Astrocyte hypoxic response is essential for pathological but not developmental angiogenesis of the retina

Abstract

Vascular/parenchymal crosstalk is increasingly recognized as important in the development and maintenance of healthy vascularized tissues. The retina is an excellent model in which to study the role of cell type-specific contributions to the process of blood vessel and neuronal growth. During retinal vascular development, glial cells such as astrocytes provide the template over which endothelial cells migrate to form the retinal vascular network, and hypoxia-regulated vascular endothelial growth factor (VEGF) has been demonstrated to play a critical role in this process as well as pathological neovascularization. To investigate the nature of cell-specific contributions to this process, we deleted VEGF and its upstream regulators, the hypoxia-inducible transcription factors HIF-1 alpha and HIF-2 alpha, and the negative regulator of HIF alpha, von Hippel-Lindau protein (VHL), in astrocytes. We found that loss of hypoxic response and VEGF production in astrocytes does not impair normal development of retinal vasculature, indicating that astrocyte-derived VEGF is not essential for this process. In contrast, using a model of oxygen-induced ischemic retinopathy, we show that astrocyte-derived VEGF is essential for hypoxia-induced neovascularization. Thus, we demonstrate that astrocytes in the retina have highly divergent roles during developmental, physiological angiogenesis, and ischemia-driven, pathological neovascularization.

Figure 1. Tissue specific deletion of floxed genes in astrocytes in the developing retina using GFAPcre-expressing mice

(A) LacZ reporter gene expression was analyzed in retinal flatmounts by X-gal staining (blue). LacZ expression in GFAPcre-negative animals (left) is absent. In GFAPcre-positive mice (right) at P3, the transgene expression is most pronounced in the central retina but is also detectable already in the periphery. (B) At the same age the growing vascular front is still limited to the immediate surroundings of the optic nerve and well within the area of LacZ expression. (C) At P14, strong LacZ expression is now observed throughout the retinal flatmounts. (D) Retinal vascular development at this age has significantly advanced, already covering the entire retina. (E) In retinal sections, LacZ expression is confined to the inner surface of the retina, consistent with the localization of astrocytes. (F) Copy numbers of a floxed allele (VHL) was quantified in retinal DNA relative to an unfloxed allele (VEGF). In GFAPcre-positive mice at P3, relative abundance of the floxed allele is significantly reduced compared to GFAPcre-negative animals, thus indicating successful recombination.

Figure 2. Conditional deletion of VEGF and HIFα-isoforms in astrocytes does not impair normal retinal vascular development in mice

(A) Respresentative picture of wildtype (left) and astrocyte specific VEGF knockout littermates (right) on P7 (upper panel) and at adult age (both >4). Knockout of VEGF does not result in apparent morphological alterations in the developing vasculature. (B) Retinal vascular development is not delayed in animals which lack VEGF, HIF-1α or HIF-2α in astrocytes as measured by distance of the growing vascular front from the optic nerve head compared to their wildtype littermates (data represents mean ± S.D.). (C,D) Total VEGF mRNA expression (D) and EPO mRNA expression (E) in retinae of P7 pups with the indicated conditional knockout in astrocytes is not significantly different compared to wildtype expression levels (data represent mean ± S.E.M.).

Figure 3. Developmental vascular defects in retinae of mice with conditional deletion of VHL in astrocytes

(A) During retinal vascular development at P7, loss of VHL in astrocytes induces extensive retinal hypervascularity which persists into adulthood (representative picture of retina flatmounts; n=3). (B) VEGF mRNA expression levels are significantly elevated in retinal tissue of GFAPcre+/VHL^+f/+f mice compared to wildtype littermates (data represents mean ± S.E.M.). (C) The excessive vessels form convoluted bundles of vessels that are ensheathed by astrocytes (scalebars 100 μm). (D) In histological sections, the abnormal vessel convolutes are located in the level of the superficial vascular plexus, locally extending into the intermediated plexus. The immunoreactivity for GFAP is increased, indicating an increased number of astrocytes that accompany the hypervascular abnormalities (scalebars: 200 μm). (E) Confocal images of the different vascular plexi in flat-mounted retinae confirm that the hypervascular changes occur predominantly in the superficial and partially in the intermediate vascular plexus, resulting in gross alterations of the vascular structure of these two plexus. In contrast, structure of the deep vascular plexus is largely preserved or rather decreased with no apparent hypervascularity (scalebars: 200 μm).

Figure 4. Pathological neovascularization in GFAPcre+/VHL+f/+f mice is driven by HIF-2α-mediated VEGF overexpression

(A) Double knockout of VHL and VEGF in astrocytes completely rescues the hypervascular phenotype observed with Vhlh-deletion (representative picture of retina flatmounts; n=3). (B) VEGF expression levels are normal in GFAPcre+/VHL^+f/+f/VEGF^+f/+f mice (data represents mean ± S.E.M.). (C) Double knockout of VHL and HIF-1α in astrocytes (left) does not change the developmental hypervascularity in the retina, and the phenotype closely resembles the one of GFAPcre+/VHL^+f/+f mice. In contrast, double deletion of VHL and HIF-2α (right) results in a normal development of the vasculature into adulthood (representative picture of retina flatmounts; n=3). (D) VEGF mRNA expression levels are reduced to wildtype levels by double deletion of VHL and HIF-2α, but not by VHL and HIF-1α (data represents mean ± S.E.M.).

Figure 5. Astrocyte-derived VEGF, controlled by HIF-2α, is a key mediator for neovascular tuft formation in OIR

(A) In a murine model of oxygen-induced retinopathy (OIR), the area of vaso-obliteration induced by hyperoxia was quantified 24 hours after oxygen treatment (P13): there are no significant differences between GFAPcre+/VEGF^+f/+f, GFAPcre+/HIF-1α^+f/+f and GFAPcre+HIF-2α^+f/+f animals and their respective wildtype littermates (data represents mean ± S.D.). (B,C) Gene expression analysis of VEGF (B) and EPO (C) in P13 retinae after hyperoxia shows that both genes are upregulated, indicating retinal hypoxia. However no genotype-specific difference of expression levels is observed between mice with conditional knockout of VEGF, HIF-1α and HIF-2α in astrocytes. (D-M) Vaso-obliteration and neovascular tuft formation in OIR was assessed at P17, 5 days after the end of the hyperoxic phase. (D-F) Loss of VEGF in astrocytes does not change vaso-obliteration but significantly reduces neovascularization. (G-I) Loss of HIF-1α in astrocytes does not affect vaso-obliteration or neovascularization in this OIR model, whereas (K-M) conditional deletion of HIF-2α significantly reduces vaso-obliteration and neovascularization relative to wildtype littermate controls (D,G,K: images show representative retina flatmounts (n>4 for all); E,F,H,I,L,M: data represents mean area relative to respective wildtype littermate controls ± S.D.).