GSK3β-mediated tau hyperphosphorylation triggers diabetic retinal neurodegeneration by disrupting synaptic and mitochondrial functions

Abstract

Background: Although diabetic retinopathy (DR) has long been considered as a microvascular disorder, mounting evidence suggests that diabetic retinal neurodegeneration, in particular synaptic loss and dysfunction of retinal ganglion cells (RGCs) may precede retinal microvascular changes. Key molecules involved in this process remain poorly defined. The microtubule-associated protein tau is a critical mediator of neurotoxicity in Alzheimer's disease (AD) and other neurodegenerative diseases. However, the effect of tau, if any, in the context of diabetes-induced retinal neurodegeneration has yet to be ascertained. Here, we investigate the changes and putative roles of endogeneous tau in diabetic retinal neurodegeneration.

Methods: To this aim, we combine clinically used electrophysiological techniques, i.e. pattern electroretinogram and visual evoked potential, and molecular analyses in a well characterized high-fat diet (HFD)-induced mouse diabetes model in vivo and primary retinal ganglion cells (RGCs) in vitro.

Results: We demonstrate for the first time that tau hyperphosphorylation via GSK3β activation causes vision deficits and synapse loss of RGCs in HFD-induced DR, which precedes retinal microvasculopathy and RGCs apoptosis. Moreover, intravitreal administration of an siRNA targeting to tau or a specific inhibitor of GSK3β reverses synapse loss and restores visual function of RGCs by attenuating tau hyperphosphorylation within a certain time frame of DR. The cellular mechanisms by which hyperphosphorylated tau induces synapse loss of RGCs upon glucolipotoxicity include i) destabilizing microtubule tracks and impairing microtubule-dependent synaptic targeting of cargoes such as mRNA and mitochondria; ii) disrupting synaptic energy production through mitochondria in a GSK3β-dependent manner.

Conclusions: Our study proposes mild retinal tauopathy as a new pathophysiological model for DR and tau as a novel therapeutic target to counter diabetic RGCs neurodegeneration occurring before retinal vasculature abnormalities.

Keywords: Diabetic retinopathy; GSK3β; Hyperphosphorylated tau; Retinal ganglion cells; Retinal neurodegeneration; Synaptic and mitochondrial dysfunction.

Fig. 1

Decreased RGCs activity in the absence of retinal microvasculopathy is associated with synaptic and axonal impairment occurring before RGCs apoptosis in HFD-induced diabetes. (a) Representative pattern electroretinography (PERG) waveforms of eyes in mice fed with regular chow (RD) or HFD for 20 and 24 weeks. Difference between peak-to-peak amplitude of P50 and N95 components (P50-N95) and the N95 peak latency were quantified. (b) Representative images of fundus fluorescein angiography. (c) Illustrative examples of retinal Evans Blue angiography. Right panels are high-power magnification of the areas indicated by the boxes. (d) Representative images of retinal immunostaining for apoptotic (TUNEL positive, green; indicated by arrow) cells. Nuclei were labeled with DAPI (blue). Scale bar, 100 μm. (e) Apoptotic RGCs were quantified and expressed as the percentage of TUNEL-positive cells to DAPI-positive cells in GCL. For each retinal section, the number of TUNEL positive cells in the GCL was counted. For each eye, results obtained from four independent sections were averaged. (f) Retinal immunofluoresence staining for synaptophysin (green, synaptophysin; blue, DAPI; scale bar, 100 μm). (g) Representative images for Golgi staining of RGCs axons at the proximal portions of optic nerves (black, indicated by arrow; scale bar, 10 μm) in longitudinal cryosections of optic nerves. Data are means ± SEM. n = 6 animals per group. ^**P < 0.01 vs age-match RD controls. GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer

Fig. 2

HFD promotes hyperphosphorylation of tau in neural retina and optic nerves. (a) Representative images of retinal double immunostaining for phospho-tau (pS396- or pT231-Tau; green) with total-tau (Tau-5; red), and non-phospho-tau (Tau-1; green) with Tau-5 (red) from mice fed with RD or HFD for 16, 20 and 24 weeks, respectively. Nuclei were counterstained with DAPI. Scale bar, 100 μm. (b) Representative images of double immunostaining for phospho-tau (pS396- or pT231-Tau; green) with total-tau (Tau-5; red) of RGCs axons at the proximal portions of optic nerves in longitudinal sections of optic nerves from mice fed with RD or HFD for 16 weeks. Scale bar, 20 μm. Higher magnifications (scale bar, 5 μm) are shown in the boxed lower panels, respectively. (c) Western blot analyses of pS396-Tau, pS404-Tau, pT231-Tau, pT205-Tau, Tau-1 and Tau-5 in total retina lysate. Intensities were quantified and normalized against the level of GAPDH and expressed as fold changes of protein abundance in the retina from HFD groups relative to their age-matched controls. Data are means ± SEM. n = 6 (a, b) or n = 4 mice (b) per group. ^*P < 0.05 and ^**P < 0.01 vs age-match RD controls. NS, no significant difference. OPL, outer plexiform layer

Fig. 3

Intravitreal delivery of si-Tau restores HFD-induced RGCs dysfunction and synapse loss. si-Tau was injected in the vitreous of the right (R) eye of mice at 20 weeks after RD (RD-R-si-Tau) or HFD (HFD-R-si-Tau), while a scramble si-sc was injected in the contralateral (left; L) eye as a control (RD-L-si-sc; HFD-L-si-sc). (a) Representative waveforms of visual evoked potential (VEP). The differences in peak amplitude (N1-P1) were quantified. (b) Representative images of double immunostaining for phospho-tau (pS396- or pT231-Tau; green) with total-tau (Tau-5; red), and for synaptophysin (green) with Tau 5 (red). Scale bar, 100 μm. (c) Representative curvilineal profile of protein immunostaining intensity from GCL to OPL across retinal depth of the images shown in B. Data are means ± SEM. n = 5 eyes per group. For each eye, data from three independent curvilineal diagrams were averaged, and the mean of five eyes was used as the representative value for each group. ^**P < 0.01 vs contralateral eye injected with si-sc. NS, no significant difference

Fig. 4

Reduced tau phosphorylation by intravitreal injection of a GSK3β inhibitor protects RGCs from HFD-induced vision and synapse loss. (a) Western blotting for IRS-1, phosphorylated-Akt (S473), total Akt, phosphorylated-GSK3β (Ser9), total GSK3β in total retina lysate. Intensities were quantified and normalized against the level of GAPDH or total proteins (Akt or GSK3β) and expressed as percentage of protein abundance in the retina from HFD groups relative to their age-matched controls. Data are means ± SEM. n = 4 mice per group. ^*P < 0.05 and ^**P < 0.01 vs age-match RD controls. (b) A GSK3β specific inhibitor TWS119 or vehicle was injected intravitreally into the right eye (HFD-R-TWS119) or left eye (HFD-L-Veh), respectively, in mice fed with HFD for 20 weeks. Representative VEP waveforms and quantification of differences in peak amplitude (N1-P1) are shown. (c) Representative double immunostaining for phopho-tau (pS396- or pT205-Tau; green) with Tau-5 (Red), and for synaptophysin (green) with pT231-Tau (red). Scale bar, 100 μm. (d) Representative curvilineal profile of protein immunostaining intensity from GCL to OPL across the image shown in c. Data are means ± SEM. n = 5 eyes (b-d) per group. ^*P < 0.05 vs contralateral eye injected with vehicle

Fig. 5

Reduced tau microtubule binding and microtubule stability are associated with synapse loss in primary RGCs upon glucolipotoxicity in a GSK3β-dependent manner. (a) Representative images of triple immunostaining for RGC-characteristic marker Thy1 (red), neuronal markers TUJ1 (green) and Map2 (blue). Scale bar, 100 μm. Primary RGCs were then exposed to conditioned medium (HG + PA) for 24 h, in the absence or presence of TWS119. (b) Representative images of subcellular expression of pT231-Tau (green) and Thy1 (red) by double immunofluorescence. Scale bar, 20 μm. (c) Representative synaptophysin (green; scale bar, 20 μm) immunostaining in RGCs. Areas boxed in are shown at higher magnification in the lower panels. (d) Western blotting for synaptophysin from whole cell lysates (Total) or synaptosome fractions (Syn). Intensities were quantified and normalized against the level of GAPDH and expressed as percentages of protein abundance under stimulation relative to control. (e) mRNAs of synaptophysin in total lysates or synaptosomes were quantified by Q-PCR. (f) Microtubule sedimentation assay. Western blotting for Tau 5 and β-tubulin in the supernatant (SN) and the microtubule pellet (pellet). Relative intensities of each protein in its respective fraction were quantified and normalized against the sum of the intensity value of that protein (total, including both supernatant and pellet fractions). MT, microtubule. (g) Western blotting for Ac-tubulin in whole cell lysates. Intensities were normalized against the level of GAPDH. (h) Representative images of double immunofluorescence for Ac-tubulin (green) and Tyr-tubulin (red). Scale bar, 40 μm. Data are means ± SEM of three independent experiments. ^*P < 0.05 and ^**P < 0.01 vs control; ^#P < 0.05 and ^##P < 0.01 vs HG + PA

Fig. 6

Mitochondrial transport and function are impaired at glucolipotoxicity-stressed RGCs in a GSK3β-dependent manner. (a) Mitochondrial axonal trafficking was determined by infecting RGCs with an adenovirus vector carrying a foreign gene for mitochondrial complex IV (Ad-GFP-Mito). Representative fluorescence images for GFP-labeled mitochondria within axons (upper panels) and kymograph images of axonal mitochondrial movement (middle and bottom panels) are shown. Traces of moving mitochondria are indicated with white arrow. The average transport speed of movable mitochondria was calculated and expressed as mito velocity (μm/min). (b) Activity of mitochondrial complex I and complex IV was measured by spectrophotometry and expressed as nmol/min/mg protein. (c) Representative images of double immunofluoresence for Tau 5 (red) and complex I (green). Scale bar, 40 μm. (d) Western blotting for complex I and complex IV from whole cell lysates (Total) or synaptosome fractions (Syn). Intensities were quantified and normalized against the level of GAPDH and expressed as percentage of protein abundance under stimulation relative to control. Data are means ± SEM of three (a, d) or four (b) independent experiments. ^*P < 0.05 and ^**P < 0.01 vs control; ^#P < 0.05 and ^##P < 0.01 vs HG + PA. NS, no significant difference

Fig. 7

Knock-down of tau rescues mitochondrial abnormalities and presynaptic loss in glucolipotoxicity-stressed RGCs. RGCs were transfected with a Cy5-labeled siRNA (si-Tau or si-sc) and treated with HG + PA. (a) Representative immunocytochemistry for Ac-tubulin (green, Ac-tubulin; red, Cy5; scale bar, 40 μm). (b) Western blotting for Ac-tubulin and Tau 5. Intensities were normalized against the level of GAPDH and expressed as fold changes of protein abundance with si-Tau relative to si-sc control. (c) Mitochondrial transport in siRNA-transfected RGCs was performed by infecting cultured RGCs with Ad-GFP-Mito and treated with HG + PA. Representative fluorescence images (upper panels; scale bar, 10 μm) for GFP-labeled mitochondria (GFP, green) within axons in siRNA-transfected cells (Cy5, red) and kymograph images of axonal mitochondrial movement (middle and bottom panels) are shown. Traces of moving mitochondria are indicated with white arrow. The average transport speed of movable mitochondria was calculated. (d) Activity of mitochondrial complex I and complex IV. (e) Western blotting for complex I, complex IV and synaptophysin from whole cell lysates (Total) or synaptosome fractions (Syn). Intensities were normalized against the level of GAPDH and expressed as fold changes of protein abundance with si-Tau relative to si-sc control. (f) Representative fluorescence images for synaptophysin (green) in siRNA-transfected cells (Cy5, red). Areas boxed in for synaptophysin immunostaining are shown at higher magnification. Scale bar, 40 μm. Data are means ± SEM of three independent experiments. ^*P < 0.05 and ^**P < 0.01 vs control si-sc