C57BL/6J mice purchased from Central Lab Animal (Seoul, Korea) were used for the animal experiments in this study. The care, use, and treatment of all animals in this study were carried out according to the Association for Research in Vision and Ophthalmology statement for the Use of Animals in Ophthalmic and Vision Research and the guidelines established by the Seoul National University Hospital Institutional Animal Care and Use Committee.

OIR was induced in newborn mice, as previously described in detail [22,24]. Briefly, newborn mice were exposed to hyperoxic conditions (75% oxygen) from day (P) 7 to P12 and return to normal room air conditions at P12. After the mice were sacrificed by cervical dislocation, the eyes were enucleated for the analyses at P17.

The six enucleated eyes were fixed in 4% paraformaldehyde for 24 hours and embedded in paraffin. Then, the paraffin blocks were trimmed and sagittally sectioned at a 4-µm-thickness, parallel to the optic nerve using a rotatory microtome (Leica RM2255, Nussloch, Germany). The sections were left to float on a 45℃ water bath for 2 to 3 minutes. After the straightening of folds in the tissue, the sections were placed on adhesive coated slides and allowed to cool on a slide holder. The dried slides were deparaffinized with xylene, dehydrated with alcohol, and stained with H&E.

Three representative slides from each of the six eyes were used for immunofluorescent staining. Eighteen sections were rinsed with phosphate-buffered saline (PBS) and blocked using DAKO serum-free protein block (X0909; Dako, Carpinteria, CA, USA). The samples were either incubated with primary antibodies at 4℃ overnight, followed by incubation with secondary antibodies for two hours at room temperature the next day, or they were incubated overnight at 4℃ with conjugated antibodies. The following antibodies were used for paraffin sections: Alexa Fluor 488 isolectin GS-IB4 conjugate (1 : 500 dilution; Life Technologies, Carlsbad, CA, USA), Anti-NG2 (1 : 500 dilution; Millipore, Bedford, MA, USA), anti-human CD235 (N-cadherin) (1 : 500 dilution; Biolegend, San Diego, CA, USA), Alexa Fluor 594 goat anti-rabbit IgG (1 : 200 dilution, Life Technologies) and Alexa Fluor 647 donkey anti-mouse IgG (1 : 200 dilution, Life Technologies). Nuclei were counterstained with DAPI (4′,6-diamidino-20 phenylindole). The sections were analyzed with a fluorescence microscope (Eclipse 90i; Nikon, Tokyo, Japan).

Microf luidic devices were fabricated using a replica molding method [25]. Briefly, a master mold with positive relief patterns of photoresist, SU-8 (MicroChem, Iselin, NJ, USA), was prepared on a silicon wafer by photo-lithography. Microf luidic devices were fabricated out of polydimethylsiloxane (Sylgard 184; Dow Corning, Midland, MI, USA). Human umbilical vein endothelial cells (HUVECs; Lonza, Basel, Switzerland) were mixed with human placental pericytes (Promocell, Heidelberg, Germany) in a 5 : 1 ratio. Then, 5 µL of the cell suspension containing 8 × 10⁶ cells/mL was added to the endothelial cell growth medium-2 (Lonza) channel of each microfluidic device and tilted vertically for 20 minutes for the cells to attach to the patterned fibrin on the lateral surface of the lateral micro channel. We used HUVECs that had been passaged less than four times, and the culture media was changed once on day two of culture.

The retinal structure was evaluated on histology with H&E stains and immunof luorescence. During the first week, the superficial plexus began to develop in a radial pattern extending from the optic nerve. By P8, the superficial plexus had completely extended into the retinal periphery, and the formation of deep plexus had started. Vertical vessels ran down from the superficial plexus to the deep plexus in the outer plexiform layer. At P12, the intermediate plexus could be found in the inner plexiform layer, and a three-layered vascular plexus continued to develop (Fig. 1A, 1B.

Successful development of the OIR model was confirmed with a whole-retina flat mount (Fig. 2A). To determine whether endothelial cells and pericytes were associated with any structural differences between normal and OIR mouse retinas, immunofluorescent staining was performed. Among the 18 sections, a total of 28 groups of neovascular tufts extending into the vitreous were observed at the mid-peripheral retina in the OIR model (Fig. 2B). Interestingly, all the neovascular tufts (28 / 28) and the three-layered vascular plexuses were both covered with pericytes in the OIR model (Fig. 2C).

Even if pericytes are present in pathological angiogenesis, it is thought that there is a difference in the interaction between endothelial cells and pericytes in normal and pathologic angiogenesis. To identify the differences in endothelial cell-pericyte interactions between normal vessels and pathological angiogenesis, expression of N-cadherin, which is known to be crucial in endothelial-pericyte adhesion, was evaluated. Although weak, we did observe an N-cadherin signal in both intraretinal vessels and neovascular tufts (Fig. 3).

We created a microfluidic model that mimics in vivo neovascularization. The device was composed of a central blood vessel channel, two adjacent media channels, and an outer fibroblast channel (Fig. 4A, 4B. After six days, we successfully created vascular networks, covered by pericytes on the albumin-containing endothelial surface (Fig. 4C). Interaction between endothelial cells and pericytes was easily visualized by immunofluorescent staining using CD-31 and anti-α-smooth muscle actin, one of the pericyte cell markers (Fig. 4D) [23]. Further immunofluorescent staining revealed co-expression of N-cadherin and α-smooth muscle actin on pericytes covering the blood vessel (Fig. 4E).