Co-inhibition of PGF and VEGF blocks their expression in mononuclear phagocytes and limits neovascularization and leakage in the murine retina

Experiments were performed with 8–10-week-old C57BL/6J mice of both sexes [20]. Animals were housed in an air-conditioned environment with 12-h light-dark schedule and had access to water and food ad libitum. All experimental procedures complied with the German animal welfare law, which is in line with the European Community law, and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The animal protocols used in this study were reviewed and approved by the governmental body responsible for animal welfare in the state of North Rhine-Westphalia, Germany (application no. 81-02.04.2017.A430).

Laser coagulation of the retina was performed by using a slit-lamp-mounted diode laser system by Quantel Medical Vitra (532-nm green laser). For laser treatment, mice were anesthetized by intraperitoneally injecting ketamine hydrochloride (100 mg/kg body weight, Ketavet; Pfizer Animal Health) and xylazine hydrochloride (5 mg/kg body weight, 2% Rompun; Bayer HealthCare) diluted in 0.9% sodium chloride. The pupils of the mice were dilated using phenylephrine 2.5%–tropicamide 0.5% before laser treatment. For fundus fluorescence angiography (FFA), immunohistochemistry (IHC), and in situ hybridization (ISH), three laser burns (energy 125 mW, duration 100 ms, spot size 100 μm) were equally placed around the optic nerve of both eyes [25]. For ELISA measurements of cytokines, the number of laser burns applied per eye was 20. To validate rupture of Bruch’s membrane, post-laser retinal structure and laser lesion size were analyzed in vivo using HRA/OCT. In case of media opacities precluding accurate laser application (pre-existing corneal scar or cataract), insufficient disruption of Bruch’s membrane, or hemorrhages, these eyes were excluded from analyses.

Animal cages were randomly allocated to the experimental groups. The following compounds (all diluted in 1× PBS) were injected intravitreally immediately after laser pulse application: 1.5 μl of either Aflibercept (10 μg/μl, Eylea, Bayer HealthCare), anti-VEGF-A (5 μg/μl, goat anti-mouse VEGF-AA IgG, AF493-NA, R&D Systems), anti-PGF (5 μg/μl, polyclonal rabbit anti-PGF antibody, ab9542, Abcam), anti-VEGF and anti-PGF combined (each 5 μg/μl), or IgG control (10 μg/μl, normal goat IgG control (AB-108-C, R&D systems). Therefore, a 34-gauge needle was inserted into the vitreous space approximately 1.5 mm below the limbus and the compounds were administered bilaterally with a NanoFil syringe (Word Precision Instruments, Sarasota, FL, USA).

Vascular leakage was analyzed 3 and 7 days after laser damage. After anesthesia of the animals and dilatation of the pupils, the vascular leakage was determined with the FA-mode of the HRA/OCT device (Spectralis™). One hundred microliters of 2.5% fluorescein (Alcon) diluted in 0.9% sodium chloride were injected intraperitoneally. Late-phase images were taken 10 min after fluorescein administration. The size of laser spots and vascular leakage was determined using the measuring tool of the Heidelberg software. The pixel intensity was quantified in two regions of interest (ROI) within and one ROI outside each laser spot using the program ImageJ. The background pixel intensity was then subtracted from the laser spot values. The data of three laser spots were averaged to obtain the mean laser-induced leakage per eye.

The eyes were enucleated and fixed in 10% neutral buffered formalin (NBF) for 2 h at room temperature. The dissected retinal and RPE/choroidal flat mounts were permeabilized overnight (5% Triton X-100, 5% Tween-20 in PBS). Unspecific antigens were blocked with BLOTTO (1% milk powder, 0.01% Triton X-100 in PBS) for 1 h at room temperature. The flat mounts were subsequently incubated in the primary antibody overnight at 4 °C (1:1000 dilution of Iba1, rabbit polyclonal, 234 003, Synaptic Systems). Flat mounts were then incubated with a 1:1000 dilution of goat anti-rabbit AlexaFluor 488 nm-conjugated secondary antibody (A11008; Life Technologies) for 1 h. In addition, RPE/choroidal flat mounts were incubated with a 1:10 dilution of primary TRITC-conjugated lectin (L5264; Sigma). After washing, retinal and RPE/choroidal flat mounts were mounted on a microscope slide and embedded with fluorescence mounting medium (S3023; DakoCytomation) [25].

Images were taken with a Zeiss Imager M.2 equipped with an ApoTome.2. The total number of Iba1-positive cells was counted for each laser spot. Cellular morphology was analyzed using a grid system to determine the mean number of grid crossing points per cell [25]. The colored pixel intensity in individual image areas of the laser spots was quantified using the Colored Pixel Counter tool for Fiji.

Areas of choroidal neovascularization in RPE/choroidal flat mounts were measured with the spline function of the graphic tool included in the ZEN software (Zeiss). Data were excluded when it came to damages to the CNV lesion during tissue processing or inability to locate a CNV lesion during imaging of an eyecup.

ISH of RPE/choroidal flat mounts was performed using the Multiplex Fluorescent Reagent Kit v2 (ACD) according to the protocol of Gross-Thebing et al. [26]. The following probes were used: Iba1 (channel C1 or C3), VEGF (channel C3), and PGF (channel C1). Images were taken with a Zeiss Imager M.2 including an ApoTome.2.

The colored pixel intensity in individual image areas of the laser spots was quantified using the Colored Pixel Counter tool for Fiji.

The concentration of cytokines in total retinal or RPE lysates were measured by ELISA according to the manufacturer’s instructions (R&D Systems). Absorbance was quantified using a TriStar² multimode plate reader LB 942 (Berthold Technologies).

All data were analyzed using GraphPad PRISM (version 7) using analysis of variance (ANOVA) and Tukey’s post-test, *P < 0.05, **P < 0.01, and ***P ≤ 0.001. The data are shown as mean ± SEM. All analyses were performed after random consecutive numbering of animals, which was only revealed after finishing all analyses.

We first compared the effects of inhibiting VEGF-A alone (aVEGF), PGF alone (aPGF), or in combinations (aVEGF plus aPGF, aflibercept) with IgG sham injections in the laser-induced mouse model of CNV. The inflammation-induced vascular leakage was assessed with fundus fluorescein angiography (FFA) at days 3 and 7 (Fig. 1a). The pixel intensities of leakage lesions (Fig. 1b, c) and their total area (Fig. 1d, e) were quantified, respectively. At 3 days after laser induction, anti-PGF alone (aPGF), the combination of aVEGF with aPGF, and aflibercept significantly reduced the pixel intensity of vascular leakage compared to IgG (Fig. 1b). At 7 days after laser, the reduction in leakage intensity was greater with all injected biologics and aflibercept showed the strongest effect with significant differences to all other conditions (Fig. 1c). When analyzing leakage area after laser damage, aflibercept was the only molecule that showed a significant effect after 3 days (Fig. 1d). Aflibercept was also most effective treatment regarding the leakage area at day 7 (Fig. 1e). Only the combined therapy of aVEGF with aPGF and aflibercept showed significant differences compared to IgG treatment, whereas single administration of either aVEGF or aPGF did not affect leakage at this time point (Fig. 1e).

We next analyzed neovessel formation by lectin staining of RPE/choroidal flat mounts at days 3 and 7 (Fig. 2). Three days after laser damage only aflibercept reduced CNV compared to the IgG control (Fig. 2b). At day 7, all treatments significantly reduced CNV formation compared to IgG, whereby the combination of aVEGF and aPGF as well as aflibercept led to the strongest decrease in CNV formation (Fig. 2c).

As the laser-CNV model also reflects an inflammation-related wound healing process [27, 28], we analyzed the overall size of the lesions at three different time points after laser (Additional file 1: Figure S1). Over time, the lesion size and the formation of a fibrotic scar decreased in all conditions and no significant effect of any treatment was observed (Additional file 1: Figure S1). This indicates that the applied compounds did not affect wound healing or fibrosis.

We next addressed which compound was effective in modulating the activity of Iba1+ microglia and macrophages in the retina (Fig. 3) and the RPE/choroid complex (Fig. 4). Quantification of Iba1⁺ cells per laser spot showed no differences between the treatments 3 days after laser damage (Fig. 3a, b). In contrast, 7 days after laser induction, inhibition of PGF either alone or in combination reduced the number of retinal microglia (Fig. 3c), with the strongest effect after aflibercept injection. These findings were confirmed by quantifying the number of Iba1 signal pixels in each area (Fig. 3d, e).

Since the morphology of microglia often correlates with their function and pro-inflammatory state, we next analyzed their ramification by counting the number of grid cross points per cell. None of the treatments affected the ramification state, indicating that the compounds mainly affected the recruitment of microglia to the laser lesion and not their overall morphology (Additional file 2: Figure S2a, b).

We next focused on the protein levels of three pro-inflammatory cytokines often rapidly produced by reactive microglia [29]. Six hours after laser damage, IL-6 was rapidly secreted in the retinal tissue, whereas IL-1β and TNF did not show laser-induced expression differences (Fig. 3f-h). Treatments with aVEGF, aVEGF combined with aPGF, and aflibercept reduced the secretion of IL-6, while aPGF showed no significant effect (Fig. 3f). Of note, 3 days after laser coagulation, the secretion of IL-6 declined to the level of naive mice (data not shown).

Staining of mononuclear phagocytes in RPE/choroidal flat mounts showed similar results than those in the retina (Fig. 4). Thus, there was no difference between all groups at day 3 (Fig. 4a, b, d), but strong effects at day 7 (Fig. 4c, e). All groups involving PGF inhibition reduced the number of Iba1+ cells in the laser spots at day 7 (Fig. 4c). Again, the analysis of grid crossing points per cell did not reveal any differences for any time point (Additional file 2: Figure S2c,d).

ELISAs of RPE/choroidal flat mounts showed increased IL-6 and IL-1β secretion 6 h after laser damage, while no TNF secretion was detected (Fig. 4f-h). Interestingly, treatment with aflibercept and aVEGF resulted in significantly reduced levels of IL-6 compared to IgG (Fig. 4f), whereas all treatments strongly inhibited IL-1β secretion (Fig. 4g).

Since trapping of PGF significantly attenuated microgliosis and macrophage numbers in the laser lesions, we next asked the question whether these cells themselves actively transcribe VEGF and PGF mRNAs using in situ hybridization (ISH) with specific amplifier probes. We first compared VEGF and PGF mRNA expression together with the marker Iba1. Therefore, ISH with RPE/choroidal flat mounts of lasered mice treated with IgG for 3 and 7 days was performed (Fig. 5a–c). The simultaneous detection of Iba1 (green) and VEGF (blue) showed that most Iba1+ mononuclear phagocytes expressed VEGF in the lesion area (Fig. 5a). Similar results were obtained with PGF (Fig. 5b). Simultaneous hybridization with probes targeting PGF and VEGF showed that the majority of cells, which were positive for VEGF, also expressed PGF at both time points (Fig. 5c). These results indicate that Iba1+ cells are capable to co-express significant amounts of VEGF and PGF mRNA.

We then analyzed the effect of the different treatments on VEGF and PGF mRNA expression in phagocytes at day 7 using ISH (Fig. 6a). Intravitreal administration of aVEGF, the combination of aVEGF and aPGF as well as aflibercept strongly reduced the color pixel intensity in the laser lesions of RPE/choroidal flat mounts hybridized with VEGF probes. Treatment with aPGF also reduced pixel intensity but significantly less than the other biologics (Fig. 6b). Hybridization with PGF probes showed that all treatments attenuated PGF mRNA levels in the laser lesion with aVEGF showing the smallest effect (Fig. 6c).

Since mRNA expression levels do not unequivocally represent translation and secretion of cytokines, we measured VEGF and PGF protein levels in RPE/choroidal flat mounts using ELISAs. Both VEGF and PGF were expressed at basal levels in RPE/choroidal flat mounts of naive mice, which strongly increased 6 h after laser damage (Fig. 6d, e). Significantly reduced VEGF and PGF levels were observed in all treatment groups (Fig. 6d, e). However, PGF protein levels were fully suppressed by aflibercept only, indicating the highest efficacy of the VEGF/PGF protein trap aflibercept compared to inhibition with single or combined antibodies (Fig. 6e).