β-Secretase (BACE1) inhibition causes retinal pathology by vascular dysregulation and accumulation of age pigment

Abstract

β-Secretase (BACE1) is a major drug target for combating Alzheimer's disease (AD). Here we show that BACE1(-/-) mice develop significant retinal pathology including retinal thinning, apoptosis, reduced retinal vascular density and an increase in the age pigment, lipofuscin. BACE1 expression is highest in the neural retina while BACE2 was greatest in the retinal pigment epithelium (RPE)/choroid. Pigment epithelial-derived factor, a known regulator of γ-secretase, inhibits vascular endothelial growth factor (VEGF)-induced in vitro and in vivo angiogenesis and this is abolished by BACE1 inhibition. Moreover, intravitreal administration of BACE1 inhibitor or BACE1 small interfering RNA (siRNA) increases choroidal neovascularization in mice. BACE1 induces ectodomain shedding of vascular endothelial growth factor receptor 1 (VEGFR1) which is a prerequisite for γ-secretase release of a 100 kDa intracellular domain. The increase in lipofuscin following BACE1 inhibition and RNAI knockdown is associated with lysosomal perturbations. Taken together, our data show that BACE1 plays a critical role in retinal homeostasis and that the use of BACE inhibitors for AD should be viewed with extreme caution as they could lead to retinal pathology and exacerbate conditions such as age-related macular degeneration.

Figure 1. Morphological analysis of BACE1 knockout mice and BACE expression in the normal retina

1. A. Representative haematoxylin/eosin staining of retinas from 4-month-old wild-type (WT) and BACE1^−/− mice. Knockout animals demonstrated shrunken retinal ganglion cells (RGCs), depopulation in the ganglion cell layer, thinning of the retina and inner nuclear layer (INL) and outer nuclear layer (ONL) degeneration. Eyes of wild-type mice (WT) were histologically normal. Transmission electron micrographs taken from the inner layer of the BACE1^−/− retina showed shrunken RGCs and neuronal cells with hyperchromatic and condensed nuclear chromatin, indicative of neuronal apoptosis (*), while cells from WT showed normal ultrastructure. Confocal micrographs of the RPE revealed an increase in lipofuscin granules in BACE1^−/− animals compared to WT. Electron micrographs of BACE1^−/− RPE/choroid appeared to contain increased lipofuscin and thinning in Bruch's membrane (BM) compared to WT.

2. B–E. Quantitative analysis showing differences in (B) retinal thickness, (C) apoptosis, (D) lipofuscin autofluorescence and (E) Bruch's membrane thickness in BACE1^−/−, BACE2^−/− and BACE1^−/− BACE2^−/− mice compared to WT (mean ± SEM, n = 5).

3. F. Representative haematoxylin/eosin staining of the RPE showing marked thinning of the RPE layer in BACE1^−/− mice compared to wild type.

4. G. Detection of BACE1 expression in the retina of wild-type mouse (4-month-old) and human eyes (39-year-old donor). Paraffin sections were immunostained using an antibody against BACE1 (red) and dual stained with agglutinin–FITC (green) to visualize the vasculature. BACE1 expression was observed in all layers of the retina, however, strongest staining was localized to the inner and outer plexiform layers and the retinal vasculature in both mouse and human. BACE1 was weakly localized in both RPE and the choroid.

5. H. mRNA levels of BACE1 and BACE2 in mouse brain, RPE choroid and retina from 4-month-old animals were analyzed by quantitative PCR. BACE expression was plotted using fold values with housekeeping gene GAPDH. The data was represented as mean ± SEM (n = 6).

Figure 2. Changes in the retinal microvasculature of BACE knockout mice and the effect of BACE1 inhibition on in vitro and in vivo angiogenesis

Animals were 4 months old and age-matched with WT.

1. BACE1^−/− retinal sections stained with agglutinin–FITC to visualize the retinal vasculature showed decreased retinal capillaries (arrows) compared to the normal organization and distribution of retinal capillaries observed in WT mice. Higher magnification images showing vascular changes and the presence of atypical, enlarged capillaries (lack of the DAPI-blue nuclear stain) in BACE1^−/− mice compared to WT. Electron micrograph of retinal vessels from WT mice revealed normal vessel filled lumens (L) with blood cells, encircled by basement membrane and adjacent pericytes while in vessels of BACE1^−/− mice there was loss of pericytes (*) and abnormal vascular endothelial cells.

2. Retinal flatmounts from animals that had been perfused with dextran-FITC confirming loss of vessels in both the superficial and deep vascular plexus in BACE1^−/− mice compared to wild type.

3. Detection of pericytes in the retina of wild-type and BACE1^−/− mice. Flat mount retinas were immunostained using an antibody against desmin (red) and dual stained with DAPI (blue). Desmin expression is greatly reduced in BACE1^−/− mice compared to wild type.

4. Quantitative analysis of the retinal vessel area per unit area of the retina revealed a significant decreases in knockout mice compared with WT (mean ± SEM, n = 6).

5. Quantitative analysis of the mean retinal area of retinal vessels revealed a significant increases in knockout mice compared with WT (mean ± SEM, n = 6).

6. Representative haematoxylin/eosin staining of the RPE and choroid showing a highly disorganized choroid in BACE2^−/− mice.

7. Quantification of length of tubule formation (mm/mm²) of microvascular endothelial cells in the Matrigel™. Cells were pretreated with VEGF (100 ng/ml) and/or PEDF (100 ng/ml), in the presence or the absence of 0.5, 1 or 5 µM β-secretase inhibitor IV (β-SI). Results are mean ± SEM of at least three independent experiments.

8. Quantitative assessment of the volume of the vascular lesions induced by laser-damage to eyes of mice receiving intravitreal injection of either PEDF, BACE1 inhibitor (βSI), a combination of the two, BACE1 siRNA, scrambled siRNA (scr) or saline vehicle control. Animals were sacrificed 14 days postlaser injury and RPE choroidal flat mounts were stained with a vascular specific marker, agglutinin-TRITC conjugate to visualize the CNV lesions and volume was determined by confocal microscopy (mean ± SEM, n = 6).

Figure 3. BACE1 is critical for ectodomain cleavage of VEGFR1

Porcine aortic endothelial cells (PAECs) were stably transfected with wild-type pVEGFR1-GFP wild-type and treated with VEGF (100 ng/ml), PEDF (100 ng/ml), PEDF + VEGF for 24 h, with or without BACE1 inhibitor (β-SI; 10 µM) or γ-secretase inhibitor DAPT (10 µM). Representative western blots from 4 independent experiments.

1. Diagram depicting the sequential stages of VEGFR1 cleavage and the fragments generated. CTF = C-terminal fragment, ICD = intracellular domain.

2. Western blot of the lysates for VEGFR1 demonstrated no major change in total VEGFR1 expression except in the presence of cycloheximide.

3. Lysates were immunoprecipitated (IP) with an antibody against the VEGFR1 N-terminal domain and immunoblotted for either the N-terminal domain of VEGFR1 or native sVEGFR1. PEDF and PEDF + VEGF treatments resulted in the appearance of a VEGFR1-NTF fragment which was blocked by BACE1 inhibition. sVEGFR1 was not affected by BACE1 inhibition. The effect of cycloheximide on sVEGFR1 levels in the culture medium confirmed that the sVEGFR1 is translationally regulated and therefore not derived from ectodomain shedding.

4. Immunoblot (IB) depicting the effect of BACE1 or γ-secretase on the generation of VEGFR1-CTF and VEGFR1-ICD.

5. Presence of full-length VEGFR1_GFP, VEGFR1_GFP-CTF and VEGFR1_GFP-ICD in membrane and cytosolic fractions determined by either using antibodies against the C-terminal of VEGFR1 or GFP.

Figure 4. BACE1 inhibition enhances accumulation of lipofusion, decreases activity of cathepsin D and elevates lysosomal pH in cultured RPE cells

1. Representative fluorescence micrographs and quantitation of autofluorescence by FACS (mean ± SEM, n = 6) in ARPE19 showed a dose-dependent increase in lipofusion granules in ARPE19 cells treated with BACE1 inhibitor for up to 14 days.

2. Representative confocal microscopy images showing intralysosomal pH changes following exposure of ARPE19 cells to different concentrations of BACE1 inhibitor. pH was monitored using lysosensor yellow/blue dextran where a shift from red to green indicates a decrease in pH. The pH values were calculated using the emission intensity ratio at 535/450 nm with excitation at 340 nm for the lysosensor demonstrated increased lysosomal pH upon BACE 1 inhibition. The data are the mean ± SEM of values from three independent experiments.

3. Representative fluorescence micrographs and quantitation of autofluorescence by FACS (mean ± SEM, n = 6) in ARPE19 showed a dose-dependent increase in lipofusion granules in ARPE19 cells treated with siRNA against BACE1 or BACE2 for up to 14 days. Vehicle only cells (Veh) and scrambled siRNA acted as the control.

4. Representative confocal microscopy images showing intralysosomal pH changes following exposure of ARPE19 cells treated with siRNA against BACE1 or BACE2. Untreated cells and scrambled siRNA acted as the control. pH was monitored using lysosensor yellow/blue dextran where a shift from red to green indicates a decrease in pH. The pH values were calculated using the emission intensity ratio at 535/450 nm with excitation at 340 nm for the lysosensor and demonstrated increased lysosomal pH upon BACE 1 inhibition. The data are the mean ± SEM of values from three independent experiments.