Localization of APN protein in the retinal vasculature
Firstly, we conducted immunofluorescence staining on retinal flat mounts to assess APN localization in the mouse retina. Figure 1A and 1B show representative immunofluorescence images for APN, alpha-smooth muscle actin (αSMA, a marker of vascular smooth muscle cells), and CD31 (a marker of vascular endothelial cells) in the primary layer and those for APN, neuron-glia antigen 2 (NG2, a marker of pericytes), and CD31 in the capillary layer, respectively. In wild-type (WT) mice, APN signal was mainly detected along retinal arterioles, but not venules, in the primary layer (Fig. 1A, upper panels), but was hardly detectable in the capillary layer (Fig. 1B, upper panels). Such APN signal was completely diminished in T-cadherin knockout (Tcad-KO) mice (Fig. 1A and 1B, bottom panels) similar to APN knockout (APN-KO) mice (Fig. 1A and 1B, middle panels). Under high magnification, APN was found to be merged with CD31, rather than with αSMA, in the retinal arterioles of WT mice (Fig. 1C and 1D). We also observed APN signal, together with CD31, in the transitional zone between arterioles and capillaries in which αSMA was no longer detectable (Fig. 1C and 1E). These results indicate that APN protein was localized on vascular endothelium of the murine retina in a T-cadherin-dependent manner.
Decreased APN accumulation in the diabetic retinal vasculature
Next, an STZ-induced diabetic mouse model was used to examine quantitative change in APN protein accumulation in the retinal vasculature under diabetic conditions (Fig. 2A). As shown in Figure 2B, immunofluorescence staining showed that APN in retinal arterioles was reduced at 5 months after the onset of diabetes, compared to non-diabetic mice. This retinal APN reduction had intensified by 9 months after the onset of diabetes, while no significant changes were observed in CD31 staining (Fig. 2B), indicating that APN accumulation in the retinal vascular endothelium progressively decreased during the course of diabetes.
Effects of dapagliflozin on retinal APN and vascular permeability in STZ-induced diabetic mice
Increased retinal vascular permeability has been reported to be observed at the early stage of DR in mice and humans, which triggers the disease progression of DR (15, 16). We thus evaluated retinal vascular leakage in STZ-induced diabetic mice by intracardiac injection of Hoechst 33258 (molecular weight of 534 Da) and fluorescein isothiocyanate (FITC)-dextran (molecular weight of 3,000-5,000 Da). In this assay, extravascular Hoechst signals could be detected when circulating Hoechst leaked out of blood vessels and was incorporated into the nuclei of perivascular cells. Under the non-diabetic state (before STZ injections), FITC-dextran was distributed within vessels and Hoechst staining was observed only in the nuclei of endothelial cells (Supple. Fig. 1A). Either Hoechst or FITC-dextran was hardly detected in the extravascular space at 1 week after the onset of diabetes, whereas, by 6 weeks, extravascular Hoechst-positive, but not FITC-dextran-positive, regions were observed mainly around the bifurcations of peripheral arterioles (Supple. Fig. 1A). These results suggest that microvascular permeability for small molecules occurred at a relatively early stage after the onset of diabetes.
Next, we investigated blood glucose-lowering effects on retinal APN accumulation and vascular permeability in diabetic mice. After the induction of diabetes, STZ-injected mice were treated with vehicle (STZ/Vehi group) or dapagliflozin, a selective sodium-glucose co-transporter 2 (SGLT2) inhibitor, (STZ/Dapa group) for 42 days (Fig. 3A). Compared to non-diabetic control mice, a similar body weight reduction was observed in both the STZ/Vehi and STZ/Dapa groups (Fig. 3B), whereas blood glucose levels were significantly reduced to below 300 mg/dL by the administration of dapagliflozin (Fig. 3C, STZ/Vehi versus STZ/Dapa). Plasma APN levels did not differ among the three groups (Fig. 3D). Six weeks of hyperglycemia was associated with a significant reduction of APN on retinal arterioles compared to control mice (Fig. 3E and 3F, Cont versus STZ/Vehi), while APN accumulation was preserved in the retina of the STZ/Dapa group (Fig. 3E and 3F, STZ/Vehi versus STZ/Dapa). As shown in immunofluorescence images (Fig. 3G), extravascular Hoechst-positive regions were increased in STZ-induced diabetic mice (Cont versus STZ/Vehi), but were reduced by dapagliflozin treatment (STZ/Vehi versus STZ/Dapa). Treatment of diabetic mice with dapagliflozin tended to reduce extravascular Hoechst-positive cells (Fig. 3H, left) and area (Fig. 3H, right). Collectively, glucose-lowering treatment with dapagliflozin prevented a decrease in retinal APN accumulation and an increase in retinal vascular permeability.
In both mice and humans, upregulation of endothelial cell adhesion molecules such as vascular cellular adhesion molecule-1 (VCAM-1) is recognized as one of causes associating with DR development (17-19). In the isolectin B4 (IB4)-labeled endothelium of retinal arterioles, VCAM-1 expression was tended to increase in diabetic mice compared to control mice (Fig. 4A and 4B, Cont versus STZ/Vehi), but was significantly suppressed by dapagliflozin (Fig. 4A and 4B, STZ/Vehi versus STZ/Dapa). In contrast to VCAM-1, expression of claudin-5, the major tight-junction molecule in endothelial cells of BRB (20), was reduced in the STZ/Vehi group (versus Cont group), but was restored in the STZ/Dapa group (versus STZ/Vehi group) (Fig. 4C).
Early progression of diabetic retinal vascular abnormalities in APN-deficient mice
Finally, since the APN reduction in the retinal endothelium accompanied early diabetic vascular changes (Fig. 3 and Fig. 4), we investigated the role of APN in the development of DR using APN-KO mice. Retinas from WT and APN-KO mice were immunohistochemically analyzed at 4 weeks after the onset of STZ-induced diabetes (Fig. 5A). Neither body weight nor plasma glucose levels differed between WT and APN-KO mice following STZ injection (Fig. 5B and 5C). Retinal vascular permeability in APN-KO mice was not altered under the non-diabetic state (Supple. Fig. 1B). However, compared to diabetic WT mice, diabetic APN-KO mice exhibited a significant increase in vascular permeability, determined by extravascular Hoechst-positive cells and area, after only 4 weeks of hyperglycemia (Fig. 5D and 5E). Additionally, hyperglycemia-induced upregulation of retinal VCAM-1 was significantly greater in retinas of APN-KO mice than in WT mice (Fig. 5F and 5G). Reflecting the increased vascular permeability, diabetic APN-KO mice showed a strong reduction in endothelial claudin-5 expression, notably in peripheral arterioles (Fig. 5H). These results indicate that APN deficiency was associated with a rapid progression of DR.