (pH 8.0), 0.5% NP-40 alternative (Calbiochem, San Diego, CA, U.S.A.) and complete EDTA-free protease inhibitors (Roche, Basel, Switzerland; 1 inhibitor tablet per 50 ml volume) at room temperature until lysed as assessed by light microscopy. Cell lysates were clarified by centrifugation at 20 000×g for 30 min at 4°C; the supernatant was removed and adjusted to 5% (v/v) glycerol, 150 mM NaCl, 7.5 mM imidazole and 20 µg/ml RNAseA (Roche) and then passed through a 0.45 µm filter before loading onto a Talon-Co2+ resin column (Clontech, Mountain View, CA, U.S.A.) pre-equilibrated in 25 mM Tris (pH 8.0), 100 mM NaCl, 5% (v/v) glycerol and 7.5 mM imidazole. The resin was washed with 10 column volumes of 25 mM Tris (pH 8.0), 150 mM NaCl, 7.5 mM imidazole and 5% glycerol. Protein was eluted from the column with 25 mM Tris (pH 8.0), 150 mM NaCl, 150 mM imidazole and 5% glycerol. Eluted protein was dialyzed overnight in 50 mM Tris (pH 8.0), 100 mM NaCl, 1 mM DTT (dithiothreitol) and 5% glycerol at 4°C. The dialyzed protein was centrifuged at 10 000×g for 10 min at 4°C to remove precipitated material, and the supernatant was desalted to 40 mM NaCl using a Pharmacia HiTrap desalting column (GE Healthcare, Little Chalfont, U.K.) according to the manufacturer's instructions. The desalted protein was applied to a MonoQ column (GE Healthcare) pre-equilibrated in 50 mM Tris (pH 8.0), 40 mM NaCl, 1 mM DTT and 5% glycerol, and fractionated over a gradient from 40 to 200 mM NaCl in the same buffer.

Molecular modelling

Ligand preparation

All ligands were created using the SKETCHER module in SYBYLX2.1.1 (Tripos) and a single 3D conformer was generated using CONCORD v6.3.1 as implemented in SYBYLX2.1.1. The ligands were then minimized using the MAXMIN module within SYBYLX2.1.1 using the Conjugate Gradient method with the Tripos force field and Gasteiger–Marsili charges. Minimization was performed until the gradient convergence cutoff of 0.05 kcal/(mol.Å) was reached. The distance-dependent dielectric function was used, with the dielectric constant set to 1.

Molecular docking

Molecular docking was performed using GOLD (Genetic Optimization for Ligand Docking) v5.2 [32]. The human PI3Kα structure (PDB entry 2RD0) was prepared for docking by modifying the side chain orientations of amino acids Gln60, Gln137, Asn145, His213, Gln269, Gln630, His701, His759 and Gln1014 based on MolProbity analysis [33]. Of these, His701 and His759 were part of the docking site, defined as an 18 Å cavity centred on the Ile800 CD1 atom. The scoring function used was ChemScore modified for use with kinases [34]. All protein and ligand atom types were generated automatically within GOLD. Protein hydrogen bond constraints were used for docking compounds 1–8, A66 and BYL719, with the backbone amide of Val851 and the side chain carboxamide group of Gln859 set as the interaction points. For GSK2126458, protein hydrogen bond constraints were used with the backbone amide of Val851 and the side chain amine of Lys802. The constraint weight was set at 10, with a minimum hydrogen bond geometry weight of 0.005. Ligand flexibility settings were kept as default, except for the Ring-NH2 and Ring-NR1R2 terms, which were set to flip. The diverse solutions option was turned on with the cluster size and RMSD (root mean square deviation) set at 1 and 1.5 Å, respectively. The Genetic Algorithm run was set at 20 for each ligand, a search efficiency of 200% was used and 20 poses were kept per ligand. All poses were then rescored with ChemScore using GOLDv5.2 and included a minimization step. The side chains of ATP-binding site amino acids Lys776, Ile800, Lys802, Asp810, Tyr836 and Asp933 were set to flexible, with movement restricted to the rotamer library available in GOLDv5.2.

Modelling of water molecules in the ATP-binding site

Two protocols were used to model water molecules in the PI3Kα active site. First, a propensity map for the PDB entry 2RD0 was generated by Superstar 2.1.2 (CCDC, Cambridge Crystallographic Data Centre) for a water oxygen probe, and a water molecule (water 4) was positioned at the map peak in the affinity site close to amino acids Tyr836, Asp810 and Asp933. Secondly, water molecules in agreement with the propensity map were extracted from PDB entries 4DK5 (HOH1312 as water 1), 3L54 (HOH30 as water 2) and 2WXF (HOH2163 as water 3; HOH2147 as water 5; HOH2242 as water 6) after superimposition onto 2RD0 using the kinase domain only and then included in the 2RD0 docking receptor. Water molecules were used in the docking calculations individually and allowed to move up to either 1 or 2 Å from the original position. Images were generated with PyMol (www.schrodinger.com).

Lipid kinase IC50 determination

IC50 values were determined using the PI3K homogenous time-resolved fluorescence (HTRF) assay (Merck Millipore, #33-016) following the manufacturer's instructions, using the supplied PIP2 as the substrate. All PI3K isoforms were titrated and used at a concentration equivalent to their EC65 [concentration at 65% of the maximum response: 65 ng/ml for PI3Kα (p110α/p85α), 30 ng/ml for PI3Kα E545K mutant, 65 ng/ml for PI3Kα H1047R mutant, 230 ng/ml for PI3Kβ (p110β/p85α), 30 ng/ml for PI3Kδ (p110δ/p85α) and 300 ng/ml for p110γ]. Inhibitors were dissolved and serially diluted in 100% DMSO (dimethylsulfoxide). DMSO was used at a final concentration of 2.5%. The results were obtained using the BioTek Synergy 2 Multi-Mode Reader.

Western blot assays of cell signalling

Cell lines were grown in αMEM (alpha-modified minimal essential growth medium) supplemented with 5% (v/v) foetal bovine serum (Invitrogen), 100 units/ml penicillin and 100 μg/ml streptomycin at 37°C with 5% CO2. Twelve-well tissue-culture plates were seeded at 500 000 cells per well. Cells were left to recover for at least 12 h and then serum-starved overnight. Cells were then treated with compounds for 60 min and stimulated with 500 nM insulin for 15 min unless otherwise stated, followed by washing with cold phosphate-buffered saline [140 mM NaCl, 8 mM Na2HPO4 and 2 mM NaH2PO4 (pH 7.4)], before the addition of lysis buffer {50 mM HEPES, 150 mM NaCl, 10 mM EDTA (pH 8.0), 10 mM Na2P2O7, 2 mM vanadate, 100 mM NaF, 1% (v/v) Nonidet P40, 10 μM leupeptin, 15 μM pepstatin A, 1 mM AEBSF [4-(2-aminoethyl) benzenesulfonyl fluoride], 0.6 μM aprotinin, 30 μM ALLN (Inhibitor of calpain I, calpain II, cathepsin B and cathepsin L) and 1 mM DTT (pH 7.4)}. Lysates were kept on ice before supernatants were collected after centrifugation at 14 000×g for 15 min at 4°C. A Pierce BCA Protein Assay Kit (Thermo Scientific, Waltham, MA, U.S.A.) was used to determine the lysate protein concentration. The lysates were then stored at −80°C for future use.

Proteins were separated by SDS–PAGE (sodium dodecyl sulphate–polyacrylamide gel electrophoresis) using a 4–15% gradient gel (Bio-Rad, CA, U.S.A.) and transferred onto polyvinylidene difluoride membranes (Pall Corporation, NY, U.S.A.). The membranes were incubated for 1 h in TBS-T (50 mM Tris, 275 mM NaCl, 5 mM KCl and 0.1% Tween) containing 3% (w/v) BSA at room temperature and overnight at 4°C in the same solution containing primary rabbit antibodies (Cell Signalling Technology, Beverly, MA, U.S.A.). Immunoreactive proteins were detected using a horseradish peroxidase-conjugated anti-rabbit goat IgG (Dako and Agilent Technologies, Santa Clara, CA, U.S.A.) and ECL® (enhanced chemiluminescence, GE Healthcare). Signals were detected using Fuji LAS4000 and analyzed with the Fuji Image Gauge software.

Results

Design and biochemical characterization of new PI3Kα inhibitors

Molecular docking was used to model the binding modes of A66, BYL719 and 1 into the PI3Kα active site using the PDB entry 2RD0. The predicted binding modes retrieved as the top-ranked poses were similar to those previously reported [17,28]. The binding mode for BYL719 docked in the absence of water had a RMSD value of 1.240 Å when compared with its observed binding mode (PDB code 4JPS) [13]. Upon closer inspection, the –CF3 group of BYL719 was mispredicted, and excluding it returned a RMSD value of 0.512 Å, indicating that the 2RD0 structure combined with molecular docking can predict ligand binding consistent with crystallographic data. Using the structure-guided inhibitor alignment of A66, PIK-75 and compound 1 (Figure 2A,B), we considered a chemotype fusion strategy (Figure 2C) that uses the pyrrolidine carboxamide of A66, the central bicyclic imidazo[1,2-a]pyridine core of PIK-75 and the cyano group of 1. This may exploit the selectivity of A66 and BYL719-like molecules, along with the linker and affinity pocket interactions of PIK-75 and 1. When docked into the 2RD0 ATP-binding site, the hybrid compound 2 could interact with Gln859, the hinge region, and fit the affinity pocket to a similar extent as predicted for 1 (Figure 2D,E).

Design strategy for compounds 2–4 based on a structure-guided alignment of A66 and compound 1.

Figure 2.
Design strategy for compounds 2–4 based on a structure-guided alignment of A66 and compound 1.

The predicted binding modes for A66 and compound 1 are shown in (A) and (B), respectively. (C) The fusion of different chemical units based on the structure-guided alignment of A66, PIK-75 and compound 1 leading to compound 2 is shown in red. Predicted PI3Kα ATP-binding site contacts are indicated by arrows. (D) The predicted docking mode for the hybrid compound 2 is shown as sticks; A66-derived component (magenta), PIK-75 and compound 1 (blue). (E) Superimposition of compound 1 (cyan carbons) and compound 2 (magenta carbons).

Figure 2.
Design strategy for compounds 2–4 based on a structure-guided alignment of A66 and compound 1.

The predicted binding modes for A66 and compound 1 are shown in (A) and (B), respectively. (C) The fusion of different chemical units based on the structure-guided alignment of A66, PIK-75 and compound 1 leading to compound 2 is shown in red. Predicted PI3Kα ATP-binding site contacts are indicated by arrows. (D) The predicted docking mode for the hybrid compound 2 is shown as sticks; A66-derived component (magenta), PIK-75 and compound 1 (blue). (E) Superimposition of compound 1 (cyan carbons) and compound 2 (magenta carbons).

Compound 2 along with a small set of analogues were then synthesized following the procedure outlined in Supplementary Methods, and tested for inhibitory activity against all Class I PI3K enzymes. Inhibition data for 2 and two halide substituted forms, 3 and 4, showed that the new series was active against PI3Kα (Table 1) and was selective for PI3Kα over other Class I isoforms. These compounds were much less potent compared with the parent compounds A66, 1 and PIK-75, with IC50 values ranging from 5 to 12 μM for wild-type PI3Kα. However, preparations of fragment-like compounds related to PIK-75 and 1 that are expected to only interact with the hinge region and affinity pocket did not show any activity at 250 μM (Supplementary Table S1), demonstrating that the addition of proposed PI3Kα-specific interactions by inclusion of a pyrrolidine carboxamide does improve the potency and selectivity towards PI3Kα within the Class I PI3K enzymes.

Table 1
IC50 values for known PI3Kα-selective inhibitors and hybrid compounds

Shown as mean ± SD (n ≤ 2).

Inhibitor Structure IC50 (nM) 
PI3Kα PI3Kβ PI3Kδ p110γ 
A66  44.3 ± 21.8 >12 500 2673 ± 709 2706 ± 1466 
PIK-75  5.6 ± 2.4 50.4 ± 17.5 171.9 ± 57.2 42.7 ± 20.8 
GSK2126458  0.8 ± 0.4 2.1 ± 1.1 0.6 ± 0.5 1.4 ± 0.1 
1  2.4 ± 1.6 50.1 ± 6.9 103.4 ± 43.6 2.4 ± 0.07 
2  5860 ± 3800 >12 500 >12 500 >12 500 
3  11 865 ± 4622 >12 500 >12 500 >12 500 
4  4479 ± 454 >12 500 >12 500 >12 500 
Inhibitor Structure IC50 (nM) 
PI3Kα PI3Kβ PI3Kδ p110γ 
A66  44.3 ± 21.8 >12 500 2673 ± 709 2706 ± 1466 
PIK-75  5.6 ± 2.4 50.4 ± 17.5 171.9 ± 57.2 42.7 ± 20.8 
GSK2126458  0.8 ± 0.4 2.1 ± 1.1 0.6 ± 0.5 1.4 ± 0.1 
1  2.4 ± 1.6 50.1 ± 6.9 103.4 ± 43.6 2.4 ± 0.07 
2  5860 ± 3800 >12 500 >12 500 >12 500 
3  11 865 ± 4622 >12 500 >12 500 >12 500 
4  4479 ± 454 >12 500 >12 500 >12 500 

Improving the fit to the affinity pocket

We next set out to improve PI3Kα ATP-binding site complementarity by improving the affinity pocket interaction and by maintaining a hydrogen bond connection with Gln859. A survey of PI3K structures available in the PDB showed that entries 3QK0, 4DK5 and 3L08 had ligands bound that were within hydrogen-bonding distance of a water molecule in the affinity pocket, while the ligand bound in 3APC interacted directly with amino acid side chains (Figure 3A). It was evident from these structures that 3-pyridyl units in different inhibitor classes can hydrogen bond to a water molecule that interacts with the side chains of affinity pocket amino acids Tyr836 and Asp810 (p110α numbering). This prompted us to consider the attachment of 3-pyridyl-containing substituents to the imidazo[1,2-a]pyridine core of 2 (Figure 3B), with the compounds synthesized following the procedure outlined in the Supplementary Methods.

Design strategy for compounds 5–8 based on structure-guided alignments with known inhibitors.
Figure 3.
Design strategy for compounds 5–8 based on structure-guided alignments with known inhibitors.

(A) Cartoon representation of PI3K X-ray crystal structures with ligands bound that project a 3-pyridyl unit into the affinity site; PDB codes are inset, and some of the polar interactions are labelled with yellow dashed lines; water molecules are shown as red spheres. (B) Incorporation of a 3-pyridyl into compound 2 making compounds 58 (top to bottom). The 3-pyridyl unit is shown in black, with substitutions shown in blue. Potential ATP-binding site interactions are indicated with arrows. (CF) Cartoon diagrams showing the overlap between the predicted binding modes for the 3-pyridyl-containing hybrid compounds and the 3-pyridyl-containing inhibitors in (A). Water from the predicted models is shown as a red sphere, and water from the crystal structure is shown as a cyan sphere. (C) Predicted binding mode for compound 5 (purple carbons) with the PDB entry 3QK0 (yellow carbons) superimposed. (D) Predicted binding mode for compound 6 (purple carbons) with the PDB entry 3APC (cyan carbons) superimposed. (E) Predicted binding mode for compound 7 (purple carbons), with the PDB entry 4DK5 (magenta carbons) superimposed. (F) Predicted binding mode for compound 8 (purple carbons) with the PDB entry 3L08 (pink carbons) superimposed.

Figure 3.
Design strategy for compounds 5–8 based on structure-guided alignments with known inhibitors.

(A) Cartoon representation of PI3K X-ray crystal structures with ligands bound that project a 3-pyridyl unit into the affinity site; PDB codes are inset, and some of the polar interactions are labelled with yellow dashed lines; water molecules are shown as red spheres. (B) Incorporation of a 3-pyridyl into compound 2 making compounds 58 (top to bottom). The 3-pyridyl unit is shown in black, with substitutions shown in blue. Potential ATP-binding site interactions are indicated with arrows. (CF) Cartoon diagrams showing the overlap between the predicted binding modes for the 3-pyridyl-containing hybrid compounds and the 3-pyridyl-containing inhibitors in (A). Water from the predicted models is shown as a red sphere, and water from the crystal structure is shown as a cyan sphere. (C) Predicted binding mode for compound 5 (purple carbons) with the PDB entry 3QK0 (yellow carbons) superimposed. (D) Predicted binding mode for compound 6 (purple carbons) with the PDB entry 3APC (cyan carbons) superimposed. (E) Predicted binding mode for compound 7 (purple carbons), with the PDB entry 4DK5 (magenta carbons) superimposed. (F) Predicted binding mode for compound 8 (purple carbons) with the PDB entry 3L08 (pink carbons) superimposed.

The enzyme inhibition data presented in Table 2 show that improving the affinity pocket complementarity increased the potency of our new compound series. Compound 5 showed a 35- to 13-fold decrease in IC50 values for wild-type PI3Kα when compared with 3 and 4, respectively. The data also showed that replacing the 3-pyridyl unit with 2-aminopyrimidine in 6, or modifying the 3-pyridyl unit by adding a 6-OMe as in 7 produced increases in potency comparable with 5, with IC50 values ranging from 322 to 213 nM, respectively, for the wild-type PI3Kα enzyme. The inhibition data for the different Class IA and IB enzymes showed that compounds 57 were also better inhibitors of the PI3Kβ, PI3Kδ and p110γ enzymes compared with those in Table 1, with IC50 values ranging from 1.4 to >12.5 μM. The selectivity between wild-type PI3Kα and the other isoforms ranged from 10- to over 37-fold for 5, 8- to 21-fold for 6 and 6- to 40-fold for 7. The data for 8 showed that it is substantially more potent than the other compounds in Table 2, with a 100-fold decrease in the IC50 value compared with 7. This compound is also not selective for PI3Kα, an outcome consistent with the inclusion of an ionizable sulphonamide unit also found in GSK2126458 (omipalisib) [35]. Compounds 1–8 were also active against the PI3Kα oncogenic mutants E545K and H1047R, and inhibited enzymatic activity of the oncogenic mutants to similar extents as the wild-type PI3Kα (Supplementary Table S2).

Table 2
IC50 values for 3-pyridyl-containing hybrid compounds

Shown as mean ± SD (n ≤ 2).

Inhibitor Structure IC50 (nM) 
PI3Kα PI3Kβ PI3Kδ p110γ 
5  340 ± 109 >12 500 4646 ± 2233 3304 ± 272 
6  322 ± 87 6689 ± 2950 2432 ± 1439 3871 ± 282 
7  213 ± 124 8476 ± 4772 1375 ± 650 2232 ± 884 
8  2.1 ± 0.2 10.2 ± 4.4 1.5 ± 0.8 2.2 ± 1.5 
Inhibitor Structure IC50 (nM) 
PI3Kα PI3Kβ PI3Kδ p110γ 
5  340 ± 109 >12 500 4646 ± 2233 3304 ± 272 
6  322 ± 87 6689 ± 2950 2432 ± 1439 3871 ± 282 
7  213 ± 124 8476 ± 4772 1375 ± 650 2232 ± 884 
8  2.1 ± 0.2 10.2 ± 4.4 1.5 ± 0.8 2.2 ± 1.5 

Modelling water ligand interactions in the PI3Kα affinity pocket

To gain some insights into the possibility of a water-mediated interaction between the 3-pyridyl unit and the PI3Kα affinity pocket contributing to the increased potency of compounds 58, water molecules were included in the docking calculations used to model the binding poses of these compounds.

As water molecules are not observed in the ATP-binding site of the PDB entry 2RD0, potential water-binding sites were identified from a propensity map generated by Superstar, and used to guide the placement of water molecules [36]. Six molecules were positioned in and around the propensity peak in the affinity pocket between Tyr836, Asp810 and Asp933 (Supplementary Figure S1), and these were either maintained in the ATP-binding site during the docking calculations, or considered displaceable, and allowed to toggle on and off and move up to either 1 or 2 Å. When ranked by the ChemScore fitness value, the pose with the highest fitness for each condition that was consistent with the binding mode of PI3Kα-selective inhibitors was analysed for the presence of water, and its interaction with the ligand and amino acid side chains (Supplementary Tables S3–S8). For compounds 3, 4 and A66, which are unable to form hydrogen bond interactions with the affinity pocket, and which also occupy the affinity pocket to different extents, a water molecule was predicted in the affinity pocket, interacting with the side chains of amino acids Tyr836 and Asp933 or Asp810 and Asp933 (Supplementary Table S9). The predicted location of the water was 2.2, 1.3 and 1.1 Å from the peak of the propensity map, and 2.0, 1.0 and 0.8 Å from a similarly placed water molecule in the superimposed BYL719 bound structure [13] for 3, 4 and A66, respectively. The data for BYL719 also indicate that the water molecule of interest is maintained in the affinity pocket in the absence of a hydrogen bond with the ligand. Taken together, these data indicate that the GOLD water docking routine supports the presence of an occupied water-binding site in the PI3Kα ATP affinity pocket.

We next applied this method to the 3-pyridyl-containing compounds 58 in Table 2, and data for the highest scoring poses consistent with a PI3Kα-selective inhibitor are presented in Supplementary Tables S3–S8, with data for the poses that had the highest fitness values found across all docking conditions summarized in Supplementary Table S9. These poses predicted that the inclusion of the 3-pyridyl unit would facilitate a hydrogen bond with a water molecule interacting with the side chain hydroxyl group of Y836 in the affinity pocket (Figure 3C–F). These poses were found under different conditions across the four inhibitors: compounds 5 and 6 used water 2, whereas 7 used water 1 and compound 8 used water 3 (Supplementary Table S9).

Within the predicted binding modes for compounds 7 and 8, Lys802 interacted with the methoxy oxygen atom of the inhibitors' methoxy-substituted 3-pyridyl unit as well as the sulphonamide group in the latter, consistent with the binding modes of other sulphonamide-containing compounds (PDB entries 3L08 and 3QK0) including GSK2126458 (Figure 3D,E).

Addition of pyrrolidine carboxamide instals PI3Kα selectivity into a PI3Kγ inhibitor

The similarity between the substructure of our new series and that of a recently described p110γ selective inhibitor 9 [37] prompted the replacement of the imidazolyl urea group from 9 with a pyrrolidine carboxamide to create the hybrid compound 10 (Figure 4). The lipid kinase IC50 data presented in Table 3 confirmed that compound 9 was a potent inhibitor of p110γ with an IC50 of 11.5 nM and selective over the other isoforms tested. It also showed that the two compounds 11 and 12, representing the central scaffold of 9, had reduced potency but were also selective for p110γ. The activity data for 10 showed that adding a pyrrolidine carboxamide group to 11 improved activity towards PI3Kα more than p110γ with IC50 values of ∼900 and 1200 nM, respectively, indicating that isoform selectivity can be switched by improving polar interactions specific to the PI3Kα ATP-binding site.

Design strategy for compound 10.
Figure 4.
Design strategy for compound 10.

An alignment of compound 5 and 9. The units fused from each compound, leading to compound 10 are coloured red and blue. Predicted PI3Kα active-site contacts are indicated for compound 5 and those indicated for 9 are based on the PDB entry 4XZ4.

Figure 4.
Design strategy for compound 10.

An alignment of compound 5 and 9. The units fused from each compound, leading to compound 10 are coloured red and blue. Predicted PI3Kα active-site contacts are indicated for compound 5 and those indicated for 9 are based on the PDB entry 4XZ4.

Table 3
IC50 values for PI3Kγ inhibitors and hybrid compounds

Shown as mean ± SD (n ≤ 2).

Inhibitor Structure IC50 (nM) 
PI3Kα PI3Kβ PI3Kδ p110γ 
9  >12 500 1564 ± 267 4347 ± 724 11.5 ± 5.5 
10  992 ± 445 >12 500 8788 ± 1059 1210 ± 681 
11  >12 500 >12 500 6664 ± 80 2561 ± 53 
12  2836 ± 999 1351 ± 165 952 ± 139 604 ± 60 
Inhibitor Structure IC50 (nM) 
PI3Kα PI3Kβ PI3Kδ p110γ 
9  >12 500 1564 ± 267 4347 ± 724 11.5 ± 5.5 
10  992 ± 445 >12 500 8788 ± 1059 1210 ± 681 
11  >12 500 >12 500 6664 ± 80 2561 ± 53 
12  2836 ± 999 1351 ± 165 952 ± 139 604 ± 60 

Inhibition of cell signalling pathways in vitro

To investigate the effect of our new series of compounds on PI3Kα cell signalling, we determined the ability of compounds 2–8 to inhibit insulin-stimulated AKT phosphorylation on Thr308 (pAKT) as a proximal marker of PI3K pathway activation, and phosphorylation of S6 Ribosomal Protein (pS6) as a distal marker of pathway activation in cell lines harbouring two different PI3Kα oncogenic mutations. These included two with the H1047R mutation (SK-OV-3, T47D) and one with the E545K mutation (MCF-7). The data presented in Figure 5 confirm previous reports [17] that AKT phosphorylation and downstream S6 Ribosomal Protein phosphorylation in SK-OV-3 and T47D cell lines were more sensitive to PI3Kα inhibition by A66 than that in the MCF-7 cell line, while all three cell lines were sensitive to treatment with the pan-PI3K inhibitor GSK2126458 at 1 μM. The data also show that, at 10 μM, compounds 2, 3 and 4 did not have a clear effect on insulin-stimulated increases in pAKT and pS6 levels in any cell line. In comparison, compound 7 clearly decreased the insulin-stimulated pAKT and pS6 signal in the SK-OV-3 and T47D cell lines, while the effect of compound 5 was less clear. Compound 8 caused a clear decrease in pAKT levels in all three cell lines, while the effect on pS6 levels was less clear in the MCF7 cell line. Compound 6 was less effective at decreasing the pAKT signal compared with 5 and 7 in both SK-OV-3 and T47D cells, despite having similar IC50 values against PI3Kα in biochemical assays.

Effects of PI3Kα-selective and pan-PI3K inhibitors on the phosphorylation of AKT and S6 ribosomal protein in cell lines.

Figure 5.
Effects of PI3Kα-selective and pan-PI3K inhibitors on the phosphorylation of AKT and S6 ribosomal protein in cell lines.

Representative western blots showing levels of AKT and S6 phosphorylation (pAKT and pS6, respectively) in (A) SK-OV-3 (H1047R PI3Kα mutant), (B) T47D (H1047R PI3Kα mutant) and (C) MCF-7 (E545K PI3Kα mutant) cell lines. The cells were serum-starved overnight, treated with the inhibitor for 60 min [10 μM for compounds 28, 1 μM for A66, PIK-75 and GSK2126458 (GSK)] and then stimulated with 500 nM insulin for 15 min. The positive control (+ ctrl) are cells stimulated with insulin in the absence of inhibitor and the negative control (− ctrl) are cells without insulin stimulation or inhibitor treatment. The cells were lysed and analyzed for AKT phosphorylation at position Thr308 [pAKT (T308)], Total AKT, S6 phosphorylation and Total S6.

Figure 5.
Effects of PI3Kα-selective and pan-PI3K inhibitors on the phosphorylation of AKT and S6 ribosomal protein in cell lines.

Representative western blots showing levels of AKT and S6 phosphorylation (pAKT and pS6, respectively) in (A) SK-OV-3 (H1047R PI3Kα mutant), (B) T47D (H1047R PI3Kα mutant) and (C) MCF-7 (E545K PI3Kα mutant) cell lines. The cells were serum-starved overnight, treated with the inhibitor for 60 min [10 μM for compounds 28, 1 μM for A66, PIK-75 and GSK2126458 (GSK)] and then stimulated with 500 nM insulin for 15 min. The positive control (+ ctrl) are cells stimulated with insulin in the absence of inhibitor and the negative control (− ctrl) are cells without insulin stimulation or inhibitor treatment. The cells were lysed and analyzed for AKT phosphorylation at position Thr308 [pAKT (T308)], Total AKT, S6 phosphorylation and Total S6.

To compare the inhibitor potency in the cell-based assays relative to the IC50 values retrieved from the enzyme assay, SK-OV-3 and T47D cell lines were treated with the three most potent inhibitors: 5 at 3.4 and 34 μM, 7 at 2.1 and 21 μM, and 8 at 20 and 200 nM, which represent 10× and 100× their IC50 values against the isolated enzyme. The western blot data presented in Figure 6A,B show that inhibition of AKT phosphorylation by 5, 7 and A66 was more pronounced at 100× IC50 in both the SK-OV-3 and the T47D cell line. In comparison, compound 8 was the least efficacious compound tested from the new series in both cell lines. However, when the inhibitor incubation time was increased from 15 to 60 min, compound 8 showed some ability to decrease the pAKT signal at 200 nM (100× IC50, Figure 6C). From these data, it appears that, within our new series, compound 7 is the most effective compound and exhibits PI3Kα-selective signalling blockade in cell-based assays.

Effects of PI3Kα-selective and pan inhibitors at 10× or 100× their IC50 values on AKT phosphorylation in SK-OV-3 and T47D cell lines.

Figure 6.
Effects of PI3Kα-selective and pan inhibitors at 10× or 100× their IC50 values on AKT phosphorylation in SK-OV-3 and T47D cell lines.

Representative western blots showing levels of AKT phosphorylation at Thr308 [pAKT (T308)] in (A) SK-OV-3 (H1047R PI3Kα mutant) and (B) T47D (H1047R PI3Kα mutant). The cells were serum-starved overnight, treated with the inhibitor for 15 min and then stimulated with 500 nM insulin for 5 min. (C) Representative western blots showing AKT phosphorylation at Thr308 in SK-OV-3 cells when treated with the inhibitor for 60 min and stimulated with 500 nM insulin for 15 min. The positive control (+ ctrl) are cells stimulated with insulin in the absence of the inhibitor and negative control (− ctrl) are cells without insulin stimulation or inhibitor treatment.

Figure 6.
Effects of PI3Kα-selective and pan inhibitors at 10× or 100× their IC50 values on AKT phosphorylation in SK-OV-3 and T47D cell lines.

Representative western blots showing levels of AKT phosphorylation at Thr308 [pAKT (T308)] in (A) SK-OV-3 (H1047R PI3Kα mutant) and (B) T47D (H1047R PI3Kα mutant). The cells were serum-starved overnight, treated with the inhibitor for 15 min and then stimulated with 500 nM insulin for 5 min. (C) Representative western blots showing AKT phosphorylation at Thr308 in SK-OV-3 cells when treated with the inhibitor for 60 min and stimulated with 500 nM insulin for 15 min. The positive control (+ ctrl) are cells stimulated with insulin in the absence of the inhibitor and negative control (− ctrl) are cells without insulin stimulation or inhibitor treatment.

Selectivity for PI3Ks within a set of AKT-activating kinases

To characterize more comprehensively the kinase selectivity of this class within the PI3K–AKT signalling pathway, 5 and 7 were tested against a selection of kinases capable of phosphorylating AKT [3846]. This included a selection of receptor tyrosine kinases along with the insulin receptor, DNA-PK, mTOR and PDK-1 (Supplementary Table S10). Neither compound showed >50% inhibition when tested at 10 μM, with the exception of 7, which inhibited mTOR (FRAP1) by 67%.

Discussion

We have described the structure-guided design of a class of 3-pyridyl-substituted imidazopyridine PI3K inhibitors that demonstrated selectivity for the PI3Kα enzyme in biochemical assays, and also have a PI3K cell signalling inhibition profile consistent with other established PI3Kα-selective inhibitors. By modelling ligand binding in a PI3Kα crystal structure using molecular docking, we predicted that the selectivity of the new series is probably mediated by hydrogen bonds to the PI3Kα-specific amino acid Gln859. This interaction is also used to great effect by the clinically used inhibitor BYL719, while potency may be influenced by a water-mediated hydrogen bond.

Structure-guided inhibitor alignment allows the identification and recombination of features from different compound classes that perform the same function when interacting with the target with those that make different interactions [47,48]. The superimposition of compounds 1 and A66 illustrated that the inhibitors used both non-overlapping and shared ATP-binding pocket sub-sites. We have shown that recombining these different interactions by merging the pyrrolidine carboxamide feature onto the central imidazopyridine core of PIK-75 resulted in compounds capable of selectively inhibiting PI3Kα, albeit with low potency. It was also clear that the potency of PIK-75 is related to the hydrazine sulphonamide and nitroaromatic ring, and its possible function of binding to the Trp-shelf. Targeting a hydrogen bond interaction with Gln859 through a carboxamide group was also used to develop the selective PI3Kα inhibitor GDC-0326 [49] on a scaffold similar to the PI3Kβ-sparing inhibitor taselisib (GDC-0032) [50].

PI3K inhibitors based on a central pyridyl-imidazopyridine scaffold [51], like the series characterized in the present study, or a related pyridyl-triazolopyridine scaffold [5254] along with its benzothiazole-based analogues [55], as well as some built on a smaller imidazopyridine core [48], are known. Targeting the affinity pocket was successfully used to improve the potency for these examples and also for PI3Kα-selective inhibitors [12,13,15,29,30]. The types of interactions used have included ionic interactions with the catalytic Lys [35,48], leading to the discovery of GSK2126458 [35], hydrogen bond interactions with donors and acceptors present in the affinity pocket [51,52], as well as hydrophobic regions in the affinity pocket [29,30] and around the P-loop [12,56]. We were also able to improve activity across all four PI3K isoforms by adding a 3-pyridyl unit to the imidazopyridine core of 2. The new pyridyl-imidazopyridine compounds were less selective for the PI3Kα enzyme than A66, but more selective than PIK-75 and compound 1. This may be a function of isoform selectivity intrinsic to the scaffold. Bell et al. [52] reported that pIC50 values for PI3Kγ, PI3Kα and PI3Kδ enzymes were influenced by the pyridyl isomer on a related triazolopyridine scaffold, with the PI3Kγ enzyme more accepting a 3-pyridyl isomer than either PI3Kα or PI3Kδ, while PI3Kα and PI3Kγ were equally tolerant of a 4-pyridyl unit. We also noted that in the absence of isoform-specific interaction motifs, some compounds containing a 3-pyridyl unit or a similar motif were selective for p110γ over the other Class I enzymes.

X-ray crystal structure data for PI3K isoforms revealed that water molecules mediate interactions between the inhibitor and enzyme affinity pocket. Inhibitor scaffolds that project a 3-pyridyl unit into the p110γ affinity pocket were able to hydrogen bond to a water located at the base of the affinity pocket [35,37,5558]. In contrast, at least two water molecules forming a hydrogen bond network were observed in the PI3Kα affinity pocket with 4-pyridyl-containing small-molecule inhibitors bound. In these structures, the ligand interacted with the water network using a different water from that involved in 3-pyridyl ligand interactions in other enzymes [13,56]. Including water molecules in molecular docking is important for predicting correct protein–ligand interactions and hydration states, and the combination of predicted and X-ray structure-derived water sites with molecular docking was used successfully for ligand-binding prediction with different types of binding sites [36,59,60]. We found that a Superstar water map combined with several water molecules around the propensity peak of interest along with the GOLD water docking method [36] was able to predict an occupied water-binding site at the base of the affinity pocket. This is consistent with the sites observed in X-ray crystal structures for p110γ and PI3Kα proteins, and binding models for our 3-pyridyl-containing compound series also predicted an interaction with the water in this site. These predicted binding models, along with the biochemical data, raise questions about the role of water in isoform selectivity. Our data indicate that while 3-pyridyl units can be used to improve affinity for Class I PI3K enzymes, the water-mediated interaction proposed in the present study may be better suited to p110γ selectivity. Introduction of a methoxy-substituted 3-pyridyl into the scaffold showed only a small improvement in potency over the unsubstituted pyridyl in compound 5, which may be related to an interaction with the catalytic lysine. Furthermore, superimposition of the best models for compounds 7 and 8, onto the BYL719 bound-PI3Kα structure, indicated that the water network may be unable to form due to a steric effect from the methoxy group (Supplementary Figure S2). Further characterization of this new ligand series will require crystallographic analysis. Crystallographic analysis of the p110γ protein with analogues of buparlisib bound identified water-mediated interactions between the ligand and conserved amino acids in the ATP-binding site that affected the activity of these compounds [61]. Direct water-mediated interactions also contribute to the selectivity profile of the BCR-Abl inhibitor bosutinib [62]; however, water displacement was proposed to have a role in the preference of an analogue of the Abl inhibitor imatinib for the tyrosine kinase c-Kit [63].

The inclusion of a larger ionizable sulphonamide group showed a more dramatic increase in potency and also quenched the selectivity between isoforms. This leap in potency might be explained by the ionic interaction between the catalytic Lys and the ionizable group, and clearly illustrates that any potential interaction between the ligand and Gln859 can be overridden by interactions elsewhere in the ATP-binding site. An ionic interaction with the analogous catalytic Lys in the PI4KIIIβ lipid kinase could also drive a potency gain and could be combined with steric hindrance to achieve selectivity across a set of lipid kinases [64].

Cell signalling data showed that some of our new compounds were able to inhibit AKT phosphorylation in cell lines harbouring the H1047R oncogenic mutant, blocking the proximal part of the signalling pathway, leading to a decrease in phosphorylation at a distal pathway marker, S6 ribosomal protein, consistent with the in vivo effects of A66 on the pathway in xenograft studies [17]. When inhibitor efficacy was explored in more detail by comparing AKT phosphorylation blockade at concentrations related to enzyme IC50 values, the PI3Kα-selective inhibitor 7 was the most efficacious of all the compounds, and was more active in the H1047R-harbouring cell lines. This profile is in agreement with previous studies with the PI3Kα-selective compound A66 [17] and is supported by our single-point inhibition data for compounds 5 and 7 against a small set of kinases that affect AKT phosphorylation. The most potent compound in the series, compound 8, was one of the least effective in cell signalling assays. One striking finding was that at a high concentration, our pan-PI3K inhibitor compound 8 was able to block AKT phosphorylation in both H1047R and E545K harbouring cell lines consistent with other pan inhibitors GSK2126458 and PIK-75. Yet, unlike PIK-75 and GSK2126458, the blockade was not obvious in downstream signalling proteins in the MCF-7 cell line.

In summary, we have used a chemotype fusion strategy to design a series of inhibitors that combine features of two known PI3Kα inhibitors from different classes. This has illustrated that the potency of PIK-75 is dependent on the nitroaromatic sulphonamide and shown that a potential interaction with Gln859 also requires more complete occupancy of the affinity pocket to achieve potent compounds. Importantly, we have shown that isoform selectivity and potency is not only controlled by interactions with isoform-specific amino acids. Molecular docking and biochemical data also predicted the potential for water molecules bound in the affinity pocket to be targeted by ligand hydrogen bond acceptors. Finally, we show that not all compounds able to inhibit the lipid kinase activity of PI3Kα can also block PI3Kα signalling in cells, with an ionizable aromatic-sulphonamide unit, rendering the most potent compound within our series much less effective in cell-based assays. Taken together, our study describes a new strategy to define the factors that are important for developing selective and potent inhibitors of the PI3Kα enzyme.

Abbreviations

     
  • DMSO

    dimethylsulfoxide

  •  
  • DTT

    dithiothreitol

  •  
  • GOLD

    Genetic Optimization for Ligand Docking

  •  
  • IC50

    half maximal inhibitory concentration

  •  
  • PDB

    protein data bank

  •  
  • PI3Ks

    phosphoinositide 3-kinases

  •  
  • RMSD

    root mean square deviation.

Author Contribution

G.Q.G., C.M.B., P.R.S. and J.U.F. developed concepts and designed and supervised experiments. J.D.K., G.W.R. and W.A.D. designed and performed the chemical synthesis. G.Q.G., J.D.K., J.M.J.D. and G.W.R. performed the experiments. G.Q.G., J.U.F., C.M.B. and P.R.S. analyzed data. G.Q.G., J.U.F., P.R.S., J.D.K., G.W.R., C.M.B. and J.M.J.D. were the main contributors in the writing of the paper.

Funding

This work was funded by the Health Research Council of New Zealand [grant no. 13-763], the Maurice Wilkins Centre for Molecular Biodiscovery and the Cancer Society Auckland Northland.

Acknowledgments

The author(s) wish to acknowledge the contribution of National eScience Infrastructure (NeSI) high-performance computing facilities to the results of this research. New Zealand's national facilities are provided by the New Zealand eScience Infrastructure and funded jointly by NeSI's collaborator institutions and through the Ministry of Business, Innovation & Employment's Research Infrastructure programme.

Competing Interests

The Authors declare that there are no competing interests associated with the manuscript.

References

References
1
Backer
,
J.M.
(
2010
)
The regulation of class IA PI 3-kinases by inter-subunit interactions
.
Curr. Top. Microbiol. Immunol.
346
,
87
114
doi:
2
Burke
,
J.E.
and
Williams
,
R.L.
(
2015
)
Synergy in activating class I PI3Ks
.
Trends Biochem. Sci.
40
,
88
100
doi:
3
Flanagan
,
J.U.
and
Shepherd
,
P.R.
(
2014
)
Structure, function and inhibition of the phosphoinositide 3-kinase p110α enzyme
.
Biochem. Soc. Trans.
42
,
120
124
doi:
4
Katso
,
R.
,
Okkenhaug
,
K.
,
Ahmadi
,
K.
,
White
,
S.
,
Timms
,
J.
and
Waterfield
,
M.D.
(
2001
)
Cellular function of phosphoinositide 3-kinases: implications for development, immunity, homeostasis, and cancer
.
Annu. Rev. Cell Dev. Biol.
17
,
615
675
doi:
5
Vadas
,
O.
,
Burke
,
J.E.
,
Zhang
,
X.
,
Berndt
,
A.
and
Williams
,
R.L.
(
2011
)
Structural basis for activation and inhibition of class I phosphoinositide 3-kinases
.
Sci. Signal.
4
,
re2
doi:
6
Vanhaesebroeck
,
B.
,
Guillermet-Guibert
,
J.
,
Graupera
,
M.
and
Bilanges
,
B.
(
2010
)
The emerging mechanisms of isoform-specific PI3K signalling
.
Nat. Rev. Mol. Cell. Biol.
11
,
329
341
doi:
7
Vanhaesebroeck
,
B.
,
Stephens
,
L.
and
Hawkins
,
P.
(
2012
)
PI3K signalling: the path to discovery and understanding
.
Nat. Rev. Mol. Cell Biol.
13
,
195
203
doi:
8
Zhao
,
L.
and
Vogt
,
P.K.
(
2008
)
Class I PI3K in oncogenic cellular transformation
.
Oncogene
27
,
5486
5496
doi:
9
Rodon
,
J.
,
Dienstmann
,
R.
,
Serra
,
V.
and
Tabernero
,
J.
(
2013
)
Development of PI3K inhibitors: lessons learned from early clinical trials
.
Nat. Rev. Clin. Oncol.
10
,
143
153
doi:
10
Samuels
,
Y.
,
Wang
,
Z.
,
Bardelli
,
A.
,
Silliman
,
N.
,
Ptak
,
J.
,
Szabo
,
S.
et al. 
(
2004
)
High frequency of mutations of the PIK3CA gene in human cancers
.
Science
304
,
554
doi:
11
Janku
,
F.
,
Tsimberidou
,
A.M.
,
Garrido-Laguna
,
I.
,
Wang
,
X.
,
Luthra
,
R.
,
Hong
,
D.S.
et al. 
(
2011
)
PIK3CA mutations in patients with advanced cancers treated with PI3K/AKT/mTOR axis inhibitors
.
Mol. Cancer Ther.
10
,
558
565
doi:
12
Bruce
,
I.
,
Akhlaq
,
M.
,
Bloomfield
,
G.C.
,
Budd
,
E.
,
Cox
,
B.
,
Cuenoud
,
B.
et al. 
(
2012
)
Development of isoform selective PI3-kinase inhibitors as pharmacological tools for elucidating the PI3K pathway
.
Bioorg. Med. Chem. Lett.
22
,
5445
5450
doi:
13
Furet
,
P.
,
Guagnano
,
V.
,
Fairhurst
,
R.A.
,
Imbach-Weese
,
P.
,
Bruce
,
I.
,
Knapp
,
M.
et al. 
(
2013
)
Discovery of NVP-BYL719 a potent and selective phosphatidylinositol-3 kinase alpha inhibitor selected for clinical evaluation
.
Bioorg. Med. Chem. Lett.
23
,
3741
3748
doi:
14
Fritsch
,
C.
,
Huang
,
A.
,
Chatenay-Rivauday
,
C.
,
Schnell
,
C.
,
Reddy
,
A.
,
Liu
,
M.
et al. 
(
2014
)
Characterization of the novel and specific PI3Kα inhibitor NVP-BYL719 and development of the patient stratification strategy for clinical trials
.
Mol. Cancer Ther.
13
,
1117
1129
doi:
15
Fairhurst
,
R.A.
,
Imbach-Weese
,
P.
,
Gerspacher
,
M.
,
Caravatti
,
G.
,
Furet
,
P.
,
Zoller
,
T.
et al. 
(
2015
)
Identification and optimisation of a 4′,5-bisthiazole series of selective phosphatidylinositol-3 kinase alpha inhibitors
.
Bioorg. Med. Chem. Lett.
25
,
3569
3574
doi:
16
Sun
,
M.
,
Hillmann
,
P.
,
Hofmann
,
B.T.
,
Hart
,
J.R.
and
Vogt
,
P.K.
(
2010
)
Cancer-derived mutations in the regulatory subunit p85α of phosphoinositide 3-kinase function through the catalytic subunit p110α
.
Proc. Natl Acad. Sci. U.S.A.
107
,
15547
15552
doi:
17
Jamieson
,
S.
,
Flanagan
,
J.U.
,
Kolekar
,
S.
,
Buchanan
,
C.
,
Kendall
,
J.D.
,
Lee
,
W.-J.
et al. 
(
2011
)
A drug targeting only p110α can block phosphoinositide 3-kinase signalling and tumour growth in certain cell types
.
Biochem. J.
438
,
53
62
doi:
18
Costa
,
C.
,
Ebi
,
H.
,
Martini
,
M.
,
Beausoleil
,
S.A.
,
Faber
,
A.C.
,
Jakubik
,
C.T.
et al. 
(
2015
)
Measurement of PIP3 levels reveals an unexpected role for p110β in early adaptive responses to p110α-specific inhibitors in luminal breast cancer
.
Cancer Cell
27
,
97
108
doi:
19
Elkabets
,
M.
,
Vora
,
S.
,
Juric
,
D.
,
Morse
,
N.
,
Mino-Kenudson
,
M.
,
Muranen
,
T.
et al. 
(
2013
)
mTORC1 inhibition is required for sensitivity to PI3K p110α inhibitors in PIK3CA-mutant breast cancer
.
Sci. Transl. Med.
5
,
196ra199
doi:
20
Berndt
,
A.
,
Miller
,
S.
,
Williams
,
O.
,
Le
,
D.D.
,
Houseman
,
B.T.
,
Pacold
,
J.I.
et al. 
(
2010
)
The p110δ structure: mechanisms for selectivity and potency of new PI(3)K inhibitors
.
Nat. Chem. Biol.
6
,
117
124
doi:
21
Hon
,
W.-C.
,
Berndt
,
A.
and
Williams
,
R.L.
(
2012
)
Regulation of lipid binding underlies the activation mechanism of class IA PI3-kinases
.
Oncogene
31
,
3655
3666
doi:
22
Knight
,
Z.A.
,
Gonzalez
,
B.
,
Feldman
,
M.E.
,
Zunder
,
E.R.
,
Goldenberg
,
D.D.
,
Williams
,
O.
et al. 
(
2006
)
A pharmacological map of the PI3-K family defines a role for p110α in insulin signaling
.
Cell
125
,
733
747
doi:
23
Zheng
,
Z.
,
Amran
,
S.I.
,
Thompson
,
P.E.
and
Jennings
,
I.G.
(
2011
)
Isoform-selective inhibition of phosphoinositide 3-kinase: identification of a new region of nonconserved amino acids critical for p110α inhibition
.
Mol. Pharmacol.
80
,
657
664
doi:
24
Hayakawa
,
M.
,
Kawaguchi
,
K.-i.
,
Kaizawa
,
H.
,
Koizumi
,
T.
,
Ohishi
,
T.
,
Yamano
,
M.
et al. 
(
2007
)
Synthesis and biological evaluation of sulfonylhydrazone-substituted imidazo[1,2-a]pyridines as novel PI3 kinase p110α inhibitors
.
Bioorg. Med. Chem.
15
,
5837
5844
doi:
25
Kendall
,
J.D.
,
Giddens
,
A.C.
,
Tsang
,
K.Y.
,
Frédérick
,
R.
,
Marshall
,
E.S.
,
Singh
,
R.
et al. 
(
2012
)
Novel pyrazolo[1,5-a]pyridines as p110α-selective PI3 kinase inhibitors: exploring the benzenesulfonohydrazide SAR
.
Bioorg. Med. Chem.
20
,
58
68
doi:
26
Liu
,
J.-L.
,
Gao
,
G.-R.
,
Zhang
,
X.
,
Cao
,
S.-F.
,
Guo
,
C.-L.
,
Wang
,
X.
et al. 
(
2014
)
DW09849, a selective phosphatidylinositol 3-kinase (PI3K) inhibitor, prevents PI3K signaling and preferentially inhibits proliferation of cells containing the oncogenic mutation p110α (H1047R)
.
J. Pharmacol. Exp. Ther.
348
,
432
441
doi:
27
Nacht
,
M.
,
Qiao
,
L.
,
Sheets
,
M.P.
,
St Martin
,
T.
,
Labenski
,
M.
,
Mazdiyasni
,
H.
et al. 
(
2013
)
Discovery of a potent and isoform-selective targeted covalent inhibitor of the lipid kinase PI3Kα
.
J. Med. Chem.
56
,
712
721
doi:
28
Kendall
,
J.D.
,
O'Connor
,
P.D.
,
Marshall
,
A.J.
,
Frédérick
,
R.
,
Marshall
,
E.S.
,
Lill
,
C.L.
et al. 
(
2012
)
Discovery of pyrazolo[1,5-a]pyridines as p110α-selective PI3 kinase inhibitors
.
Bioorg. Med. Chem.
20
,
69
85
doi:
29
Fairhurst
,
R.A.
,
Gerspacher
,
M.
,
Imbach-Weese
,
P.
,
Mah
,
R.
,
Caravatti
,
G.
,
Furet
,
P.
et al. 
(
2015
)
Identification and optimisation of 4,5-dihydrobenzo[1,2-d:3,4-d]bisthiazole and 4,5-dihydrothiazolo[4,5-h]quinazoline series of selective phosphatidylinositol-3 kinase alpha inhibitors
.
Bioorg. Med. Chem. Lett.
25
,
3575
3581
doi:
30
Gerspacher
,
M.
,
Fairhurst
,
R.A.
,
Mah
,
R.
,
Roehn-Carnemolla
,
E.
,
Furet
,
P.
,
Fritsch
,
C.
et al. 
(
2015
)
Discovery of a novel tricyclic 4H-thiazolo[5',4':4,5]pyrano[2,3-c]pyridine-2-amino scaffold and its application in a PI3Kα inhibitor with high PI3K isoform selectivity and potent cellular activity
.
Bioorg. Med. Chem. Lett.
25
,
3582
3584
doi:
31
Zheng
,
Z.
,
Amran
,
S.I.
,
Zhu
,
J.
,
Schmidt-Kittler
,
O.
,
Kinzler
,
K.W.
,
Vogelstein
,
B.
et al. 
(
2012
)
Definition of the binding mode of a new class of phosphoinositide 3-kinase α-selective inhibitors using in vitro mutagenesis of non-conserved amino acids and kinetic analysis
.
Biochem. J.
444
,
529
535
doi:
32
Jones
,
G.
,
Willett
,
P.
,
Glen
,
R.C.
,
Leach
,
A.R.
and
Taylor
,
R.
(
1997
)
Development and validation of a genetic algorithm for flexible docking
.
J. Mol. Biol.
267
,
727
748
doi:
33
Chen
,
V.B.
,
Arendall
, III,
W.B.
,
Headd
,
J.J.
,
Keedy
,
D.A.
,
Immormino
,
R.M.
,
Kapral
,
G.J.
et al. 
(
2010
)
Molprobity: all-atom structure validation for macromolecular crystallography
.
Acta Crystallogr. D. Biol. Crystallogr.
66
,
12
21
doi:
34
Verdonk
,
M.L.
,
Berdini
,
V.
,
Hartshorn
,
M.J.
,
Mooij
,
W.T.M.
,
Murray
,
C.W.
,
Taylor
,
R.D.
et al. 
(
2004
)
Virtual screening using protein-ligand docking: avoiding artificial enrichment
.
J. Chem. Inf. Comput. Sci.
44
,
793
806
doi:
35
Knight
,
S.D.
,
Adams
,
N.D.
,
Burgess
,
J.L.
,
Chaudhari
,
A.M.
,
Darcy
,
M.G.
,
Donatelli
,
C.A.
et al. 
(
2010
)
Discovery of GSK2126458, a highly potent inhibitor of PI3K and the mammalian target of rapamycin
.
ACS Med. Chem. Lett.
1
,
39
43
doi:
36
Verdonk
,
M.L.
,
Chessari
,
G.
,
Cole
,
J.C.
,
Hartshorn
,
M.J.
,
Murray
,
C.W.
,
Nissink
,
J.W.M.
et al. 
(
2005
)
Modeling water molecules in protein-ligand docking using GOLD
.
J. Med. Chem.
48
,
6504
6515
doi:
37
Collier
,
P.N.
,
Messersmith
,
D.
,
Le Tiran
,
A.
,
Bandarage
,
U.K.
,
Boucher
,
C.
,
Come
,
J.
et al. 
(
2015
)
Discovery of highly isoform selective thiazolopiperidine inhibitors of phosphoinositide 3-kinase γ
.
J. Med. Chem.
58
,
5684
5688
doi:
38
Mahajan
,
K.
and
Mahajan
,
N.P.
(
2015
)
ACK1/TNK2 tyrosine kinase: molecular signaling and evolving role in cancers
.
Oncogene
34
,
4162
4167
doi:
39
Leroy
,
C.
,
Ramos
,
P.
,
Cornille
,
K.
,
Bonenfant
,
D.
,
Fritsch
,
C.
,
Voshol
,
H.
et al. 
(
2016
)
Activation of IGF1R/p110β/AKT/mTOR confers resistance to α-specific PI3K inhibition
.
Breast Cancer Res.
18
,
1
13
doi:
40
Mackenzie
,
R.
and
Elliott
,
B.
(
2014
)
Akt/PKB activation and insulin signaling: a novel insulin signaling pathway in the treatment of type 2 diabetes
.
Diabetes Metab. Syndr. Obes.: Targets Ther.
7
,
55
64
doi:
41
Yao
,
J.-J.
,
Sun
,
J.
,
Zhao
,
Q.-R.
,
Wang
,
C.-Y.
and
Mei
,
Y.-A.
(
2013
)
Neuregulin-1/ErbB4 signaling regulates Kv4.2-mediated transient outward K+ current through the Akt/mTOR pathway
.
Am. J. Physiol. Cell. Physiol.
305
,
C197
C206
doi:
42
Taniyama
,
Y.
,
Ushio-Fukai
,
M.
,
Hitomi
,
H.
,
Rocic
,
P.
,
Kingsley
,
M.J.
,
Pfahnl
,
C.
et al. 
(
2004
)
Role of p38 MAPK and MAPKAPK-2 in angiotensin II-induced Akt activation in vascular smooth muscle cells
.
Am. J. Physiol. Cell Physiol.
287
,
C494
C499
doi:
43
LoPiccolo
,
J.
,
Blumenthal
,
G.M.
,
Bernstein
,
W.B.
and
Dennis
,
P.A.
(
2008
)
Targeting the PI3K/Akt/mTOR pathway: effective combinations and clinical considerations
.
Drug Resist. Updat.
11
,
32
50
doi:
44
Song
,
G.
,
Ouyang
,
G.
and
Bao
,
S.
(
2005
)
The activation of Akt/PKB signaling pathway and cell survival
.
J. Cell Mol. Med.
9
,
59
71
doi:
45
Carmona
,
F.J.
,
Montemurro
,
F.
,
Kannan
,
S.
,
Rossi
,
V.
,
Verma
,
C.
,
Baselga
,
J.
et al. 
(
2016
)
AKT signaling in ERBB2-amplified breast cancer
.
Pharmacol. Ther.
158
,
63
70
doi:
46
Stronach
,
E.A.
,
Chen
,
M.
,
Maginn
,
E.N.
,
Agarwal
,
R.
,
Mills
,
G.B.
,
Wasan
,
H.
et al. 
(
2011
)
DNA-PK mediates AKT activation and apoptosis inhibition in clinically acquired platinum resistance
.
Neoplasia
13
,
1069
1080
doi:
47
Brough
,
P.A.
,
Barril
,
X.
,
Borgognoni
,
J.
,
Chene
,
P.
,
Davies
,
N.G.M.
,
Davis
,
B.
et al. 
(
2009
)
Combining hit identification strategies: fragment-based and in silico approaches to orally active 2-aminothieno[2,3-d]pyrimidine inhibitors of the Hsp90 molecular chaperone
.
J. Med. Chem.
52
,
4794
4809
doi:
48
Hughes
,
S.J.
,
Millan
,
D.S.
,
Kilty
,
I.C.
,
Lewthwaite
,
R.A.
,
Mathias
,
J.P.
,
O'Reilly
,
M.A.
et al. 
(
2011
)
Fragment based discovery of a novel and selective PI3 kinase inhibitor
.
Bioorg. Med. Chem. Lett.
21
,
6586
6590
doi:
49
Heffron
,
T.P.
,
Heald
,
R.A.
,
Ndubaku
,
C.
,
Wei
,
B.
,
Augistin
,
M.
,
Do
,
S.
et al. 
(
2016
)
The rational design of selective benzoxazepin inhibitors of the α-isoform of phosphoinositide 3-kinase culminating in the identification of (S)-2-((2-(1-isopropyl-1H-1,2,4-triazol-5-yl)-5,6-dihydrobenzo[f]imidazo[1,2-d][1,4]oxazepin-9-yl)oxy)propanamide (GDC-0326)
.
J. Med. Chem.
59
,
985
1002
doi:
50
Ndubaku
,
C.O.
,
Heffron
,
T.P.
,
Staben
,
S.T.
,
Baumgardner
,
M.
,
Blaquiere
,
N.
,
Bradley
,
E.
et al. 
(
2013
)
Discovery of 2-{3-[2-(1-isopropyl-3-methyl-1H-1,2-4-triazol-5-yl)-5,6-dihydrobenzo[f]imidazo[1,2-d][1,4]oxazepin-9-yl]-1H-pyrazol-1-yl}-2-methylpropanamide (GDC-0032): a β-sparing phosphoinositide 3-kinase inhibitor with high unbound exposure and robust in vivo antitumor activity
.
J. Med. Chem.
56
,
4597
4610
doi:
51
Pecchi
,
S.
,
Ni
,
Z.-J.
,
Han
,
W.
,
Smith
,
A.
,
Lan
,
J.
,
Burger
,
M.
et al. 
(
2013
)
Structure guided optimization of a fragment hit to imidazopyridine inhibitors of PI3K
.
Bioorg. Med. Chem. Lett.
23
,
4652
4656
doi:
52
Bell
,
K.
,
Sunose
,
M.
,
Ellard
,
K.
,
Cansfield
,
A.
,
Taylor
,
J.
,
Miller
,
W.
et al. 
(
2012
)
SAR studies around a series of triazolopyridines as potent and selective PI3Kγ inhibitors
.
Bioorg. Med. Chem. Lett.
22
,
5257
5263
doi:
53
Bergamini
,
G.
,
Bell
,
K.
,
Shimamura
,
S.
,
Werner
,
T.
,
Cansfield
,
A.
,
Muller
,
K.
et al. 
(
2012
)
A selective inhibitor reveals PI3Kγ dependence of TH17 cell differentiation
.
Nat. Chem. Biol.
8
,
576
582
doi:
54
Sunose
,
M.
,
Bell
,
K.
,
Ellard
,
K.
,
Bergamini
,
G.
,
Neubauer
,
G.
,
Werner
,
T.
et al. 
(
2012
)
Discovery of 5-(2-amino-[1,2,4]triazolo[1,5-a]pyridin-7-yl)-N-(tert-butyl)pyridine-3-sulfonami de (CZC24758), as a potent, orally bioavailable and selective inhibitor of PI3K for the treatment of inflammatory disease
.
Bioorg. Med. Chem. Lett.
22
,
4613
4618
doi:
55
Collier
,
P.N.
,
Martinez-Botella
,
G.
,
Cornebise
,
M.
,
Cottrell
,
K.M.
,
Doran
,
J.D.
,
Griffith
,
J.P.
et al. 
(
2015
)
Structural basis for isoform selectivity in a class of benzothiazole inhibitors of phosphoinositide 3-kinase γ
.
J. Med. Chem.
58
,
517
521
doi:
56
Han
,
W.
,
Menezes
,
D.L.
,
Xu
,
Y.
,
Knapp
,
M.S.
,
Elling
,
R.
,
Burger
,
M.T.
et al. 
(
2016
)
Discovery of imidazo[1,2-a]-pyridine inhibitors of pan-PI3 kinases that are efficacious in a mouse xenograft model
.
Bioorg. Med. Chem. Lett.
26
,
742
746
doi:
57
D'Angelo
,
N.D.
,
Kim
,
T.-S.
,
Andrews
,
K.
,
Booker
,
S.K.
,
Caenepeel
,
S.
,
Chen
,
K.
et al. 
(
2011
)
Discovery and optimization of a series of benzothiazole phosphoinositide 3-kinase (PI3K)/mammalian target of rapamycin (mTOR) dual inhibitors
.
J. Med. Chem.
54
,
1789
1811
doi:
58
Smith
,
A.L.
,
D'Angelo
,
N.D.
,
Bo
,
Y.Y.
,
Booker
,
S.K.
,
Cee
,
V.J.
,
Herberich
,
B.
et al. 
(
2012
)
Structure-based design of a novel series of potent, selective inhibitors of the class I phosphatidylinositol 3-kinases
.
J. Med. Chem.
55
,
5188
5219
doi:
59
Cappel
,
D.
,
Wahlström
,
R.
,
Brenk
,
R.
and
Sotriffer
,
C.A.
(
2011
)
Probing the dynamic nature of water molecules and their influences on ligand binding in a model binding site
.
J. Chem. Inf. Model.
51
,
2581
2594
doi:
60
de Graaf
,
C.
,
Pospisil
,
P.
,
Pos
,
W.
,
Folkers
,
G.
and
Vermeulen
,
N.P.E.
(
2005
)
Binding mode prediction of cytochrome P450 and thymidine kinase protein–ligand complexes by consideration of water and rescoring in automated docking
.
J. Med. Chem.
48
,
2308
2318
doi:
61
Bohnacker
,
T.
,
Prota
,
A.E.
,
Beaufils
,
F.
,
Burke
,
J.E.
,
Melone
,
A.
,
Inglis
,
A.J.
et al. 
(
2017
)
Deconvolution of Buparlisib's mechanism of action defines specific PI3K and tubulin inhibitors for therapeutic intervention
.
Nat. Commun.
8
,
14683
doi:
62
Levinson
,
N.M.
and
Boxer
,
S.G.
(
2013
)
A conserved water-mediated hydrogen bond network defines bosutinib's kinase selectivity
.
Nat. Chem. Biol.
10
,
127
132
doi:
63
Robinson
,
D.D.
,
Sherman
,
W.
and
Farid
,
R.
(
2010
)
Understanding kinase selectivity through energetic analysis of binding site waters
.
ChemMedChem.
5
,
618
627
doi:
64
Rutaganira
,
F.U.
,
Fowler
,
M.L.
,
McPhail
,
J.A.
,
Gelman
,
M.A.
,
Nguyen
,
K.
,
Xiong
,
A.
et al. 
(
2016
)
Design and structural characterization of potent and selective inhibitors of phosphatidylinositol 4 kinase IIIβ
.
J. Med. Chem.
59
,
1830
1839
doi: