Therapeutic intervention in neuroinflammation for neovascular ocular diseases through targeting the cGAS-STING-necroptosis pathway

A published RNA-seq data (GSE160306↗) of retinal tissues from healthy controls, diabetic patients, patients with non-proliferative diabetic retinopathy (NPDR), and PDR patients was downloaded and re-analyzed for screening the regulatory factors involved [11]. The data of macular and peripheral retinal tissues were used for analysis. DESeq2 (v1.30.1) was applied for identification of differentially expressed genes (DEGs) with the criteria of absolute fold change > 2 and adjusted P-value < 0.05. All the DEGs were used to conduct Gene Ontology (GO) enrichment analysis by clusterProfiler (v4.0.2). GO terms with adjusted P-value < 0.05 were considered significantly. Gene set enrichment analysis (GSEA) was conducted using the GSEA software (http://www.gsea-msigdb.org/gsea/index.jsp↗) with 1000 times gene-set permutations and cutoff values of FDR < 0.25. A power analysis was performed and statistical analysis was conducted using one-way ANOVA with Turkey’s post hoc test for comparisons of these four groups. Pearson’s correlation test was employed for correlation analysis between the expression of STING and the average expression from GO terms.

All animal experiments were approved by Institutional Animal Care and Use Committee of Zhongshan Ophthalmic Center, Sun Yat-sen University (Ethic ID: 2020 − 102). The Sting^gt (#017537) mice were obtained from Jackson Laboratory and C57BL/6J mice were purchased from GemPharmatech Company. CNV model was established as previously described [12]. Briefly, laser photocoagulation was performed using an argon laser (Lumenis, Inc., Santa Clara, CA, USA) with a center wavelength of 532 nm, an incident power of 200mW, a spot size of 100 μm, and a pulse duration of 100ms to induce rupture of Bruch’s membrane. Intravitreal injections were performed using a 5-µL Hamilton syringe with a 33-gauge needle. A concentration gradient of STING inhibitors and activator was established to determine the optimal dosage. The following concentrations were set for the experiments: C-176 at 1 mM, 5 mM, and 25 mM; SN-011 at 0.8 mM, 4 mM, and 20 mM; and diABZI at 0.1 mM, 0.5 mM, and 2.5 mM. Fundus photography and fluorescein angiography by intraperitoneal injection of 2% fluorescein sodium solution (Alcon laboratories, TX, USA) (5 µl/g) were captured using the Micron IV retinal imaging system (Phoenix Research Laboratories, Pleasanton, CA, USA). OIR model was established as previously described [13]. These mice received a single intravitreal injection of STING inhibitors and activator at P12 and the extent of retinal angiogenesis was determined at P17.

Myeloid cells from retinae or choroidal-scleral complexes were isolated using the EasySep CD11b + Cell Isolation Kit (STEMCELL, Canada). Tissues (n = 5–7) were initially isolated on ice, mechanically fragmented, and digested in a solution containing 25 µl papain (5 µl/ml, Worthington) and 35 µl Dnase (7 µl/ml, Sigma-Aldich) at 35 °C for 8 min. The resulting single-cell suspension underwent EasySep Cell Isolation Kit protocol, including cell labeling, antibody incubation, magnetic bead binding, and cell separation. A purified CD11b + myeloid cell population (> 85% purity) was obtained for subsequent experiments (Fig. S1).

The scRNA-seq data for the retinae of control mice and OIR mice was retrieved from the GEO database (GSE152928↗) [13]. The filtering, clustering, and subclustering analysis were performed as described previously [13]. We further scored all subtypes of microglia using the irGSEA package (v1.1.2) for pathways such as ROS, IFN-α, Angiogenesis, and generated a graphical representation of the scores.

Bv2, N9, Raw264.7 or bEnd.3 cells were cultured in DMEM/F12 containing 10% fetal bovine serum and 1% Penicillin-Streptomycin (Invitrogen). Hypoxic condition was produced in a Forma 3111 Series II Water Jacketed Incubator (Thermo Fisher Scientific; Waltham, MA, USA) with 5% CO2, 1.5% O2, and balance nitrogen gas. The cells were pre-stimulated with either C-176 (10µM) or vehicle before cultured in hypoxic condition for 24 h.

The qPCR was performed as previously described [14]. The 20 µl reaction mix included 2 µl cDNA, 10 µl 2×SYBR Premix Ex Taq, 7 µl ddH2O, and 10 µmol/l of primer pairs. Roche LightCycler480 II was used for qPCR, sequences of the primers used in this study were listed in the supplementary Table 1. The western blotting was performed as previously described [14], and the primary antibodies were listed in the supplementary Table 2.

The immunofluorescence on whole-mounts and cryosections were carried out as previously described [13]. The primary antibodies were listed in the supplementary Table 2. The TUNEL assay was performed following the manufacturer’s protocol (Roche Diagnostics, Indianapolis, IN, USA). Confocal microscope images were captured using LSM980 (Carl Zeiss). H&E staining was conducted and the thickness of CNV lesions was measured as previously described [12, 13].

The representative results were showed in the figures. Data quantification was presented as mean ± standard error of measurement (SEM). Statistical analysis was conducted using one-way ANOVA with Turkey’s post hoc test for comparisons of three or more groups, or a 2-tailed Student t-test for two-group comparisons. A p-value < 0.05 was considered statistically significant. (*p < 0.05; **p < 0.01; ***p < 0.001).

To investigate the regulatory factors involved in the pathological process of retinal neovascularization, we downloaded and re-analyzed the published RNA-seq datasets of human retina [11]. Generally, the differentially expressed genes were enriched in the late DR stage of PDR but not other stages of diabetes and NPDR (Fig. 1A), which was coincident with original study by Beker et al. [11]. Specifically, we focused on the differentially expressed genes in the PDR stage. The GO analysis revealed that the upregulated genes in PDR mainly involved in the myeloid leukocyte activation, regulation of inflammatory response, innate immune response, type I interferon (IFN) production, and so on (Fig. 1B). In addition, GSEA analysis also showed genes related to positive regulation of type I IFN production and viral induced cytoplasmic pattern recognition receptor signaling pathway were up-regulated in the PDR patients compared with healthy controls (Fig. 1C). Furthermore, the expressions of CGAS and STING, key molecules of cytoplasmic DNA sensing pathway that induces the production of type I IFN and activates the innate immune system, were significantly elevated in PDR patients compared to others, whereas no significant difference was observed between diabetic or NPDR patients and healthy controls (Fig. 1D). Then the relationships between STING and the angiogenesis or inflammatory responses were analyzed. The expression of STING is positively correlated to genes associated with angiogenesis or inflammatory responses (Fig. 1E), indicating the possible involvement of STING signaling in the angiogenic process.

Next, we examined the cGAS-STING pathway in two classical models of retinal/choroidal angiogenesis, the CNV and OIR models (Fig. 2A). The laser-induced CNV is a self-limiting model of angiogenesis, with lesions peaking at D7 and then gradually regressing [15]. Another classical angiogenesis model, OIR, demonstrates the development of retinal neovascular tufts from P12 to P17, reaching their peak at P17, followed by a gradual regression of retinal angiogenic tufts [16]. As illustrated in Fig. 2B, the levels of cGAS and phosphorylated-STING (p-STING) in the choroid-scleral complex were notably up-regulated on D3 post laser administration and gradually declined thereafter, but still be elevated on D7. The downstream molecules, TBK1 and IRF7 also exhibited heightened activation on D3 and D7 (Fig. 2B). The expressions of p-γH2A.X, a marker of double-strand break (DSB) damage response signaling, p-STING, and p-TBK1 were significantly elevated on D3 and D7 post-CNV induction (Fig. 2C and S2A, B), indicating the involvement of the cGAS-STING axis in the induction of CNV. Similarly, the expressions of p-STING and p-TBK1 in OIR retinae were significantly increased as early as P13, with a prominent peak on P15, followed by a gradual decline (Fig. 2D). On P17, the peak time-point of retinal angiogenesis development in OIR, the cGAS-STING pathway was highly activated (Fig. 2E), indicating the contribution of the cGAS-STING pathway in the progression of pathological retinal angiogenesis. These findings collectively suggest that the activation of the cGAS-STING signaling contributes to the development of retinal/choroidal pathological angiogenesis in mice.

To further evaluate the effect of cGAS-STING pathway on the development of pathological angiogenesis, we established the CNV and OIR models utilizing Sting^gt mice, which harbor a mutant missense allele in the Sting gene resulting in the lack of STING protein [17]. As shown in the representative fundus images of Fig. 3A, laser burns were successfully induced in both WT and Sting^gt mice. The Bruch’s membrane was damaged by focal laser photocoagulation and the new formed and leaky choroidal blood vessels invaded into the retina in the CNV model [15]. When monitored by FFA, The Sting^gt-CNV mice present relatively less leakage area (Fig. 3A). Choroid-sclera flat mounts and histological sections stained with HE also revealed that Sting^gt-CNV mice displayed a smaller area of neovascularization and a thinner lesion thickness in both the transverse and longitudinal sections of the eyeball (Fig. 3B). Similarly, areas of retinal neovascular tufts and avascular zone were remarkably decreased in the Sting^gt-OIR mice (Fig. 3C), and neovascular cells anterior to the internal limiting membrane (ILM), hallmarks of angiogenesis in OIR, were significantly reduced in the Sting^gt-OIR mice (Fig. 3C). The pharmacological activation and inhibition of STING were also utilized to further evaluate the role of STING on retinal angiogenesis. We conducted dose-response experiments to determine the optimal dosage of C-176, SN-011, and diABZI. As shown in Fig. S3A-C, concentrations of 5 mM for C-176, 4 mM for SN-011, and 0.5 mM for diABZI were selected as the lowest effective concentrations for further study. Intravitreal injection of the STING inhibitor C-176 resulted in effects similar to the Sting^gt group, while administration of diABZI, a STING agonist that binds to the cGAMP binding pocket of STING, led to STING activation and larger area of neovascular tufts, indicating the essential role of STING activation for promoting pathological angiogenesis (Fig. 3D). In conclusion, these results demonstrated that the cGAS-STING signaling is required for the development of pathological angiogenesis.

To uncover the potential mechanisms underlying cGAS-STING activation during retinal/choroidal angiogenesis, we re-analyzed our previously published single-cell data in OIR on P17 [13]. Interestingly, the genes related to cGAS-STING signaling were enriched in the retinal myeloid cells (Fig. S4A), consistent with previous studies that STING is mainly expressed in myeloid cells located in central nervous system (CNS) [18]. Immunofluorescence staining of cryosection showed that the cGAS was specifically expressed on the Iba-1 + microglia that adjacent to neovascular tufts, while lack of cGAS expression was observed in the Iba-1 + microglia cells away from neovascular tufts, indicating the involvement of cGAS-STING signaling in the microglia activation and migration towards the neovascular tufts (Fig. 4A). The choroid-sclera flat-mounts staining showed that Iba-1 + myeloid cells around the vessels in WT-CNV mice exhibited the amoeboid activation morphology and expressed TNF-α and IFN-β, of which production were triggered by cGAS-STING signaling. By contrast, the Sting^gt mice presented much less TNF-α and IFN-β expressed on the Iba-1 + myeloid cells (Fig. 4B). Similarly, the Sting^gt-OIR mice also exhibited less amoeboid Iba-1 + microglia around the CD31 + tufts, and the TNF-α and IFN-β expressed on the Iba-1 + microglia were significantly reduced (Fig. S4B) indicating the STING activation contributed to the microglia activation and the inflammatory responses during angiogenesis. In addition, the qPCR analysis demonstrated lower secretion of pro-angiogenic factors and reduced activation of interferon-stimulated genes (ISGs) in the Sting^gt group (Fig. 4G). Collectively, these results demonstrated STING activity in myeloid cells directed inflammatory responses and angiogenesis in the CNV and OIR models.

Given that cytoplasmic DNA serves as a trigger for cGAS-STING signaling [19, 20], we aim to further investigate whether hypoxic stress induces cytoplasmic DNA leakage and downstream STING signaling activation in myeloid cells. Two microglia cell lines, Bv2 and N9, as well as a macrophage cell line, Raw264.7, were cultured under the hypoxic condition for 24 h. As expected, upregulation of p-γH2A.X was observed in all three cell lines in response to hypoxic insults (Fig. 4C, D). In addition, the cGAS-STING signaling was activated with up-regulation of cGAS, p-STING, and p-TBK1 in the hypoxia-stimulated myeloid cells. However, no such response was observed in the bEnd.3 endothelial cell line under hypoxic condition (Fig. 4C, D). Consistently, immunofluorescence staining revealed that hypoxia led to the accumulation of cytoplasmic DNA, possibly as the micronuclei from nuclear DNA leakage or mtDNA leakage from mitochondria (Fig. 4E, F and S5A-C). These cytoplasmic DNA were co-stained with p-STING, indicating the STING activation upon the recognition of cytosolic DNA in myeloid cells.

Activation of cGAS-STING pathway leads to increased expression of downstream inflammatory factors, including I-IFN and TNF-α, which serves as a classical inducer of cell necroptosis. Our previous studies have found that microglia promoted the retinal angiogenesis through RIP3-mediated necroptosis [13]. Here, we hypothesize that the cGAS-STING signaling in ocular myeloid cells would induce their necroptosis, resulting in the development of angiogenesis. To validate this hypothesis, we firstly performed the GSEA analysis on the scRNA-seq data of the CD11b + cells. Among the five different CD11b clusters (Fig. 5A), sMG2 (small microglia cluster 2), a previously identified subset associated with necroptosis [13], was highly responded to hypoxia and exhibited increased IFN-α responses (Fig. 5A), indicating the involvement of cGAS-STING signaling in myeloid cells necroptosis. Furthermore, the expression of key proteins involved in the necroptosis pathway was examined in enriched CD11b + myeloid cells from both the CNV and OIR models. As expected, the p-RIP3, p-MLKL, and FGF2 were highly expressed in the myeloid cells enriched from WT-CNV and WT-OIR, which were abrogated by Sting deficiency (Fig. 5B), indicating the requirement of STING activity for necroptosis. Immunofluorescence staining and 3D co-localization analysis also showed that the Iba-1 + myeloid cells in Sting^gt-CNV and Sting^gt-OIR mice presented with less p-MLKL expression and reduced TUNEL positivity in comparison to WT groups (Fig. 5D, E). In vitro, hypoxia induced necroptosis with high expression of RIP1, p-RIP3, and p-MLKL in N9 and Raw264.7, which was suppressed by STING inhibitor of C-176, indicating the essential role of STING activation on necroptosis in myeloid cells (Fig. 5C). Taken together, these findings suggested that activation of cGAS-STING triggered necroptosis in myeloid cells, resulting in the development of pathological neovascularization.

Finally, we explored the therapeutic potential of targeting cGAS-STING signaling in neovascular ocular diseases. Two STING inhibitors, C-176 and SN-011 [21, 22], were employed to block the cGAS-STING axis through intravitreal injections in OIR and CNV mice. Both C-176 and SN-011 showed remarkable suppression of pathological vascular lesions, along with reduced FFA leakage areas in the CNV model (Fig. 6A, B). Moreover, combination therapy of C-176 or SN-011 with anti-VEGF neutralizing antibody resulted in the least leakage area (Fig. 6A, B), further highlighting the therapeutic potential of targeting the cGAS-STING axis in neovascular ocular diseases. Similarly, the intravitreal injection of C-176 or SN-011 could significantly inhibit the neovascular tufts areas and their combination therapy with anti-VEGF neutralizing antibody also present with least neovascular tufts areas in the OIR model (Fig. 6C, D). Collectively, our findings demonstrated that targeting STING by C-176 or SN-011 could effectively suppress retinal/choroidal pathological angiogenesis, and combination therapy with anti-VEGF could achieve enhanced efficacy.

Below is the link to the electronic supplementary material.

Supplementary Material 1

Supplementary Material 2

Supplementary Material 3