Preferential Inhibition of Akt and Killing of Akt-Dependent Cancer Cells by Rationally Designed Phosphatidylinositol Ether Lipid Analogues

INTRODUCTION

The PI3k/Akt signal transduction pathway is activated by many types of cellular stimuli or toxic insults and once active, can regulate fundamental cellular functions such as transcription, translation, proliferation, growth, and survival (1 , 2) . The serine/threonine kinase Akt (or PKB) is a crucial kinase in this pathway. Akt was first described as the cellular homologue of the product of the v-akt oncogene (3, 4, 5) , and it has three isoforms, Akt1, Akt2, and Akt3 (or PKB-α, PKB-β, and PKB-γ). Activation of all three isoforms is similar in that phosphorylation of two sites, one in the activation domain and one in the COOH-terminal hydrophobic motif, are necessary for full activity. For Akt1, phosphorylation of Thr-308 in the activation domain by phosphoinositide-dependent kinase-1 (PDK-1) is dependent on the products of PI3k, phosphatidylinositol 3,4 bisphosphate (PIP₂), and phosphatidylinositol 3,4,5 trisphosphate (PIP₃; Refs. 1 and 6 ). Levels of PIP₂ and PIP₃ are controlled by the tumor suppressor, dual-phosphatase PTEN, whose tumor suppressor function is due to dephosphorylation of PIP₂ and PIP₃ at the 3′ position (7) . The mechanism of Ser-473 phosphorylation is less clear. Kinases potentially responsible for Ser-473 phosphorylation include PDK-1 (8) , integrin-linked kinase (ILK) or an ILK-associated kinase (9 , 10) , Akt itself (11) , or a plasma membrane lipid raft-associated kinase (12) . In addition, a recent study has suggested that tyrosine phosphorylation may also be important for Akt activation (13) .

Once active, Akt exerts many cellular effects through the phosphorylation of downstream substrates such as Bad, caspase 9, ASK1, and MDM2 that regulate the apoptotic machinery (14, 15, 16, 17, 18) ; the cdk inhibitors p21 and p27 that control cell cycle progression (19, 20, 21, 22) ; forkhead transcription family members and inhibitor of nuclear factor-κB kinases that control gene expression (23 , 24) ; the kinase glycogen synthase kinase (GSK)-3 that controls glucose metabolism, cell cycle, and apoptosis (25) ; and tuberin and the mammalian target of rapamycin that control protein translation (26, 27, 28, 29) . The cumulative effect of altering these cellular processes can promote transformation in vitro or tumorigenesis and tumor growth in vivo.

Many lines of evidence demonstrate that Akt is a critical player in the development, growth, and therapeutic resistance of cancers. First, studies in transgenic mice that overexpress constitutively active Akt show that Akt, in combination with other genetic alterations, contributes to tumor formation in many tissues (30, 31, 32) . Tumor formation is also characteristic of PTEN heterozygous knockout mice that have increased levels of active Akt (2 , 6) . Second, activation of Akt is an early response to carcinogen exposure in vitro and in vivo and may have a permissive role for the development of tobacco-related cancers (33) . Third, active Akt has been detected in over eight types of human cancers in vivo 3 and has been functionally linked with poor clinical outcomes (34, 35, 36) . Fourth, Akt activity promotes resistance to chemotherapy and radiation (37 , 38) . Collectively, these studies suggest that inhibiting the Akt pathway might have therapeutic value for patients with cancer, and they have formed the basis for widespread efforts to develop approaches that inhibit Akt.

Despite the potential value of inhibiting Akt and concerted efforts by industry and academia to develop Akt inhibitors, no small molecule Akt inhibitors exist. One approach has used molecular modeling of the interaction of PIP₂ with the pleckstrin homology (PH) domain of Akt to guide the synthesis of phosphatidylinositol ether lipid analogues (PIAs) that were designed to inhibit this interaction (39) . The most effective PIA from this set, DPIEL, contained a single 3-deoxy substitution on the inositol ring and inhibited platelet-derived growth factor-induced Akt activation and cell growth (40) . To improve potency and metabolic stability, we recently modified the inositol ring to create 2-modified 3-deoxy PIAs. Evaluation of these compounds showed that two 2-modified 3-deoxy PIAs were able to inhibit constitutive Akt phosphorylation in cancer cell lines but that DPIEL was ineffective (41) .

In this paper, we extend these studies and provide a biological analysis of a series of chemically modified PIAs that bear different sets of two substitutions on the inositol ring. We identify five PIAs that inhibit Akt and downstream signaling without inhibiting other kinases upstream of Akt, and we show that these compounds preferentially induce apoptosis in cancer cell lines with high levels of Akt activity. Our data support the concept of selectively targeting Akt to treat cancer and identify PIAs as lead compounds for possible development as small molecule Akt inhibitors.

MATERIALS AND METHODS

Materials

The synthesis of the PIAs has been described previously (41) . All antibodies (except anti-α-tubulin, anti-p85, and anti-cyclin D1) and the Akt kinase assay kit were purchased from Cell Signaling Technologies (Beverly, MA). Anti α-tubulin was purchased from Sigma Chemical Co. (St. Louis, MO). Antibodies against the p85 subunit of PI3k (anti-p85) were obtained from Upstate Biotechnology (Lake Placid, NY). Antibodies to cyclin D1 were from Santa Cruz Biotechnology (Santa Cruz, CA). Protease inhibitor mixture was obtained from Sigma Chemical Co., and protein assay materials were from Bio-Rad (Hercules, CA). All cell culture reagents were purchased from Life Technologies, Inc. (Rockville, MD). Protran pure nitrocellulose membranes were purchased from Schleicher & Schuell (Dassel, Germany). Adenoviruses containing β-galactosidase or myristolated-Akt1 were generous gifts from Dr. Kenneth Walsh (Boston University, Boston, MA) and have been described previously (42) .

Methods

Cell Culture.

NSCLC lines were provided by H. Oie or Dr. F. Kaye at NCI/Navy Medical Oncology (Bethesda, MD). All cell lines were maintained in 75-cm² flasks in DMEM and supplemented with 10% (v/v) fetal bovine serum (FBS), 100 units/ml penicillin, and 100 μg/ml streptomycin. Cells were incubated in a 37°C and 7.0% CO₂ atmosphere incubator. The stock cultures were split on a weekly basis at a 1:5 or 1:10 ratio. Breast cancer cell lines were obtained from Dr. S. Lipkowitz at the NCI/Navy Medical Oncology and were maintained in RPMI 1640 supplemented with 10% (v/v) FBS, 100 units/ml penicillin, and 100 μg/ml streptomycin in an incubator calibrated to 37°C and 6% CO₂ in 75-cm² flasks. Stock flasks were split on a weekly basis at a 1:4, 1:10, or 1:20 ratio. BEAS2B cells were provided by Dr. Curtis C. Harris (NCI, Bethesda, MD) and were grown in LHC8 medium (BioSource, Camerillo, CA) supplemented with 100 units/ml penicillin in an incubator at 37°C and 3.5% CO₂ in 75-cm² flasks. Stock flasks were split on a weekly basis at a 1:4 ratio.

Pharmacological Treatments.

NSCLC cells were plated 2–2.5 × 10⁵ cells/well in 6- or 12-well plates in DMEM containing 10% FBS and incubated for 24 h. The medium was then changed to DMEM with 0.1% FBS, and the cells were incubated overnight. After overnight incubation, cells were treated with PIAs dissolved in DMSO (10 μm) for 2 h (immunoblotting/kinase assays), 18 h (immunoblotting experiments), or 24 h (apoptosis studies). In all experiments, DMSO was added to control samples and had no effect on Akt activity. After incubation with PIAs, the cells were harvested for immunoblot analysis or for assessment of apoptosis as described below. For dose response studies, NSCLC cells were plated 2–2.5 × 10⁵ cells/well in 6-well plates. After attachment, the medium was changed to DMEM containing 0.1% FBS overnight. Cells were treated with the indicated doses of PIAs for 2 h, and the cells were harvested for immunoblot analysis. Quantification of band density was performed using NIH Image software. Dose curves were obtained by quantifying ratios of P-Ser-473 to total Akt under each condition and setting the ratio of P-Ser-473/total Akt for DMSO-treated cells equal to 1.

Immunoblotting.

Cell extracts were prepared by washing cells with PBS and adding 100 μl of 2× Laemmli sample buffer supplemented with 2 μl protease inhibitor mixture/well as described previously (43) . Lysates were sonicated for 15 s with a Vibra Cell sonicator. The protein yield was quantified using the Bio-Rad Dc protein assay kit. Equivalent protein was loaded, and the lysates were separated by SDS-PAGE and then transferred to nitrocellulose membranes. Equivalent loading was confirmed by staining membranes with fast green as described previously (44) . The membranes were blocked for 1 h in blocking buffer (1× Tris-buffered saline, 5% milk, and 0.20% Tween 20) and placed in primary antibody (1× Tris-buffered saline, 5% milk, 0.10% Tween 20, and 1:1000 antibody) overnight at 4°C. Nitrocellulose membranes were washed three times in wash buffer (0.10% NP40, 0.10% Tween 20, and 1× Tris-buffered saline). Primary antibody was detected using horseradish peroxidase-linked goat antimouse or goat antirabbit IgG antibodies and visualized with the enhanced chemiluminescent detection system (Super Signal; Pierce, Rockford, IL). Immunoblot experiments were performed at least three times.

PI3k Assays.

H1703 or H157 cells were plated at a density of 2 × 10⁶ cells/100-mm dish. The following day, the medium was changed to DMEM containing 0.1% FBS. After overnight incubation, cells were treated with 10 μm PIAs 5, 6, 7, 23, 24, 25, or DMSO control for 2 h. Cells were lysed in 1% NP40 lysis buffer containing 137 mm NaCl, 20 mm Tris-HCl (pH 7.4), 1 mm CaCl₂, 1 mm MgCl₂, 1% NP40, 1 mm phenylmethylsulfonyl fluoride, and 0.1 mm sodium orthovanadate. Cellular extracts containing 900-1100 μg of protein were incubated with anti-p85 for 1 h at 4°C, followed by protein A-agarose beads (Santa Cruz Biotechnology) for an additional 1 h at 4°C. The reaction to demonstrate PI3k activity in PI3k immunoprecipitates was performed as described previously (45) . In brief, immunoprecipitates were incubated with kinase reaction buffer mix [10 nm Tris-HCl (pH 7.4), 150 nm NaCl, 5 mm EDTA, and 0.1 mm sodium orthrovanadate added fresh], 100 mm MgCl₂, 2 mg/ml phosphatidylinositol (Avanti Polar Lipids, Alabaster, AL), and 30 μCi of [γ-³²P]ATP (Redivue ATP; Amersham Biosciences, Piscataway, NJ) for 10 min at 37°C. Reaction products were separated by TLC on 1% potassium oxalate-treated TLC plates (Whatman LK6DF silica gel 60). ³²P-labeled 3′-phosphoinositides were visualized by autoradiography. Kinase assays were repeated at least three times.

Akt Kinase Assays.

Akt kinase assays of cells treated with PIAs were performed as described previously (38) , using the manufacturer’s recommendations (Cell Signaling Technologies) with modifications described below. Cells were plated at a concentration of 2–4 × 10⁵ and treated with PIAs as described above for 2 h. Cells were washed once with ice-cold PBS, and 200 μl of ice-cold lysis buffer [20 mm Tris (pH 7.5), 150 mm NaCl, 1 mm EDTA, 1 mm EGTA, 1% Triton X-100, 2.5 mm sodium PP_i, 1 mm β-glycerol phosphate, 1 mm Na₃VO₄, 1 mm phenylmethylsulfonyl fluoride, and 1 mm leupeptin] were added to the cells for 10 min. Lysates were cleared and allowed to immunoprecipitate for 2–3 h at 4°C with anti-Akt antibody. Immunoprecipitates were washed twice with lysis buffer and twice with kinase buffer. Kinase reaction was performed for 30 min at 30°C in kinase buffer supplemented with 200 μm ATP and 1 μg of GSK-3α/β fusion protein. Reaction was terminated with 3× SDS buffer. The samples were heated at 100°C for 5 min and loaded into a 12% SDS-polyacrylamide gel. Additional kinase assays were performed on Akt isolated from untreated H157 cells as follows. H157 cells were grown in T-75 flasks until ∼80% confluent. Cells were harvested with ice-cold lysis buffer, and cleared lysate was divided into three equal volumes. Aliquots were immunoprecipitated for 2–3 h at 4°C with anti-Akt antibody. Immunoprecipitates were washed twice with lysis buffer and twice with kinase buffer and incubated with 10 μm PIA 7, PIA 5, or DMSO for 30 min at room temperature. After incubation, kinase reactions were performed and processed as described above.

Apoptosis Assays.

Cells were treated with PIAs for 18–24 h as described above. Floating cells were collected, and adherent cells were harvested by trypsinization and then centrifuged at 1000 × g for 5 min. Cells were fixed in ice-cold 70% methanol added dropwise and then incubated at −20°C for 30 min. Cells were centrifuged and incubated with propidium iodide (25 μg/ml) supplemented with RNase A (30 μg/ml) for 30 min at room temperature. Quantification of sub2N DNA was determined by flow cytometry analysis using a Becton-Dickinson FACSort and by manual gating using CellQuest software. Apoptosis experiments were performed in triplicate and repeated at least three times.

“Washout” Experiments.

Additional apoptosis assays were performed as described above but with the following modifications. Cells were treated with PIAs for 2, 4, 6, or 24 h. After PIA treatment, media were changed to fresh DMEM media with 0.1% FBS without PIAs for the remainder of the experiment. At 24 h, cells were collected and prepared for flow cytometric analysis as described above. Apoptosis experiments were performed in triplicate and repeated at least three times.

Adenoviral Infections.

BEAS2B cells were plated at a density of 2.5 × 10⁵ cells/well in 6-well plates. After attachment, cells were washed with PBS and infected with media containing adenoviral constructs for β-galactosidase or myristolated Akt (MOI 100) for 24 h. Infected cells were then treated with PIA 24 or DMSO, and cells were harvested and analyzed for apoptosis or for immunoblot analysis as described above.

Translocation Assays.

H157 or A549 cells were transfected with 2 μg of green fluorescent protein (GFP)-Akt-PH construct using Fugene (Roche, Indianapolis, IN) transfection reagent according to the manufacturer’s instructions. Transfection efficiencies were ∼30% at 36 h, as assessed by fluorescent microscopy. Cells were placed into 0.1% FBS DMEM for 5 h and pretreated or not with PIA5 (10 μm) for 2 h, with or without insulin-like growth factor I (IGF-I; 50 ng/ml) for 10 min. After IGF-I stimulation, cells were washed once with PBS and then fixed for 10 min at −20°C in a mixture of methanol:acetone (1:1). After fixation, GFP and 4′,6-diamidino-2-phenylindole fluorescence was assessed using an Axioscope 2 microscope (Zeiss, Thornwood, NY).

Molecular Modeling.

Docking and scoring studies of the interactions of PIAs with Akt were performed using a virtual screening package from Tripos (46) . For calculation purposes, the long lipophilic tail of the PIAs was truncated leaving only glycerol fragments where the hydroxyl groups were substituted with methyl groups. Because of this simplification, the binding mode of inositol 1,3,4,5-tetraphosphate (IP4), the head group of PIP₃, was compared with the binding modes of PIAs predicted by FlexX. The ligands PIA 5, PIA 6, PIA 23, PIA 24 phosphorylated at positions 4 and 5, and PIA 25 phosphorylated at position 3e were docked to the binding site of Akt and scored using FlexX (47 , 48) and CScore (46) modules in Sybyl, respectively. The 30 best conformations generated after running FlexX were clustered according to the pose of the inositol ring. The most populated poses of PI analogues were saved and depicted.

RESULTS

Identification of PIAs That Inhibit Akt.

Using molecular modeling studies of the PH domain of Akt, a series of phosphatidylinositol analogues were rationally designed, synthesized, and modified. The chemical modifications of the PIAs are shown in Fig. 1A ⇓ . Cell-based assays were conducted to screen the effectiveness of the different PIAs on decreasing Akt activation. Activation of Akt was assessed using phospho-specific antibodies directed against Ser-473 in immunoblotting experiments. These phospho-specific antibodies only recognize Akt in an active state. Cancer cell lines with high levels of endogenous Akt activity that have mutant PTEN (H157 and MB468) or wild-type PTEN (H1703) were treated for 2 h with PIAs (10 μm). PIA 7 serves as a negative control because it is composed of the ether lipid backbone and lacks an inositol moiety. As shown in representative immunoblots (Fig. 1B) ⇓ , a subset of PIAs were able to decrease Akt phosphorylation without affecting total levels of Akt in two lung cancer cell lines that differ in status of p53, K-ras, and PTEN (H1703 and H157; Ref. 38 ). The cumulative data from screening 24 PIAs in three cancer cell lines is shown in Fig. 2A ⇓ . PIAs 5, 6, 23, 24, and 25 exhibited the most complete and consistent inhibition of Akt phosphorylation in the three cell lines tested. PIAs 10, 13, 15, and 16 exhibited cell line-specific inhibition of Akt phosphorylation. Interestingly, the lead compound from an earlier generation of PIAs, DPIEL, was completely ineffective in these assays. Because PIAs 5, 6, 23, 24, and 25 inhibited Akt phosphorylation in these three cell lines, they were chosen for additional evaluation. The structures of these PIAs are shown in Fig. 2B ⇓ .

To confirm that inhibition of Akt phosphorylation was indicative of inhibition of kinase activity and to show that PI3k activity was not affected, in vitro kinase assays were performed (Fig. 2C) ⇓ . Similar to the inhibition of Akt phosphorylation by PIAs observed in the immunoblotting experiments, treatment of H1703 cells with PIA 5, 6, 23, 24, and 25 decreased Akt phosphorylation that is indicative of Akt activity, as determined by the decrease in phosphorylation of an exogenously added, GSK-3 peptide substrate (Fig. 2C ⇓ , top left panels). PIAs 10 and 7 (the ether lipid moiety used as a negative control) did not inhibit Akt. When active Akt was immunoprecipitated from untreated H157 cells and subsequently incubated with PIAs 7 and 5, there was no inhibition of Akt activity (Fig. 2C ⇓ , top right panels). These studies show that inhibition of Akt by PIAs requires intact cells. PI3k activity was not affected by PIAs 5, 6, 23, 24, 25, and 7 (Fig. 2C ⇓ , lower panels), consistent with earlier observations that DPIEL did not inhibit PI3k (49) . Similar results were also obtained when PI3k was immunoprecipitated using anti-p110 or antiphosphotyrosine antibodies (data not shown). Together, these studies indicate that PIAs 5, 6, 23, 24, and 25 inhibits the phosphorylation and activity of Akt in intact cells without affecting PI3k.

Modeling of the Interaction of PIAs with the PH Domain of Akt.

To characterize how PIAs might interact with the PH domain of Akt, molecular modeling of the head group of PIP₃, IP4, and the biologically active PIAs was performed (Fig. 2D) ⇓ . Previously, crystallographic studies had shown that IP4 in the phosphoinositide-binding site of the PH domain of Akt is positioned so that its 2-hydroxyl group is pointed inside the binding cleft and that the network of the hydrogen bonds in the binding site of Akt recognizes only three phosphate groups bound to the equatorial 1-, 3-, and 4-hydroxyl groups of the inositol ring and the axial hydroxyl group in position 2 (50 , 51) . Docking studies revealed that the 2-hydroxy substitutions in PIA5, PIA23, and PIA24 (Fig. 2D ⇓ , panels A–C) cannot be accommodated by the binding site unless they are positioned outside of the binding groove. This sterically driven change in facial orientation of the inositol ring within the binding site leads to rotation of the ring to maximize binding interactions between the polar phosphate groups and the complementary residues. For PIA6 and PIA25, from which a second hydroxyl group has been deleted, additional reorientation of the inositol ring takes place to maximize hydrogen-bonding interactions (Fig. 2D ⇓ , panels D and E). Assuming that the X-ray structure of Akt cocrystallized with IP4 correctly represents the orientation of PIP₂ and PIP₃ when bound to Akt, our calculated binding results for the PIAs suggest that the orientation of their lipophilic side chains will be different from that observed for natural substrates PIP₂ or PIP₃ when bound to Akt. Given that the binding of PIP₃ to the PH domain of Akt causes a conformational change that allows Akt to be activated by PDK-1 (52) , it is possible that PIAs inhibit Akt through changing the conformational state of the PH domain. One prediction from this model is that PIAs inhibit activation of Akt by inhibiting its translocation to the plasma membrane.

Growth Factor-Stimulated Translocation of Akt Is Inhibited by PIA Treatment.

To test whether PIAs inhibit Akt translocation, H157 cells were transfected with a plasmid containing the PH domain of Akt fused to GFP (GFP-PH-Akt), and the cells were treated with IGF-I with our without pretreatment with PIA5 (Fig. 2E) ⇓ . In untreated cells, GFP-PH-Akt is predominantly cytoplasmic. When cells were treated with IGF-I, punctate membranous staining was observed, indicating translocation of Akt. In the presence of PIA5, IGF-I-induced translocation of GFP-PH-Akt was inhibited. PIA5 treatment induced rounding of the cells but did not affect subcellular localization of GFP-PH-Akt. Similar inhibition of translocation of Akt was observed when these experiments were performed with A549 cells (data not shown).

PIAs Inhibit Akt in a Dose- and Time-Dependent Manner.

Because the identification of PIAs 5, 6, 23, 24, and 25 as inhibitors of Akt was determined at a given dose (10 μm) and time of exposure (2 h), additional studies were conducted to determine the dose dependence and time dependence of Akt inhibition by PIAs. Fig. 3A ⇓ shows the dose response of H1703 cells treated with different concentrations of PIAs 5, 6, and 24 for 2 h. Inhibition of Akt was virtually complete at 5 μm. The calculated IC₅₀s for PIAs 5, 6, and 24 in H1703 cells were 4.13, 4.28, and 2.49 μm, respectively (Fig. 3A ⇓ , top panels). Similar dose-dependent responses were observed in the H157 cells treated with PIAs 5, 6, and 24 (data not shown).

To determine the time course of Akt inhibition by PIAs, H1703 or H157 cells were treated with PIAs 5 and 6 (10 μm) for different times, and Akt phosphorylation was assessed (Fig. 3B) ⇓ . An inhibitor of PI3k, LY294002 (10 μm), was added as a positive control, and PIA 7, the ether lipid lacking an inositol ring, served as a negative control. In H1703 cells, PIAs 5 and 6 and LY294002 completely inhibited Akt phosphorylation within 15 min. Complete inhibition was maintained until the 2-h time point, when cells treated with PIAs 5 and 6 began to restore Akt phosphorylation. At 18 h, PIAs continued to inhibit Akt activation, but the level of inhibition was less than that observed with LY294002. Time-dependent inhibition of Akt was also observed when H157 cells were treated with PIA 5 or PIA 6 or LY294002. Although Akt phosphorylation was decreased within 15 min after treatment with PIAs 5 and 6, LY294002 treatment completely attenuated Akt phosphorylation within 15 min. After 2 h of incubation, PIAs 5 and 6 completely inhibited Akt phosphorylation. Interestingly, Akt phosphorylation was still attenuated by PIAs 5 and 6 at 18 h, but Akt phosphorylation had returned to baseline in the LY294002-treated cells. The basis for the different kinetics of inhibition by PIAs or LY294002 is unknown. The results of the dose dependence and time course experiments show that PIAs inhibit Akt within minutes at low micromolar concentrations.

PIAs Selectively Inhibit Downstream Targets of the Akt Pathway.

To determine the specificity of PIAs, activation state-specific antibodies against other kinases and downstream substrates were used in immunoblotting experiments in H1703 and H157 cells. These experiments evaluated the activation state of PDK-1, a kinase upstream of Akt that activates Akt; nine downstream proteins whose phosphorylation is increased in response to Akt activation—tuberin, 4EBP-1, p70S6K, AFX, FKHR, GSK-3, c-Raf, ASK-1, and MDM2; and a member of the mitogen-activated protein kinase superfamily, extracellular signal-regulated kinase (ERK), whose phosphorylation is not affected by Akt activation or administration of LY294002 in these cell lines (data not shown). Of the downstream substrates, tuberin, AFX, FKHR, GSK-3, c-Raf, ASK-1, and MDM2 are direct substrates of Akt, whereas 4EBP-1 and p70S6K are indirectly phosphorylated as a consequence of Akt activation.

Treatment of H1703 cells with PIAs 5, 6, 23, 24, and 25 altered the activation of Akt and many downstream substrates (Fig. 4A) ⇓ , but did not alter phosphorylation of the upstream kinase, PDK-1, at a site necessary for PDK-1 activity. PIAs decreased Akt phosphorylation at Ser-473 and Thr-308 without affecting levels of total Akt protein. Ser-473 phosphorylation was inhibited to a greater extent than Thr-308 phosphorylation, and this may be related to better recovery of Thr-308 phosphorylation at 18 h or to the fact that PDK-1, the Thr-308 kinase, is commonly constitutively active and is not inhibited by PIAs. Many downstream Akt substrates showed decreased phosphorylation after PIA treatment. Phosphorylation of tuberin, 4EBP-1, and p70S6K, substrates that control protein translation, was inhibited by PIA treatment, as was phosphorylation of forkhead family members AFX and FKHR that control transcription. Phosphorylation of GSK-3 and c-Raf was attenuated by PIAs 5, 6, 23, 24, and 25. Inhibition of ASK-1 phosphorylation was only observed with PIAs 23, 24, and 25 in H157 cells, but phosphorylation of MDM2 was not affected by PIA treatment. To further demonstrate selectivity of PIAs, we assessed the phosphorylation status of a kinase downstream of Ras, ERK, which is a member of the mitogen-activated protein kinase kinase superfamily. ERK phosphorylation was not inhibited by PIAs 23, 24, and 25 and was slightly increased by PIAs 5 and 6. Collectively, these data show that although no PIAs decreased PDK-1 phosphorylation, PIAs inhibited the phosphorylation of Akt and of many substrates downstream of Akt.

Similar inhibitory effects of PIAs were observed in H157 cells. When PIAs were added to H157 cells, phosphorylation of Akt was diminished, but phosphorylation of PDK-1 was not (Fig. 4B) ⇓ . Downstream of Akt, PIAs inhibited phosphorylation of three substrates that control translation (tuberin, 4EBP-1, and p70S6K). The virtual complete inhibition of p70S6K phosphorylation by PIAs 5, 6, 23, 24, and 25 in H157 cells is in contrast to the results observed in H1703 cells. Phosphorylation of forkhead family members was also decreased by PIA treatment, as was phosphorylation of GSK-3 and c-Raf. Inhibition of ASK-1 phosphorylation or MDM2 phosphorylation was not observed with PIA treatment in H157 cells. In general, PIAs 5 and 6 were overall more effective inhibitors of downstream substrates in H157 cells. PIA-specific inhibition of downstream substrates was less commonly observed in H157 cells, although within the group of PIAs 23, 24, and 25, PIA 23 exerted the greatest inhibition. Finally, PIAs 5 and 6 increased phosphorylation of ERK, but PIAs 23, 24, and 25 did not affect ERK phosphorylation.

Additional support for selective effects of PIAs on Akt and downstream substrates was provided by analysis of two other cell lines with high endogenous levels of Akt activity (MB468 and ZR751). When tested in these cells, PIAs 5, 6, 23, 24, and 25 uniformly inhibited phosphorylation of Akt, p70S6K, 4EBP-1, and GSK-3, but not PDK-1 (data not shown). Together, these studies show that PIAs 5, 6, 23, 24, and 25 selectively inhibit Akt activation and phosphorylation of components downstream of Akt without inhibiting ERK activation.

PIAs Preferentially Induce Apoptosis in Cell Lines with High Levels of Active Akt.

Because PIAs inhibited Akt downstream components in H157 or H1703 cells that have high levels of constitutively active Akt that depend upon Akt for survival (37 , 38) , we assessed induction of apoptosis by PIAs. For these experiments, we tested PIAs in lung cancer or breast cancer cell lines that maintain high (H1703, MB468, H157, and H1155) or low (H1355, A549, MCF-7, and MB231) endogenous levels of Akt activity under conditions of serum starvation (37 , 38) . Cells were treated with PIAs 5, 6, 23, 24, and 25 for 24 h, and apoptosis was assessed morphologically and by measuring the generation of sub2N DNA formation using flow cytometry. When measured quantitatively in cells with high levels of active Akt, apoptosis increased 20–30-fold with the administration of PIAs 5, 6, 23, 24, and 25, as compared with DMSO-treated cells or cells treated with PIA 7 (Fig. 5B) ⇓ ⇓ . In contrast, PIAs were less effective in cell lines with low levels of constitutively active Akt, as manifest by fewer morphological changes and a less robust induction of apoptosis (4–5-fold). In general, individual PIAs induced apoptosis similarly, with the one exception of PIA 24 in the H1155 cells, where apoptosis caused by PIA 24 was approximately one-half of that induced by PIAs 5, 6, 23, and 25. Similar effects of PIA-induced apoptosis were also observed when apoptosis was measured using an ELISA-based assay that measures histone release (data not shown). The fact that PIAs increased apoptosis by 4–5-fold in cell lines with low levels of Akt activity probably reflects the fact that these cell lines have low, but not absent, levels of active Akt (37 , 38) . Indeed, Fig. 5B ⇓ ⇓ shows that PIA treatment does inhibit Akt activation in two cell lines with low levels of endogenous Akt activity that exhibited less of an apoptotic response to PIA treatment (A549 and H1355 cells). This suggests that the ability of PIAs to induce apoptosis is related to the dependence of the cells on Akt rather than to the degree of inhibition of Akt.

Previous experiments from our laboratory using traditional cytotoxic chemotherapies have shown that the NSCLC and breast cancer cell lines used in these experiments typically require 24–48 h for apoptosis to occur (37 , 38) . To assess the time of exposure to PIAs that is needed to commit these cells to apoptosis, H157 cells were treated with PIAs for 2, 4, or 6 h, then the medium containing PIAs was removed, and apoptosis was assessed 24 h after the initiation of the experiment (Fig. 5C) ⇓ ⇓ . Small increases in apoptosis were observed after only 2 h of exposure. After 6 h of exposure, levels of apoptosis were similar to those observed for continuous 24 h of exposure for three of four PIAs tested. These studies show that PIAs cause an early commitment to apoptosis.

To demonstrate that the cytotoxicity of the PIAs was dependent upon the ability of PIAs to inhibit Akt, we compared induction of apoptosis by PIAs 5 and 8 that are structurally identical (except for the linker group; Fig. 5D ⇓ ⇓ ) but that differ greatly in ability to inhibit Akt (Fig. 2A) ⇓ . PIA 5 contains a phosphate linker and inhibits Akt (Fig. 1B) ⇓ . PIA 8 contains a carbonate linker and did not inhibit Akt phosphorylation (Fig. 1B) ⇓ . When administered to H157 cells, only PIA 5 was cytotoxic (Fig. 5C) ⇓ ⇓ , thereby correlating inhibition of Akt and induction of apoptosis. Collectively, the results of these cellular experiments show that PIAs 5, 6, 23, 24, and 25 increase apoptosis preferentially in cells lines with constitutively high levels of Akt activity. Moreover, these data suggest that cytotoxicity of PIAs correlates with the ability to inhibit Akt.

To confirm the role of Akt modulation in controlling the cytotoxic effects of PIAs, BEAS2B lung epithelial cells were infected with adenoviruses expressing a constitutively active form of Akt (MyrAkt) or β-galactosidase. (Of note, these studies were attempted with the NSCLC and breast cancer cell lines used above, but these cell lines were not infectable with these adenoviruses.) Administration of PIA24 increased apoptosis and decreased Akt phosphorylation in BEAS2B cells infected with β-galactosidase (Fig. 5E) ⇓ ⇓ . In contrast, PIA24 was less cytotoxic and caused less inhibition of Akt phosphorylation in BEAS2B cells infected with MyrAkt. These studies confirm the role of modulation of Akt activation as the main biological effect underlying PIA cytotoxicity.

DISCUSSION

In the present study, we analyzed a series of modified PIAs for their ability to inhibit Akt activity and to induce apoptosis in cancer cell lines with constitutively active Akt. Of the initial 24 compounds tested in our cell panel, PIAs 5, 6, 23, 24, and 25 markedly inhibited Akt at 2 h. The common structural features of these effective PIAs were a phosphate linker between the ether side chain and the inositol ring and modifications at positions 2 and 3 or 4 and 5 of the inositol ring. The 2,3-modified analogues included: a 2-methoxy,3-deoxy analogue (PIA 5); a 2,3-di-deoxy analogue (PIA 6); a 2-isobutyoxy,3-deoxy analogue (PIA 23); and a 2-cyclohexylmethoxy,3-deoxy analogue (PIA 24). The 4,5-modified analogue (PIA 25) was deoxygenated at the 4- and 5-positions of the inositol ring. Interestingly, when the phosphate group linking the ether lipid component to the inositol ring of the active PIAs was replaced by a carbonate linker, the inhibitory activity of the PIAs was lost (e.g., compare PIAs 8 versus 5 or 9 versus 6). Other PIAs with carbonate linkers that were ineffective at inhibiting Akt included PIAs 12, 14, and 15. Dimeric PIAs (PIAs 17, 18, 19, 20, 21, and 22) were also ineffective at inhibiting Akt. However, some PIAs with phosphate linkers and different ring modifications (PIAs 10, 13, and 16) were able to inhibit Akt in a cell line-specific manner. The basis for cell line specificity is presently unclear but may be partially related to differences in uptake of PIAs because when PIAs 10 and 16 were administered to H157 or H1703 cells for 24 h (instead of 2 h), inhibition of Akt and induction of apoptosis was observed (data not shown).

Inhibition of Akt by PIAs was dose dependent and time dependent. The IC₅₀s for PIAs 5, 6, and 24 were in the low micromolar range, which might be achievable in vivo but is a concern regarding additional development of these compounds. When PIAs 5 and 6 were analyzed in H1703 cells for time-dependent inhibition of Akt activation and were compared with a PI3k inhibitor, LY294002, similar rapid times of onset of inhibition were observed (within 15 min), but the inhibition of Akt was maintained longer with LY294002 treatment. In contrast, the onset of inhibition of Akt by PIAs 5 and 6 in H157 cells lagged inhibition by LY294002, but inhibition by PIAs 5 and 6 was maintained longer than with LY294002. These cell lines differ in many ways, but one relevant molecular difference might be PTEN status. Although the activity of PIAs is likely PTEN independent because PIAs can inhibit Akt phosphorylation in cancer cells with wild-type PTEN (H1703) or mutant PTEN (H157 and MB468; Fig. 2A ⇓ ), the delayed onset of inhibition in cells with mutant PTEN suggests that PTEN status may determine the kinetics of Akt inhibition by PIAs. Future studies will address this issue.

Inhibition of Akt phosphorylation and activity by PIAs was not due to inhibition of upstream kinases such as PI3k or PDK-1. The lack of inhibition of PI3k by earlier PIAs such as DPIEL was observed in two prior studies (40 , 49) . Additional evidence that PIAs did not inhibit PDK-1 includes the fact that PIAs did not inhibit the activity of purified PDK-1 in a high-throughput kinase assay (data not shown) and that PIAs did not affect the phosphorylation of another PDK-1 substrate, protein kinase C-δ, at its PDK-1 site (T505; data not shown).

In contrast to the lack of inhibition of upstream components, PIAs 5, 6, 23, 24, and 25 inhibited phosphorylation of downstream substrates that control key cellular functions. Differences in PIA-induced inhibition of phosphorylation of downstream substrates might be related to differences in stability or metabolism of the various PIAs, differences in kinetics of Akt inhibition, or phosphorylation of these substrates by other kinases that are not inhibited by PIAs. The apparent greater inhibition of phosphorylation of downstream substrates by PIAs compared with inhibition of Akt phosphorylation might be related to the fact that some Akt consensus sites in downstream substrates can also be phosphorylated by other kinases. If these “off-target” kinases were to be affected as an indirect result of PIA-induced Akt inhibition for 18 h, then one might observe greater inhibition of downstream substrates than Akt itself. Among substrates directly phosphorylated by Akt, GSK-3 phosphorylation was decreased with PIA administration. Because GSK-3 activity can promote apoptosis by altering glucose metabolism, inhibiting antiapoptotic molecules such as heat shock factor 1 and heat shock protein 70, as well as inhibiting cell cycle regulatory molecules such as cyclin D1 and p21 (53) , PIA-induced decreased GSK-3 phosphorylation could have an overall effect of increasing or maintaining GSK-3 activity, inhibiting cell cycle progression, and/or promoting apoptosis. Of note, cyclinD1 protein expression was not altered by PIA administration in H157 or H1703 cells (data not shown). PIAs did not alter cell cycle distribution in the eight cell lines tested in Fig. 5B ⇓ ⇓ , although LY294002 has been previously shown to induce a G₁-G₀ arrest in these cell lines (37 , 38) .

Other direct Akt substrates that were inhibited by PIAs in H1703 and H157 cells included members of the Forkhead family of transcription factors, AFX and FKHR, which regulate apoptosis through altering transcription of genes such as BIM, p27, and Fas ligand (54) . PIA treatment also decreased phosphorylation of three substrates downstream of Akt that control the initiation phase of protein synthesis—tuberin, 4EBP-1, and p70S6K. Tuberin is a direct substrate of Akt, and 4EBP-1 and p70S6K are indirectly phosphorylated as a consequence of Akt activation. Phosphorylation of tuberin and 4EBP-1 was inhibited by PIA treatment in both H1703 and H157 cells, but p70S6K phosphorylation was inhibited to a much greater extent in H157 cells. The likely cumulative effect of modulating phosphorylation of these three proteins is diminished protein synthesis. This may be relevant because the initiation of protein translation is associated with tumorigenesis (55) . Interestingly, when we evaluated the effects of PIAs against two direct Akt substrates that more directly control the apoptotic process, ASK-1 and MDM2, only ASK-1 phosphorylation was slightly inhibited by PIAs 23, 24, and 25 in H157 cells. Collectively, these data suggest that decreased phosphorylation of downstream substrates of Akt may modulate many cellular processes including metabolism, transcription, protein translation, and/or apoptosis.

We also tested PIAs for inhibition of activation of other kinases downstream of Ras. Phosphorylation of c-Raf, a direct Akt substrate, was inhibited by PIA treatment in H1703 or H157 cells. Concomitant with decreased c-Raf phosphorylation, ERK phosphorylation was increased in H1703 and H157 cells by PIAs 5 and 6 but not PIAs 23, 24, and 25. Increased ERK phosphorylation after PIA treatment may be a consequence of decreased c-Raf phosphorylation at Ser-259, because phosphorylation of c-Raf by Akt at Ser-259 has been shown to inhibit the mitogen-activated protein/ERK/ERK pathway (56) . The lack of correlation between inhibition of c-Raf phosphorylation and increased ERK phosphorylation with PIAs 23, 24, and 25 may indicate that this group of PIAs has a different profile of “off-target” activities that affect the Raf/mitogen-activated protein/ERK/ERK pathway. Taken together, these data show that PIAs inhibit Akt and phosphorylation of downstream substrates without inhibiting upstream kinases or ERK.

In addition to decreasing Akt phosphorylation and activity and inhibiting the phosphorylation of several downstream targets, PIAs induced apoptosis preferentially in cell lines with high levels of Akt activity. Based on historical comparisons of experiments performed in our laboratory using standard chemotherapy agents, PIAs induced more apoptosis at 24 h than did cisplatin (20 μm), paclitaxel (2.5 μm) etoposide (50 μm), trastuzumab (10.5 μm), or gemcitabine (100 μm) at 48 h (38) . Moreover, the induction of apoptosis was related to Akt inhibition, because only PIAs that inhibited Akt activity effectively induced apoptosis, and a constitutively active form of Akt that does not depend upon binding of PIP₃ to the PH domain for translocation bypassed the inhibitory effects of PIAs.

The clinical implications of these apoptosis assays could be important, because Akt is becoming a highly validated therapeutic target in cancer, a disease characterized by dysregulated apoptosis. Inhibition of Akt might have great potential benefit for patients with cancer, and support for the concept of targeting Akt comes from many observations. First, over 54% of human cancers have active Akt that is detectable in situ. 3 Thus, small molecule Akt inhibitors could have wide applicability as cancer drugs. Second, inhibition of the PI3k/Akt pathway by biochemical or genetic means increases the efficacy of chemotherapy and/or radiation in vitro and in vivo (37 , 38 , 57 , 58) . Finally, several standard chemotherapeutic agents and chemopreventive agents inhibit the PI3k/Akt pathway as a consequence of administration in vitro, and in some cases inhibition of Akt is directly responsible for cytotoxicity of these agents (59) .

Despite the widely acknowledged need for Akt inhibitors, however, none are commercially available. Based on the properties we have described for PIAs 5, 6, 23, 24, and 25, we believe that PIAs are the best characterized Akt inhibitors to date and that PIAs compare favorably with other inhibitors of the PI3k/Akt pathway such as wortmannin or LY294002, earlier generation PIAs, or nonspecific small molecule kinase inhibitors engineered for greater specificity toward Akt. The PI3k inhibitors wortmannin and LY294002 may have limited clinical utility due to specificity concerns, adverse side effects, and solubility issues. Wortmannin, in addition to inhibiting PI3k, inhibits myosin light chain kinase, phospholipase C, phospholipase D, phospholipase A₂, and DNA-dependent protein kinase (59) . LY294002 inhibits PI3k as well as the aryl hydrocarbon receptor, which is a ligand-activated transcription factor (60) . In vivo use of LY294002 in mice has been associated with many adverse side effects, including death (61) . Furthermore, wortmannin and LY294002 are only soluble in organic solvents, which may additionally limit their clinical application. Although DMSO was used as a universal solvent for all PIAs in our studies because initial evaluation revealed that PIAs with carbonate linkers were insoluble in water (data not shown), PIAs 5, 6, 23, 24, and 25 are in fact soluble in water. Therefore, although wortmannin and LY294002 inhibit the PI3k/Akt pathway, their lack of specificity, potential side effects, and poor solubility raise doubts as to their suitability as lead compounds for additional drug development.

The biggest advantage of PIAs 5, 6, 23, 24, and 25 over early generation PIAs such as DPIEL is greater efficacy. Although molecular modeling led to the synthesis of DPIEL, a first generation 3-deoxy PIA that inhibited platelet derived growth factor-induced Akt activity and cell growth in vitro (40) , DPIEL did not inhibit Akt in our cellular assays. This might reflect that fact that the inhibition of endogenous Akt activity in cancer cell lines that we used for screening is probably a more rigorous test for Akt inhibition, because Akt is likely to be activated through multiple mechanisms in these cell lines, including activation of multiple growth factor receptors, activation of Ras, and inactivation of PTEN. Moreover, in xenograft studies, oral administration of DPIEL resulted in low bioavailability due to acid lability of the compound, and i.v. administration resulted in massive hemolysis and death (62) . Structural similarities between DPIEL and PIAs raise the possibility that poor bioavailability and toxicity will also be characteristic of PIA administration in vivo, which would preclude development of PIAs as drugs. However, preliminary acute toxicology experiments in mice suggest that PIAs may be better tolerated than DPIEL (data not shown). Additional toxicologic and pharmacokinetic experiments will address these issues. Nevertheless, DPIEL lacks the efficacy and safety profile for an appropriate Akt inhibitor lead compound.

In a final comparison, PIAs 5, 6, 23, 24, and 25 are more fully characterized and have greater specificity than a recently described, nonspecific small molecule kinase inhibitor that was engineered to lose activity against other kinases but retain activity against Akt. Chemically modifying H89, an inhibitor of protein kinase A, Reuveni et al. (63) synthesized a compound called NL-71-101 that had diminished activity toward protein kinase A but retained modest inhibition of growth factor-induced Akt activity. NL-71-101 did induce apoptosis in one cell line with high endogenous Akt activity (63) . However, PIAs may be better lead compounds than NL-71-101, because NL-71-101 was less specific, less potent, and slower acting (thus raising the possibility of indirect activity) and did not inhibit endogenous Akt activity. Another potential problem with the design of NL-71-101 is that it functions as a competitive inhibitor for ATP binding. Creating kinase inhibitors that target the ATP binding site of a kinase can be fraught with specificity problems because all kinases possess ATP binding sites. This has been perhaps best observed with STI-571 (Gleevec), a competitive inhibitor of the ATP-binding site of many kinases (64) . Interestingly, the wide clinical application of Gleevec is partially due to the fact that it inhibits many kinases including bcr-abl, platelet-derived growth factor receptors, and c-Kit (65, 66, 67) . Whether specific inhibition of Akt would be clinically valuable is unknown, but specific inhibition of Akt will be required to rigorously validate Akt as a molecular target.

The fact that rationally designed PIAs inhibit Akt and promote apoptosis in cancer cells with constitutively high levels of Akt appears to fulfill the need for development of small molecule Akt inhibitors that are specific and effective. Not only is this of potential importance for cancer therapy and/or prevention, but also for other diseases such as rheumatoid arthritis, AIDS, and other infectious diseases whose pathogenesis may depend upon Akt activation (68, 69, 70, 71, 72, 73) . These studies raise many questions that will be addressed in ongoing and future experiments. Do other kinases contribute to the biological effects of PIAs? Can PIAs be effectively combined with chemotherapy or radiation therapy? Are PIAs effective in vivo? Will this generation of PIAs have better bioavailability and stability, with less toxicity in vivo? Will one PIA emerge from the group of five as a lead compound? The answers to these questions will help to determine whether PIAs can be advanced to realize the potential clinical benefit of inhibiting Akt.