AZD2171: A Highly Potent, Orally Bioavailable, Vascular Endothelial Growth Factor Receptor-2 Tyrosine Kinase Inhibitor for the Treatment of Cancer

Inhibition of vascular endothelial growth factor-A (VEGF) signaling is a promising therapeutic approach that aims to stabilize the progression of solid malignancies by abrogating tumor-induced angiogenesis. This may be accomplished by inhibiting the kinase activity of VEGF receptor-2 (KDR), which has a key role in mediating VEGF-induced responses. The novel indole-ether quinazoline AZD2171 is a highly potent (IC₅₀ < 1 nmol/L) ATP-competitive inhibitor of recombinant KDR tyrosine kinase in vitro. Concordant with this activity, in human umbilical vein endothelial cells, AZD2171 inhibited VEGF-stimulated proliferation and KDR phosphorylation with IC₅₀ values of 0.4 and 0.5 nmol/L, respectively. In a fibroblast/endothelial cell coculture model of vessel sprouting, AZD2171 also reduced vessel area, length, and branching at subnanomolar concentrations. Once-daily oral administration of AZD2171 ablated experimental (VEGF-induced) angiogenesis in vivo and inhibited endochondral ossification in bone or corpora luteal development in ovary; physiologic processes that are highly dependent upon neovascularization. The growth of established human tumor xenografts (colon, lung, prostate, breast, and ovary) in athymic mice was inhibited dose-dependently by AZD2171, with chronic administration of 1.5 mg per kg per day producing statistically significant inhibition in all models. A histologic analysis of Calu-6 lung tumors treated with AZD2171 revealed a reduction in microvessel density within 52 hours that became progressively greater with the duration of treatment. These changes are indicative of vascular regression within tumors. Collectively, the data obtained with AZD2171 are consistent with potent inhibition of VEGF signaling, angiogenesis, neovascular survival, and tumor growth. AZD2171 is being developed clinically as a once-daily oral therapy for the treatment of cancer.

Vascular endothelial growth factor-A (VEGF) is a pivotal stimulus of physiologic and pathologic angiogenesis, including the sustained neovascularization required to support solid tumor growth (1). Tumor blood vessels furnish solid malignancies with oxygen and nutrients, facilitating further dysregulated growth, and provide a route for metastatic dissemination. New vascular capillaries are derived via a complex, orchestrated series of events, involving activation of endothelial and perivascular cells and modifications to the surrounding basement membrane and extracellular matrix. Although many diverse stimuli are implicated in this process, signaling induced by VEGF is considered rate limiting. A dependency on VEGF is attributed to its ability to regulate key processes throughout the angiogenic cascade, including endothelial cell migration, proliferation and protease expression (2–4), microvascular integrin expression (5), recruitment of endothelial cell precursors (6), capillary tube formation (7), neovascular survival (8), interactions with mural cells (9), and enhanced vascular permeability (10). The production of VEGF can be disproportionately up-regulated in tumors via oncogene activation (11), loss of tumor suppressor function (12, 13), or changes in oxygen or glucose status (14). Tumor capillaries induced by overexpression of VEGF are tortuous and dilated (15); an architecture that is reminiscent of immature vessels during early vascular remodeling.

VEGF binds to the second and third immunoglobulin-like domains of its specific transmembrane receptors Flt-1 (VEGFR-1) and KDR (VEGFR-2) on endothelial cells, initiating receptor homodimerization or heterodimerization (16, 17). The conformational change induced in the receptor complex stimulates intrinsic kinase activity which transphosphorylates tyrosine residues within the cytoplasmic domains. These phosphorylated peptide sequences serve as recognition sites for Src homology 2 domain-binding proteins that subsequently propagate intracellular signaling. Activated KDR has been shown the major stimulator of angiogenesis and vascular permeability (18, 19). KDR signaling responses include (i) mitogenic signaling via activation of a phospholipase C-γ/protein kinase C/Raf/mitogen-activated protein kinase (MAPK) pathway (20), (ii) motogenic signaling through phosphorylation of FAK and paxillin (21), and (iii) survival signaling, in a complex with VE-cadherin, β-catenin, and phosphoinositide 3′-kinase (22).

With the exceptions of longitudinal bone extension during growth and cyclical changes in the female reproductive tissues, angiogenesis does not occur in healthy adults. Chronic inhibition of this process may therefore be tolerated and provide a means by which to prevent tumor progression (23). Given that KDR transduces the angiogenic effects of VEGF, and tissue expression of this receptor is largely confined to the endothelium, inhibition of KDR signaling should provide a means for achieving selective therapeutic intervention.

A number of small molecule approaches to inhibit the intrinsic tyrosine kinase activity of KDR have been described previously, with a range of nanomolar potencies, selectivities, and pharmacokinetic properties (24–27). Here we describe AZD2171, a highly potent (subnanomolar IC₅₀) inhibitor of KDR tyrosine kinase and VEGF-induced signaling in endothelial cells. This compound has pharmacokinetic properties that make it suitable for chronic once-daily oral dosing. AZD2171 prevents VEGF-induced angiogenesis in vivo and shows dose-dependent activity in a range of human tumor xenografts in mice; statistically significant inhibition of tumor growth being evident with doses as low as 0.75 to 1.5 mg per kg per day. Furthermore, we show that inhibition of tumor growth is not merely a consequence of preventing new vessel formation; vascular regression can be observed in tumors following AZD2171 treatment.

AZD2171. 4-[(4-Fluoro-2-methyl-1H-indol-5-yl)oxy]-6-methoxy-7-[3-(pyrrolidin-1-yl)propoxy]quinazoline (AZD2171; Fig. 1) was synthesized according to the processes described in the International Patent Application Publication Number WO 00/47212, in particular those described in Example 240 of WO/47212. The free base of AZD2171 was used in all preclinical studies, with a molecular weight of 450.51. For all in vitro assays, AZD2171 was prepared initially as a 10 mmol/L stock solution in DMSO and diluted in the relevant assay media. All in vivo studies were conducted by once-daily oral gavage. For studies in mice, AZD2171 was suspended in 1% (w/v) aqueous polysorbate 80 (polyoxyethylene; ref. 20; sorbitan mono-oleate in deionized water) and dosed at 0.1 mL/10 g of body weight. For studies in rats, AZD2171 was suspended in a 0.5% (w/v) hydroxypropyl methylcellulose solution containing 0.1% (w/v) aqueous polysorbate 80 and given at 5 mL/kg body weight.

Kinase inhibition. Flt-4 tyrosine kinase was obtained from ProQinase GmbH (Freiburg, Germany) and c-Kit tyrosine kinase from Upstate Biotechnology, Inc. (Lake Placid, NY). Each additional kinase used was generated as a cell lysate, following infection of insect cells with recombinant baculoviruses containing kinase domains. All enzyme assays were run at, or just below, the respective K_m for ATP (0.2-30 μmol/L). The inhibitory activity of AZD2171 was determined against a range of recombinant tyrosine kinases [KDR, Flt-1, Flt-4, c-Kit, PDGFR-α, PDGFR-β, CSF-1R, Flt-3, FGFR1, Src, Abl, epidermal growth factor receptor (EGFR), ErbB2, Aur-A, and Aur-B] using ELISA methodology described previously (28). Selectivity versus CDK2 and CDK4 serine/threonine kinases was examined using scintillation proximity assays with a retinoblastoma substrate and [γ-³³P]ATP (29). Activity versus the dual specificity kinase MAPK kinase (MEK), was determined with a MAPK substrate, [γ-³³P]ATP, and paper capture/scintillation counting (29). Microcal Origin software (vs. 3.78, Microcal Software, Inc., Northhampton, MA) was used to interpolate IC₅₀ values by nonlinear regression.

Inhibition of growth factor–stimulated receptor phosphorylation. Inhibition of receptor phosphorylation within cells was determined using ELISA (KDR and EGFR) or Western blotting (c-Kit, PDGFR-α and PDGFR-β, or Flt-3) methods, or methods incorporating a fluorescent immunocytochemical end point (CSF1-R or ErbB2). Parental cell lines were purchased from the American Type Tissue Collection (ATCC, Manassas, VA), except for MonoMac6 cells, which were obtained from the German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany). With the exception of experiments examining erbB2, cells were serum starved overnight, incubated with AZD2171 for 60 to 120 minutes, and stimulated, where required, with the relevant ligand: VEGF₁₆₅ (100 ng/mL) for 7 minutes, SCF (50 ng/mL) for 10 minutes, PDGF-AA or PDGF-BB (50 ng/mL) for 5 minutes, EGF (1 μg/mL) for 3 minutes, or CSF-1 (40 ng/mL) for 10 minutes. VEGF, SCF, EGF, and CSF-1 were obtained from R&D Systems, Inc. (Abingdon, United Kingdom) and PDGF-AA and PDGF-BB from Sigma-Aldrich (Poole, United Kingdom). Inhibition of KDR phosphorylation was determined in human umbilical vein endothelial cells (HUVEC) by ELISA at ProQinase. EGFR phosphorylation was examined using a sandwich ELISA in KB human oral squamous tumor cells, with an sc-120 capture antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) and detection via horseradish peroxidase immunoconjugated to a 4G10 phosphotyrosine antibody (Upstate Biotechnology).

Western blotting was done using standard SDS-PAGE methods, loading 50 to 75 μg of protein per lane, with detection by enhanced chemiluminescence. Total and phosphorylated c-Kit was measured in NCI-H526 human small cell lung tumor cells, using antibodies to c-Kit (Santa Cruz Biotechnology) and phospho-c-Kit (Cell Signaling Technology, Inc., Beverly, MA). PDGFR-α and PDGFR-β phosphorylation was examined in MG63 human osteosarcoma cells, using antibodies to total (R&D Systems) and phosphorylated (Santa Cruz Biotechnology and Cell Signaling Technology) receptors. Inhibition of Flt-3 phosphorylation was investigated in MonoMac6 human acute myeloid leukemia cells, which contain an activating mutation in the juxtamembrane sequence of the receptor, using Flt-3 (Cell Signaling Technology) and phospho-Flt-3 (Cell Signaling Technology) antibodies.

ErbB2 phosphorylation was examined in MCF-7 breast carcinoma cells stably expressing constitutively activated wild-type erbB2 receptor. Cells were incubated with AZD2171 for 4 hours at 37°C, fixed in 3.3% formaldehyde/PBS, and immunostained with a phospho-erbB2 primary antibody (Santa Cruz Biotechnology) followed by an Alexa-Fluor 488 secondary antibody (Molecular Probes, Paisley, United Kingdom). ErbB2 phosphorylation was quantitated using an Acumen Explorer fluorescence laser-scanning plate reader (TTP Labtech Ltd., Royston, United Kingdom). CSF-1R phosphorylation was detected in 3T3 murine fibroblasts stably transfected with human CSF-1R. Cells were fixed in 4% formaldehyde and immunostained using a primary antibody to phosphorylated CSF-1R (Cell Signaling Technology). CSF-1R phosphorylation was quantitated using a Cellomics ArrayScan HCS Reader (Cellomics, Inc., Pittsburgh, PA).

Inhibition of KDR phosphorylation was also examined in HUVEC by Western blotting. HUVEC were serum starved overnight, incubated with AZD2171 for 90 minutes and stimulated with VEGF (50 ng/mL) for 8 minutes. Human VEGF₁₆₅ was generated as described previously (29). Antibodies against pKDR (rabbit polyclonal generated to phosphorylated Y1214), KDR (Santa Cruz Biotechnology), pMAPK (Cell Signaling Technology), or MAPK (Cell Signaling Technology) were used.

Inhibition of growth factor–mediated cellular proliferation. HUVEC proliferation in the presence and absence of growth factors was evaluated following a 4-day incubation by ³H-thymidine incorporation, as described previously (28). Proliferation of MG63 osteosarcoma cells was induced by PDGF-AA, which selectively activates PDGFR-α homodimer signaling. Cells were cultured in DMEM without phenol red (Sigma-Aldrich) containing 1% charcoal stripped FCS, 2 mmol/L glutamine, and 1% nonessential amino acids (Invitrogen, Paisley, United Kingdom) for 24 hours. AZD2171 or vehicle was added with PDGF-AA ligand (50 ng/mL; Sigma-Aldrich) and plates reincubated for 72 hours. Cellular proliferation was determined using a bromodeoxyuridine ELISA (Roche Diagnostics Ltd., Lewes, United Kingdom).

Inhibition of tumor cell proliferation in vitro. The human tumor cell lines, Calu-6 (lung carcinoma), SW620 (colorectal carcinoma), MDA-MB-231 (mammary gland adenocarcinoma), PC-3 (prostate adenocarcinoma), and SKOV-3 (ovarian adenocarcinoma) were obtained from ATCC. Cell proliferation was determined over a 3-day period as described previously (30).

Human in vitro angiogenesis assay. HUVECs and human diploid fibroblasts were obtained (day 0) as cocultures in 24-well plates (AngioKit, TCS CellWorks Ltd., Buckinghamshire, United Kingdom) and the medium immediately replaced with MCDB 131 medium (Life Technologies, Paisley, United Kingdom) containing 2% FCS, 1% glutamine, and 1% penicillin/streptomycin (Sigma, Poole, United Kingdom). AZD2171 or vehicle (0.01% DMSO in MCDB 131) was added to the cocultures. Medium, with AZD2171 or vehicle, was replenished on days 3, 6, and 8. Tubule formation was examined at day 10 following fixing and staining of tubules for CD31 (platelet/endothelial cell adhesion molecule 1) according to the manufacturer's instructions. Comparative treatment with either an antihuman VEGF neutralizing antibody or a control immunoglobulin G antibody (30 μg/mL, R&D Systems) was also examined.

To quantify tubule growth, a novel whole-well image analysis method was developed using a Zeiss KS400 3.0 Image Analyser (Imaging Associates Ltd., Bicester, United Kingdom) and a bespoke computer program. Tubule formations within each well were measured excluding a rim of 100-μm depth to avoid edge retraction artifact. The computer program segmented the images using a gray level threshold tool to select the stained cells of interest. The resultant binary images were skeletonized and nodal junctions (branch points) removed to determine the total length of an individual tubule. The branch points were counted and the total area of CD31 staining determined from the original binary image but with a correction for cells that had not migrated (labeled cells with an area of <20 μm).

Inhibition of vascular endothelial growth factor–induced angiogenesis in vivo. A Matrigel plug assay done by Vasculogen (Minneapolis, MN) was used to examine VEGF-induced angiogenesis directly. Briefly, Matrigel (0.5 mL) containing heparin (20 units/mL; Sigma Chemicals, St. Louis, MO) was diluted 3:1 in HBSS or recombinant human VEGF₁₆₅ (150 ng/mL; R&D systems) and injected s.c. into female, BALB/c athymic (nu/nu genotype) mice of 6 to 8 weeks of age (five per group). AZD2171 (1.5 or 6 mg per kg per day) or vehicle was given orally, from the day of Matrigel implantation, for 7 days. Matrigel plugs were recovered on day 8, one half being snap-frozen in liquid nitrogen using Tissue-Tek, ornithine carbamyl transferase compound (Fisher Scientific, Pittsburgh, PA) and the other fixed in neutral formalin and embedded in paraffin for H&E staining. Frozen sections were stained with a rat monoclonal antibody reactive to mouse CD31 conjugated to phycoerythrin (BD Biosciences PharMingen, San Diego, CA). Two sections were cut from the Matrigel samples at different levels and analyzed for vessel density. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) and immunofluorescent images (n = 7-10) of CD31 staining obtained from each sample at random. Each area was photographed under both red and UV filters. Morphometric analysis of vessel density and architecture was carried out after converting the images into a binary image using Adobe Software. Images were skeletonized using a program that converts binary data into single pixel density tracings (31). This enabled the number of vessel ends, vessel branch points (nodes), and total vessel length to be quantified with an Image Processing Toolkit (RGI, Inc., Raleigh, NC).

Effect of AZD2171 on bone growth and luteal development. Young female Alderley Park rats (6 weeks of age, Wistar derived, n = 5) were dosed orally, once daily for 28 days with AZD2171 (1.25-5 mg per kg per day) or vehicle. Additional rats (five per group) were treated with AZD2171 (5 mg per kg per day) or vehicle for 28 days and maintained for a further 28 days without treatment, to examine the effect of compound withdrawal. Histologic paraffin wax sections of the femorotibial joints and ovaries were stained with H&E. Morphometric image analysis of femorotibial sections was done (29), with growth plate areas from both the femur and tibia in each joint being combined for an analysis of the effect of compound treatment. The area of corpora lutea in H&E-stained ovary sections was similarly determined by morphometric analysis (Joyce-Loebl Magiscan Image Analyser, Applied Imaging Ltd., Newcastle upon Tyne, United Kingdom).

In vivo tumor models. Protocols for establishing s.c. PC-3, Calu-6, SKOV-3, MDA-MB-231, and SW620 tumors in female nude (nu/nu genotype) mice were as described previously (29, 30). When tumors reached a volume of 0.1 to 0.5 cm³, mice were randomized (6-12 per group) and AZD2171 (0.75-6 mg per kg per day) or vehicle given once daily by oral gavage. Tumor volumes were assessed by bilateral Vernier caliper measurement at least twice weekly and calculated using the formula (length × width) × √ (length × width) × (π/6), where length was taken to be the longest diameter across the tumor and width the corresponding perpendicular. Growth inhibition was calculated from the start of treatment by comparison of the mean change in tumor volume for control and treated groups. To remove any size dependency before statistical evaluation (the variance in mean tumor volume data increases proportionally with volume and is therefore disproportionate between groups), data was log-transformed before statistical evaluation using a one-tailed two-sample t test.

Histologic assessment of tumor vasculature in response to AZD2171 therapy. Mice bearing established Calu-6 human lung tumor xenografts (0.2 ± 0.01 cm³, mean volume ± SE) were selected (day 0) and treated chronically with AZD2171 (6 mg per kg per day, p.o.) or vehicle. Tumors were collected (6-15 per group) 4 hours after the last dose of AZD2171 or vehicle, on days 1, 2, 7, 14, and 21. CD31 was then detected in sections using a chromagen end point or fluorescent immunostaining. CD31 staining with a chromagen end point was done on tumor specimens fixed in zinc fixative (PharMingen) using methodology described previously (29). These were analyzed blind to treatment assignment using a KS400 instrument (Imaging Associates). CD31-positive vessel number and total CD31 staining area/5,000 μm² viable tumor area were calculated for each section. For CD31 fluorescent immunostaining, formalin-fixed sections were incubated with a CD31 antibody (Santa Cruz Biotechnology) in serum block followed by an immunoglobulin G conjugated to Alexa Fluor 488 (Molecular Probes) and counterstained using ProLong Gold anti-fade reagent with DAPI (Molecular Probes). Fluorescent CD31 staining was visualized using an Axiovert S100 fluorescent microscope (Carl Zeiss SMT, Inc., Thornwood, NY) and images analyzed using MetaMorph version 6.1 software (Universal Imaging Co., Downingtown, PA).

AZD2171 is a highly potent inhibitor of KDR tyrosine kinase and shows selectivity versus a range of additional kinases. AZD2171 is a highly potent inhibitor of recombinant KDR tyrosine kinase activity in vitro (IC₅₀ < 1 nmol/L; Table 1). Additional activity is observed against the kinase associated with Flt-1 (IC₅₀ = 5 nmol/L) and the VEGF-C and VEGF-D receptor Flt-4 (IC₅₀ ≤ 3 nmol/L). The inhibitory activity of AZD2171 was also examined against each recombinant PDGFR-related kinase in vitro because of their structural similarity to the VEGF family of receptors. The IC₅₀ values for inhibition of c-Kit and PDGFRβ tyrosine kinase (2 and 5 nmol/L, respectively) were in a range similar to that measured versus Flt-1 and Flt-4. However, when compared with KDR kinase inhibition, AZD2171 selectivity versus the remaining PDGFR-related members ranged from >36-fold (PDGFR-α) to >1,000-fold (Flt-3). Excellent selectivity for KDR was evident versus a range of unrelated tyrosine and serine/threonine kinases, including EGFR (>1,600-fold selectivity) and MEK (>10,000-fold selectivity; Table 1). Furthermore, in addition to the data presented, no inhibition of enzyme activity was detected when 10 μmol/L of AZD2171 was examined with 100 μmol/L ATP against AMPK, ChK1, c-jun NH₂-terminal kinase, MAPK2, MSK-1, PKA, Akt/PKB, PKCα, Rock II, SAPK2b, SAPK2c, SGK, CSK, and PI 3-kinase (data not shown).

AZD2171 inhibits vascular endothelial growth factor–induced KDR phosphorylation in human endothelial cells. In HUVEC, AZD2171 inhibited VEGF-stimulated phosphorylation of KDR in a dose-dependent manner with an IC₅₀ value of 0.0005 μmol/L (Table 2). The phosphorylation of MAPK, a downstream marker of VEGF signaling, was inhibited concomitantly (Fig. 2). To assess AZD2171 selectivity more critically, particularly between the PDGFR family of receptors, a range of cell lines were used to examine inhibition of receptor phosphorylation at a cellular level (Table 2). In comparison with inhibition of KDR phosphorylation, the AZD2171 IC₅₀ for inhibition of c-Kit phosphorylation in NCI-H562 cells was found only 2-fold higher, whereas 10- and 16-fold higher concentrations, respectively, were required for comparable inhibition of PDGFR-α and PDGFR-β phosphorylation in MG63 osteosarcoma cells. In accordance with the kinase selectivity profile, AZD2171 showed a high degree of selectivity (420- to >20,000-fold) for inhibition of KDR phosphorylation versus the remaining PDGFR family members, CSF-1R and Flt-3, and versus the erb family members, EGFR and erbB2 (Table 2).

AZD2171 inhibits vascular endothelial growth factor–induced proliferation potently and shows selectivity versus mitogenesis induced by other growth factors. Consistent with the activity shown against VEGF-stimulated KDR autophosphorylation in HUVEC, AZD2171 inhibited VEGF-stimulated HUVEC proliferation potently with an IC₅₀ value of 0.0004 ± 0.0002 μmol/L (Table 3). AZD2171 was found more potent against VEGF-induced HUVEC proliferation (IC₅₀ = 0.4 nmol/L) than other VEGFR tyrosine kinase inhibitors that have entered clinical development, including the pthalazine PTK787 (IC₅₀ = 0.008 ± 0.001 μmol/L, mean ± SE; n = 4), the indolinone SU11248 (IC₅₀ = 0.04 ± 0.02 μmol/L, mean ± SD; n = 2) and the isothiazole CP-547,632 (IC₅₀ = 0.06 ± 0.001 μmol/L, mean ± SD; n = 2; data not shown). To address the functional relevance of inhibiting receptor phosphorylation in human cells, AZD2171 selectivity was examined further against proliferation induced by different growth factors. In HUVEC, AZD2171 showed selectivity of 275-fold for inhibition of VEGF-induced proliferation versus basic fibroblast growth factor (bFGF)–induced proliferation, and 1,250-fold selectivity versus an effect on EGF-induced proliferation (Table 3). The effect of AZD2171 on PDGFR-α–dependent cellular proliferation (stimulated by PDGF-AA) was examined in MG63 cells, and an IC₅₀ value of 0.04 μmol/L determined. This concentration is 100-fold greater than that required for comparable inhibition of VEGF-induced proliferation in HUVEC. The data indicate that AZD2171 can selectively inhibit VEGFR-dependent proliferation, in contrast to proliferation mediated by FGFR1 or EGFR, and that appreciable functional selectivity is also evident versus PDGFRα.

Micromolar concentrations of AZD2171 are required to inhibit tumor cell proliferation directly in vitro. The ability of AZD2171 to inhibit tumor cell growth directly was examined in vitro to ascertain whether subsequent antitumor activity in vivo could be ascribed to a direct antiproliferative effect or an indirect effect from inhibiting VEGF signaling. AZD2171 only inhibited tumor cell proliferation in vitro at comparatively high concentrations. The IC₅₀ values (mean ± SE) determined from three to five independent experiments against tumor cells were 3.0 ± 0.4 μmol/L (SKOV-3), 3.8 ± 0.5 μmol/L (MDA-MB-231), 5.8 ± 0.2 μmol/L (PC-3), 6.4 ± 0.6 μmol/L (Calu-6), and 7.4 ± 0.7 μmol/L (SW620). These concentrations are between 7,500- and 18,500-fold greater than required for comparable inhibition of VEGF-stimulated HUVEC proliferation and were not attained within in vivo tumor xenograft experiments (data not shown). The ability of AZD2171 to inhibit tumor growth in vivo can therefore be attributed to inhibition of VEGF signaling.

AZD2171 inhibits tubule sprouting in vitro at subnanomolar concentrations and prevents VEGF-induced angiogenesis in vivo. Although the complex angiogenic phenotype cannot be completely replicated in vitro, assays can be used to model vascular sprouting. We modified a commercially available fibroblast/endothelial cell coculture system (AngioKit, TCS CellWorks) so that exogenous bFGF was omitted from the incubation medium. Under these conditions, tubule growth is regulated more directly by VEGF, production of which increases during the course of the assay.⁵

When compared with treatment with an isotype control antibody (Fig. 3B,, iv) that did not affect tubule growth, a VEGF-neutralizing antibody inhibited vessel development significantly (Fig. 3B,, v). Consistent with potent activity against VEGF signaling, AZD2171 inhibited vessel branching, length, and area (Fig. 3) with IC₅₀ values (mean ± SE; five independent experiments) of 0.0001 ± 0.00004, 0.0001 ± 0.00003, and 0.0002 ± 0.00007 μmol/L, respectively. To examine inhibition of angiogenesis in vivo that was selectively driven by VEGF, Matrigel plugs containing VEGF were implanted s.c. in mice and vessel development examined over the course of 8 days. AZD2171 completely abolished VEGF-induced vessel formation. The data obtained with 6 mg per kg per day AZD2171 versus number of vessel ends, branch points and total vessel length was not significantly different to that observed with 1.5 mg per kg per day (Fig. 4; P > 0.05 by one-tailed t test). Collectively, these data show that AZD2171 is a potent inhibitor of VEGF-induced angiogenesis.

AZD2171 induces hypertrophy in bone growth plate and inhibits luteal development in ovary: physiologic processes that are dependent upon angiogenesis. Angiogenesis and VEGF signaling are essential events in endochondral ossification. An inhibitor of VEGF receptor tyrosine kinase activity should therefore prevent further ossification in the epiphyseal growth plates of growing animals (growth plates are closed in adults) and increase the zone of hypertrophy. AZD2171 produced a dose-dependent increase in the hypertrophic chondrocyte zone of the tibial and femoral growth plates in female rats (Fig. 5A and B). Administration of 1.25, 2.5, or 5.0 mg per kg per day AZD2171 (p.o.) increased the combined epiphyseal growth plate area by 36%, 283%, and 481%, respectively. A further 28-day period without treatment (subsequent to administration of 5 mg per kg per day AZD2171 for 28 days) resulted in complete reversal of this phenotype (Fig. 5C).

The rapid development and growth of the ovarian corpus luteum is also critically dependent upon vascular growth. This angiogenic process is predominantly regulated by VEGF. An ovarian cycle time of only 4 days in rat means that histologic sections of an ovary reveal a dynamic picture of preovulatory and regressing follicles. Chronic administration of AZD2171 (5 mg per kg per day p.o. for 28 days) afforded a marked reduction in luteal area when compared with the ovaries of control (vehicle treated) animals (67% inhibition: Fig. 5D and E).

The effect of AZD2171 treatment on bone growth plate and ovary is supportive of inhibition of VEGF-induced angiogenesis within complex physiologic settings.

AZD2171 shows broad-spectrum activity in human tumor models at doses that are well tolerated. Once-daily oral administration of AZD2171 (1.5 mg per kg per day) inhibited growth of all human tumor xenografts examined in nude mice, irrespective of the histologic type (Table 4; Fig. 6). Statistically significant growth inhibition was obtained with 0.75 mg per kg per day AZD2171 in three of five tumor models examined, and the growth of each tumor model inhibited by >90% following administration of 6 mg per kg per day. AZD2171 treatment was well tolerated; following continuous administration of 6 mg per kg per day for 3 weeks (data accumulated from all experiments over 21 days), body weights were comparable with pretreatment values and only 1.5 % different to those of control animals (Fig. 6B). This is particularly notable when the weight of the additional tumor burden in control animals is considered. The broad-spectrum antitumor activity observed with AZD2171 may be attributed to a common effect on tumor vasculature, and its good tolerability to a comparatively selective profile in vivo.

AZD2171 induces vascular regression in human lung tumor xenografts. To determine whether AZD2171 affects the survival and morphology of tumor vasculature, mice bearing established Calu-6 tumors were randomized and treated chronically with AZD2171 (6 mg per kg per day) or vehicle. Mice were removed from each group at intervals to compare changes in tumor vascular area and vessel number (Fig. 7A). The mean control Calu-6 tumor volume was found to increase by ∼6-fold over the 21-day period examined, and AZD2171 treatment was found to inhibit this tumor growth by 68% (Fig. 7A).

Immunohistochemical analysis of control tumors indicated that when the CD31-positive area or number of microvessels were normalized to tumor area, the values remained relatively constant throughout the duration of the experiment, irrespective of the accompanying volume change (Fig. 7B and C). Whereas an apparent difference (P = 0.02, one-tailed t test) was found between CD31 area (per 5,000 μm²) on day 1 versus day 21, the CD31 area measured on day 1 was higher than in all subsequent groups, including day 2 where tumor volumes were directly comparable. The CD31 area on day 21 was not significantly different (P = 0.3) from that measured on day 2 (Fig. 7B), and the vessel density (number of vessels per 5,000 μm²) on day 21 was not significantly different from that on day 1 (P = 0.3; Fig. 7C).

With acute AZD2171 treatment (day 1: two doses of AZD2171 given 28 and 4 hours before sampling) neither vascular variable was affected significantly. However, following an additional dose of AZD2171 (i.e., day 2: AZD2171 given at 52, 28, and 4 hours before sampling), a 47% reduction in vessel number was observed (P = 0.02), accompanied by a 40% reduction in CD31 area that was just outside statistical significance (P = 0.06). By day 21, vessel number and CD31 area were reduced by ≥70% (P < 0.005). These data indicate that AZD2171 can cause significant vascular regression in tumors. The rapid onset of vessel regression is most probably due to a direct effect of AZD2171 on tumor endothelium, which is likely to be derived from potent inhibition of VEGF signaling.

Inhibition of VEGF signaling is being pursued avidly as a therapeutic strategy in oncology, to inhibit angiogenesis, neovascular survival, and vascular permeability in tumors. These VEGF-induced responses are thought dependent upon the KDR signal transduction in endothelial cells (18, 19, 32). AZD2171 may therefore have potential as an antitumor therapy, by virtue of its potent inhibitory activity against KDR tyrosine kinase. Subnanomolar IC₅₀ values were obtained when AZD2171 was examined against recombinant KDR kinase activity and VEGF-induced KDR phosphorylation in human endothelial cells, with concomitant inhibition of VEGF-induced phenotypes.

AZD2171 also showed activity at low nanomolar concentrations in enzyme assays, versus the kinases associated with the two other VEGF receptor family members, Flt-1 and Flt-4. The exact role of Flt-1 signaling in physiologic and pathologic angiogenesis remains unclear. Flt-1 has a relatively weak signaling capacity in vitro and, in contrast to KDR, a number of reports suggest that its stimulation has minimal effects on mitogenic, motogenic, or permeability responses in endothelial cells (33–35). Flt-1 with a deleted kinase domain has also been shown to support normal vascular development in mice, suggesting that Flt-1–stimulated angiogenic sprouting may be mediated by ligand sequestration, to regulate the spatial availability of VEGF, rather than via an intrinsic signaling response (36). Despite these observations, Flt-1 has been shown to co-operate directly with KDR via heterodimerization, and Flt-1 homodimers may transactivate KDR through cross-talk (37). Furthermore, Flt-1 alone can bind two additional VEGF homologues, PlGF and VEGF-B, and the former has been shown, under experimental in vitro conditions, to induce differential signaling through Flt-1 when compared with VEGF itself (37). These findings support the possibility that direct inhibition of Flt-1 tyrosine kinase may be beneficial. In contrast to Flt-1, Flt-4 does not bind VEGF, PlGF, or VEGF-B but only binds the homologues VEGF-C and VEGF-D. This receptor has a critical role in lymphangiogenesis (38) and a prognostic link with expression of VEGF-C and/or VEGF-D and nodal metastasis has been identified for different tumor types (39, 40). Experimentally, lymph node metastasis can be promoted by expression of VEGF-C (41) and inhibited with an antibody to Flt-4 (42). Direct inhibition of Flt-4 signaling may therefore also have therapeutic benefit in limiting subsequent tumor dissemination. However, it has recently been suggested that KDR heterodimerization (induced by VEGF-C or VEGF-D) is required for ligand-stimulated phosphorylation of Flt-4 and its associated cellular signaling (43); a dependency that would be perturbed directly by a KDR tyrosine kinase inhibitor. Determining the full relevance of AZD2171 activity against Flt-1 and Flt-4 kinase will necessitate additional investigation of receptor signaling in cells.

The selectivity of an ATP-competitive kinase inhibitor in vivo is dependent upon the comparative potency of the compound against a target, the degree of inhibition required to prevent a given phenotypic effect and the level of plasma/tissue exposure attained at the evaluated dose. Collectively, the enzyme, receptor phosphorylation, and cellular proliferation data obtained with AZD2171 indicate that it has selectivity for inhibition of VEGF signaling but also suggest that it may have relevant activity versus c-Kit tyrosine kinase which could provide added therapeutic benefit in the treatment of c-Kit–dependent tumors (44, 45). That large concentrations of AZD2171 were required to inhibit the growth of tumor cells directly in vitro, and that AZD2171 was particularly well tolerated in tumor xenograft models at doses that proved highly efficacious, is in further support of a selective inhibitory profile.

AZD2171 combines potent activity versus KDR tyrosine kinase and selectivity with pharmacokinetic properties that are appropriate for oral once-daily administration. In female rat, studies examining i.v. (5 mg/kg) and oral (30 mg/kg) dosing of AZD2171 revealed a terminal plasma half-life of 9 hours, a relatively low clearance of 0.8 L per hour per kg, an oral bioavailability of 60%, and an unbound plasma fraction of 6% (data not shown). For this reason, all preclinical in vivo work was conducted by daily oral gavage.

AZD2171 treatment inhibited physiologic processes that are critically dependent upon VEGF signaling and angiogenesis. During endochondral bone formation, VEGF expression by hypertrophic chondrocytes regulates metaphyseal angiogenesis (46), with capillaries that invade the cartilage expressing both Flt-1 and KDR (47). This vascular invasion is critical to enable the terminal differentiation and apoptosis of growth plate chondrocytes, matrix resorption by osteoclasts, and mineralization by osteoblasts. The survival of osteoclasts (48) and migration and differentiation of osteoblastic cells (49, 50), both of which express Flt-1 and KDR, can also be influenced directly by VEGF signaling. Furthermore, chondrocyte survival has recently been suggested to have a VEGF dependency (51). Consistent with a fundamental regulatory role for VEGF signaling in bone morphogenesis, AZD2171 perturbed endochondral ossification in growing rats significantly and produced a marked growth plate hypertrophy; a phenotype also induced by VEGF sequestration (52) or conditional knockout of VEGF in cartilage (53). Whereas this phenotype has also been reported with other VEGF receptor tyrosine kinase inhibitors (26, 30), the doses of AZD2171 that produce significant effects are comparatively much lower. The magnitude of hypertrophy induced by AZD2171 has also not been shown previously. This phenotype was reversed completely following an additional 28-day period of AZD2171 withdrawal, indicating that chronic inhibition of VEGF signaling is required to maintain an effect in vivo. In addition to inhibiting ossification in bone growth plate, AZD2171 also markedly reduced corpora luteal area in rat ovary. The cyclic corpus luteum of the ovary is considered the site of strongest physiologic angiogenesis, with rapid luteal development involving distinct phases of VEGF-dependent blood vessel growth and maturation, followed by vessel regression during luteolysis that involves endothelial cell detachment (54). This is the first study to report the effects of a VEGF receptor kinase inhibitor on follicular development, although a soluble VEGF receptor construct has been previously found to have inhibitory effects on corpora luteal angiogenesis and development in mice (55) and a KDR blocking antibody shown to significantly delay follicular development in rhesus monkey (56). Analogous changes in bone growth plate and ovary have also been observed in primate following treatment with the VEGF-neutralizing antibody, bevacizumab (57).

AZD2171 (1.5 and 6 mg per kg per day) completely abolished VEGF-dependent angiogenesis in vivo, in a s.c. Matrigel plug assay, suggesting that it might also inhibit tumor growth at comparatively small doses. Doses of 6 mg per kg per day AZD2171 and lower were examined for activity in a panel of established, histologically distinct (colon, lung, prostate, breast, and ovary), human tumor xenograft models in athymic mice. Statistically significant inhibition was obtained with 1.5 mg per kg per day AZD2171 in all models and with 0.75 mg per kg per day in a number of models. The broad-spectrum antitumor profile observed with AZD2171 is consistent with an effect on tumor vasculature, a common growth-limiting target in all models, as opposed to tumor cell targets that are subject to variable expression and dependency. AZD2171 was efficacious at doses that are significantly lower than reported with other VEGFR tyrosine kinase inhibitors, which require administration within the 20 to 100 mg per kg per day range to achieve significant inhibition of tumor growth in mice (24–26).

To investigate temporal changes in tumor vessel growth and survival following AZD2171 treatment, a Calu-6 lung tumor xenograft model was examined, which is known to be growth inhibited by VEGF inhibition but does not regress in response to treatment (29). Vascular density in control Calu-6 tumors did not change significantly during 21 days of growth. In contrast, a comparison of vascular density in AZD2171-treated tumors with corresponding time-matched controls revealed a 47% reduction after 52 hours of treatment and a 73% reduction following 3 weeks of treatment. These data indicate that progressive vascular regression occurs in human tumor xenografts treated chronically with AZD2171, demonstrating that inhibition of VEGF signaling can have a potent antivascular effect. A number of preclinical studies with VEGF sequestering agents have also shown evidence of reduced vascular density in tumors following treatment. These include regression of existing tumor vessels with a VEGFR-Fc fusion construct in tumor bearing RIP-Tag2 transgenic mice (58), a KDR blocking antibody in squamous cell tumor xenografts (59), and a VEGF-neutralizing antibody in LS174T colon tumor xenografts (60). Primitive vessels are believed to be susceptible to VEGF blockade, because their survival is highly dependent upon KDR mediated Akt/PKB signaling (8). These data serve to emphasize that the consequences of inhibiting VEGF signaling in pathologic conditions may extend beyond simply preventing new vessel growth.

That inhibition of VEGF signaling can provide clinical benefit to patients with solid tumors, has been confirmed recently by bevacizumab (Avastin, Genentech, South San Francisco, CA), a monoclonal antibody to VEGF-A. In combination with irinotecan, 5-fluorouracil and leucovorin, bevacizumab imparts a significant prolongation of survival and response variables in patients with first-line metastatic colorectal cancer (61). An additional small preoperative biomarker study has also suggested antivascular effects in rectal carcinoma patients, following bevacizumab treatment (62). Direct inhibition of KDR tyrosine kinase activity with a small molecule, such as AZD2171, is an alternative therapeutic approach that may afford advantages to sequestration of VEGF. Inhibiting the kinase domain of KDR is mechanistically different, in that it should prevent receptor signaling irrespective of the activating ligand. In addition to VEGF-A, the fully proteolytically cleaved forms of the related VEGF gene family members VEGF-C and VEGF-D can also bind to and activate KDR (63, 64). In particular, VEGF-C can induce proliferative, migratory, tubulogenic, and permeability responses through KDR (63, 65). Chronic administration of a small molecule to generate steady state plasma levels may also provide a pharmacokinetic advantage, because the distribution and clearance of a large biopharmaceutical may possibly be influenced more highly by heterogeneous tumor anatomy and variable tumor blood flow (66).

In summary, AZD2171 is a highly potent inhibitor of KDR tyrosine kinase that has pharmacokinetic properties suitable for once-daily oral administration. VEGF-induced angiogenesis, neovascular survival, and growth of human tumor xenografts, are inhibited significantly by AZD2171, at comparatively low doses. AZD2171 is currently being evaluated in clinical trials as a once-daily oral therapy for the treatment of a variety of malignancies. Encouragingly, these clinical studies indicate that AZD2171 retains a good pharmacokinetic profile in man, with dose-proportional increases in plasma exposure and a mean terminal plasma half-life of 20 hours (67).

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 Karen Malbon, Claire Barnes, Minwei Ye, Stephen Brightwell, John Stawpert, and Sandra Oakes for the technical assistance in selectivity testing; Kevin Randal for morphometric image analysis; David Blowers, Ian Taylor, Hazel Weir, and Rick Davies for protein production and purification; Mick Shaw for the isolation and supply of HUVECs; and Michel Vautier for doing the initial synthesis of AZD2171.