A commercial bevacizumab solution (Avastin; Genentech, South San Francisco, CA) was frozen and lyophilized as the API. Under aseptic conditions, the dried material was combined with a PLGA solution (Resomer, RG 755 S; Evonik, Parsippany, NJ) in proportions to make 5% bevacizumab in the PLGA and dissolved in methylene chloride. The mixture was homogenized to make a solid–oil–water dispersion in multiple steps. Bevacizumab–PLGA microparticles were recovered by centrifugation, washed, and freeze dried. Following a controlled proprietary technique (Sustained Nano Systems), the bevacizumab–PLGA microparticles were processed into active DMCs. Aliquots of dry, active DMCs were placed in 2-mL, snap-top Eppendorf centrifuge tubes for storage. In a similar manner, control DMCs were prepared without an API.
For in vitro tests, approximately 100 mg of dried DMCs were suspended in 2 mL of phosphate-buffered saline (PBS). At predetermined time points, the tubes were centrifuged and the supernatant removed to assay for protein content representing the released amount of the bevacizumab API. The removed diluent was then replaced with an equal volume of PBS, washed, and refilled again with PBS, and the tube was stored for the next time period. This sampling was repeated at intervals for up to 12 months from the initial bevacizumab–DMC preparation. Protein content (as bevacizumab) was determined using a bicinchronic acid assay (QuantiPro BCA assay; MilliporeSigma, St. Louis, MO) and read at 562 nm with a spectrophotometer. The entire in vitro study was performed in triplicate.
Random samples of the supernatant were selected from the same PBS suspensions at nine time periods to assess in vitro bevacizumab content as determined by enzyme-linked immunosorbent assay (ELISA) (Bevacizumab ELISA Assay Kit; Eagle Biosciences, Nashua, NH). Samples were read at 450 nm with a spectrophotometer and compared to a previously prepared standard plot that used a fresh, authentic bevacizumab solution at known concentrations.
Purity and molecular integrity were assessed on additional samples using size-exclusion chromatography–high-performance liquid chromatography (SEC-HPLC). Supernatant samples were extracted in PBS as described above and placed on an Agilent 1100 HPLC device (Agilent Scientific Instruments, Santa Clara, CA) using a TSKgel G3000SWXL (7.8 × 300 mm, 5 µm) silica column (Tosoh, Tokyo, Japan). Readings were made at 280 nm and compared to the authentic reference bevacizumab sample.
An in vivo assessment of anti-angiogenic bioactivity over time was conducted using the rabbit corneal model for suppression of neovascular encroachment from the limbus in response to corneal suture injury.
7,8 All animal studies were reviewed and approved by the Institutional Animal Care and Use Committee of Stony Brook University and met the guidelines of the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research.
Twelve adult male New Zealand white rabbits (Charles River Laboratories, Wilmington, MA) were randomly assigned to one of two groups: six active and six control. One pair, consisting of one active and one control, was followed for 3 months and then euthanized. Their eyes were enucleated, fixed in 10% neutral buffered formalin, processed, paraffin embedded, and sectioned, and the specimens were stained with hematoxylin and eosin for histological evaluation. Four pairs were studied for 5.5 months to observe and photograph the neovascularization response, and one pair was followed for 12 months. Under general anesthesia, a single, radial 9-0 silk suture (Ethicon, Somerville, NJ) was placed in clear cornea (mid-stromal) 2 to 3 mm from the limbus to stimulate corneal neovascularization in the study eye. The fellow eye was not used. Reconstituted in approximately 0.5 mL of sterile normal saline, the control or bevacizumab–active DMC suspension was administered as a single injection of 30 mg subconjunctivally using a 27-gauge needle (targeting the sub-Tenon space) in the same meridian as the suture near the limbus. To ensure continuous stimulation of the angiogenic response, the corneal silk sutures were replaced when necessary without additional DMC administration. Photography of the conjunctiva, cornea, and sclera was performed at set intervals with a standardized digital camera (Nikon D3300 with a 105-mm Micro-NIKKOR f/2.8G lens; Nikon, Tokyo, Japan). Images were subsequently matched for color balance, exposure, apparent magnification, and size (DxO PhotoLab 4.1.2, Boulogne-Billancourt, France) with the goal to focus on the limbal area at the same meridian as the suture and injection. Sets of images were presented at the end of the study in a randomized, masked manner (in terms of both sequence and treatment group) to a group of eight clinical ophthalmologists as graders, to assess corneal neovascularization and assign a standardized, semiquantitative grade based on the intensity of neovascularization for each image. The in vivo data analysis was performed using hierarchical linear models to account for the clustered nature of the data (i.e., each animal had more than one score over time); each animal could be scored differently at different time points. The fixed effects were treatment group and time (in days). Random effects were animals, graders, and time. To test for differences between treatment groups, this variable was entered in the model to predict the scores given by all graders over time. To test for the trajectories (slopes) of scores over time, we tested the interaction term treatment group × time. A positive slope reaching statistical significance means that the scores increased significantly between the two time points, for example. A non-significant slope (negative or positive) means that the scores did not change significantly. Statistical significance was defined as P < 0.05 (type 1 error). Computerized statistical analyses were performed with Stata 14.2 (StataCorp, College Station, TX).