Platelet-Derived Growth Factor-D Activates Complement System to Propagate Macrophage Polarization and Neovascularization

Platelet-derived growth factor-D (PDGF-D) is highly expressed in immune cells. However, the potential role of PDGF-D in immune system remains thus far unclear. Here, we reveal a novel function of PDGF-D in activating both classical and alternative complement pathways that markedly increase chemokine and cytokine responses to promote macrophage polarization. Pharmacological targeting of the complement C3a receptor using SB290157 alleviated PDGF-D-induced neuroinflammation by blocking macrophage polarization and inhibited pathological choroidal neovascularization. Our study thus suggests that therapeutic strategies targeting both PDGF-D and the complement system may open up new possibilities for the treatment of neovascular diseases.


INTRODUCTION
Tissue inflammation is a cellular response initiated by various factors, such as invasion of foreign material and microbes and clearance of damaged cellular debris to maintain tissue homeostasis (Medzhitov, 2008). Low levels of inflammatory responses also help maintain normal homeostasis (Chen and Xu, 2015). Dysfunction or hyperactivation of inflammation results in various inflammatory neurodegenerative disorders, such as age-related macular degeneration (AMD), Alzheimer's disease, Parkinson's disease, and uveitis (Tan et al., 2020;Yang et al., 2020). AMD remains a significant cause for progressive loss of central vision, if uncontrolled, leading to legal blindness (DeAngelis et al., 2017). Microglia, resident immune cells in the retina, choroidal macrophages (Yu et al., 2020) and retinal pigment epithelial (RPE) cells play a central role during inflammation by secreting various chemokines, cytokines, growth factors and elements of the complement system (Holtkamp et al., 2001;Shi et al., 2008;de Oliveira Dias et al., 2011). Additionally, several studies have identified alterations of the complement pathway in AMD pathogenesis (Wu and Sun, 2019). Small menagerie of complement proteins can activate the complement system via classical, lectin and alternative pathways, converging on the critical complement component C3, to generate C3a, C5a and membrane-attacking complex (MAC) C5b-C9, acknowledged in the drusen of AMD patients (Crabb et al., 2002;Mullins et al., 2014;Kim et al., 2020). Increasing evidence suggests that persisting inflammatory milieu supports the classical macrophage activation (M1 polarization) to generate tissue-destructing proinflammatory signals, and alternative macrophage activation (M2 polarization) generates antiinflammatory signals to promote pathological angiogenesis (Apte, 2010). However, currently, it is not well understood how the complement system in CNV is activated, although its components are found in neovascular lesions of wet AMD patients.
Platelet-derived growth factor-D (PDGF-D), a member of the PDGF family, has been shown to exert diverse functions under physiological and pathological conditions (Bergsten et al., 2001;Kazlauskas, 2017;Folestad et al., 2018;Kumar and Li, 2018). Several studies have implicated PDGF-D in the promotion of inflammation and in the increased migration of monocytes and macrophages under pathological conditions. During intracerebral hemorrhage, PDGF-D promotes neuroinflammation and enhances macrophage infiltration (Yang et al., 2016). Adipocyte-derived PDGF-D promotes adventitial fibrosis with inflammation, thereby contributing to the development of aortic aneurysm in obesity (Zhang et al., 2018). Ostendorf et al. (2006) and Boor et al. (2007) demonstrated that the administration of human mAb CR002, a PDGF-D antibody, reduced glomerular infiltration of monocytes/macrophages and prevented epithelialmesenchymal transition. In our previous work, we have shown that Pdgfd knockdown by shRNA inhibited the macrophage infiltration and reduced choroidal neovascularization . These studies highlight the effects of PDGF-D on inflammatory cell infiltration under pathological conditions. However, the mechanism by which PDGF-D induces inflammatory cell activation and migration is not well understood. Notably, a recent study has shown that PDGF-D inhibits tumor growth by binding to natural killer (NK) cell receptor NKp44, leading to the production of tumorsuppressive cytokines by NK cells (Barrow et al., 2018), in contrast to many previous reports showing angiogenic and oncogenic effects of PDGF-D (Li et al., 2003;Kumar et al., 2010;Kumar and Li, 2018). Thus, more in-depth studies are warranted to verify the functions of PDGF-D and the underlying mechanisms.

Animals
All animal experiments were approved by the Animal Use and Care Committee of Zhongshan Ophthalmic Center at the Sun Yat-sen University, Guangzhou, People's Republic of China. C57BL/6J (6-8 weeks old) mice were purchased from Pengyue Company (Shandong, China). All mice were maintained on a 12-h light/dark cycle with water and chow provided ad libitum and were housed in an SPF facility in the Ophthalmic Animal Laboratory of Zhongshan Ophthalmic Center at the Sun Yat-sen University. Five minutes after intraperitoneal injection of 4% chloral hydrate (10 ml/kg body weight), mice were anesthetized before treatment or euthanized directly by cervical dislocation.

Proliferation Assay
RAW264.7 and dTHP1 cell proliferation assays were performed using a Cell Counting Kit-8 assay (CCK8, Dojindo, Japan). Cells were seeded in 96-well culture plates (5,000 cells/well for RAW264.7 cells and 10,000 cells/well for dTHP1 cells, respectively) and 450 nm absorption values were recorded after treatment with the CCK-8 reagent.

Migration Assay
RAW264.7 cells were seeded in the cell culture inserts (Ibidi, Germany, cat: 80209) at 1 × 10 5 cells/chamber and the inserts were removed after 12 h to generate cell-free gaps. dTHP1 cells were seeded in 48-well plates (3 × 10 5 cells/well) and the cell-free area was produced by scratching the wells with a 200 µl pipette tip. Images were obtained at 0 and 24 h, respectively, after treatment and were analyzed using the ImageJ software.

Construction of Adeno-Associated Virus (AAV) for RPE-Specific PDGF-D Overexpression
The CMV promoter of the AAV vector pAV-CMV-C-FH (Vigenebio, China, cat: pAV100001-OE) was replaced with the human VMD2 promoter (-598 bp upstream to 378 bp downstream of the transcription start site), known to drive efficient and specific transgene expression in RPE (Alexander and Hauswirth, 2008). The multi-cloning site of the vector was replaced by the coding sequence of human PDGFD gene (NM_025208). The AAV was packaged by Vigene Biosciences (Shandong, China) and stored at -80 • C.

Subretinal AAV Injection in Mice
Mouse pupils were dilated by topical application of tropicamide. An intraperitoneal injection of 4% chloral hydrate (10 ml/kg body weight) was performed for anesthesia. Topical anesthesia was performed on the cornea using procainamide. Carboxymethyl cellulose sodium was used to avoid the development of cornea xerosis. Subretinal injection was performed using a sterilized 5 µl syringe (Hamilton, cat: 7633-01) with a 33-gauge blunt needle (Hamilton, cat: 7803-05, 33/15mm/3) through a puncture hole of 0.2 mm in diameter behind the cornea limbus, and AAV (1 µl/eye, 5 × 10 13 vg/ml) was injected. A successful subretinal injection was indicated by the visualization of the semicircular retinal detachment around the injection site under a microscope or by fundus imaging.

Intraperitoneal Injection of C3a Antagonist
The 20 mg/ml stock solution of the C3a antagonist SB290157 (MCE, cat: HY-101502A) was prepared by dissolving the compound in a minimal volume of sterile dimethyl sulfoxide (DMSO, Sigma, cat: D4540). SB290157 was further diluted using corn oil to a final concentration of 2 mg/ml for injection at a dose of 30 mg/kg body weight. Three weeks after AAV injection, SB290157 or vehicle was injected intraperitoneally every 2 days for 3 times. In the laser-induced CNV model, the antagonist was intraperitoneally injected 3 times after laser photocoagulation.
For analysis of flat-mounted retinas, 1 week after AAV-PDGF-D subretinal injection, SB290157 (MCE, cat: HY-101502A) was injected intraperitoneally (30 mg/kg body weight) every 2 days. After 1 week, the mice were sacrificed, and the eyes harvested and fixed in 4% PFA for the analysis of flat-mounted retinas. The anterior segment of the eye and lens were removed. The retinas were separated from the sclera and flattened on a glass slide and dissected by making four radial cuts. Flat-mounted retinas were stained with I isolectin B4 (IB4, Thermo Fisher, cat: 121411) and analyzed using a fluorescent microscope AX10 imager.Z2 (ZEISS). The vascular branch points were analyzed using Angio Tool (version 0.5).

RNA Sequencing and Transcriptomic Analysis
Four weeks after AAV injection, the mice were sacrificed. The eyeballs were removed and the fascial tissues and muscles around the eyeballs dissected on ice. Choroidal tissue was quickly dissected and put into the reagent for RNA isolation. RNA was sent to Shanghai Pharmaceutical Kangde Co., Ltd. (Shanghai, China) for RNA sequencing. Sequenced raw reads were mapped to mouse mm10 reference genome using STAR (v2.4.2a). After alignment, RSEM (V1.2.29) was used to generate FPKM values for known gene models. Differentially expressed genes were identified using DESeq2 (v1.22.2). Fold-changes were estimated according to each sample's FPKM. Differentially expressed genes were selected using the following filter criteria: P-value ≤ 0.1, FDR ≤ 0.1, fold-change ≥ 1.5, mean FPKM ≥ 1. Gene set functional enrichment analysis was performed using cluster Profiler (v3.17.0) with gene set size set to 5-500, P-value cutoff set to 0.05, and adjusted P-value set to 0.25. Volcano plots were generated using ggplot2 (v3.3.0). Heatmap plots were generated using pheatmap (v1.0.12). Gene-function networks for differentially expressed chemokine genes and related biological processes were visualized using Cytoscape (v3.6.1).

RNA Isolation and Real-Time Quantitative PCR
Total RNA was isolated using the TRNzol reagent (TIANGEN, cat: DP424) and converted to cDNA using the Fast King RT Kit (TIANGEN, cat: KR116) according to the manufacturer's instructions. Real-time quantitative PCR was carried out in a 10 µl reaction containing the SYBR Select Master Mix (Vazyme, cat: Q331) in technical quadruplicate using a Quantstudio 6K Flex system (Life Technologies). Results were analyzed using the Quantstudio 12 K Flex Software v1.2.2 (Thermo Fisher Scientific). Relative mRNA levels were calculated based on the 2 − CT method, using the 18S rRNA as references.

Choroid Explant Assay
C57BL/6J mice at postnatal day eight were sacrificed, and eyes were enucleated and kept in ice-cold phosphate buffered saline before dissection. After removing the cornea and lens from the anterior of the eye, the peripheral choroid-scleral complex was separated from the retina and cut into pieces (approximately 1 mm × 1 mm). Choroid pieces were transferred into growth factor-reduced Matrigel (BD, Cat: 356231) and seeded in 24well plates followed by Matrigel solidification for 10 min. A volume of 500 µl of medium was added to each well and incubated at 37 • C with 5% CO 2 for 48 h. The medium was changed every 48 h. Individual explants were imaged daily using an inverted microscope. Areas of choroidal sprouts were quantified using ImageJ.

Laser Induced CNV
The laser-induced CNV mouse model was performed as described previously (Zhang et al., 2009). Briefly, 8 weeks old female mice were anesthetized by intraperitoneal injection of 4% chloral hydrate (10 ml/kg body weight), and eyes were dilated by topical application of tropicamide. Four laser spots were made by laser photocoagulation (90 mV power, 75 ms duration, 75 µm spot size, Oculight Infrared Laser System 810 nm, IRIDEX Corporation) at an equal distance from the optic nerve in each eye for CNV. The cornea of mice were treated with antibiotic tobramycin ointment locally after laser photocoagulation and mice were placed on a 37 • C electric blanket until wake. After 7 days, the eyecups were flat-mounted and the immunohistochemical staining were performed as described (Zhang et al., 2009).

Statistical Analysis
Gene expression analysis by Q-PCR are expressed as means ± SD. While other results are expressed as means ± SEM. The statistical significance between the control and PDGF-D, or between AAV-GFP and AAV-PDGF-D groups were assessed with the unpaired student's two-tailed t test. Multiple group comparisons were performed with ordinary one-way ANOVA test. Differences between groups were tested with GraphPad Prism software (version 7.04) and considered statistically significant for P < 0.05.

PDGF-D-Induced Retinal Epithelial Cell Secretome Promotes Macrophage Migration
Under pathological conditions, PDGF-D has been shown to promote macrophage migration (Uutela et al., 2004). However, the mechanism by which PDGF-D promotes macrophage migration is not well understood. To address this, we stimulated murine macrophages (RAW264.7) and differentiated human macrophages (dTHP1) with PDGF-D. Murine macrophages expressed PDGFR-β and the PDGF-D co-receptor NRP1 but not PDGFR-α ( Figure 1A and Supplementary Figure 1A). The human monocytic cell line (THP1) did not express PDGF receptors ( Figure 1A and Supplementary Figure 1B). However, upon differentiation to macrophages by PMA, they expressed PDGFR-α, PDGFR-β and NRP1 ( Figure 1A

PDGF-D Overexpression in Mouse Retinal Epithelial Cells
Recombinant adeno-associated virus (AAV) has been shown to be effective for retinal gene therapy due to their efficient transduction of RPE cells with low toxicity (Vandenberghe and Auricchio, 2012). To over-express PDGF-D in mouse RPE cells, we constructed AAV type 8 (AAV8) vector expressing human PDGF-D (AAV-PDGF-D) driven by the retinal pigment epithelium specific VMD2 promoter (Figure 2A). PDGF receptors are expressed in both mouse retinae and choroid ; Supplementary Figures 3A-D). Four weeks after subretinal injection of AAV-PDGF-D or AAV-GFP (as a control), overexpression of PDGF-D mRNA and protein in the RPE-choroid complex were detected (Figures 2B-D). Moreover, immunofluorescence staining identified PDGF-D or GFP in the

Activation of the Complement Pathway Revealed by Transcriptomic Analysis of PDGF-D-Overexpressing RPE-Choroids
To identify PDGF-D-induced downstream pathways, we performed unbiased transcriptomic analysis using the PDGF-D-overexpressing RPE-choroid complex. A total of 2,486 differentially expressed genes (DEGs) were identified. Among them, 1697 were up-regulated and 789 were down-regulated ( Figure 3A). Biological function enrichment gene ontology analysis showed that the most enriched biological processes were related to the regulation of immune system (Figure 3B), particularly, the complement pathway ( Figure 3C). Other pathways included chemokine and its receptor (Figure 3D and Supplementary Figure 4B), cytokine signaling (Supplementary  Figures 4C,E) and regulation of extracellular matrix and growth factors (Supplementary Figures 4D,F). Importantly, both gene (Supplementary Figure 4A) and protein analysis confirmed the activation of both classical ( Figure 3E) and alternative complement pathways (Figure 3F). Immunofluorescence staining further identified C1q expression in both RPE and choroids of PDGF-D-overexpressing samples (Figure 3G), where PDGF-D overexpression led to increased accumulation of IA/IE + macrophages in the choroids, which were also positive for C1q staining (Figures 3G,H). Further functional network analysis of DEGs demonstrated pathways related to activation of the complement system, immune cell migration and activation of immune responses in PDGF-D-overexpressing RPE-choroids ( Figure 3I). These findings thus underscore PDGF-D overexpression-induced inflammation by activating the complement pathway and chemokine/cytokine signaling.

PDGF-D-Induced Complement Activation Promotes Macrophage Polarization
The complement component C1qa has been implicated in the promotion of the of M2 microphage polarization to mitigate tissue inflammation (Spivia et al., 2014), while complement anaphylatoxin C3a and C5a are acknowledged to promote tissue inflammation by activating monocytes and macrophages (Bohlson et al., 2014). Interestingly, PDGF-D overexpression upregulated both M1 polarization markers, such as Tnfα, Il1β, Nos2 and Cxcl10, and M2 polarization markers, such as Arg1, Il10, and Chi3i3 ( Figure 4A). Furthermore, immunofluorescence staining revealed that PDGF-D promoted IBA1 + macrophages in both retinae and choroids, which were also positive for CD16/32 staining, an M1 polarization marker ( Figure 4B), with higher percentages in the PDGF-D overexpressing retinae and choroids ( Figure 4C). In addition, IBA1 + macrophages were positive for CD206 staining, an M2 polarization marker ( Figure 4D), with higher proportions in PDGF-D overexpressing choroids and retinae ( Figure 4E). Together, these findings underline the presence of both pro-and anti-inflammatory milieu triggered by PDGF-D overexpression.  (Kumar et al., 2014). Since PDGF-D induced marked macrophage activation, we examined whether this affected retinal and choroidal blood vessel. In the PDGF-D overexpressing RPE-choroid complex, mRNA levels of genes encoding proangiogenic and extracellular matrix (ECM) regulators, such as Tgfβ1, Fgf2, Mmp9, Mmp12 and Mmp2, were upregulated ( Figure 5A). The PDGF family members are known to promote proliferation and recruitment of mural cells (Li et al., 2003;Uutela et al., 2004). PDGF-D overexpression increased CD31 + endothelial cell density in both retinae and choroids, while αSMA + smooth muscle cell coverage increased only in choroids (Figures 5B,C). Thus, increased PDGF-D expression levels promoted retinal and choroidal blood vessel growth and maturation.

Inflammatory Pathological Angiogenesis Triggered by PDGF-D
Choroidal blood vessels and RPE cells possess a unique symbiotic relationship by nourishing each other. The disruption of this association results in RPE degeneration or choroidal neovascularization (Lutty et al., 1999). Since PDGF-D overexpression increased choroidal endothelial cell density, we tested the effect of PDGF-D on choroids by stimulating choroidal explants with PDGF-D. PDGF-D promoted robust choroidal endothelial cell sprouting (Figures 6A,B) in a dosedependent manner (Supplementary Figures 5A,B). Moreover, injury of the PDGF-D-overexpressing RPE cells by laser treatment augmented abnormal growth of IB4 + neovessels with more IBA1 + macrophages (Figures 6C,D). Furthermore, CD31 + pathological neovessels intermingled with IBA1 + macrophages (Figures 6E,F). These data indicated that the inflammatory milieu elicited by PDGF-D nurtured pathological choroidal neovascularization.

Pharmacological Inhibition of the Complement Cascade Alleviates Inflammation and Pathological Neovascularization
Since the induction and activation of the complement pathway by PDGF-D overexpression led to more macrophages and higher blood vessel density in the retinae and choroids, we tested whether blocking complement activation could inhibit pathological neovascularization and treated PDGF-D-stimulated macrophages and RPE with a C3a-receptor antagonist SB290157, which is known to block complement activation (Hutamekalin et al., 2010). SB290157 treatment inhibited migration of mouse and human macrophages ( Figure 7A and Supplementary Figure 6A) in vitro. Furthermore, intraperitoneal injection of SB290157 to the RPE-specific PDGF-D overexpressing mice markedly reduced infiltration of IBA1 + macrophages in the retinae and choroids (Figures 7B,C) and suppressed the expression of both types of macrophage polarization markers (Supplementary Figure 6B). Additionally, immunofluorescence staining of endothelial and mural cells showed a marked reduction in blood vessel density and smooth muscle cell coverage (Figures 7D,E) with concomitant reduction of angiogenic and chemokine gene signatures (Figure 7F and Supplementary Figure 6C). Moreover, analysis of flat-mounted retinas confirmed the above findings by showing that PDGF-D overexpression increased retinal vascular branch points, which was abolished by SB290157 treatment (Supplementary  Figures 7A,B). Since complement components are found in CNV lesions of AMD patients, the complement pathway was inactivated by SB290157 treatment in a mouse CNV model. Importantly, intraperitoneal injection of SB290157 inhibited CNV by reducing immune cell density in neovascularization lesions (Figures 7G,H). Together, these observations suggested that PDGF-D-induced activation of the complement pathway   is critical for the promotion of macrophage infiltration and CNV formation.

DISCUSSION
The pathogenesis of AMD is associated with degenerative conditions in the neural retina, and dysfunctional RPE cells or choroids. A growing body of evidence has demonstrated that many inflammatory signals, such as chemokines, cytokines, growth factors and alterations in the complement pathway, can modify the functions of neuronal, vascular, glial, and immune cells to promote macular degeneration and pathological neovascularization. Our current work, for the first time, provides evidence that increased PDGF-D expression activated complement pathway to orchestrate tissue milieu by altering the expression of chemokines and cytokines responses to initiate macrophage activation and triggering neuroinflammatory conditions that exacerbate the pathogenesis of AMD.
The migration and activation of infiltrating inflammatory macrophages is detrimental to tissue function. Apart from exogenous stimuli, endogenously secreted molecules, such as cytokines, chemokines and growth factors, are essential to initiate macrophage migration (Lu et al., 1998;Apte, 2010;Qian et al., 2011;Jenkins et al., 2013;Gordon et al., 2014). PDGF-D expression in the synovial membranes of patients with rheumatoid arthritis and osteoarthritis is escorted with accumulation of synovial fibroblasts and macrophages (Pohlers et al., 2006). In chronic atherosclerotic lesions, PDGF-D was found to be colocalized with macrophages and promoted migration of THP1 monocytes in a dose-dependent manner (Wågsäter et al., 2009). Intriguingly, in our work, PDGF-D  failed to promote migration of mouse or human macrophages directly. While we do not know the exact reasons for this discrepancy, it could be attributed to different experimental conditions. However, intriguingly, while PDGF-D did not promote macrophage migration by itself, conditioned medium from PDGF-D-stimulated RPE cells did promote the migration of mouse and human macrophages. Since PDGF-D is a potent stimulant of RPE proliferation and migration (Li et al., 2007;Kumar and Li, 2018), and RPE can secrete a wide array of immunomodulatory cytokines (Holtkamp et al., 2001) and chemokines (Ma et al., 2009) that can regulate macrophage response, PDGF-D-induced RPE secretome hence can promote the migration and activation of macrophages.
Over the last decade, mounting evidence has shown that the complement system plays an important role in the pathogenesis of AMD (Sparrow et al., 2012). With the initial finding of complement factor H (CFH) as a high-risk factor for AMD, additional studies identified C3a, C5a and the membrane-attacking complex C5b-9 in the drusen of AMD patients (Nozaki et al., 2006;Mullins et al., 2014). Indeed, transcriptomic analysis of PDGF-D-overexpressing RPE-choroid samples indicated that the presence of classical complement component C1qa and its activated product C3a. C1q expression was strongly associated with RPE cells and IA/IE + macrophages, and C3a protein expression was seen in both the retinae and choroids. Interestingly, it has been shown that inhibition of PDGF-D using the monoclonal antibody CR002 decreased C5b-9 deposition in an experimental glomerulonephritis model (Ostendorf et al., 2006), supporting a role of PDGF-D in the regulation of the complement system. Several reports have also shown that activation of the complement pathway and C5b-9 can regulate the expression of numerous chemokines and cytokines (Kilgore et al., 1996;Selvan et al., 1998;Risnes et al., 2003). Additionally, under inflammatory conditions, RPE cells can generate a myriad of cytokines that can activate macrophages (Holtkamp et al., 2001). Indeed, engagement of both the classical (C1qa) and alternative (C3) complement pathways by PDGF-D overexpression markedly increased the levels of numerous chemokines and cytokines, and triggered polarization of both M1 and M2 macrophages. In fact, C1qa can downregulate inflammasome activation and promote the polarization of inflammation-resolving M2-like macrophages to engulf atherogenic lipoproteins (Bohlson et al., 2014;Spivia et al., 2014). While enhanced C3a signaling facilitates M1 polarization to exacerbate renal interstitial fibrosis, C3 gene deletion increased neovascularization in a mouse retinopathy model (Langer et al., 2010;Cui et al., 2019). PDGF-D has been demonstrated to have pleiotropic effects on vascular and nonvascular cells to stimulate pathological angiogenesis (Li et al., 2003;Kumar et al., 2010). Here, our data highlight yet another VEGF-A-independent function of PDGF-D by regulating the complement system and immune cells to promote CNV formation. Thus, PDGF-D imparted its effects by modulating the complement pathway to polarize macrophages, thereby promoting pathological neovascularization. Targeting the complement system has been shown to protect mice from accumulating inflammatory mononuclear phagocytes in the subretinal space to maintain tissue homeostasis (Calippe et al., 2017). Complement fragments C3a, C5a, and C5b-9 are generated from C3 during complement activation. Treatment with SB290157, a potent and selective C3areceptor antagonist, suppressed inflammation by blocking macrophage activation in animal models (Ames et al., 2001;Lim et al., 2013;Rowley et al., 2020). Indeed, we also observed that SB290157 treatment constrained macrophage polarization and blocked infiltration of macrophages by suppressing the expression of various chemokines, cytokines and growth factors, thus decreasing inflammatory neovascularization in the eye.
Collectively, our data reveal that increased PDGF-D levels activate the complement pathway, subsequently leading to marked macrophage activation and inflammation, the key pathologies of neovascular AMD. Therapeutic strategies targeting PDGF-D signaling and complement-mediated inflammation may provide new possibilities for the treatment of neovascular diseases.

DATA AVAILABILITY STATEMENT
The authors confirm that the RNA-seq data supporting the findings of this study are accessible in the GEO repository (Accession Number: GSE164972).

ETHICS STATEMENT
The animal study was reviewed and approved by the Zhongshan Ophthalmic Center at the Sun Yat-sen University, Guangzhou, People's Republic of China.

AUTHOR CONTRIBUTIONS
ZX and QW designed and performed the experiments, analyzed the data, and wrote a part of the manuscript. WL, LH, JiZ, and JuZ performed the experiments and analyzed the data. BX, SW, HK, XCL, and CL provided critical experimental tools and suggestions. XRL and AK designed the experiments, provided resources and supervision, analyzed the data, and wrote the manuscript. All authors contributed to the article and approved the submitted version.