Test Materials and Cell Lines.

SF0166-FA, (S)-3-(6-(difluoromethoxy) pyridin-3-yl)-3-(oxo-2-3-(3-(5,6,7,8-tetrahydro-1,8-naphthyridin-2-yl)propyl)imidazolidin-1-yl)propanoic acid (mol. wt. of 475 g/mol), and SF0166, the 2-amino-2-methyl-2-propanol salt (mol. wt. of 565 g/mol), were manufactured by Anthem Biosciences (Bangalore, India) (Askew et al., 2014). Both forms of the test article are referred to as SF0166 in this article. The free acid form (SF0166-FA) was used in all studies except the VEGF-induced neovascularization in Dutch-Belted rabbits and the ocular distribution analyses in rabbits.

Human dermal microvascular endothelial cells (HMVEC) were from Lonza (Basel, Switzerland); rabbit aortic endothelial cells (RAEC) were from Cell Biologics (Chicago, IL); rat lung microvascular endothelial cells (RLMVEC) were from Vec Technologies (Rensselear, NY); K562 (leukemia) and HT29 (adenocarcinoma) were from American Type Culture Collection (Manassas, VA); and canine aortic endothelial cells (CnAOEC) were from Cell Applications (San Diego, CA).

Animals.

This research was carried out in accordance with the Guide for the Care and Use of Laboratory Animals as adopted and promulgated by the National Institutes of Health and was approved by the clinical research organization’s review board. Female 129SVE newborn mouse pups with mothers were obtained from Charles River Laboratories (Wilmington, MA). Male naive Dutch-Belted rabbits weighing approximately 1.5–2.0 kg were obtained from Covance (Denver, PA) or Western Oregon Rabbit (Philomath, OR). Food and water were supplied ad libitum.

Determination of Physiochemical Properties.

A two-phase system plate was used for octanol/buffer partitioning to determine log D values. Octanol in equilibrium with universal buffer (0.15 M NaCl and 0.100 M each of phosphoric, boric, and acetic acids) was adjusted with NaOH to pH 2.0, 4.0, 7.4, and 9.0. SF0166 and standards in 10 mM dimethylsulfoxide (DMSO) were added to the plates for a final concentration of 10% DMSO. The plates were sealed, vortexed, and centrifuged to aid in phase settling. The assay was conducted on an ADW workstation using chemiluminescent nitrogen detection (Analiza, Cleveland, OH). Calculated log D values were corrected for background nitrogen.

To determine pKa values, SF0166 in 10 mM DMSO was diluted with 2 mM HCl and methanol to achieve final concentrations of 60% methanol, 100 µM SF0166, and 1% DMSO. The assay was conducted in 96-well plates using a pKa PRO Analyzer and Estimator software (AATI, Ames, IA). The average pH spacing between buffer points was 0.4 pH U, covering a pH range of 1.7–11.2. Four consecutive runs were conducted starting with 60% cosolvent and decreasing to 30% cosolvent buffers.

To assess permeability, SF0166 in 10 mM DMSO was diluted with 1× phosphate-buffered saline (PBS) for a final concentration of 200 µM SF0166, loaded onto a BD Gentest precoated PAMPA plate (BD Biosciences, San Jose, CA), and incubated 5 hours in the dark at ambient temperature. Samples were then diluted 20-fold with a 50:50 mixture of mobile phase components (0.1% formic acid in water:0.1% formic acid in acetonitrile) and analyzed using an Agilent Technologies (Santa Clara, CA) 1100 high-performance liquid chromatography coupled with a CTC HTC-PAL autosampler (CTC Analytics, Lake Elmo, MN) and AQUASIL C18 column (3 µM, 50 × 2.1 mm; Thermo Fisher Scientific, Waltham, MA). Time-of-flight mass spectometry (TOF MS) data were acquired using an Agilent 6538 Ultra High Accuracy TOF MS in extended dynamic range (m/z 100–1000) using generic MS conditions. Following data acquisition, exact mass extraction and peak integration were performed using MassHunter Software (Agilent Technologies).

Binding to α_vβ₃, α_vβ₅, α_vβ₆, and α_vβ₈ Integrins.

SF0166 was assessed for binding to α_vβ₃, α_vβ₅, α_vβ₆, and α_vβ₈ integrins using a previously published in vitro assay (Ludbrook et al., 2003). Briefly, α_vβ₃, α_vβ₅, α_vβ₆, and α_vβ₈ integrins (R&D Systems, Minneapolis, MN) were coupled to Dynabeads M-270 Epoxy (ThermoFisher Scientific), and conditions for the binding of the coupled Dyna beads to the integrins’ respective ligands were optimized. Ligands were as follows: vitronectin (R&D Systems) for α_vβ₃ and α_vβ₅; latency-associated peptide transforming growth factor-β1 (R&D Systems) for α_vβ₆ and α_vβ₈. Each integrin–ligand pair was treated with SF0166 at concentrations of 3.9–500 nM in 0.08% DMSO (triplicate measures). After a 3-hour incubation, the beads were washed, labeled with a primary antibody to the ligand (R&D Systems) and a secondary antibody conjugated with fluorescein isothiocyanate (Sigma-Aldrich, St. Louis, MO), and analyzed by flow cytometry. The method was validated by screening the effect of echistatin, cilengitide, and CWHM12 (synthesized by Anthem BioSciences) (Henderson et al., 2013) on the appropriate integrin–ligand pair. IC₅₀ values were determined using Prism software (GraphPad, La Jolla, CA).

Testing of Antiadhesion Activity.

The activity of SF0166 in blocking adhesion to vitronectin was assessed in HMVEC, RAEC, and RLMVEC, and in blocking adhesion to vitronectin and fibronectin was assessed in CnAOEC. (Adhesion to vitronectin is mediated by α_vβ₃ and α_vβ₅; adhesion to fibronectin is predominantly mediated by α₅β₁.)

HMVEC were used at passages 9–14, and RAEC and RLMVEC cells were used at passages 4–14. Cells were grown in T175 tissue culture flasks and dislodged by gentle 3-minute treatment with Accutase. After washing, the cells in suspension in RPMI were loaded with calcein-AM (5 µM) for 30 minutes at 37°C and resuspended into RPMI without phenol red containing 10% fetal bovine serum. Assays were performed in 96-well plates, which were coated with vitronectin in PBS, pH 7.4, by incubating 50 µl 10 µg/ml solution for 1.5 hours at room temperature or overnight at 4°C. The plates were then blocked with 1% bovine serum albumin in PBS (30 minutes at room temperature) and washed with PBS. Cell suspensions were plated at a density of 5.0 × 10⁴ (HMVEC and RAEC) or 1.0 × 10⁵ (RLMVEC). SF0166 was added at the same time as the cells. Two lots of SF0166 were tested. The compound was tested using a maximum concentration of 1 μM with half-log dilutions. The plates were incubated for 1.5 hours at 37°C. The plates were gently washed with two cycles of aspiration of the supernatant and addition of 100 µl prewarmed fresh Dulbecco’s PBS (DPBS).

CnAOEC were used at passages 4–6. Cells were grown in T75 or T175 tissue culture flasks in canine endothelial cell growth medium (Cell Applications); cells were then washed with DPBS containing 0.5 M EDTA, and then incubated 5–10 minutes with enzyme-free cell dissociation solution (Sigma-Aldrich). After washing, the cells were diluted to 400 cells/µl in canine endothelial cell growth medium. Assays were performed in 96-well plates, which were coated with vitronectin (50 µl 10 µg/ml solution; 500 ng/well) or fibronectin (1 µg/ml; 50 ng/well) in DPBS by incubating for 2 hours at room temperature. The plates were then blocked with 1% bovine serum albumin in PBS (30 minutes at room temperature) and washed with PBS. SF0166 was tested using a maximum concentration of 1 μM with half-log dilutions. After addition of SF0166, cell suspensions were added at a density of 2.0 × 10⁴ cells/well. The plates were incubated for 1 hour at 37°C. The plates were gently washed with two cycles of aspiration of the supernatant and addition of 100 µl prewarmed fresh canine endothelial cell growth medium. An equal volume of alamarBlue (0.1 mM resazurin solution in DPBS) was added to each well and incubated for 1 hour at 37°C.

The activity of SF0166 was compared with L-000845704 (Murphy et al., 2005), a positive control, for inhibiting the binding to vitronectin and fibronectin of HMVE, RLMVEC, K562 (leukemia), and HT29 (adenocarcinoma) cell lines. HMVEC were used at passages 9–14; RLMVEC and HT29 cells were used at passages 4–10; and K562 cells were used at passages 3–8. Cell culture and assays were conducted as described above with the following differences. Cell suspensions were plated at a density of 2.0 × 10⁵. SF0166 was tested at 10 concentrations with a half-log dilution schedule. Cell adhesion to vitronectin was tested in duplicate or triplicate at compound concentrations ranging from 316 nM to 0.317 pM. Cell adhesion to fibronectin was tested in duplicate or triplicate at compound concentrations ranging from 200 μM to 20 nM for HMVEC and K562 cell lines and from 316 nM to 0.317 pM for HT29. The plates were gently washed with one cycle of aspiration of the supernatant and addition of 100 µl prewarmed fresh DPBS.

Fluorescence of adherent cells was measured using a multimode plate reader (Victor 2V; PerkinElmer, Waltham, MA) at excitation/emission wavelengths of 485/535 nm. IC₅₀ values were calculated with Prism software by fixing the bottom of the curves to a value of blank for empty wells.

Antiangiogenic Activity Using Chick Chorioallantoic Membrane Assay—Basic Fibroblast Growth Factor–, VEGF-, and Platelet-Derived Growth Factor–Induced Models.

The ovo antiangiogenic activity of SF0166 was assessed using the chorioallantoic membrane (CAM) model, a model of angiogenesis that is widely used for evaluation of antiangiogenic therapies (Ponce and Kleinman, 2003). Fertile hen eggs were procured from a hatchery and were cleaned and decontaminated with alcohol. One milliliter of albumin was removed by syringe and incubated for 8 days. CAM surfaces were grafted with gelatin sponges impregnated with 0.5–5 µg SF0166, dissolved in PBS along with 50 ng basic fibroblast growth factor, VEGF, or platelet-derived growth factor (PDGF). A total of 1 µg staurosporine was used as a positive control. The CAMs were further incubated to day 12, and on day 12 the CAMs were fixed with 4% formaldehyde in PBS, dissected, and imaged. Imaging was performed under constant illumination and magnification using a stereo microscope fitted with a digital camera. Image analysis was performed by drawing a ring around each graft and counting the number of vessels crossing the ring (blood vessel score). Data were analyzed using Excel 2007.

Ocular Distribution.

The ocular uptake and ocular tissue distribution of SF0166 were assessed after a single topical ocular administration into both eyes in Dutch-Belted rabbits. The vehicle was boric acid, hydroxypropyl-β-cyclodextrin, EDTA, and benzalkonium chloride. SF0166 was administered at a concentration of 5% and a volume of 50 µl/eye (2.5 mg/eye). Plasma and the ocular tissue samples (aqueous humor, vitreous humor, sclera, and retina–choroid plexus) were collected at 1, 4, 8, and 12 hours postdose [n = 3 animals (six eyes) per timepoint] and stored frozen until analysis. Working solutions were prepared in 50:50 acetonitrile:water. Working solutions were combined with plasma, a 1:1 mixture of aqueous and vitreous humor, or a mixture of blank homogenized sclera, retina, and choroid to make eight-point calibration curves. Plasma samples were analyzed using a validated liquid chromatography with tandem mass spectrometry procedure. Briefly, 100 µl aliquots were mixed with 0.75 ml acetonitrile (Spectrum, New Brunswick, NJ) containing 10 ng SF0166-d3 internal standard. After shaking for 5 minutes, the sample was centrifuged to remove precipitated proteins, and 200 µl of the resulting supernatant was transferred to an autosampler tube, diluted with 0.5 ml American Society for Testing and Materials type I water, and vortex-mixed for instrumental analysis. For the determination of SF0166 in vitreous and aqueous humor, study specimens were homogenized in a shaker for 1 hour and centrifuged, and aliquots were taken for extraction and analysis. Ocular tissue specimens were finely cut, weighed, and treated with 30 parts [extraction volume (microliter) to sample weight (milligram); three parts for lens tissue] of American Society for Testing and Materials type I water and shaken for approximately 0.5 hour. Sample extraction was repeated after the addition of 10 parts of acetonitrile (one part for lens tissue) and another 0.5 hour of shaking. The samples were then centrifuged, and aliquots were taken for extraction and analysis. The specimen aliquots (100 µl) were treated with 200 µl acetonitrile-containing internal standard and shaken for approximately 0.5 hour. The samples were then centrifuged to remove precipitated proteins and diluted with 200 µl water for instrumental analysis. Samples were analyzed using liquid chromatography with tandem mass spectrometry. Ocular pharmacokinetic parameters were determined by noncompartmental analysis using R (version 3.3.1) with the package PKNCA.

OIR Mouse Model.

The efficacy of SF0166 was evaluated in the OIR mouse model. On postnatal day 7, mice were subjected to hyperoxia (75% oxygen) to inhibit retinal vessel growth and induce retinal vessel loss (vaso-obliteration). At postnatal day 12, mice were returned to a normoxic environment to induce retinal neovascularization. On postnatal days 13–17, mice received bilateral topical instillation of vehicle or 5% SF0166 twice daily (BID). At postnatal day 18, retinal tissue was collected, processed, and analyzed to measure vaso-obliteration and neovascularization.

Retinas were stained with isolectin B4 to selectively label retinal vasculature. Stained retinal flat mounts were imaged and analyzed to quantify both vaso-obliteration, defined as the avascular area central to the optic nerve, and neovascularization. Statistical analyses were performed using Prism software using an unpaired t test.

Laser-Induced CNV Model in Dutch-Belted Rabbits.

In the laser-induced CNV model, the safety and efficacy of SF0166 were evaluated over 5 weeks using clinical ophthalmic exams, ocular coherence tomography (OCT), fundus photography, and fluorescein angiography. Prior to placement on study, animals underwent prescreening ophthalmic examinations, including slit-lamp biomicroscopy and indirect ophthalmoscopy. Ocular findings were scored as shown in Supplemental Table 1. Only animals with a score of 0 in all categories were included in the study. Animals were anesthetized for CNV induction, and one drop of topical proparacaine hydrochloride anesthetic (0.5%) was placed in each eye before procedures. CNV was induced by laser treatment on day −3. An external diode laser was applied to the retina using a laser contact lens and a slit-lamp biomicroscope. Both eyes of each animal underwent laser photocoagulation treatment (18–23 spots per eye sized 20–100 µm). Following laser treatment, 50 μl of a 25 μg/ml VEGF solution (1.25 μg dose) was injected intravitreally into each eye, which prolongs the viability of CNV induced by laser injury (Gum et al., 2005). Topical antibiotic ointment was applied after the laser CNV induction procedure. Overall safety was assessed through daily general health and gross ocular observations.

The test articles were SF0166, bevacizumab (Avastin) as positive control, and vehicle. The SF0166 consisted of 2.5 wt% SF0166 dissolved in PBS containing 5 wt% sulfobutyl ether β-cyclodextrin. SF0166 (1.25 mg/eye) and vehicle control were administered topically in 50 μl into each eye BID starting on day 1, except on day 35 when the terminal ocular distribution study was performed. Bevacizumab (1.25 mg/eye) was administered by intravitreal injection once on day 1.

The Heidelberg Spectralis OCT (software version 5.4) was used to capture both the OCT and fluorescein angiography images during the study. Lesion dimensions (widths and heights measured in micrometers) were measured using the OCT images and utilizing the algorithms of the Heidelberg Spectralis OCT software. Lesion sites that were more uniform in shape and away from any anomalies associated with the initial CNV induction were selected for measurements. Four sites were chosen for each eye, and the same four sites were assessed at each timepoint.

The areas of the previously measured lesions were analyzed by individually tracing the overall perimeter of the CNV anomaly using the algorithms of the Heidelberg Spectralis OCT software. Statistical analysis was performed on the area of the lesions for each eye over the course of the study, and the data were analyzed using analysis of variance. Differences among groups were analyzed using Dunnett’s multiple comparison procedure.

VEGF-Induced Neovascularization Model in Dutch-Belted Rabbits.

The VEGF-induced neovascularization and vascular leakage model study evaluated the safety, efficacy, and distribution of SF0166. Prior to placement on study, each animal underwent an ophthalmic examination (slit-lamp biomicroscopy and indirect ophthalmoscopy). Ocular findings were scored as shown in Supplemental Table 2. The acceptance criteria for placement on study were scores of 0 for all variables. Neovascularization and vascular leakage were evaluated by fluorescein angiography, and the images were graded to determine the fluorescein angiography score.

Test article administration to the eyes of six Dutch-Belted rabbits per group began on day 1 of the study [2 days prior to VEGF injection (vascular leakage/neovascularization induction)]. SF0166 was dissolved in 5% hydroxypropyl β-cyclodextrin and polyethylene glycol 6000 in borate buffer. SF0166 at concentrations of 1% to 5% (0.5–2.5 mg/eye) and vehicle control were administered topically in 50 μl to each eye once daily or BID, as shown in Supplemental Table 3. Bevacizumab (1.25 mg/eye) was administered by intravitreal injection once on day 1.

Vascular leakage was induced on day 3 via intravitreal injection of VEGF (500 ng/eye). Clinical ophthalmic examinations and fluorescein imaging were performed at baseline (day −2) prior to the initiation of dosing and prior to termination on day 8. Body weights were recorded prior to the start of dosing and prior to termination. Overall safety was assessed through daily general health observations.

The fluorescein angiography images were evaluated according to the scale shown in Supplemental Table 2 to provide a mean fluorescein angiography score.