Human target validation of phosphoinositide 3-kinase (PI3K)β: effects on platelets and insulin sensitivity, using AZD6482 a novel PI3Kβ inhibitor


Sven Nylander, AstraZeneca R&D Mölndal, 431 83 Mölndal, Sweden.
Tel.: +46 31 7762149; fax: +46 31 7763761.


Nylander S, Kull B, Björkman JA, Ulvinge JC, Oakes N, Emanuelsson BM, Andersson M, Skärby T, Inghardt T, Fjellström O, Gustafsson D. Human target validation of phosphoinositide 3-kinase (PI3K)β:effects on platelets and insulin sensitivity, using AZD6482 a novel PI3Kβ inhibitor. J Thromb Haemost 2012; 10: 2127–36. See also Jackson SP, Schoenwaelder SM. Antithrombotic phosphoinositide 3-kinase β inhibitors in humans – a ‘shear’ delight! This issue, pp 2123–6.

Summary.  Background:  Based on in vitro and animal data, PI3Kβ is given an important role in platelet adhesion and aggregation but its role in insulin signaling is unclear.

Objective:  To strengthen the PI3Kβ target validation using the novel, short-acting inhibitor AZD6482.

Methods and results:  AZD6482 is a potent, selective and ATP competitive PI3Kβ inhibitor (IC50 0.01 μm). A maximal anti-platelet effect was achieved at 1 μm in the in vitro and ex vivo tests both in dog and in man. In dog, in vivo AZD6482 produced a complete anti-thrombotic effect without an increased bleeding time or blood loss. AZD6482 was well tolerated in healthy volunteers during a 3-h infusion. The ex vivo anti-platelet effect and minimal bleeding time prolongation in the dog model translated well to data obtained in healthy volunteers. AZD6482 inhibited insulin-induced human adipocyte glucose uptake in vitro (IC50 of 4.4 μm). In the euglycemic hyperinsulinemic clamp model, in rats, glucose infusion rate was not affected at 2.3 μm but reduced by about 60% at a plasma exposure of 27 μm. In man, the homeostasis model analysis (HOMA) index increased by about 10–20% at the highest plasma concentration of 5.3 μm.

Conclusions:  This is the first human target validation for PI3Kβ inhibition as anti-platelet therapy showing a mild and generalized antiplatelet effect attenuating but not completely inhibiting multiple signaling pathways with an impressive separation towards primary hemostasis. AZD6482 at ‘supratherapeutic’ plasma concentrations may attenuate insulin signaling, most likely through PI3Kα inhibition.


Phosphoinositide 3-kinases (PI3Ks) are lipid kinases that phosphorylate the D3 hydroxyl group of membrane phosphoinositides (PIs). They are divided into three distinct classes (Class I, II and III) based on their primary structure, mode of regulation and substrate specificity. Class I PI3Ks are heterodimers with a catalytic (p110α, β, γ and δ) and a regulatory subunit (p85 or p101). They are involved in signal transduction through tyrosine kinase receptors and G protein-coupled receptors and are responsible for agonist-induced production of the second messenger phosphatidylinositol (3,4,5)trisphosphate (PIP3) by phosphorylation of the 3-position in the inositol ring of phosphatidylinositol (4,5)bisphosphate (PIP2) using ATP as a phosphate donor [1]. PI3Kα and β are ubiquitously expressed whereas PI3Kδ is mainly expressed in leukocytes and PI3Kγ in leukocytes, platelets and cardiomyocytes.

PI3Ks have been attributed a central role in cell signaling from a number of receptors involved in proliferation, motility, immune system function etc. [2]. Insulin signaling is also modulated through PI3Ks, which promotes glucose uptake and lipogenesis. The literature indicates that the effect on insulin signaling is mediated via PI3Kα but a minor role for PI3Kβ cannot be excluded [3–6]. In spite of the ubiquitous distribution of PI3Kβ, its role in non-platelet cellular effects has not been discussed much apart from a recently described role in tumor cell proliferation [5–7].

Platelets contain all class I isoforms (α, β, γ and δ) although the levels of PI3Kδ are very low. The role of the individual PI3K isoforms in regulating platelet functional responses has started to be defined during the recent years and PI3Kβ has been attributed the most prominent role of the four. Available data describe a role for PI3Kβ in ADP-induced P2Y12 signaling, GPVI and integrin αIIbβ3 outside-in signaling during adhesion and aggregation [8–14]. Inhibition of PI3Kβ has been shown to inhibit thrombus formation in different in vivo models without significantly affecting primary hemostasis [8,14,15]. For a recent review on the roles and functions of PI3Kβ see Gratacap M-P 2011 [1].

AZD6482 is a novel isoform-selective inhibitor of PI3Kβ that blocks the interaction of the enzyme with ATP. It was originally developed by Jackson et al. as a racemic mixture named KN-309 [16]. The aim of the present study was to use AZD6482, the active enantiomer, as a tool compound and characterize it with regards to PI3K isoform selectivity, platelet aggregation, thrombus formation, bleeding and effects on insulin signaling. A special emphasis was given to the translational aspects and AZD6482 was evaluated in healthy volunteers to get an initial human target validation.

Materials and methods


AZD6482 Data S1. ((-) 2-[(R)-1-(7-Methyl-2-morpholin-4-yl-4-oxo-4H-pyrido[1,2-a]pyrimidin-9-yl)-ethylamino]-benzoic acid (also known as KIN-193) [17], AZ12502334 (the S-enantiomer), AZ12379678 (the racemate, KN-309) [18], AZ12649385 [19] and AZ12312604 [20] were from AstraZeneca (Table 1).

Table 1.   Chemical structures of used PI3K inhibitors Thumbnail image of

Methods: in vitro

Selectivity  PI3Kα/β/γ/δ enzyme isoform selectivity was evaluated in vitro using human recombinant enzyme and an AlphaScreen assay to measure PIP3 production. AZD6482 was assessed in functional assays against 88 protein kinases including the closely related PI3K-like protein kinases; DNA-dependent protein kinase (DNA-PK), ataxia telangiectasia mutated protein (ATM) and mammalian target of rapamycin (mTOR).

Akt phosphorylation  Inhibition of Akt phosphorylation during a 2-h incubation with AZD6482 was measured by a decrease in the number of cells (breast adenocarcinoma MDA-MB-468) detected by a Akt(Ser 473)-specific antibody.

Platelet aggregation assays  Citrate anti-coagulated blood was collected by venipuncture from healthy donors. Platelet-rich plasma (PRP) light transmission aggregometry (LTA) was evaluated using the PAP-8 (Biodata, Horsham, PA, USA) aggregometer and an in-house developed 96-well plate assay. Whole blood impedance aggregometry was evaluated using the Multiplate (Dynabyte, Munich, Germany) impedance aggregometer. Whole blood shear induced platelet adhesion and aggregation was evaluated using the cone and plate analyzer (CPA), Impact-R (DiaMed, Yokneaam, Israel).

Insulin-induced glucose uptake by human adipocytes  Primary human insulin sensitive adipocytes (from patients or obtained from Zen-Bio, Durham, NC, USA) were grown to confluence and insulin-starved for 24 h before insulin-induced uptake of radio-labeled glucose.

Methods: in vivo rat and dog

All animal experiments in this study were approved by the ethical committee for animal research at the University of Göteborg, Sweden.

Arterial thrombosis and bleeding  A modified Folts’ model [21,22] was used and anesthetized dogs received vehicle (saline) followed by consecutive doses of AZD6482 intravenously (i.v.) over 30-min periods (bolus 0.03–1.3 μg kg−1 and infusion 0.005–0.24 μg kg min−1). The main parameters were blood flow (cyclic flow reductions, CFRs), bleeding time, blood loss, ex vivo platelet aggregation (Multiplate) and drug exposure.

Insulin action  Overnight fasted rats were infused with three consecutive escalating doses of AZD6482 and three alternative PI3K inhibitors (n = 4/group) during 30 min. A blood sample at the end of each dose interval gave plasma exposure, as well as glucose and insulin levels, the product of which was used to calculate the relative homeostasis model analysis (HOMA) insulin resistance index. Based on the relative HOMA index vs. exposure data, two doses of AZD6482 were selected for more rigorous assessment of insulin action in rats under conditions of hyperinsulinemic–euglycemic glucose clamp [23]. The steady-state glucose infusion rate (GIR) needed to maintain euglycemia was used as an effect parameter and three groups of rats (n = 4–5/group) were studied: vehicle, low dose (1000 μg kg−1 + 50 μg kg min−1) and high dose (6000 μg kg−1 + 300 μg kg min−1) AZD6482.

Methods: healthy volunteers

Healthy male subjects gave informed consent for participation in the study. This study was performed in compliance with the Declaration of Helsinki, ICH/Good Clinical Practice and it was approved by the independent ‘Regional Ethics Committee’ (Gothenburg, Sweden).

This was a randomized, double-blind, placebo-controlled, single-dose escalation study conducted at a single center with 40 male volunteers (18–36 years) randomized into the study. Subjects were randomized to receive either AZD6482 or placebo at all study sessions and each individual participated in a maximum of two study sessions. There were 7 dose-escalation steps (cumulative 3-h i.v. dose of AZD6482 0.9, 4.5, 13.5, 40.4, 121.5, 364.5 and 455.8 mg). There was a full clinical examination 30 days prior to entry and a follow-up visit, 7–10 days after completion of the last study session with a physical examination, laboratory screen and questions about adverse events (AE). The main parameters measured in the study were cutaneous bleeding time (modified IVY-technique [24]), ex vivo platelet aggregation (ADP-induced PRP LTA [PAP-8], ADP and CD9 antibody-induced whole blood impedance aggregometry [Multiplate] and whole blood shear-induced platelet adhesion [Impact-R]), HOMA insulin resistance index ([Fasting Plasma Glucose (mm) * Fasting Serum Insulin (μU mL−1)]/22.5 [25]), blood pressure, heart rate, drug exposure and laboratory values. The plasma concentrations of AZD6482 were determined by protein precipitation followed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Pharmacokinetic analysis was carried out by non-compartmental calculations using WinNolin® Enterprise (Pharsight, Sunnyvale, CA, USA).

Data presentation  In vitro experiments and animal data are given as mean ± standard error of the mean (SEM) and data from healthy volunteers as mean values ± standard deviation (SD) or 95%, two-sided confidence intervals (CI). Statistical methods are given in the Results and Data S1.


In vitro studies

Potency and selectivity  AZD6482 is a potent PI3Kβ inhibitor (Table 2). The racemate AZ12379678 and the S-enantiomer AZ12502334 were about three-fold (IC50 0.03 μm, n = 2) and > 200-fold (IC50 2.3 μm, n = 3) less potent compared with the R-enantiomer AZD6482, respectively. The potency of AZD6482 decreased with increasing ATP concentrations evidenced by a linear increase in IC50 for all PI3K isoforms (Fig. 1A–C), which indicates an ATP-competitive mode of action. AZD6482 potently inhibited Akt(Ser 473) phosphorylation in adenocarcinoma MDA-MB-468 cells (IC50 0.04 μm, n = 14).

Table 2.   AZD6482 in vitro potency and selectivity vs. recombinant human enzyme activities
EnzymeIC50m)SEM N Selectivity ratio vs. PI3Kβ
  1. PI3Kα/β/γ/δ enzyme activity was evaluated in vitro using human recombinant enzyme and an AlphaScreen assay (measuring formation of PIP3). DNA-dependent protein kinase (DNA-PK) was evaluated in a functional assay with protein phosphorylation read out by ELISA.

Figure 1.

 Concentration response of AZD6482 on recombinant human PI3Kβ enzyme activity in the presence of different ATP substrate concentrations; 4 (▼), 16 (♦), 64 (■), 250 (▲) and 1000 μm (•) (A). IC50 values for AZD6482 vs. PI3Kβ calculated from data in Figure 1A plotted vs. ATP concentration (B). IC50 values for AZD6482 vs. PI3Kα (•)/γ(■)/δ(▲) plotted vs. ATP concentration (C). PI3Kα/β/γ/δ enzyme activity, PIP3 production, was evaluated in vitro using human recombinant enzyme and an AlphaScreen assay (mean ± standard error of the mean [SEM], n = 4).

The selectivity of AZD6482 was assessed against 88 protein kinases (DATA S1). The selective ratio was high (> 200-fold) with the exception of the other PI3K isoforms and one related protein kinase (DNA-PK), 8–108-fold (Table 2), whereas ATM and mTOR displayed high selectivity in spite of also being PI3K like.

Inhibition of platelet aggregation  The role of PI3Kβ in human platelet function was evaluated using a 96-well plate PRP assay (Fig. 2A–F). For all tested agonists (ADP, collagen and thrombin receptor activating peptide [TRAP]), AZD6482 inhibited platelet aggregation induced by low agonist concentrations, but this effect was overridden at higher agonist concentrations regardless of the AZD6482 concentration. TRAP-induced aggregation proved to be the least dependent on PI3Kβ. Maximal inhibition of all three agonist-induced aggregation responses was achieved at about 1 μm AZD6482. The maximal effect achieved was equivalent to complete PI3Kα/β/γ/δ inhibition by the pan PI3K inhibitor wortmannin [26], 5 μm, indicating that PI3Kβ is the main PI3K isoform in platelet function.

Figure 2.

 ADP (A, D), collagen (B, E) and thrombin receptor activating peptide (TRAP) (C, F)-induced aggregation (undiluted human PRP, 96-well plate assays). Agonist concentration response in the absence (+) and presence of 5 μm wortmannin (○) or 1 μm AZD6482 (Δ) (A–C). AZD6482 concentration response vs. ADP-induced aggregation, 1 (+), 5 (○) and 20 μm (Δ) (D); vs. collagen-induced aggregation, 0.3 (+), 1 (○) and 3 μg mL−1 (Δ) (E); vs. TRAP-induced aggregation, 5 (+), 10 (○) and 20 μm (Δ) (F). Data are mean, n = 4.

The human anti-platelet potency of AZD6482 was evaluated in a number of in vitro assays using several agonists (Table 3). The most potent effect was seen on shear induced platelet adhesion and aggregation which is visualized in Fig. 3. In this assay, 5 μm AZD6482 appears not to inhibit primary adhesion but has a strong effect on aggregate formation.

Table 3.   AZD6482 in vitro potency in human platelet aggregation assays
AssayIC50 (μm)SEM n
  1. LTA in PRP induced by ADP used the PAP-8 aggregometer. Whole blood impedance platelet aggregometry induced by ADP and Collagen was evaluated using the Multiplate aggregometer. Whole blood shear induced platelet adhesion and aggregation was evaluated using the CPA, Impact-R.

  2. SEM, standard error of the mean; LTA, light transmission aggregometry; CPA, cone and plate analyzer.

 5 μm ADP0.620.264
 20 μm ADP2.931.364
Impedance aggregometry
 6.4 μm ADP0.270.0428
 3.3 μg mL−1 Collagen0.210.103
 Average size0.120.054
Figure 3.

 Representative images of whole blood shear (1800 s−1)-induced platelet adhesion and aggregation from a representative cone and plate analyzer (ImpactR) experiment comparing vehicle, dimethylsulfoxide (DMSO) (A) or 5 μm AZD6482 (B). Plasma proteins including von Willebrand factor (VWF) and fibrinogen were allowed to adhere to the well surface by pre-incubating with whole blood for 15 s before applying the shear.

The racemate, AZ12379678, had about 25% the potency of AZD6482, close to the theoretical 50%, using ADP as an agonist (impedance aggregometry) whereas the S-enantiomer, AZ12502334, was inactive with no inhibition detected at concentrations ≤ 10 μm (data not shown).

AZD6482 was also evaluated in vitro in dog and rat blood using ADP-induced impedance aggregometry. AZD6482 was more potent in human blood than in dog and rat blood with IC50 values of 0.27, 1.4 and 1.8 μm in human, dog and rat blood, respectively. Protein binding in man, dog and rat was 8%, 10% and 11% free unbound, respectively.

Insulin-induced glucose uptake  AZD6482 concentration-dependently inhibited insulin-induced glucose uptake by human adipocytes in vitro with an IC50 value of 4.4 ± 0.8 μm (n = 4).

Animal studies

Arterial thrombosis and bleeding  AZD6482 (Fig. 4) induced a concentration-dependent anti-thrombotic effect in vivo in the dog (abolition of the CFRs) and the blood flow could be completely restored (mean EC50 0.35 μm). In parallel, a reduction of ADP-induced platelet aggregation ex vivo occurred (impedance aggregometry, mean EC50 0.13 μm). A more than 80% inhibition of thrombus formation and platelet aggregation was seen at plasma concentration of 1 μm as predicted from the in vitro platelet function data. At this plasma concentration there was no detectable increase in bleeding time or blood loss. A further nine-fold increase in drug exposure only induced a modest (1.7-fold) increase in bleeding time.

Figure 4.

 Effects of AZD6482 on arterial thrombosis (○), ex vivo ADP-induced platelet aggregation (Multiplate) (Δ), bleeding time (□) and blood loss (x) in the modified dog Folts’ model. The antithrombotic effect was measured as, % restoration of blood flow in the damaged stenosed femoral artery (mean ± SEM, n = 6).

Insulin action  A relative HOMA index of three times the baseline level, as indicated by the dashed line in Fig. 5A, was considered to represent the smallest reliable increase in insulin resistance: achieved at ∼0.3 μm for AZ12312604 and 10–30 μm for AZD6482 and AZ12379678. AZ12649385 exhibited the least propensity to deteriorate insulin action, staying below the minimal relative HOMA threshold up to the highest plasma drug level tested (39 μm). The plasma drug concentrations inducing three-fold increases in the relative HOMA index correlated with the in vitro IC50 values for inhibition of PI3Kα but not PI3Kβ as can be seen in Table 4. The insulin resistance effect of a high dose of AZD6482 was rapidly reversible with normalization of the HOMA index paralleling the decline in plasma drug concentration after cessation of an infusion (Fig. 5B).

Figure 5.

 Effects of four PI3K inhibitors, AZD6482 (▲), AZ12312604 (•), AZ12379678 (◊) and AZ12649385 (▪) on the relative homeostasis model analysis (HOMA) index in the rat (A, mean ± SEM, n = 4/group). Decline in AZD6482 plasma exposure (▪) and normalization of the increased relative HOMA index (○) vs. time after discontinuation of an infusion of AZD6482 in the rat model (B, n = 2/group).

Table 4.   Effect of AZD6482 and three alternative PI3K inhibitors in the relative HOMA index model in the rat
CompoundPlasma exposure at relative HOMA index of 3PI3Kα IC50m)PI3Kβ IC50m)
  1. Infusion of a compound in anesthetized rats with the measurement of glucose, insulin and drug concentration in plasma. The relative the homeostasis model analysis (HOMA) index is the product of glucose and insulin corrected for vehicle and time-dependent changes (n = 4/group).

AZ12312604≈ 0.3 μm0.010.11
AZD6482≈ 10 μm0.870.01
AZ12379678≈ 30 μm1.370.03
AZ12649385> 40 μm22.80.16

Hyperinsulinemic-euglycemic clamps in rats consisted of an initial baseline period (Clamp1) followed by a vehicle or AZD6482 infusion period (Clamp2). The results show that the animals of the various groups were successfully clamped at the target plasma glucose levels (Table 5). In the vehicle control group there was a tendency for GIR to be slightly higher during the Clamp2 vs. Clamp1 phases although this did not reach statistical significance (P = 0.057). AZD6482 at 300 μg kg min−1 reaching a plasma exposure of 26.8 ± 2.6 μm, but not at 50 μg kg–1 min–1 with a exposure of 2.3 ± 0.1 μm, drastically reduced GIR (P < 0.0001), in spite of increased total- and human-insulin concentrations.

Table 5.   Effects of AZD6482 in the hyperinsulinemic–euglycemic glucose clamp model in the rat
  GIR (μmol min−1 per kg)Glucose (mm)*Human insulin (nm)*Total insulin (nm)*AZD6482 (μm)*
  1. In anaesthetized rats the steady state glucose infusion rate (GIR) needed for euglycemia during infusion of human insulin was assessed during two consecutive clamps, the latter (Clamp2) during which vehicle or AZD6482 was infused (mean ± standard error of the mean [SEM], n = 4–5/group). *Plasma levels. Rat plus human insulin. P < 0.05 vs. Clamp1.

VehicleBasal8.1 ± 0.70.3 ± 0.1
Clamp1124 ± 48.4 ± 0.81.7 ± 0.12.2 ± 0.1
Clamp2138 ± 48.1 ± 0.71.8 ± 0.12.2 ± 0.10.00
AZD6482Basal7.6 ± 0.20.3 ± 0.1
50 μg kg min−1Clamp1105 ± 27.3 ± 0.21.8 ± 0.12.2 ± 0.1
Clamp2107 ± 17.2 ± 0.31.8 ± 0.12.2 ± 0.12.3 ± 0.1
AZD6482Basal7.7 ± 0.10.4 ± 0.1
300 μg kg min−1Clamp1106 ± 107.6 ± 0.31.9 ± 0.12.3 ± 0.1
Clamp238 ± 117.8 ± 0.52.3 ± 0.23.4 ± 0.526.8 ± 2.6

Heart rate and blood pressure  In response to administration of AZD6482 to rats in the relative HOMA index experiments the BP and HR were largely unchanged. In the hyperinsulinemic glucose clamp experiments, however, BP and HR remained unchanged during the initial control periods but after AZD6482 administration there was a clear, partly transient and dose-related increase in BP (approximate increases in BP were 6 and 15 mmHg at the 50 and 300 μg kg min−1 dose, respectively) but no change in HR.

Healthy volunteers

Study subjects  All subjects were males (Caucasians). They were well balanced with regard to demographic characteristics (placebo [n = 14] vs. AZD6482 [n = 26]; age [years] 23 ± 2 vs. 26 ± 4, weight [kg] 74 ± 10 vs. 79 ± 8, BMI [kg m−2] 23 ± 3 vs. 24 ± 2). One subject (in dose group 121.5 mg) was excluded from the per protocol analysis set. Appearance of a skin reaction on the infusion arm of this subject led to the decision to terminate the experiment. This subject received no other dose. In addition, PK/PD data is missing for one subject in the 364.5-mg dose as this subject had a low pre-dose hemoglobin level and was not dosed.

Pharmacokinetics  The plasma concentration of AZD6482 at the end of the infusion, the 3-h sample, referred to as the steady-state concentration (Css) (Fig. 6A), declined rapidly. The mean effective elimination half-life was calculated to approximately 20 min (range 5–43 min) in the four highest dose groups (calculation not possible for the lower doses due plasma levels below LLOQ 0.005 μm). The AUC 0-∞ and the Css increased in a dose-proportional manner. The CI for all dose groups was 71 ± 26 L h−1 ranging from 40 to 90 L h−1 with one outlier with a clearance of 226 L h−1. The median (range) Vss across the evaluated dose levels was 48 (40–68) L. Urine was collected (0–24 h) after the 121.5- and 364.5-mg doses and the renal clearance of AZD6482 was low, 0.09 ± 0.04 L h−1.

Figure 6.

 Data from healthy volunteers in response to a 3-h intravenous infusion of AZD6482. Plasma concentration vs. time after single ascending doses of AZD6482 (0.9 to 455.8mg) (A). Decline in AZD6482 plasma exposure (Δ) and normalization of ADP-induced impedance aggregation (○) (Multiplate) vs. time after discontinuation of an infusion of 121.5 mg (B). Within subject prolongation of bleeding time vs. plasma concentration of AZD6482 (C). The effects of placebo (○) and ascending doses of AZD6482 (13.5–455.8 mg) on the homeostasis model analysis (HOMA) index relative to predose value vs. time since the start of the experiment. Subjects fasted from 10 h before the start of the experiment until the end of the experiment (D). Increase in the HOMA index relative to the predose value vs. plasma concentration of AZD6482 (E). Values in (A), (B) and (D) are mean, n = 6/dose except 364.5 mg which is n = 12. The linear regression models in (C) and (E) were based on log-transformed data for both parameters.

Inhibition of platelet function  The effect on platelet function was evaluated ex vivo by three different tests: (i) light transmission aggregometry (LTA) in PRP induced by 5 and 20 μm ADP; (ii) impedance aggregometry in whole blood induced by 6.4 μm ADP and 2 μg mL−1 CD9 antibody; and (iii) shear induced platelet adhesion/aggregation in whole blood. Blood was sampled pre-dose and at 2 h after the start of the infusion. The calculated EC50 and maximal effect (Emax) from these ex vivo tests are listed in Table 6.

Table 6.   AZD6482 potency in platelet aggregation assays in healthy volunteers ex vivo
Assay E max (%)95% CIEC50m)95% CI
  1. Ex vivo platelet aggregation models were the same as those used and described in Table 3 except for CD9 antibody used as agonist in place of collagen. N = 7 for all doses except for the second highest dose, n = 12.

  2. LTA, light transmission aggregometry; CI, confidence interval; CPA, cone and plate analyzer.

 5 μm ADP10126–1760.31−0.27 to 0.89
 20 μm ADP5742–721.00.25–1.7
Impedance aggregometry
 6.5 μm ADP8675–980.200.11–0.29
 2 μg mL−1 anti-CD97264–810.090.04–0.14
 Average size6257–670.040.03–0.04
 Surface coverage6859–770.100.02–0.18

Blood samples were also collected at 4, 6, 8 and 24 h for ADP-induced impedance aggregometry in order to monitor the reversibility of the effect. Fig. 6B illustrates the decay of the platelet aggregation at the 121.5-mg dose level together with AZD6482 plasma exposure. Sixty minutes after the end of the infusion the exposure of AZD6482 had declined from about 1.2 to 0.1 μm and the inhibition of platelet aggregation from about 90% to 50% and after 3 h the plasma level was approximately 0.025 μm and the platelet function was normalized.

Bleeding time  Cutaneous bleeding time was evaluated for all but the two lowest doses of AZD6482. Bleeding time at 2 h after the start of the infusion (relative to the pre-dose value) vs. plasma concentration of AZD6482 is shown in Fig. 6C. A significant but shallow plasma concentration dependent relationship with prolongation of bleeding time was observed (P < 0.005). The highest observed plasma concentration (at 2 h after start of infusion) of AZD6482, 6.2 μm, resulted in a significant mean prolongation of bleeding time of 60% (33–92.3%, CI) relative to baseline.

Insulin action  The mean HOMA index relative to pre-dose vs. time after the start of the infusion, broken down by dose of study drug, are graphically depicted in Fig. 6D. There was a tendency for the three highest doses (121.5–455.8 mg) to increase the HOMA index as compared with the baseline. However, this possible effect was small and rapidly reversible with normalization 30 min after infusion stop. Fig. 6E shows that there is a significant (P < 0.05) but shallow plasma concentration-dependent relationship with increasing the HOMA index. However, the predicted increase is probably not > 10–20%, for the highest dose.

Safety  AEs were infrequent, mild and reversible. One subject on AZD6482 was discontinued as a result of itching and a rash at the infusion site. The subject showed full recovery within 1 h of discontinuation. Another subject on placebo reported pain at the infusion site 40 min before the start of the infusion and was discontinued owing to a hematoma. The most common AEs, during the study session in subjects on AZD6482 were: dry mouth and paresthesia (three subjects for each AE). In the AZD6482 group, two subjects experienced epistaxis during the study session and five subjects during the wash out period. The epistaxis among AZD6482-treated subjects occurred mostly (6/7) at time points after normalization of platelet function.

Clinical laboratory evaluation, blood pressure and electrocardiography  There were no clinically relevant treatment-related changes or trends in any laboratory variables in subjects on AZD6482 during the study and no obvious difference between the subjects in the AZD6482 and placebo group. Supine systolic and diastolic blood pressures were largely unchanged and similar in the groups. All subjects were considered to have electrocardiographs (ECGs) within the normal physiological range.


AZD6482 proved to be a potent inhibitor of the enzyme activity of recombinant human PI3Kβ. The activity of AZD6482 is enantiomer selective as AZD6482 is at least twice as potent compared with the racemate, AZ12379678, and > 200-fold as potent as the S-enantiomer AZ12502334. The selectivity ratio of AZD6482 against a variety of enzymes and receptors is high (> 200-fold) with the exception of PI3Kα/γ/δ (86-, 108- and 8-fold, respectively) and one related protein kinase DNA-PK (> 40-fold). This selectivity profile is similar to previously reported data [17]. Evaluation of potential off target effects in PI3Kβ null mice would have been optimal; however, we failed to establish these mice as a result of male infertility probably because of testicular atrophy. A better alternative not available to use is probably the megakaryocyte/platelet-specific PI3Kβ null mice and this should be evaluated in future studies. The suboptimal selectivity vs. DNA-PK is not surprising given that DNA-PK belong to a family of proteins that contain a catalytic domain with motifs that are typical of the PI3Ks. DNA-PK is involved in the repair of double-stranded DNA breaks and in V(D)J recombination, in which the T-cell receptor and immunoglobulin genes are rearranged [27]. Double-stranded DNA breaks also occur during cancer chemo- or radiotherapy and inhibition of DNA-PK has been proposed as a target for cancer therapy. PI3Kδ, together with PI3Kγ, regulate immune cell signaling and combined PI3Kδ/γ inhibition has been reported to be required for potent inflammatory marker suppression [28] this is why the limited selectivity of AZD6482 vs. PI3Kδ may be tolerable as PI3Kγ selectivity is high.

AZD6482 was shown to be a potent inhibitor of Akt(Ser473) phosphorylation in adenocarcinoma MDA-MB-468 cells confirming prior data in different cells lines in vitro as well as inhibition of Akt phosphorylation in vivo in xenograft tumors [17]. Data on Akt phosphorylation in platelets is lacking; however, AZD6482 was shown to be a potent inhibitor of platelet aggregation in multiple platelet function tests involving several receptors and signaling pathways. Maximal anti-platelet effect was achieved at and above 1 μm in the in vitro and ex vivo tests both in dog and man. Inhibition of PI3Kβ shifted the concentration-response curve of the tested agonists to the right yielding full inhibition at lower agonist concentrations. However, the effect of the PI3Kβ inhibition could be completely overridden at high agonist concentrations. The rightward shift proved to be largest for collagen followed by ADP whereas only a minor shift of TRAP-induced aggregation was observed. This fits well with the described role of PI3Kβ in collagen-induced GPVI and ADP-induced P2Y12 signaling [8–14]. AZD6482 showed the highest potency in the shear-induced platelet adhesion and aggregation assay on a surface of plasma proteins including fibrinogen and von Willebrand factor (VWF). As both fibrinogen and VWF will induce αIIbβ3 outside-in signaling and eventually platelet ADP release, this may fit with the described role of PI3Kβ in αIIbβ3 outside-in signaling together with a secondary role for ADP and P2Y12 [8,12,14].

The rapid decline in plasma exposure in man was paralleled by a rapid normalization of platelet function which was complete 180 min post end of infusion.

In vivo in dog, AZD6482 produced a complete anti-thrombotic effect without significantly compromising hemostasis as no increase in bleeding time or blood loss was seen at plasma exposure that achieved a full anti-thrombotic effect, 1 μm. More so there was only a limited effect on bleeding time (1.7-fold increase) at nine times this plasma exposure. In man at ‘therapeutic’ plasma concentration (≈ 1 μm) there was no increase in bleeding time while at 5–6 times this plasma exposure bleeding time was 1.6-fold increased. This finding supports previous published data in rat and mouse [8,12,14,15] with different PI3Kβ inhibitors as well data from PI3Kβ null mice. We are aware of only one report with pronounced increased bleeding [29]. The reason for the different data in this study could be because of methodological differences as proposed by the authors including the genetic background of mice and the time allowed to measure re-bleeding.

The anti-platelet effect measured ex vivo in dog directly mirrors the antithrombotic effect in vivo and full inhibition of both occurred at approximately 1 μm plasma exposure. Thus, the ADP-induced impedance aggregometry ex vivo platelet function test well predicts the in vivo anti-thrombotic efficacy of PI3Kβ inhibition by AZD6482. The ex vivo anti-platelet effect and the bleeding time prolongation in the dog model translated well to man as the corresponding data from the phase I study almost superimposed those from the dog when plotting the exposure vs. effect curves from both species together (Fig. 7). This is the first report on target validation of PI3Kβ inhibition as an anti-platelet therapy in man showing an impressive separation between the anti-platelet effect ex vivo in man and effects on primary hemostasis as evaluated by cutaneous bleeding time. Compared with prior reported separation data for P2Y12 antagonists using the same Folts dog model [21], as used here, the separation for AZD6482 is unique. A speculative explanation is that AZD6482 leaves the shear induced primary platelet aggregation intact but inhibits secondary platelet aggregation (Fig. 3). A contributing factor could also be that high agonist concentrations can override the platelet inhibition by AZD6482 (Fig. 2). It is also possible that, soluble agonist concentrations are higher in a bleeding cavity as agonists are not washed away to the same extent as in the intravascular compartment.

Figure 7.

 The effects of AZD6482 on inhibition of ex vivo whole blood ADP-induced impedance aggregation (Muliplate) in man (•) and dog (Δ) and fold increase (vs. vehicle and pre-dose, respectively) in bleeding time in man (x) and dog (□). Values are ± SEM, human data, n = 6/dose except 364.5 mg which is n = 12, dog data n = 6.

PI3Kα has been attributed a major role in insulin signaling in the literature but a minor role for PI3Kβ has not been completely rolled out [3–6]. Thus, a PI3Kβ inhibitor with a limited (86-fold) selectivity vs. PI3Kα such as AZD6482 may attenuate insulin signaling through PI3Kα inhibition but it cannot be excluded that PI3Kβ inhibition may contribute to the response. For these reasons the potential to influence insulin signaling was evaluated both in vitro using human adipocytes, in vivo in two rat models and in vivo in man. AZD6482 inhibited insulin-induced human adipocyte glucose uptake with a mean IC50 of 4.4 μm. In rats, AZD6482 increased the relative HOMA index by about nine times at a plasma exposure of about 20 μm with no significant effect at about 3 μm. In the more fine-tuned euglycemic hyperinsulinemic clamp model in rats, GIR was reduced by about 60% at a plasma exposure of 27 μm with no significant effect at 2.3 μm. These in vitro and animal in vivo data translated into a weak but significant plasma concentration-dependent relationship with the HOMA index in man. However, the estimated increase was minor and in the range of 10–20%, for the highest dose tested (455.8 mg) corresponding to a mean steady state plasma exposure of 5.7 μm. As for the disappearance of platelet inhibition the effect on the HOMA index in rat and man rapidly declined in parallel to the changes in plasma concentration. In the rat, compounds with a range of PI3Kα potency was evaluated and their ability to increase the relative HOMA index correlated with the IC50 values for inhibition of PI3Kα but not PI3Kβ (Table 4). There is a consistency in these data between animal and man. Taken together, it indicates that inhibition of PI3Kα and not PI3Kβ is primarily responsible for the effects on insulin signaling which is in line with previous reports. However, the effect of chronic exposure needs to be evaluated in future clinical trials. For instance, mice with catalytically inactive PI3Kβ have been shown to develop mild insulin resistance with age [5]. Efforts in our labs to discover orally active PI3Kβ inhibitors with further improved selectivity towards PI3Kα will be reported in a series of papers elsewhere [26].

In conclusion, AZD6482 is a potent and selective PI3Kβ inhibitor with a mild and generalized antiplatelet effect attenuating but not completely inhibiting multiple signaling pathways. A wide separation between the anti-thrombotic effect and bleeding in a dog model, translated well to man providing the first human target validation of PI3Kβ inhibition. AZD6482 was well tolerated in man as no significant drug-related AEs were found as a result of the 3-h infusion. AZD6482 had a short plasma half-life owing to a high clearance and a relatively small distribution volume. With this profile, AZD6482 may be useful as a parenteral antiplatelet agent in situations where a low bleeding risk is desirable such as during acute stroke and cardiothoracic surgery. During short-term infusion, any small effect on insulin signaling is of no clinical importance. Thus PI3Kβ is an attractive anti-platelet target and further clinical investigations with AZD6482 or other inhibitors are warranted.


Individual authorship contributions: S.N.; overall project leader and lead bioscientist, B.K.; in vitro PI3K, J.A.B.; Folts model, J.C.U.; in vitro platelet function, N.O.; Insulin signaling, B.M.E.; Human pharmacokinetics, M.A.; Statistics, T.I. and O.F.; Chemistry, T.S.; Study physician and DG; Senior scientist. S.N. and D.G. wrote the paper.

Other contributions: Brian Bryzinski, contribution to clinical study design, Therese Hagstedt, hyperinsulinemic-euglycemic glucose clamp. Helen Zachrisson, Folts model. Göran Wahlund, rat HOMA index. Malin Enerbäck, Britt-Marie Wissing, Agnes Leffler, Åke Jägervall, Fredrik Wågberg, Anna-Karin Asztély, and Helena Lindmark in vitro platelet function. Anita Dellsén, in vitro PI3K. Ninni Herling-Svensson and Annika Johansson, ex vivo platelet function and in vivo bleeding in healthy volunteers. Angela Menschik-Lundin, DNA-PK. Eva-Marie Andersson, Johanna Ehnebom and Tony Clementz, insulin induced glucose uptake.

Disclosure of Conflict of Interest

The authors are employees of AstraZeneca.