Phosphoinositide 3-kinase γ/δ inhibition limits infarct size after myocardial ischemia/reperfusion injury

Identification of PI3K Isoform-Selective Inhibitors.

Screening a novel family of pteridines for activity against class IA (α, β, δ) and IB (γ)PI3K isoforms identified panisoform inhibitors (e.g., TG100713), as well as more selective compounds [e.g., TG101110, which spares PI3Kα; supporting information (SI) Table 1]. The most promising compound, TG100-115, inhibited PI3Kγ and -δ (IC₅₀ values of 83 and 235 nM, respectively), whereas both PI3Kα and -β were relatively unaffected (IC₅₀ values >1 μM). As a gauge of general specificity, TG100-115 was also assayed against a 133 protein kinase panel, none of which were inhibited at IC₅₀ values <1 μM (SI Table 2).

Modeling studies indicated that the freedom of conformational rotation permitted by ring substituents governs isoform selectivity (SI Fig. 6). Although TG100-115, TG100713, and TG101110 all allow for substituent ring energy minima at 30–40°, each compound pays differing penalties as angles diverge from this range. TG100-115 exhibits the greatest barrier to rotation (most conformations ≫50 kcal) and consequentially is the most isoform selective, TG100713 with the most conformational flexibility (ring A being almost freely rotatable with maxima ≈5 kcal; 1 kcal = 4.18 kJ) is the least selective, and TG101110 occupies an intermediate position (several conformations ≤10 but some ≫50 kcal).

Regarding kinase isoforms, PI3Kγ is most tolerant of ring conformational variations followed by PI3Kδ, supporting a model in which these two kinases are more readily fit by all three compounds; PI3Kα and -β being less tolerant lockout TG100-115 and partially TG101110 (Fig. 1). For TG100-115, modeling supports that two strong hydrogen bonds can form with the PI3Kγ hinge region: the N-3 nitrogen of the pteridine ring serving as an acceptor to the backbone NH of Val-882 and the 2-primary amine group serving as donor to the same residue's backbone carbonyl. Additionally TG100-115's 4-amino group is positioned adjacent to the backbone carbonyl of Glu-880; however, bond formation geometry is poor. These hydrogen bond patterns are similar to those for ATP as seen in the published ATP/PI3Kγ complex 1E8X, supporting a model for TG100-115 binding within the catalytic domain (Fig. 1c). Interactions are also formed outside the ATP-binding site, where a pocket accommodates TG100-115's 6′ substituted metaphenol, and Asp-841 (on the kα3 helix at the pocket's rear) forms a hydrogen bond with the OH group.

To place these data in context, two reagents commonly used as PI3K inhibitors were also profiled (SI Table 1). LY294002 was revealed as a relatively weak inhibitor (IC₅₀ values against PI3Kδ and -β of 561 and 858 nM, respectively, and no meaningful activity against PI3Kα or -γ); wortmannin, by contrast, displayed activity in the same general potency range as TG100-115 (55–147 nM) but with a panisoform profile.

PI3Kγ/δ Inhibition Selectively Impacts Growth Factor Signaling.

PI3K isoforms can differently regulate cellular processes, as with mitogenesis falling under PI3Kα and -β control (18). We therefore compared the compounds profiled in SI Table 1 for their effects on cell proliferation. As expected, panisoform inhibitors (TG100713 and wortmannin) strongly inhibited EC proliferation (Fig. 2a). TG101110 and LY294002 displayed similar efficacy, suggesting that either PI3Kα or -β inhibition is sufficient for blocking mitogenesis. TG100-115, by contrast, had no effects on EC proliferation even at relatively high concentrations (up to 10 μM). Additionally, TG100-115 did not block VEGF-induced angiogenesis in vivo (using a Matrigel implant model; data not shown). In agreement with these data, this compound was also found to have no influence on VEGF-stimulated ERK phosphorylation (Fig. 2b), a signaling event in this growth factor's mitogenic pathway (25). It did, however, interrupt other VEGF signaling pathways, such as those that culminate in VE-cadherin phosphorylation, the culmination of a signaling cascade that underlies VEGF's proedema activity (26). These data indicate that, whereas the non-ERK-related pathway(s) leading to VE-cadherin phosphorylation depend upon intact PI3Kγ and/or -δ activity, those that control cell proliferation are not.

PI3Kγ/δ Inhibition Reduces Edema and Inflammation in Vivo.

Because VE-cadherin phosphorylation triggers a reduction in endothelial barrier function (26), TG100-115's ability to block this cellular event should translate into an antiedema effect. This was directly demonstrated in Miles assay studies, where this compound inhibited VEGF-induced vascular permeability (Fig. 3). TG100-115 also blocked histamine-induced permeability, as predicted based on PI3K's role in GPCR signaling. Varying the time between compound administration and agonist challenge demonstrated that a single i.v. administration reduced edema formation for at least 4 h (data not shown).

The Miles assay models a primarily endothelial-based response that develops rapidly (within minutes). As a comparator, we next used a rodent hindpaw model, where more complex inflammatory reactions develop over several hours. Agonists included PAF, a GPCR ligand that activates both EC and leukocytes, and dextran, a phagocytic stimulus for mast cell and leukocyte activation (27). Histology revealed quite clearly that TG100-115 strongly antagonized both the edema and leukocyte infiltration induced by these two mediators (Fig. 3 c–f). Paw volume (a more quantitative measure of edema and inflammation) was reduced by 62% and 78% in response to PAF or dextran, respectively (n = 8; P < 0.001). Finally, to correlate these in vivo responses with the molecular target of interest, we monitored PI3K pathway signaling through Western blot analyses of Akt phosphorylation (a PI3K-mediated event). VEGF injection i.v. in mice induced a rapid Akt phosphorylation readily detectable in lung lysates and, as expected, pretreatment with TG100-115 blocked this response (SI Fig. 7). Blockade was seen with TG100-115 doses as low as 0.5 mg/kg and persisted over a period of several hours.

PI3Kγ/δ Inhibition Limits Infarct Development and Improves Myocardial Functioning in Rodents.

We previously documented in rodents the vascular changes, such as edema and neutrophil activation, which contribute to infarct development (26). Given the antiinflammatory actions of our PI3Kγ/δ inhibitor, therefore, we tested this compound for possible cardioprotective activities. In a rodent model of MI, TG100-115 delivered as a single i.v. bolus 60 min after reperfusion routinely reduced infarct size by ≥40%, with maximal efficacy reached by a dose of 0.5 mg/kg (Fig. 4a. [This model initiates with a 60-min coronary artery occlusion followed by complete reperfusion and then infarct measurement at 24 h; in control animals, the ischemic zone typically covers 30–45% of the total left ventricle (LV) with 55–70% infarction of this area. Pilot studies revealed that longer ischemic periods do not produce larger infarcts and that infarcts do not reach a fixed size until ≈6 h after reperfusion.] Immunohistochemistry as well as EM revealed similar patterns of monocyte and neutrophil infiltration in hearts from TG100-115 vs. vehicle-treated animals (data not shown). Although it was obvious that infarcts were smaller in TG100-115 animals, inflammatory infiltrates were present to an equivalent degree in infarcted myocardium from both treatment groups. Inflammation was not detectable, by contrast, in viable tissue (i.e., myocardium not showing morphologic signs of cardiomyocyte or vascular damage). These observations are consistent, therefore, with an action by TG100-115 to reduce the overall area in which inflammation occurs and thus the final extent of infarction.

To better define the available therapeutic window for this cardioprotective effect, TG100-115 was administered at various times during the reperfusion period (Fig. 4b). Delivery (as a single bolus) between 0 and 3 h after reperfusion produced statistically significant but equivalent reductions in infarct development (in this case 58–67% reductions vs. vehicle-treated controls). Repeat dosing at both 0 and 3 h also provided equivalent efficacy, suggesting that a single administration is sufficient for maximal efficacy. Finally, to confirm the functional benefit of infarct reduction, myocardial contractility was assessed by echocardiography 4 weeks after infarct induction. The percent LV long axis fractional shortening observed in animals that had received vehicle placebo was 27.2 ± 1.9 (mean ± SEM), whereas animals dosed with TG100-115 as a single 0.5 mg/kg bolus 1 h after reperfusion had a 24% improvement in this measure (33.6 ± 2.0, n = 12–13; P = 0.03). These data therefore confirm a long-lasting functional benefit to PI3Kγ/δ inhibition during myocardial reperfusion injury.

PI3Kγ/δ Inhibition Limits Infarct Development in a Porcine MI Model.

In a final series of studies, MIs were modeled in the pig, because this species better approximates human coronary anatomy and responses to myocardial I/R injury. With aggressive model parameters of a 90-min ischemic period, therapeutic dosing 30 min after reperfusion, and infarct measurement at 24 h, we typically observed ischemic zones representing 20–30% of the total LV and 20–30% infarction of this area in control animals. In initial dose-ranging studies, generally equivalent responses were observed using TG100-115 doses of 0.5–10 mg/kg (data not shown), and we therefore elected to conduct a statistically powered test at the lowest dose. Animals dosed with TG100-115 as a single 0.5 mg/kg i.v. bolus 30 min after reperfusion developed smaller infarcts vs. vehicle-treated controls (Fig. 5). Measuring infarct area as percent of total LV ischemic area, infarct size was reduced by 35% (P = 0.04). Viable tissue within the ischemic zone was increased by 37% (P = 0.04), directly demonstrating the cardioprotective effect of PI3Kγ/δ inhibition. Finally, taking advantage of the larger porcine heart to make more detailed measurements, infarct areas within the LV free wall only (as opposed to the free plus attached wall measurements reported up to this point) were also determined, reasoning that free wall infarcts are of greatest relevance to overall myocardial function. As anticipated, TG100-115 reduced free wall infarct size by 38% vs. vehicle-treated controls (P = 0.05).

Kinase Inhibitors.

TG100-115 [3-[2,4-diamino-6-(3-hydroxyphenyl)pteridin-7-yl]phenol], TG100713 [3-(2,4-diamino-pteridin-6-yl)-phenol], and TG101110 [6-(1H-indol-4-yl)-pteridin-2,4-diamine] were designed by and synthesized at TargeGen; Wortmannin and LY294002 were purchased from Calbiochem (San Diego, CA). Compounds were prepared for in vitro assays as DMSO stocks and for in vivo assays as either PEG or sulfobutyl ether β-cyclodextrin (CyDex, Lenexa, KS) formulations.

Molecular Modeling.

Protein coordinates for human PI3Kγ were taken from RCSB Protein Data Bank entry 1E8Z; additional structures (1E7U, 1E7V, 1E8W, and 1E8Y) were superimposed to examine potential flexibility and binding modes. Compound structures were docked into PI3K active sites using interactive modeling and automated docking software, then minimized in a molecular mechanics program using the consistent force field (CFF) with a distance-dependent dielectric (Insight II and Discover; Accelrys, San Diego, CA). For optimization, all atoms were allowed movement in residues containing at least one atom within 8 Å of any ligand atom. Ribbon models were generated by using YASARA (Yasara Biosciences, Graz, Austria). Complexes were overlaid by using a method that superimposes corresponding carbon α atoms (43), and renderings generated by using Insight II.

Kinase Assays.

PI3K reactions were constructed by using recombinant human kinases, 3 μM ATP, phosphatidylinositol substrate, and cofactors, and reaction progression measured by using a luminescent-based detection system to quantify ATP consumption (see SI Supporting Text). Protein kinase assays were performed by using commercial screening services (Upstate Biotechnology, Charlottesville, VA; Invitrogen, Carlsbad, CA).

Western Blots.

Serum-starved human umbilical vein EC (Cambrex BioScience, Walkersville, MD) were treated for 15 min with test agents (10 μM) or vehicle, followed by a 5-min treatment with 20 ng/ml recombinant human VEGF (PeproTech, Rocky Hill, NJ) or vehicle. Lysates were then processed as Western blots to detect phosphorylated VE-cadherin and ERK1/2 or total ERK2 (see SI Supporting Text). For detection of in vivo phospho-Akt, BALB/c mice were first dosed i.v. with TG100-115 or vehicle and then injected i.v. at the desired time with 20 ng of VEGF or saline. Lungs were explanted after 5 min and processed to lysates, and Western blots performed as described.

Cell Proliferation Assays.

Human umbilical vein EC plated in 96-well cluster plates (5,000 cells/well) were cultured in assay medium (containing 0.5% serum and 50 ng/ml VEGF) in the presence or absence of test compounds (10 μM), and cell numbers were quantified by XTT assay (Promega, Madison, WI) 24, 48, or 72 h later.

Miles Assay.

All animal studies followed current NIH Guidelines for the Use of Laboratory Animals and were performed according to Institutional Animal Care and Use Committee-approved protocols. Sprague–Dawley rats (175–200 g) were dosed i.v. with either TG100-115 (1 mg/kg) or vehicle, and 1–4 h later Evans blue dye (Sigma, St. Louis, MO) was administered i.v. as 500 μl of a 2% sterile saline solution. Immediately after dye injection, animals were injected intradermally on each shaved flank with 100 μl of saline, VEGF (2 μg/ml stock), or histamine (10 μg/ml stock). Thirty minutes later, injection sites were photographed.

Hindpaw Inflammation.

Sprague–Dawley rats (275–300 g) were dosed i.v. with either TG100-115 (5 mg/kg) or vehicle and then injected 30 min later in the plantar hindlimb surface with 100 μl of dextran suspension (Sephadex G-75 suspended at 80 mg/ml in saline and autoclaved) or PAF (16 pM c-PAF stock; Biomol International, Plymouth Meeting, PA). Animals were administered buprenorphine (0.01 mg/kg, s.c.) to control pain, and 3 h later paw dimensions were measured to the nearest 0.1 mm using calipers. Volume was then calculated and expressed as mm³; injection sites were harvested and processed to H&E-stained paraffin sections.

Animal MI Models.

Rodent and porcine MI models are fully described in SI Supporting Text. Briefly, Sprague–Dawley rats were subjected to 60 min of complete left anterior descending coronary artery (LAD) occlusion followed by complete reperfusion; TG100-115 or vehicle was delivered i.v. at 0–3 h after reperfusion. Twenty-four hours later, LAD occlusion was repeated, animals were injected with Evans blue dye to delineate perfused from ischemic areas, and then heart sections were stained in triphenyltetrazolium chloride to delineate viable tissue from infarct and photographs were processed using an image analysis software program to quantify infarct area. Alternatively, LV function was assessed using echocardiography at 4 weeks by determining percent fractional shortening from long axis images (by personnel blinded as to treatment identity). For the porcine model, farm swine were subjected to 90 min of complete LAD occlusion followed by complete reperfusion; TG100-115 or vehicle was delivered i.v. 30 min after reperfusion, and infarct area was determined as in the rodent model at 24 h.

Statistics.

Two group comparisons were made using unpaired Student's t tests, and multiple group analyses using one-way ANOVA with post hoc Dunnett's tests (SigmaStat software; SPSS, Chicago, IL). If Dunnett's tests showed statistical differences for more than one group compared with controls, a second ANOVA for intergroup differences was performed using the Student–Newman–Keuls method as the post hoc test. Statistical significance was defined as P < 0.05.