Extensive evidence implicates activation of the lipid phosphatidylinositide 3-kinase (PI3K) pathway in the genesis and progression of various human cancers. PI3K inhibitors thus have considerable potential as molecular cancer therapeutics. Here, we detail the pharmacologic properties of a prototype of a new series of inhibitors of class I PI3K. PI103 is a potent inhibitor with low IC50 values against recombinant PI3K isoforms p110α (2 nmol/L), p110β (3 nmol/L), p110δ (3 nmol/L), and p110γ (15 nmol/L). PI103 also inhibited TORC1 by 83.9% at 0.5 μmol/L and exhibited an IC50 of 14 nmol/L against DNA-PK. A high degree of selectivity for the PI3K family was shown by the lack of activity of PI103 in a panel of 70 protein kinases. PI103 potently inhibited proliferation and invasion of a wide variety of human cancer cells in vitro and showed biomarker modulation consistent with inhibition of PI3K signaling. PI103 was extensively metabolized, but distributed rapidly to tissues and tumors. This resulted in tumor growth delay in eight different human cancer xenograft models with various PI3K pathway abnormalities. Decreased phosphorylation of AKT was observed in U87MG gliomas, consistent with drug levels achieved. We also showed inhibition of invasion in orthotopic breast and ovarian cancer xenograft models and obtained evidence that PI103 has antiangiogenic potential. Despite its rapid in vivo metabolism, PI103 is a valuable tool compound for exploring the biological function of class I PI3K and importantly represents a lead for further optimization of this novel class of targeted molecular cancer therapeutic. [Cancer Res 2007;67(12):5840–50]
The lipid kinase family of phosphatidylinositide 3-kinases (PI3K) are responsible for the generation of 3-phosphorylated inositides, including the important second messenger PtIns(3,4,5)P3, resulting in activation of signal transduction pathways implicated in many physiologic processes (1–4). The class I enzymes, the most well characterized of the PI3K to date, are divided into class IA (p110α, p110β, and p110δ) and class IB (p110γ) according to their structure and interaction with p85 and p55 regulatory subunits (1, 5). Deregulation of the PI3K pathway has been reported in a significant proportion of cancers. For example, overexpression of p110α has been shown in ovarian, small cell lung, cervical, glioblastoma, head and neck, and bladder cancers (6–10). Activating mutations of the PIK3CA gene encoding p110α have been identified in a large number of cancers, including ovarian, breast, colon, and glioblastoma; in some cases, these were shown to be oncogenic (11–14). In addition to a role for p110α, recent studies have shown that oncogenic transformation can be induced by all other class I isoforms (15). An alternative form of pathway activation results from the mutation of phosphatase and tensin homologue deleted on chromosome 10 (PTEN), the gene that encodes the phosphatase that dephosphorylates the 3-phosphoinositides. This is the second most frequently mutated tumor suppressor gene after p53 (16). Downstream of the lipid phosphorylation step regulated by PI3K and PTEN, AKT is amplified and overexpressed in ovarian, breast, and pancreatic cancers (17, 18), and TSC2 mutations inhibit the formation of the tumor suppression complex TSC1-2 (19). Upstream of PI3K, growth factor receptors are mutated, overexpressed, or activated in many cancers (20). Furthermore, there is accumulating evidence to implicate PI3K-regulated signaling in angiogenesis and metastasis (21–24). Altogether, the frequency of deregulation of this pathway suggests that the lipid kinases that produce PtIns(3,4,5)P3 are attractive targets for cancer drug discovery (25). A number of chemical inhibitors of the PI3K family have been developed, but these either lack potency or specificity or exhibit unsuitable pharmaceutical properties (26). The most frequently used inhibitors to date are the natural product wortmannin and the flavone LY294002. Recently, however, more drug-like inhibitors have been described. PWT-458 is a pegylated derivative of wortmannin, an unstable and broad-spectrum natural product PI3K inhibitor, that has been shown to suppress growth of U87MG glioma in vitro and in vivo with a concomitant decrease in phosphorylation of AKT (27). A series of semisynthetic derivatives of the natural product viridins have also been described, of which PX-866 has shown antitumor efficacy in A549 lung carcinoma and OVCAR-3 ovarian carcinoma xenografts and also potentiated the effects of gefitinib (28, 29). A triazine derivative, ZSTK474, was described as a potent PI3K inhibitor with antitumor activity in A549 lung adenocarcinoma, PC3 prostate adenocarcinoma, and WiDr colorectal carcinoma xenografts following oral administration (30).
We previously reported the discovery of three new series of PI3K inhibitors (31–33). Here, we describe the detailed pharmacologic properties of a novel synthetic small molecule of the pyridofuropyrimidine class, PI103, which is a potent and selective inhibitor of class I PI3K (31). Inhibition of human cancer cell growth in vitro by PI103 is associated with pharmacodynamic biomarker effects consistent with target inhibition. PI103 shows antitumor activity in vivo in several human tumor xenograft models, including cancers with documented abnormalities in the PI3K pathway, in which pharmacodynamic effects were also observed. In addition, PI103 inhibits tumor cell invasion both in vitro and in vivo and also blocks endothelial cell function and vascularization of human tumor xenografts. As this study was being completed, PI103 was also shown in a chemical biology and screening approach to exhibit selective inhibition of the PI3K pathway and to inhibit proliferation of human glioma cells (34–36). PI103 represents an exploratory compound for investigating the therapeutic relevance of PI3K inhibitors in cancers with a variety of genetic aberrations in the pathway and is a lead for further optimization of a novel class of targeted molecular cancer therapeutic.
Materials and Methods
Compound supply. PI103 was synthesized originally at Astellas Pharmaceuticals (32). Recent batches were provided by PIramed Ltd. All batches had similar potency and selectivity.
Enzyme assays. Enzyme assays were done using human recombinant enzymes. A scintillation proximity radiometric assay was done in the presence of 1 μmol/L ATP (31, 32). IC50 values were determined by fitting a sigmoidal dose-response and represent means of three experiments. The selectivity against a panel of 70 kinases was assessed at a concentration of 10 μmol/L (Upstate).
Cell culture. The human cell lines U87MG, HCT116, HT29, PC-3, MDA-MB-468, MDA-MB-435, A549, SKOV-3, MDA-MB-435, IGROV-1, OVCAR-3, and Detroit 562 were obtained from the American Type Culture Collection. A2780 cells were obtained from the European Collection of Cell Cultures. All cell lines were grown in DMEM (Invitrogen Ltd.) containing 2 mmol/L glutamine (Life Technologies) with 100 units/mL penicillin and 100 μg/mL streptomycin (Life Technologies) and supplemented with 10% fetal bovine serum (PAA) in 5% CO2 in air at 37°C. Human umbilical vein endothelial cells (pooled HUVEC) and their appropriate growth medium and supplements were obtained from TCS CellWorks. Cells were cultured according to the supplier's instructions and used at passages 3 to 8. Cell viability was routinely greater than 90%, as judged by trypan blue exclusion. All cell lines routinely tested negative for Mycoplasma by PCR. The GI50 values were determined using a sulfurhodamine blue assay following 96 h of continuous exposure (37).
Immunoblotting and immunoassay. Cells were treated with 1× or 5× GI50 PI103 for various times, and protein extracts were prepared by washing the cells once in ice-cold PBS and then lysing with cell lysis buffer (Cell Signaling Technologies), supplemented with protease inhibitors (Boehringer/Roche). Determination of protein concentration was done using BCA Protein Assay kit (Pierce). Tumors were snap frozen using liquid nitrogen immediately following excision and stored at −80°C until processed. They were homogenized into a fine power under liquid nitrogen using a model 6750 freezer mill (Glen Crestor). Ground powder was immediately resuspended in lysis buffer [150 mmol/L NaCl, 1 mmol/L EDTA, 50 mmol/L Tris (pH 7.4), 1% (v/v) Triton X-100, 1 mmol/L NaF, 1 mmol/L sodium orthovanadate, 10 mg/mL TLCK, 5 μmol/L fenvalerate, 5 μmol/L bpVphen, 1 mmol/L phenylmethylsulfonyl fluoride, protease cocktail (1:100), phosphatase cocktail I and II (1:50)] by aspirating through a 23-gauge needle. Samples were incubated on ice and centrifuged at 1,500 × g for 10 min, and the supernatants were recovered and aliquoted.
Equal amounts of protein were separated by electrophoresis through NuPage Bis-Tris 4% to 12% gradient gels (Invitrogen); proteins were transferred onto nitrocellulose or polyvinylidene difluoride (0.45 μm) pore membranes using the NuPage system and protocol from Invitrogen. Antibodies were obtained from the following suppliers: phosphorylated AKT (p-AKT) Ser473, p-AKT Thr308, p-GSK3β Ser9, GSK3β, cleaved poly(ADP-ribose) polymerase (PARP; Cell Signaling Technologies), cyclin D1 (NeoMarkers), glyceraldehyde-3-phosphate dehydrogenase (Chemicon International). All antibodies were used as recommended by the manufacturer, and specific antigen-antibody interaction was detected with a horseradish peroxidase–conjugated secondary antibody IgG using enhanced chemiluminescence Western blotting detection reagents (Amersham Biosciences).
Electrochemiluminescent immunoassay. A Meso Scale Discovery system was used to measure the levels of AKT activation in tumors following exposure to PI103. Multi-spot p-AKT Ser473/total AKT assays were performed with 10 μg protein in duplicate. The protocol used was as recommended by the manufacturer except that samples were incubated on the plates overnight before addition of secondary antibody.
Flow cytometry. DNA histograms were generated by flow cytometry analysis. Cells were washed in PBS and fixed in 70% ethanol for 30 min at 4°C. Fixed cells were washed once with PBS, resuspended in PBS containing 100 mg/mL RNase A and 40 mg/mL propidium iodide, and incubated for 30 min at 37°C. All samples were analyzed on a Beckman Coulter EPICS Elite ESP flow cytometer. Data were analyzed, and histograms were plotted using winMDI version 2.5 flow cytometry application software (Scripps Institute).
Measurement of caspase activation. Activation of apoptotic effector caspases that cleave the DEVD (Asp-Glu-Val-Asp) motif was assayed using a non-fluorescent caspase substrate Z-DEVD-R110 that releases fluorescent rhodamine-110 when cleaved by caspases that recognise the DEVD motif. Briefly, cells were seeded at 4,000 per well in a 96-well plate and treated 24 h later with inhibitors or vehicle control for 1 to 3 days. Caspase activity was assessed by adding an equal volume of reagent and substrate (Promega), incubating for 16 h, and measuring rhodamine-110 fluorescence at 535 nm following excitation at 485 nm using a Victor 1420 plate reader (Perkin-Elmer).
Localization assays. Redistribution assays of green fluorescent protein (GFP)–tagged proteins were done at Bioimage.6
Pharmacokinetics and metabolism. Microsomes (mouse or human) were obtained from Tebu-Bio and incubated for 30 min with 1 or 10 μmol/L inhibitor as described (38). All animal experiments were done in accordance with local requirements for animal experimentation and national United Kingdom Co-ordinating Committee on Cancer Research guidelines (39). Female BALB/c mice (6–8 weeks old; Charles River) were dosed i.v. with 2.5 mg/kg PI103 in 10% DMSO, 0.5% Tween 20 in saline, or 50 mg/kg PI103 in 20% 4-hydroxypropyl β-cyclodextrin in saline at 0.1 mL/10 g i.p. Blood was collected after serial bleeding and centrifuged, and the plasma was frozen at −80° C. Tissues were snap frozen in dry ice and kept at −80°C until analysis.
Quantitative analysis was done by liquid chromatography tandem mass spectrometry (LC-MS/MS) using multiple reaction monitoring on a ThermoFinnigan TSQ 700 equipped with an electrospray ionization source. Chromatographic separation was achieved by a Waters system (Wisp 717 autosampler and 600 MS pump and gradient former) on a Discovery C18 column (Sigma-Aldrich; 5 cm × 4.6 mm inner diameter) with a 4-min 20 to 100 linear gradient of methanol in 0.1% formic acid following a 3-min isocratic period at 0.6 mL/min. The following transitions were monitored (341.9 > 261.9 at −35 V, 554 > 362 at −40 V). Quantitation was done versus external calibration at the level of 2 to 20,000 nmol/L (8 points). Quality controls were included at concentrations of 15, 750, and 5,000 nmol/L. Aliquots (100 μL) of mouse plasma were added to 10 μL of internal standard (an analogue of PI103, used at concentrations of 5,000 nmol/L) and precipitated with 3 volumes of methanol. Following centrifugation at 2,500 × g for 5 min, 20 μL supernatant was analyzed. Tissues were homogenized in 2 or 3 volumes PBS and treated as plasma. Quantitation was done on Excalibur software. The assay was linear over the range of concentrations analyzed (r > 0.99).
Qualitative measurements (metabolite detection and characterization) were done by LC-MS/MS on a Thermoseparation AS3000 autosampler, P4000 quaternary pump, and UV1000 detector attached to a LCQ ion trap (ThermoFinnigan). Samples were separated using the same column as above with a slower gradient of 10% to 90% methanol over 6 min after 0.5-min isocratic at 1 mL/min. The scan range was 50 to 1,000 amu. Pharmacokinetic calculations were done using WinNonlin noncompartmental analysis model 201 for plasma following i.v. administration and model 200 for tissues and plasma following i.p. administration (40).
Human tumor xenograft efficacy and pharmacodynamic studies. For s.c. tumor xenograft models, 2 × 106 cells were injected s.c., bilaterally, into female (or in the case of PC3 prostate tumors, male) NCr athymic mice 6 to 8 weeks of age, bred in-house. The breast carcinomas were inoculated into mammary fat pads. Dosing commenced when tumors were established (∼5 mm mean diameter) for the period indicated. PI103 was prepared in 20% 4-hydroxypropyl β-cyclodextrin and injected i.p. Tumors were measured across two perpendicular diameters, and volumes were obtained from the following formula: V = 4/3π [(d1 + d2) / 4]3 (41).
When the experiment was terminated, mice were bled, plasma samples were prepared, and tumors were excised and weighed; % treated/control (T/C) values were calculated from final tumor weights or, in the case of HCT116 carcinoma, from tumor volumes on the last day of therapy because tumors were allowed to grow out to observe recovery. Tumor samples were snap frozen for pharmacokinetic or pharmacodynamic analysis and, in some cases, cryopreserved or fixed for histologic examination. In the latter case, samples were sectioned and stained with H&E using standard techniques. For examination of vasculature, tumors were stained with rat anti-mouse endothelial antibody MECA 32 with goat anti-rat horseradish peroxidase–conjugated secondary antibody as previously described (42).
OVCAR-3 ovarian cancer cells were passaged in vivo, and 5 × 106 cells were injected i.p. into female Ncr nu mice. Therapy commenced 15 days later and continued daily until day 30. Mice were sacrificed 15 min, and 4 h after the last drug dose, tumor burden was calculated, and samples taken for pharmacodynamic analysis by Western blotting. Visceral organs were examined for evidence of tumor invasion both macroscopically and by histology. In separate pharmacodynamic studies, mice bearing established OVCAR-3 orthotopic ascites tumors or s.c. U87MG glioblastoma xenografts were given two doses of 100 mg/kg PI103 in β-cyclodextrin vehicle i.p. before harvesting of tumor samples for p-AKT electrochemiluminescent immunoassay analysis at intervals after the final dose.
Cell migration and invasion assays. Cells were labeled by incubation with the green fluorescent dye Celltracker green 5-chloromethylfluorescein diacetate (Invitrogen) at 1 μmol/L for 1 h. Cells (2 × 104 in 350 μL serum-free media) were then placed into the upper well of an 8-μm pore Fluoroblok membrane insert in 24-well companion plates (BD Biosciences) in DMEM vehicle or various concentrations of PI103. Medium (800 μL) containing 5% heat-inactivated FCS was added to the lower chamber as a chemoattractant. The assay plates were incubated at 37°C in 5% CO2 for 16 to 24 h (tumor cells) or 4 h (HUVEC). An inverted fluorescent microscope and digital camera (Olympus LX70) were used to obtain images of the migrated cells on the underside of the membrane. Three different fields of view were scored for each well. The number of migrated cells was calculated using Image Pro-Plus 4.0 software (Media Cybernetics). Invasion was assayed in a similar fashion using inserts coated with Matrigel. The observation period was extended to 36 h for tumor cells and 18 h for HUVEC.
PI103 is a selective class I PI3K inhibitor. PI103 was synthesized following a high-throughput screen to identify inhibitors of the enzymatic activity of recombinant PI3K p110α (32). PI103 potently and competitively inhibited all class IA isoforms, α, β, and δ, with IC50 values of 2, 3, and 3 nmol/L, respectively. It is ∼200-fold more potent than the commonly used broad-spectrum PI3K inhibitor LY294002 (IC50 = 500, 580, and 600 nmol/L against p110α, p110β, and p110δ, respectively). PI103 exhibited 5- to 7-fold less activity against the class IB p110γ isoform (IC50 = 15 nmol/L) and DNA-PK (IC50 = 14 nmol/L), a member of the PI3K superfamily, as well as 15- to 20-fold less activity against the class II PI3K C2β (IC50 = 43 nmol/L). In addition, there was at least 1,300-fold selectivity over the class III PI3K Vsp34 (IC50 = 3,950 nmol/L). PI103 inhibited TORC1 (83.9% inhibition at 0.5 μmol/L), consistent with recent data (33). There was no inhibition of PKCα or PKA at concentrations of 1 μmol/L. PI103 was also screened against a panel of 70 protein kinases that are representative of the human kinome and containing a number of oncogenic proteins such as cyclin-dependent kinases, receptor tyrosine kinases, and kinases involved in the regulation of signal transduction, including AKT. Little or no inhibitory activity was seen at 10 μmol/L PI103, indicating a high degree of selectivity for the PI3K superfamily (data not shown).
PI103 inhibits the PI3K pathway and proliferation of cancer cells in vitro. Figure 1A shows the cellular GI50 values of PI103, evaluated in a panel of human cancer cell lines comprising prostate, lung, ovary, colon, and breast cancer, glioblastoma, and oropharyngeal carcinoma and following 96 h of continuous treatment with PI103. These cell lines have different genetic abnormalities that can result in activation of the PI3K pathway (e.g., PTEN deletion or a PI3K p110α mutation within the kinase or other regulatory domains; refs. 11, 13, 43–46). PI103 exhibited a range of GI50 values from 0.06 to 0.93 μmol/L compared with 4.60 to 12.10 μmol/L for LY294002. Thus, the greater potency of PI103 on the molecular target was also carried over into enhanced growth inhibition in cancer cells. PI103 showed activity in cell lines with wild-type and mutant PTEN and retained potency in cells with mutant PIK3CA. These included the Detroit 562 oropharyngeal carcinoma cell line that harbors the mutant PIK3CA H1047R, the HCT116 colorectal carcinoma line with an H1047R mutation in the kinase domain, the A2780 ovarian carcinoma line with a C2 domain mutation, and the IGROV-1 ovarian carcinoma with a p85 binding domain mutation and stop 1069W mutation resulting in extension of the protein by four amino acids (A2780 and IGROV; data not shown; ref. 11).
Similar to LY294002, treatment with PI103 resulted in a time- and concentration-dependent decrease in phosphorylation of AKT Ser473 in all cell lines (Fig. 1A and B; Supplementary Fig. S1A and B). Figure 1B shows decreased phosphorylation of AKT Ser473 following treatment of HCT116 human colorectal carcinoma cells with increasing PI103 (1–10× GI50) or of U87MG glioblastoma and PC3 prostate cancer cells (1× and 5× GI50; Supplementary Fig. S1A and B). The effects of PI103 on the downstream components of the PI3K pathway were rapid, with inhibition of phosphorylation of AKT Ser473 and Thr308 and GSK3β Ser9 occurring within 15 min of treatment (Fig. 1C). This was followed by an equally rapid recovery 15 min after PI103 was removed and the cells were transferred to drug-free medium.
The effects of PI103 on the redistribution of a forkhead transcription factor (FKHR), PDK1, AKT1, AKT2, or AKT3 following stimulation with IGF-1 were measured in U20S human osteosarcoma cells expressing FKHR fused to GFP or CHO cells expressing PDK1, AKT1, AKT2, and AKT3 fused to GFP. Activation of PI3K and production of PIP3 result in recruitment of PDK1 and AKT to the plasma membrane via their PH domains (45). Activated AKT subsequently phosphorylates a number of substrates (including FKHR) that prevents their translocation from the cytoplasm to the nucleus (46). Consistent with inhibition of PI3K signaling, PI103 induced the redistribution of GFP-tagged FKHR from the cytoplasm to the nucleus (IC50 = 30 nmol/L) in U2OS cells (Fig. 1D,, left) and inhibited the localization of GFP-AKT1, GFP-AKT2, GFP-AKT3 and GFP-PDK1 to the plasma membrane in CHO cells with IC50 values of 17, 13, 11, and 66 nmol/L respectively (Fig. 1D , right, AKT2; Supplementary Fig. S1C). PI103 had a similar potency to wortmannin in all of these assays; in contrast, LY294002 was less potent, with 200-fold higher IC50 values (data not shown).
PI103 also decreased cyclin D1 expression as early as 4 h posttreatment in U87MG cells (Fig. 2A). This was consistent with a time- and concentration-dependent G1 cell cycle arrest that was detected by flow cytometry in all the cell lines listed in Fig. 1A. Figure 2B and C illustrate typical cell cycle data for U87MG glioblastoma cells. Note that there was no evidence for a sub-G1 peak associated with the induction of apoptotic cell death in U87MG cells (Fig. 2C) or any other of the cell lines listed in Fig. 1A (data not shown). The similar molecular signature and cell cycle effects of PI103 in PC3 cells are presented as supplemental data (Supplementary Fig. S1D). Treatment of the same human cancer cell line panel with LY294002 at 5× and 10× GI50 concentrations frequently resulted in apoptosis as detected by cleavage of PARP, a caspase substrate. However, similar increases in PARP cleavage were not apparent following PI103 treatment (Fig. 1B). The induction of apoptosis by LY294002, but not PI103, was confirmed in HCT116 cells by direct measurement of the activation of the apoptotic effector proteases (caspases) following 1 to 3 days of exposure of up to 10× GI50 (Fig. 2D).
Pharmacokinetics and metabolism of PI103. PI103 was metabolized by >70% after 30 min of incubation with human and mouse microsomes under standard conditions (Fig. 3A). The main metabolite was the glucuronide of PI103 formed on the phenol ring, representing on average 57% of the total metabolites. In addition, a hydroxylated product on the morpholino moiety and a hydroxylated glucuronide were also formed. This resulted in opening of the morpholino ring, which we observed previously with LY294002 and other benzochromenone DNA-PK inhibitors (47, 48).
The concentration versus time curves in plasma, liver, kidney, and spleen following 2.5 mg/kg PI103 given i.v. to female BALB/C mice are shown in Fig. 4B. Pharmacokinetics of PI103 showed a high volume of distribution at steady state (Vss = 0.0516 L). This was consistent with a very rapid tissue distribution, resulting in Cmax occurring 5 min after dosing in liver, kidney, and spleen (Fig. 3B). The liver to plasma ratio was 1.3 versus 0.7 for kidney and spleen. The compound cleared from plasma with a rate of 0.19 L/h, which exceeds liver blood flow (Fig. 3C). Rapid tissue clearance was also seen.
Following i.p. administration of 50 mg/kg PI103, the apparent clearance was similar to that observed following i.v. dosing (0.12 L/h), but the concentrations were sustained in plasma and remained above GI50 levels for longer than 12 h (Fig. 3D). This contrasts with previous observations with LY294002, where plasma levels remained above active levels for only 1 h with comparable tissue distribution to that of PI103 (47). These encouraging results prompted us to investigate pharmacokinetic, pharmacodynamic, and efficacy effects in tumor-bearing animals.
Following 14 daily doses of 40 or 70 mg/kg PI103 i.p. in 20% 4-hydroxypropyl β-cyclodextrin to mice bearing U87MG glioblastoma xenografts, measurement of the peak area ratio of parent to glucuronide in both plasma and tissue showed extensive glucuronidation with 20-fold more glucuronide than parent compound in plasma (Fig. 4A). Despite this extensive metabolism, tumor levels of parent compound were found to be above GI50 levels for 2 h (Fig. 4B).
PI103 decreases AKT phosphorylation in vivo. Following two doses, 12 h apart, of 100 mg/kg PI103 i.p. in female Ncr nude mice bearing U87MG human glioblastoma xenografts, there was a reproducible and significant drop in phosphorylation of AKT on Ser473 15 min to 2 h after dosing, consistent with inhibition of the PI3K pathway. The ratio of p-AKT Ser473/total AKT measured by the electrochemiluminescent immunoassay reached a nadir at 30 to 60 min (falling to ∼50% of control levels) and returned to control levels by 4 h (Fig. 4C). This is consistent with intratumoral levels of PI103 measured over the same period in this xenograft model (Fig. 4B).
Antitumor activity in vivo. Given the pharmacokinetic and pharmacodynamic data described above, we next sought evidence of therapeutic efficacy in human tumor xenografts selected for evidence of activation of the PI3K pathway by differing genetic events. As shown in Fig. 5, PI103 exhibited therapeutic activity at well-tolerated doses (<15% weight loss) in a number of human tumor models, including the PTEN null U87MG glioblastoma, PC3 prostate, and MDA-MB-468 breast carcinomas (Fig. 5A and B) and the PIK3CA mutant HCT 116 colon carcinoma (Fig. 5C). In the case of the HCT116 colon carcinoma, dosing was terminated on day 14 when the T/C, calculated from tumor volumes, was 68% and 28% for animals dosed once or twice daily with 30 mg/kg PI103, respectively. On day 21, when the experiment was terminated, the tumors in the former group had recovered to control values, but in the latter group, the T/C was still 48% (Fig. 5D).
In vivo efficacy of PI103 was also assessed in three further xenograft tumors. These were SKOV3 and IGROV-1 ovarian carcinomas that have p85 binding domain and kinase domain mutations of PIK3CA, respectively, as well as the MDA-MB-435 breast carcinoma that overexpresses ERB-B2 (Fig. 5D). Statistically significant tumor efficacy (P < 0.05) was obtained in two of three of these experiments and overall in seven tumor models described in this section (Fig. 5D).
Inhibition of tumor cell and endothelial cell migration and invasion in vitro and in vivo. In a Transwell assay, PI103 inhibited chemomigration and invasion of a variety of tumor cells [e.g., HCT116 colon carcinoma, PC3 prostate carcinoma, MDA-MB-468 breast carcinoma, and Detroit 520 squamous cell carcinoma (Fig. 6A) and also U87MG glioblastoma (44% inhibition at 50 nmol/L, data not shown)]. HUVEC invasion of Matrigel was also significantly inhibited and to a greater extent than with LY294002 at equipotent concentrations based on proliferation GI50 (Fig. 6B).
Inhibition of invasion and angiogenesis were also observed in human breast tumor xenografts. Control tumors showed extensive invasion of underlying muscle and also of distended i.t. blood vessels (Fig. 6C, top left and middle, arrows), whereas tumors treated with 30 mg/kg PI103 twice daily remained encapsulated and noninvasive (Fig. 6C,, top right and middle). In control tumors, there was an extensive vascular network, whereas in treated tumors, the vessels were fewer and smaller (Fig. 6C , bottom).
Finally, we tested the effects of PI103 in a xenograft model of late-stage ovarian carcinoma using i.p. implantation of OVCAR-3 cells. Therapy was initiated once tumors were well established and continued for 15 days. Daily doses of 15 mg/kg or 30 mg/kg, which are well below toxic levels, reduced tumor burden (T/C, 61–62%) in the peritoneal cavity (Supplementary Fig. S2A), and harvested tumor cells showed decreased levels of p-AKT Ser473 at 15 min and 4 h after dosing, although in the latter case, there was evidence of recovery of p-AKT Ser473 in one of the two samples (Supplementary Fig. S2B). We also noted a substantial decrease in i.p. tumor spread with undetectable liver, diaphragm, kidney, and ovary invasion and reduced incidences of mesenteric and lymph node metastasis in PI103-treated animals, as assessed macroscopically and by histologic examination (Supplementary Fig. S2C). In a separate pharmacodynamic study, we showed inhibition of p-AKT Ser473 in OVCAR-3 ascites and solid tumor deposits measured by electrochemiluminescent immunoassay at 30 min and 2 h following two doses of 100 mg/kg PI103 (Supplementary Fig. S2D).
In this study, we describe the detailed pharmacologic characterization of the pyridofuropyrimidine PI3K inhibitor PI103. We discovered this inhibitor by elaboration of a high-throughput screening hit (32). As also suggested in other recent reports, we find PI103 to be a highly potent inhibitor of PI3K (31, 34). We show here that PI103 exhibits a high degree of selectivity for class I PI3K versus a panel of 70 protein kinases. PI103 also inhibits DNA-PK and mammalian target of rapamycin (mTOR). In addition, we show the activity of PI103 in a range of tumors in vitro and in xenograft models, together with evidence of inhibition of invasion and angiogenesis.
PI103 is especially potent against the class IA enzymes α, β, and δ, with IC50 values of 2 to 3 nmol/L. The p110γ isoform is inhibited, with an IC50 of 15 nmol/L. This profile may be important therapeutically because all four class I enzymes have been shown to induce oncogenic transformation (15). It is valuable to have a potent and selective inhibitor to establish proof of concept for the effects of inhibition of a specific molecular target in cells. We have shown that the potency against PI3K translates well into cells, and that PI103 is very active in U87MG glioma and PC3 prostate carcinoma cells, both of which are PTEN null. In addition, PI103 was active both in vitro and in vivo in cells that harbor PIK3CA mutations in both kinase and regulatory domains (A2780, Detroit 562, HCT116, IGROV-1, and SKOV3). Such mutations have been shown to be frequent in human cancers and result in constitutive activation of the kinase (11). Taken together, these results suggest that PI103 is active in cells exhibiting deregulation of the PI3K pathway by various means, and that several of the oncogenic mutations thus far reported in PIK3CA are unlikely to confer resistance to this inhibitor. Further work is required to investigate in more detail the relationship between sensitivity to PI103 and variables such as PI3K mutations/alterations and levels of p-AKT in cancer cells. No obvious relationship is apparent from the present results (Fig. 1A). For example, there is a wide range of sensitivity for the cell lines with PTEN abnormalities. However, our studies in an isogenic ovarian cancer model have indicated that a dominant-negative AKT, which shows a decrease in p-AKT Ser 473 is 2-fold less sensitive to PI103 compared with the wild-type cell line.7
Bhattacharya et al., unpublished data.
The cellular potency of PI103 is associated with a decrease in p-AKT Ser 473 in all the human cancer cell lines studied as shown here in the U87MG tumor by Western blot analysis and by an electrochemiluminescence immunoassay. This molecular biomarker modulation is consistent with inhibition of the signaling pathway downstream of PI3K and has also been reported recently by others (35). In addition, we were able to show further potent downstream effects on cellular localization of PDK1, AKT1, AKT2, AKT3, and FKHR at concentrations similar to wortmannin in CHO or U20S human osteosarcoma cells stimulated with IGF. In all cell lines examined, treatment with PI103 resulted in a robust growth arrest in the G1 phase of the cell cycle that was preceded by a decrease in cyclin D1 expression. This is consistent with reports that G1 cell cycle progression and expression of G1 cyclins are regulated by PI3K/AKT/mTOR signaling (49) and was also observed in a panel of glioma cell lines (35).
Also of particular interest was our observation that potent and selective inhibition of class I PI3K with PI103 did not result in apoptosis, even at several multiples of the GI50 concentrations that clearly resulted in marked inhibition of AKT phosphorylation and complete cytostasis. This lack of apoptosis was exemplified here with HCT116 colon carcinoma tumor cells and was also observed in a wide range of human cancer cell lines tested. It should be noted that the cancer cells used were competent to undergo apoptosis as shown following treatment with LY294002. These data are somewhat surprising given the frequent association of the PI3K pathway with survival signaling and apoptosis (50). However, the expectation of apoptosis upon PI3K inhibition is based heavily on the use of LY294002. In addition to inhibiting other members of the PI3K superfamily and casein kinase 2 (51), it is known that LY294002 induces cell death by producing reactive oxygen species independent of the PI3K pathway in human prostate carcinoma LNCaP cells (52). Interestingly, ZSTK474 was reported as a weak inducer of apoptosis in certain cancer cell lines, an activity more similar to LY294002 than PI103 (30).
In addition to showing potency and selectivity, together with encouraging cytostatic properties in cancer in vitro, we wished to show proof of principle for antitumor activity in animal models. To do this, it was necessary to demonstrate that PI103 exhibited suitable pharmacokinetic and pharmacodynamic properties in vivo. Although subject to high plasma clearance, PI103 was able to achieve tumor levels above those required for in vitro tumor cell growth inhibitory activity for 2 to 4 h with doses of 40 or 70 mg/kg once or twice daily.
PI103 given i.p. at doses of 30 to 100 mg/kg/d showed >50% growth inhibition in a number of human tumor xenograft models selected for deregulation of the PI3K pathway. Significant efficacy was achieved in the PTEN null U87MG glioma, PC3 prostate carcinoma, and MDA-MB-468 breast carcinoma; SKOV-3 ovarian carcinoma and MDA-MB-435 breast carcinoma that overexpress ERB-B2 receptors; and also the SKOV3 ovarian and HCT116 colon carcinoma that express a mutant PIK3CA. Fan et al. (34) observed antitumor activity at doses of 5 mg/kg in xenografts derived from U87MG cells transfected with constitutively activated epidermal growth factor receptor (EGFR). We were unable to detect PI103 in U87MG xenografts following administration of 5 mg/kg in vivo in the same vehicle as used by Fan et al. (50% DMSO in saline), and we saw no down-regulation of p-AKT Ser473 by the sensitive electrochemiluminescent immunoassay at that dose (data not shown). However, we observed 89% and 46% T/C following daily treatment of U87MG xenografts with 40 and 70 mg/kg PI103, respectively. This was not surprising given the plasma levels we have observed for PI103. The results suggest that the antitumor activity observed by Fan et al. may be due to the EGFR dependence of the modified U87MG cells used in their study. Those authors also argued, based on the properties of a panel of PI3K inhibitors, that combined activity against p110α and mTOR was important for activity in glioma cell lines. We were able to detect a decrease in AKT phosphorylation in vivo in the U87MG glioma tumor and orthotopic OVCAR-3 ovarian carcinoma model both by Western blot and by the more sensitive and quantitative electrochemiluminescent immunoassay. Overall, our in vivo antitumor studies show growth inhibition rather than regression in a range of human tumor xenografts with different histologies and molecular pathologies, consistent with the in vitro data that indicate a G1 cell cycle arrest rather than apoptosis.
The antitumor effects were observed at well-tolerated doses, despite the fact that the pharmacokinetic properties of PI103 are suboptimal. This compound is relatively insoluble due to its tricyclic core structure. In addition, it has a number of metabolic hotspots, including the phenol ring, which we have shown is subject to glucuronidation, and the morpholino ring, which is susceptible to hydroxylation. These properties result in high plasma and tissue clearance. We have previously reported similar metabolic hotspots in LY294002 and in NU7026, a DNA-PK inhibitor (47, 48). In particular, it seems that the morpholino ring, which has been reported to bind to the hinge region in PI3Kγ (53–55), is a highly labile moiety and is susceptible to oxidation. Although no significant effect was seen on circulating glucose levels with this agent, and only very limited weight loss was observed (data not shown), it cannot be excluded that a compound with better pharmaceutical properties may be more toxic as well as potentially more effective. Despite the pharmaceutical limitations of PI103, our results show proof of principle for antitumor effects with this class of PI3K inhibitor.
In addition to antiproliferative activity, we showed that PI103 significantly decreased tumor cell chemomigration and invasion. This was seen in vitro in four different tumor cell types (including those with mutant PIK3CA) and in vivo in the MDA-MB-435 breast carcinoma and OVCAR-3 orthotopic ovarian carcinoma model that mimics late-stage clinical disease. The results are consistent with previous studies implicating PI3K in cell motility and invasion (22, 56, 57). The precise isoform(s) involved are still debated and may vary between cell types (57, 58), but PI103 inhibits three class IA PI3K isoforms with equal potency that may contribute to its anti-invasive effects.
Finally, we showed that activated human endothelial cells are also sensitive to PI103, and that greater inhibition of invasion through the Matrigel is obtained compared with LY294002 at equipotent concentrations (based in proliferation GI50 values). In addition, we showed that the vascularity of orthotopic MDA-MB-435 breast carcinomas was reduced in PI103-treated mice. Thus, there is a potential for inhibitors of this type to have antitumor activity via effects on angiogenesis as well as through direct action on tumor cells.
In conclusion, we have identified a potent inhibitor of class I PI3K. This compound shows potency in vitro and antitumor activity in vivo with evidence of target modulation. PI103 represents a valuable tool for exploring the biological functions of class I PI3K enzymes and is a lead for further optimization of this novel class of molecular therapeutics.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
P. Workman is a Cancer Research UK Life Fellow.
S. Guillard and L. Bjerke were recipients of studentships from The Institute of Cancer Research.
Current address for B. Nutley: Cantest Ltd., 4606 Canada Way, Bunaby, British Columbia, V5G1K5, Canada.
Current address for L. Kelland: Cancer Research Technology, Wolfson Institute for Biomedical Research, University College London, Gower Street, London CE1 6BT, United Kingdom.
Grant support: Cancer Research UK grants C309/A2187 and C309/A8274.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Drs. Khatereh Ahmadi and Yoshitaka Yoneyama for excellent assistance with project management.