All animal experiments were conducted following the guidelines of the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Animal Care Committee of Kyungpook National University (Approval No. 2019-0104-01). C57BL/6J mice purchased from The Jackson Laboratory were used for all in vivo experiments. Mice were allowed free access to standard laboratory chow and tap water in a climate-controlled room with a 12 h light/dark cycle. Anesthesia was induced by i.p. injection of 250 mg/kg 2,2,2-tribromoethanol (Sigma–Aldrich, St. Louis, MO, USA) at a dosage of 250 mg/kg in survival procedures and 400 mg/kg in non-survival procedures.

For in vivo studies, the minocycline treatment regimen was modified from a previously described study (Scholz et al., 2015). Mice received i.p. injections of minocycline (45 mg/kg; Sigma–Aldrich, M9511), SB-3CT (10 mg/kg; EMD Millipore, Temecula, CA, USA, S1326), a selective MMP-2, 9 inhibitor, or normal saline vehicle with two initial twice-daily injections starting 1 day before CNV induction (Day -1), and one daily injection for 7 days beginning on the day of CNV induction (Day 0).

Eight-week-old male C57BL/6J mice were used for experiments. A 532 nm OcuLight GLx Laser System (IRIDEX Corporation, Mountain View, CA, USA) was used to generate lesions. According to previous studies (Yanai et al., 2014; Hasegawa et al., 2017), four lesions were induced for neovascular leakage and CNV size measurements, and 10 lesions were induced for gelatin zymography. The laser was set to the following parameters: 100 mW power, 100 μm spot size and 0.1-s pulse duration. The appearance of a bubble at the site of photocoagulation signified disruption of the Bruch’s membrane. After sacrificing mice, eyes were enucleated and fixed for 30 min. The whole retinas were then separated from the underlying RPE/choroid for whole-mount. For gelatin zymography, unfixed retina and RPE/choroid tissues were snap-frozen in liquid nitrogen for later analysis.

At 7 days after CNV induction (Day +7), eyes were enucleated and fixed in 4% paraformaldehyde for 30 min. Retinas were then removed from the choroid and sclera to generate choroidal flat mounts. Alexa Fluor 488-conjugated to isolectin B4 (1:100; Invitrogen, Carlsbad, CA, USA, I21411) was used to stain the eyecups overnight at 4°C. Flat-mount images were captured using an LSM 800 fluorescence microscope with an Airyscan detector (Carl Zeiss, Oberkochen, Germany). ImageJ software was used to measure CNV lesion size, according to protocols described in previous studies (Yanai et al., 2014; Hasegawa et al., 2017).

At 6 h and 1 day post-CNV induction, enucleated mouse eyes were fixed in 4% paraformaldehyde for 1 h at room temperature, cryoprotected in 30% sucrose at 4°C overnight, and embedded in OCT compound. Eyes were cryosectioned to 15 μm thickness. For immunofluorescence, sections were blocked in a blocking buffer (0.5% Triton, 0.2% BSA, and 5% donkey serum in PBS) for 1 h at room temperature and subsequently incubated with primary antibodies overnight at 4°C. After washing, samples were incubated with secondary antibodies for 1 h at room temperature. Goat anti-Iba1 antibody (1:250; FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan, 011-27991), rabbit anti-TMEM119 (1:500, Synaptic Systems, Göttingen, Germany, 400 002), rat anti-CD31 (1:100, BD Biosciences, NJ, USA, 553370), and rabbit anti-MMP-9 antibody (1:100; Abcam, Cambridge, MA, USA, ab38898) were used for primary antibodies, and Alexa Fluor 488-conjugated donkey anti-goat antibody (1:500; Thermo Fisher Scientific, Waltham, MA, USA, A11055), Alexa Fluor 594-conjugated donkey anti-rabbit antibody (1:500; Thermo Fisher Scientific, A21207), and Alexa Fluor 647-conjugated donkey anti-rat antibody (1:500; Jackson Immuno Research Laboratories, INC., PA, USA, 712-605-15) were used for secondary antibodies.

A Micron IV Retinal Imaging Microscope (Phoenix Technology Group, Pleasanton, CA, USA) was used to capture fluorescein images. After anesthesia and pupil dilation, images were obtained 3–5 min (early phase) and 7–10 min (late phase) after i.p. injection of 0.1 ml 2% fluorescein sodium (Akorn, Lake Forest, IL, USA). A previously described scheme was used to grade the hyperfluorescent lesions: faint or mottled fluorescence without leakage was scored as 0 (no leakage); hyperfluorescence without any increase in size or intensity between early and late phases was scored as 1 (mild leakage); hyperfluorescence with constant size but increasing intensity was scored as 2A (moderate leakage); and hyperfluorescence with increasing size and intensity was scored as 2B (clinically significant leakage) (Yanai et al., 2014; Hasegawa et al., 2017).

Retinal and RPE/choroid tissues were homogenized in lysis buffer (25 mM Tris-HCl buffer, 100 mM NaCl, 1% Nonidet P-40 (NP-40), and Complete Mini EDTA-free Protease Inhibitor Cocktail tablets; Sigma–Aldrich (11836170001) and centrifuged at 14,000 g for 10 min. Aliquots of supernatants containing equal amounts of protein (30 μg) were loaded without heating onto 10% SDS-polyacrylamide gels containing 0.1% gelatin (Sigma–Aldrich). Following electrophoresis, gels were washed twice for 30 min in renaturing buffer at room temperature. Gels were then incubated for 48 h in developing buffer at 37°C. After incubation, gels were stained with 0.5% Coomassie blue (Sigma–Aldrich) for 30 min and then destained for 1 h.

Samples were lysed using radioimmunoprecipitation assay (RIPA) lysis buffer (Thermo Scientific, Waltham, MA, USA) with protease and phosphatase inhibitor cocktail (Roche Holding AG, Basel, Switzerland) and centrifuged at 4°C and 14,000 g for 20 min. The supernatant was collected, and protein concentration was determined using a BCA kit (Thermo Scientific). The same amount of protein for each sample (20 μg) was loaded on a polyacrylamide gel (10%). Proteins were separated by SDS-PAGE and electro-transferred to PVDF membranes (Bio-Rad). Membranes were blocked using 5% skim milk in PBS with 0.25% Tween-20 (PBST), and subsequently incubated at 4°C overnight with anti-MMP9 (1:1,000; Abcam, ab38898) or β-actin (1:5,000; Thermo Scientific, MA5–15739) antibodies. After washing, membranes were bound to HRP-conjugated secondary antibodies [Donkey IgG anti-rabbit (1:2,000; Cell Signaling Technology, Beverly, MA, USA, 7074s) or anti-mouse (1:2,000; Cell Signaling Technology, 7076s)] and incubated at room temperature for 1 h. Blots were developed using enhanced chemiluminescence (ECL) Western blotting detection reagent (Thermo Fisher Scientific) and analyzed using a MicroChemi system (DNR Bio-imaging Systems, Neve Yamin, Israel).

For flow cytometric studies, eight retinas were pooled and minced for digestion with 0.05% TrypLE (Thermo Fisher Scientific, 12604021) for 5 min at 37°C. The samples were filtered through a 40 μm cell strainer (SPL) and stained with Zombie Aqua Kit (1:500; Biolegend, San Diego, CA, USA, 423101). The cells were washed and suspended in cold 0.2% BSA (Thermo Fisher Scientific, A1000801) and blocked with Fc Block^™ (0.01 μg/ml; BD, Franklin Lakes, NJ, USA, 553142) for 5 min at 4°C. Cell surface staining with CD45 V450 (0.3 μg/ml; BD, 560501) and CD11b PerCP-Cy5.5 (0.2 μg/ml; BD, 550993) was performed in the dark for 60 min at 4°C in buffer. Cells were detected using a BD LSR X-20 flow cytometer (BD) and analyzed with FlowJo^™ V10 (BD). According to previous studies (Greter et al., 2015; Lee et al., 2019), CD11b and CD45 were used to identify the following cells: CD45^low resting microglia (CD11b⁺CD45^low), CD45^intermediate activated microglia (CD11b⁺CD45^int), and CD45^high infiltrating leukocytes (CD11b⁺CD45^high).

Institutional Review Board approval from the Kyungpook National University School of Medicine was obtained for all patient samples used in this study (Approval No. KNUH 2012-04-006-020). The study was conducted following the tenets of the Declaration of Helsinki. Informed consent was obtained from all study participants. Aqueous humor samples were collected from consenting patients undergoing cataract surgery in the control group and treatment-naïve patients with neovascular AMD in the disease group.

Statistical analyses were performed using Prism 6.0 (GraphPad Software, San Diego, CA, USA). Results are expressed as mean ± SEM. Comparison of two groups was performed using an unpaired non-parametric Mann–Whitney U-test. Multiple-group comparison was performed by Tukey’s multiple comparison test. Statistical significance was defined as P < 0.05.

In our first series of experiments, the activities of MMP-9 and MMP-2 in aqueous humor samples from six patients with neovascular AMD and six age-matched control subjects were analyzed (Figure 1). Active MMP-9 (Figure 1B, left panel) and active MMP-2 (Figure 1C, left panel) levels were increased in the neovascular AMD group relative to control. Furthermore, pro-MMP-9 (Figure 1B, right panel) was increased in the neovascular AMD group relative to control, but pro-MMP-2 was unchanged (Figure 1C, right panel).

Next, we investigated the spatial-temporal enzymatic activities of MMP-2 and -9 in the CNV mouse model using gel zymography. In the RPE/choroid, the band intensities of pro-MMP-9 and active MMP-9 dramatically increased 6 h after CNV induction, which was maintained through Day 1, followed by a subsequent decrease of active MMP-9 to basal levels by Day 3 post-CNV induction (Figure 2C). Active MMP-2 increased in RPE/choroid tissue from CMV mice at Days 3 and 7, although pro-MMP-2 was not significantly changed (Figure 2D). In retinas from CNV mice, no active MMP-9 or MMP-2 was observed (Supplementary Figure 1). Although MMP-9 Western blots of RPE/choroid tissue exhibited similar time-dependent patterns to gelatin zymography (Supplementary Figure 2), separate pro- and active MMP-9 bands could not be detected, as with gelatin zymography, which could be why prior studies used gelatin zymography instead of Western blotting in this context (Lambert et al., 2002; Manabe et al., 2005; Kim et al., 2018).

Previous studies have demonstrated that MMP-9 amplifies pathological angiogenesis (Pepper, 2001) and inflammation (Nagase and Woessner, 1999), and microglia are a known source of MMPs, contributing to the inflammatory process (Lively and Schlichter, 2013; Rojewska et al., 2014). Furthermore, recruitment of activated microglia appears as early as 1 h after CNV induction (Karlstetter et al., 2015). If microglia contribute to the pathogenesis of CNV by secreting MMP early in the CNV disease process, microglial inhibitors such as minocycline could suppress MMP-9 activities, suppressing the development of CNV. Thus, we confirmed the effect of minocycline over time in the CNV mouse model (Figures 2C,D). In the RPE/choroid, minocycline significantly suppressed both pro and active MMP-9 (Figure 2C), as well as active MMP-2 (Figure 2D, lower panel) at Days 3 and 7.

In addition to microglia, MMPs are also present in leukocytes (Nuttall et al., 2007), although in the laser-induced CNV model, leukocyte infiltration peaks at 2–3 days after CNV induction (Sakurai et al., 2003; Tsutsumi et al., 2003). Thus, we evaluated time-dependent changes in a retinal population of activated microglia and infiltrating leukocytes from CNV mice (Figure 3A). Resident resting microglia (CD11b⁺CD45^low) were abundant in control mice. The proportion of activated microglia (CD11b⁺CD45^int) began to increase very early, beginning by 1 h post-CNV induction, peaking at 6 h post-CNV, and maintaining elevation until Day 7 (Figure 3B). The above change was similar to our finding that active MMP-9 increased in the RPE/choroid. However, the proportion of infiltrating leukocytes (CD11b⁺CD45^high) in CD11b⁺ cells increased at Day 1 and Day 3 and subsequently decreased to basal levels by Day 7 (Figure 3C). This finding is similar to a previous study demonstrating that recruitment of macrophages to CNV lesions occurs 3 days post-CNV induction and subsequently decreases (Jawad et al., 2013).

Next, to evaluate the localization of activated microglia with upregulated MMP expression, we conducted immunofluorescence staining of cryosections at 6 h (Figure 4) and 1-day post-CNV induction (Supplementary Figures 3, 4). Activated microglia with amoeboid morphology were aggregated on CNV lesions, and MMP-9 was increased in CNV lesions. Activated microglia in CNV lesions were decreased in the minocycline group relative to the vehicle group. Furthermore, intracellular MMP-9 was present in Iba-1⁺ cells in CNV lesions, which were CD31⁻ (Figure 4). Also, all Iba-1⁺ cells localized to CNV lesions were positive for TMEM119, a specific microglial marker (Bennett et al., 2016; Supplementary Figure 4). Taken together, these findings suggested that microglial activation and MMP-9 expression occurred soon after CNV induction.

To determine if MMP inhibition could alleviate CNV, we evaluated the effects of SB-3CT, a selective MMP-9 and MMP-2 inhibitor (Shin et al., 2016). To evaluate whether early MMP-9 inhibition could affect CNV severity, the SB-3CT administration started 1 day before CNV induction and was maintained until Day 7 (Figure 5A). The proportion of grade 2B lesions, defined as exhibiting clinically significant vascular leakage, decreased in CNV mice treated with SB-3CT relative to vehicle-treated CNV mice (Figures 5B,C). Furthermore, CNV lesion size decreased in mice treated with SB-3CT compared with control mice, as assessed with choroidal flat mounts (Figures 5B–D). This suggested that SB-3CT alleviated neovascular leakage and vascular outgrowth, which could have been modulated by early suppression of MMP-9 from microglia. Similarly, minocycline, a known microglial inhibitor (Yrjanheikki et al., 1999), also decreased neovascular leakage and vascular outgrowth in CNV lesions (Supplementary Figure 5).

These findings suggested that suppression of MMP-9 secreted from microglia could attenuate neovascular lesion formation in CNV. In particular, MMP-9 is involved in the breakdown of the parenchymal basement membranes surrounding blood vessels (Agrawal et al., 2006). This process can damage the Bruch’s membrane and blood-retinal barrier, with accompanying neovascular outgrowth which occurs similarly in the blood-brain barrier (Rosenberg, 2002). Thus, modulation of microglial MMPs could be a targeted approach to CNV treatment.