1. EPCR is highly expressed in retinal ECs during RNV.
The mouse model of oxygen-induced retinopathy (OIR) serves as a proxy for human pathological RNV, including PDR and ROP.23 The neovascular response peaked on postnatal day 17 (P17) when the mice were returned to room air for five days (Fig. 1A). To identify the potential genes involved in pathological neovascularization, we performed transcriptomic analyses of the retinas of P17 mice of OIR (GSE241239). To minimize the RNA-seq bias, we further analyzed transcriptomic data from two other independent studies (GSE194176 and GSE158799).24,25 A combined analysis revealed 149 upregulated genes (Fig. 1.B-C). Among these candidate genes, we decided to focus on Epcr because it was reported to be specifically expressed on the surface of vascular ECs and stem cells.13,14 Further data confirmed that EPCR expression was upregulated in the retinas of P17 mice in the OIR (Fig. 1.D-E). We sorted ECs from the mouse retina to evaluate the expression of EPCR in the retinal vasculature. Endothelial EPCR expression was much higher in the OIR group than that in the control group (Fig. 1F-G).
Next, we examined the expression pattern of EPCR in the retina. Immunofluorescence staining with an anti-EPCR antibody showed that EPCR was selectively expressed in the retinal ECs of P17 mice under physiological conditions (FigS1A). In the OIR model, the expression of EPCR was detected only in retinal ECs, and the fluorescence intensity of EPCR signals was much higher in the neovascular tufts (FigS1B-C). Epcr-CreERT2 (also known as Procr-CreERT2) mice were previously used to mark proliferative ECs in the retinal vasculature.21 Here, we used Epcr-CreERT2 and Rosa26tdT mice to perform genetic lineage tracing of EPCR-positive cells in the retina of P17 mice. Tamoxifen (TAM) was administered for three consecutive days at P10, P11, and P12, and the retina was analyzed at P17 (FigS1D). Under physiological conditions, a very small proportion of EPCR + cells were observed (Fig. 1H). While there was a large proportion of EPCR + cells in the retina of OIR mice, and EPCR + cells mainly colocalized with neovascular tufts (Fig. 1H-I). These data indicated that hypoxia significantly increased the transcription of EPCR. Hypoxia is also known to play a vital role in the retinal vascular development process.26 Therefore, the expression pattern of EPCR in the retinal vasculature of postnatal mice were then assessed. Immunofluorescence staining with anti-EPCR showed that the levels of EPCR were much higher in the margin of the retinal vasculature (FigS1E). Consistent data were acquired using Epcr-CreERT2 and Rosa26tdT mice (FigS1F-G). Collectively, these data suggest that EPCR is specifically expressed in ECs of the retina and is upregulated in neovascularization in the OIR model and during retinal vascular expansion.
EPCR activates downstream signaling pathways by binding to its ligands. Thus, we assessed the concentrations of EPCR ligands in the serum of OIR mice. The data showed that the levels of aPC in the serum of OIR mice were higher than those in control mice (Fig. 1J). To evaluate whether the findings in mice hold promise for human patients with RNV, we investigated the levels of aPC in the vitreous humor and serum in patients with PDR. The concentrations of aPC in the vitreous humor and serum were significantly higher in patients with PDR than in control with macular hole or idiopathic epiretinal membrane (Fig. 1K-L).
2. EPCR contributes to RNV in vivo and in vitro.
To determine the role of EPCR in RNV, we intercrossed the Cdh5-CreERT2 mouse with Epcrflox/flox mouse to generate EC-specific Epcr KO mice (EpcriΔEC). The expression of EPCR was almost completely ablated in ECs sorted from EpcriΔEC mice (FigS2A-B). To assess the role of EPCR in pathological RNV, we examined the changes in the retinal vasculature of EpcriΔEC mice of OIR at P17. Our data showed that the area of neovascular tufts in the retina of EpcriΔEC mice was smaller than that in EpcrWT mice (Fig. 2A-B). In contrast, the avascular area in EpcriΔEC mice’ retina was larger than that in EpcrWT mice (Fig. 2A, 2C). Furthermore, we used 5-ethynyl-2’-deoxyuridine (EdU) to label proliferative cells in the retina and used E26 transformation-specific (ETS)-related gene (ERG) as an ECs marker. The result showed that EdU+ ECs in the retina decreased significantly in EpcriΔEC mice compared to EpcrWT mice (Fig. 2D-E). These in vivo results suggest that upregulated EPCR in pathological vasculature drives EC growth and angiogenesis in OIR.
Next, we determined the effect of Epcr deletion on retinal vascular expansion in postnatal retinas. EC-specific deletion of Epcr led to a sparse vascular network and reduced radial expansion of the superficial retinal vascular plexus (FigS2C-D). Tip cells and branch points were decreased in the retina of Epcr KO mice (FigS2E-G). Meanwhile, ECs proliferation was also significantly inhibited in EpcriΔEC mice (FigS2H-I).
To further evaluate the impact of elevated EPCR on angiogenesis, we used an adenovirus to overexpress EPCR and assessed the angiogenic behavior of ECs in vitro. EPCR adenovirus treatment significantly increased the protein levels of EPCR in ECs, which promoted ECs proliferation in an EdU staining assay (Fig. 2F-G, FigS2J). A three-dimensional endothelial spheroid assay showed the number and length of endothelial sprouts were increased in the EPCR adenovirus treatment group (Fig. 2H-I). Moreover, a wound scratch assay revealed that EPCR adenovirus-treated ECs were more motile than control adenovirus treated ECs (Fig. 2J-K). Furthermore, the knockdown of EPCR using siRNA suppressed ECs proliferation, migration and sprouting (Fig. 2L-Q, FigS2K). Together with in vivo results, these data indicated that EPCR promotes vascular growth.
3. Pharmacological inhibition of EPCR ameliorates pathological retinal neovascularization.
To evaluate the translational potential of targeting EPCR in anti-angiogenic therapy, we assessed the effects of a neutralizing antibody against EPCR on pathological angiogenesis in vitro and in vivo. We found that a single dose of EPCR antibody significantly reduced the area of neovascular tufts compared to the vehicle and anti-IgG groups (Fig. 3A-B). Meanwhile, the avascular area increased in the EPCR antibody treatment group, which was consistent with that in Epcr KO mice (Fig. 3A, C). Suppressed proliferation of ECs in retinal neovascular tufts was also observed after treatment with the EPCR antibody (Fig. 3D-E). Similar results were obtained in vitro as EPCR antibody treatment significantly inhibited ECs proliferation, migration and sprouting (Fig. 3F-L). These results were consistent with our findings using EpcriΔEC mice, suggesting the translational potential of anti-EPCR in retinal neovascularization.
4. EPCR regulates the expression of HO-1 through Nrf2/Keap1
To reveal the possible molecular mechanisms underlying EPCR-regulated angiogenesis, we performed transcriptomic analyses of EPCR siRNA- or control siRNA-treated ECs (GSE249130, FigS3A). These data were then combined and analyzed with the mice retina results with OIR (FigS3B). 26 genes were upregulated in the retina of OIR mice and downregulated in EPCR-depleted ECs (Fig. 4A). Among these genes, HO-1 was the most significantly affected (Fig. 4A). Increased mRNA and protein levels of HO-1 were confirmed in the retinas of OIR mice (FigS3C-D). Further data showed that the expression of HO-1 was suppressed at the mRNA and protein levels in ECs treated with EPCR siRNA (Fig. 4B-C). Additionally, the expression of HO-1 was upregulated at the mRNA and protein levels in ECs treated with EPCR adenovirus (Fig. 4D-E). The expression of HO-1 in retinal ECs of OIR mice was then evaluated. We sorted ECs from the retina of OIR mice using flow cytometry. The results indicated that the expression of HO-1 in ECs was upregulated in the OIR group (FigS3E-F). This was further confirmed by immunofluorescence staining. The expression of HO-1 in physiological retina did not show apparent pattern, while it was highly expressed in neovascular tufts of OIR retinas (FigS3G-H). Further data showed that the expression of HO-1 in ECs sorted from the retina of EpcriΔEC mice decreased compared with that in Epcrwt mice (Fig. 4F-G). Collectively, these results suggested that HO-1 was a downstream molecule of EPCR in regulating angiogenesis.
AKT is recognized as a target of EPCR signaling.27 High EPCR expression or stimulation with aPC usually leads to AKT activation.16,19 We found that overexpression of EPCR using an adenovirus promoted the activation of AKT, whereas silencing EPCR with siRNA inhibited AKT phosphorylation (Fig. 4H-I). Nuclear factor erythroid 2-related factor 2 (NRF2) is a ubiquitous transcription factor directly regulating HO-1.28 NRF2 is tightly regulated by Kelch-like ECH-associated protein 1 (KEAP1) through ubiquitination and proteasome-dependent degradation.29 Therefore, we assessed the effects of EPCR on the expression of KEAP1 and NRF2. Overexpression of EPCR in ECs promoted the levels of NRF2 and HO-1 and inhibited the expression of KEAP1 (Fig. 4I). Decreased expression of NRF2 and HO-1, and increased expression of KEAP1 were observed after EPCR knockdown in ECs (Fig. 4H). Meanwhile, aPC treatment stimulated the activation of AKT, and increased the expression of NRF2 and HO-1 while suppressing the expression of KEAP1 (FigS3I).
AKT activates NRF2 in multiple cells.30,31 Our data showed that inactivating AKT with LY294002 (a PI3K inhibitor) blocked the effect of EPCR overexpression on NRF2, KEAP1, and HO-1 (Fig. 4J). Consistent results were observed when ECs were treated with aPC (FigS3J). These results suggested that AKT plays an important role in mediating EPCR signaling. Furthermore, we observed that depleting the expression of KEAP1 using siRNA rescued the downregulation of NRF2 and HO-1 caused by EPCR siRNA (Fig. 4K, FigS3K). Consistent results were obtained after treatment with Ki696 (a KEAP1 inhibitor) (Fig. 4L). These data suggested that EPCR controlled HO-1 expression via the AKT-KEAP1-NRF2 pathway.
5. Genetic ablation of endothelial Ho-1 attenuates angiogenesis
To evaluate the role of HO-1 in retinal angiogenesis, we bred Cdh5-CreERT2 mice with Ho-1flox/flox mice to generate EC-specific Ho-1 KO mice (Ho-1iΔEC). The deletion efficiency was determined using flow cytometry-sorted ECs, and the expression of HO-1 in ECs sorted from Ho-1iΔEC mice was significantly decreased (FigS4A). The role of HO-1 in pathological retinal angiogenesis was then determined in the OIR mice. Our data showed that the area of neovascular tufts in the retina of Ho-1iΔEC mice was smaller, while the avascular area in the retina of Ho-1iΔEC mice was larger than that in Ho-1WT mice (Fig. 5A-C). Furthermore, our results showed that EdU+ ECs decreased significantly in the retina of Ho-1iΔEC mice than in Ho-1WT mice (Fig. 5D-E).
Physiological retinal vascular development was also affected in Ho-1iΔEC mice. Deleting endothelial Ho-1 reduced radial expansion of the superficial retinal vascular plexus (FigS4B, C). Meanwhile, the vascular density, tip cells and branch points were decreased in the retina of Ho-1iΔEC mice (FigS4D-F). Additionally, EdU+ ECs decreased significantly in the retina of Ho-1iΔEC mice than in Ho-1WT mice (FigS4G-H).
The role of HO-1 in angiogenesis was further assessed in vitro. First, the knockdown efficiency of HO-1 siRNA was evaluated. The EdU assay showed that silencing HO-1 inhibited ECs proliferation induced by the EPCR adenovirus (Fig. 5F-G). A three-dimensional endothelial spheroid assay was used to evaluate the angiogenic potential of ECs treated with EPCR adenovirus and HO-1 siRNAs. Fewer and shorter sprouts were observed in HO-1 siRNAs-treated ECs than EPCR adenovirus treated ECs (Fig. 5H-I). Besides, ECs treated with HO-1 siRNAs exhibited impaired migration ability compared to the EPCR adenovirus treatment group (Fig. 5J-K). We also assessed the effects of HO-1 knockdown on ECs. The results showed that silencing HO-1 suppressed ECs proliferation, migration, and sprouting (FigS4J-O). These results collectively revealed the vital role of HO-1 in developmental angiogenesis and RNV.
6. Endothelial heme catabolism impacts angiogenesis.
HO-1 catalyzes heme degradation into CO, biliverdin, and iron.32 Catabolism of cytotoxic labile heme and the generation of CO exert cytoprotective effects.33 Thus, we further assessed whether the angiogenic role of HO-1 was mediated by its enzymatic function. Here, HO-1 enzymatic inhibitor, zinc protoporphyrin Ⅸ (ZnPPⅨ), was used to determine their effect on pathological RNV. The data showed that ZnPPⅨ treatment decreased the area of neovascular tufts in the retina and increased the avascular area (Fig. 6A-C). Besides, ZnPPⅨ suppressed ECs proliferation as EdU + retinal ECs decreased significantly in mice treated with ZnPPⅨ (Fig. 6D-E).
Next, we determined the effect of ZnPPⅨ on retinal vascular expansion. ZnPPⅨ treatment led to a sparse vascular network and reduced radial expansion of the superficial retinal vascular plexus (FigS5A-B). The tip cells and branch points were decreased in the retina of mice treated with ZnPPⅨ (FigS5C-E). Meanwhile, ZnPPⅨ significantly suppressed ECs proliferation in the retina (FigS5F-G).
Further data showed that EPCR-induced cell proliferation, migration and sprouting of ECs in vitro were inhibited by ZnPPⅨ (Fig. 6F-K). Additionally, we evaluated the effect of ZnPPⅨ on ECs treated with HO-1 siRNA. The result showed that ZnPPⅨ did not affect ECs proliferation, migration, and sprouting after HO-1 knockdown (FigS5H-M). These data indicated that endothelial heme catabolism contributes to EPCR-mediated RNV.
7. Heme catabolism-derived CO contributes to pathological angiogenesis.
Endogenous CO is a signaling molecule that promotes tumor growth and angiogenesis.34,35 Heme catabolism in macrophages increases tumor angiogenesis through CO production, suggesting a role for CO in regulating angiogenesis.33 Here, we showed that pathological retinal neovascular tufts in carbon monoxide releasing molecule-3 (CORM3) treated Ho-1iΔEC mice significantly increased compared with that in Ho-1iΔEC mice, while the avascular area was decreased (Fig. 7A-C). Furthermore, the data showed that EdU+ ECs also increased significantly in the CORM3-treated retina of Ho-1iΔEC mice than that in Ho-1WT mice (Fig. 7D-E). Pathological retinal neovascular tufts in CORM3-treated EpcriΔEC mice also significantly increased, and avascular area decreased compared with that in EpcriΔEC mice (FigS6A-C). Consistent data were acquired for the number of EdU+ ECs (FigS6D-E).
We assessed the effect of CORM3 on postnatal retinal vasculature in Ho-1iΔEC and EpcriΔEC mice. Treatment with CORM3 could rescue the sparse vascular network and reduce the radial expansion of superficial retinal vascular plexus in Ho-1iΔEC mice (FigS7A-B). Compared with Ho-1iΔEC mice, tip cells and branch points increased in the retina treated with CORM3 (FigS7C-E). Consistent results were acquired in CORM3-treated EpcriΔEC mice (FigS7F-J).
Furthermore, CORM3 rescued HO-1 silence-caused inhibition of ECs proliferation, migration, and sprouting (Fig. 7F-K). Additionally, the effect of CORM3 on EPCR-deleted ECs was evaluated. The results showed that EPCR silencing inhibited ECs proliferation, migration, and sprouting, which was partially reversed after treatment with CORM3 (FigS8A-F). These results suggested that CO was an important contributor to EPCR dependent endothelial heme catabolism-regulated angiogenesis.