Tau accumulation in the retina promotes early neuronal dysfunction and precedes brain pathology in a mouse model of Alzheimer’s disease

The 3xTg mice bearing the human mutations in the genes encoding presenilin 1 (PS1M146V), amyloid precursor protein (APPSwe), and tau (MAPTP301L) [51], tau knockout mice (strain Mapt-tm1[EGFP]Klt/J), and age-matched littermate wild-type controls were purchased from Jackson Laboratories (Bar Harbor, ME) and maintained in our animal facility. Experiments were performed using 3 or 6 month-old female mice because they exhibit a more severe disease phenotype than their age-matched male counterparts [52]. All animal procedures were approved by the Centre de Recherche du Centre Hospitalier de l’Université de Montréal (CRCHUM) Animal Care Committee (approved protocol #N14024ADPs), and followed the guidelines of the Canadian Council on Animal Care.

Whole retinas, optic nerves or brains (frontal cortex and hippocampus) were isolated and homogenized in ice-cold lysis buffer: 50 mM Tris pH 7.4, 1 mM EDTA, 150 mM NaCl, 1% NP-40, 5 mM Na fluoride, 0.25% Na deoxycholate, and 2 mM NaVO₃ supplemented with protease and phosphatase inhibitors. Protein homogenates were centrifuged at 18,000 g for 5 min, and the supernatants were removed and resedimented to yield soluble extracts. Samples in Laemmli buffer were boiled for 5 min, resolved in 4–15% SDS polyacrylamide gradient gels (Bio-Rad Life Science, Mississauga, ON), and transferred onto nitrocellulose membranes (Bio-Rad Life Science). Blots were incubated in blocking solution (10 mM Tris pH 8.0, 150 mM NaCl, 0.1% Tween-20 and 5% milk) for 1 h at room temperature, followed by overnight incubation at 4 °C with each of the following primary antibodies: total tau (K9JA, 1 μg/ml, Dako North America, Carpinteria, CA), phospho-tau S396-S404 (PHF1, 1:100, gift of P. Davies, Albert Einstein College of Medicine, Bronx, NY), phospho-tau S199 (PS199, 1 μg/ml, Invitrogen, Burlington, ON), phospho-tau S202-T205 (AT8, 0.8 μg/ml, Thermo-Fisher Scientific, Waltham, MA), MC-1 (1:100, gift of P. Davies), ALZ-50 (1:100, gift of P. Davies), or β-actin (0.5 μg/ml, Sigma-Aldrich, Oakville, ON). Membranes were washed and incubated in peroxidase-linked donkey anti-mouse or donkey anti-rabbit antibodies (0.5 μg/ml, GE Healthcare, Mississauga, ON). Blots were developed with a chemiluminescence reagent (ECL, Amersham Biosciences, Arlington Heights, IL) and exposed to X-OMAT imaging film (Eastman Kodak, Rochester, NY). Densitometry was performed on scanned autoradiographic films using the ImageJ software (http://​imagej.​nih.​gov/​ij/​). Films were obtained from at least three independent western blots each carried out using retinal samples from different groups.

Animals were perfused with 4% paraformaldehyde and the eyes and optic nerves were rapidly dissected. Tissue was embedded in optimal cutting temperature compound (Tissue-Tek, Miles Laboratories, Elkhart, IN), and retinal (16 μm) or optic nerve (12 μm) cryosections were collected onto Superfrost Plus microscope slides (Thermo-Fisher Scientific). The following primary antibodies were added to retinal or optic nerve sections in blocking solution and incubated overnight at 4 °C as described [53]: total tau (K9JA, 2 μg/ml, Dako), tubulin isoform βIII (TUJ1, 2.5 μg/ml; Sigma-Aldrich), or neurofilament H (NF-H, 20 μg/ml, Sternberger Monoclonals Inc., Lutherville, MA). For whole-mounted retinas, tissue was permeabilized overnight at 4 °C in blocking solution, rinsed and incubated for 5 days at 4 °C in the following primary antibodies: total tau (K9JA, 2 μg/ml, Dako), RNA-binding protein with multiple splicing (RBPMS, 1:1000, PhosphoSolutions, Aurora, CO), or NF-H (20 μg/ml, Sternberger Monoclonals Inc.). Sections or whole retinas were washed and incubated with secondary donkey anti-rabbit or anti-mouse Alexa Fluor 594 and 488 (2 μg/ml, Life Technologies, Eugene, OR). Fluorescent labeling was observed using a Zeiss Axio Observer (Carl Zeiss, Canada) or a Leica SP5 confocal microscope (Leica Microsystems Inc., Concord, ON). All retinal and optic nerve images were acquired under identical conditions using the same illumination intensity, time exposure, and magnification, with careful attention to avoid signal saturation and/or bleaching. The areas sampled were selected using an unbiased stereological sampling method as described (http://​www.​stereology.​info).

Total RNA was isolated from individual retinas using the RNEasy Mini kit (Qiagen Inc., Valencia, CA). cDNAs were generated from 1 μg of total RNA using the QuantiTect Reverse Transcription Kit (Qiagen Inc.). Real-time PCR was performed using TaqMan probes and primers that target exon 5, expressed by all tau isoforms (pan-tau, catalog # Rn01495715), exon 4a specific to big tau (catalog # Rn01495711), or β-actin RNA as control (catalog # 4331182) (Applied Biosystems, Waltham, MA). Amplification was performed using the 7900HT Fast Real-Time PCR System (Applied Biosystems) with the following cycle conditions: 95 °C for 15 s, 60 °C for 1 min, 72 °C for 1 min. Reactions were run in triplicates for each sample and the 2^-ΔΔCt formula was used for the calculation of differential gene expression.

Anterograde axonal transport was assessed by injection of cholera toxin β subunit (CTβ) conjugated to Alexa Fluor 488 (Molecular Probes, Life Technologies, Eugene OR) as described previously [54, 55]. CTβ is a reliable marker of active transport and has been consistently used to assess RGC anterograde transport to the superior colliculus [56, 57]. CTβ (1% diluted in sterile PBS, total volume: 1 μl) was injected intravitreally using a custom-made sharpened microneedle generated from a borosilicate glass capillary tube (5 μl, World Precision Instruments, Sarasota, FL) as previously described by us [58]. Briefly, the glass capillary was pulled using a two stage needle puller (PC-10, Narishige International, Amityville, NY) to produce thin microneedles of 6 cm in length. Under a dissecting microscope, a sharp blade was used to carefully create an opening at the tip of the microneedle. The resulting opening had an elliptical shape with a major and minor axis diameter of approximately 190 μm and 70 μm, respectively. The tip of the microneedle was then beveled using a micropipette bevelling system until the tip was sharp and the edges were flat and smooth [58]. For intravitreal injections, animals were anesthetized with isoflurane (2%, 0.8 L/min). The upper eyelid was gently retracted and the tip of the glass microneedle positioned at a 45° angle relative to the ocular surface. A light pressure was exerted to insert the microneedle through the conjunctiva, sclera and retina into the vitreous cavity. The tracer was injected and the needle was retracted slowly to avoid reflux. This route of administration avoided injury to ocular structures, including the iris and the lens. A small drop of antibiotic was applied topically after the surgery (Vigamox, 0.5%, Alcon Canada Inc., Mississauga, ON), and there were no signs of post-operative infection or inflammation. Animals were perfused transcardially with 4% paraformaldehyde three days after CTβ administration; brains were removed and embedded in optimal cutting temperature compound (Tissue-Tek, Miles Laboratories). Serial coronal cryosections of the entire superior colliculus from each animal were obtained using a cryostat (50-μm thickness). Seven sections per superior colliculus, from rostral to caudal, were selected using an unbiased stereological sampling method. Sections were photographed digitally using the Zeiss Axio Observer fluorescent microscope with Apotome (Carl Zeiss) and the area of the CTβ signal in each section was measured using the Imaris MeasurementPro module (Bitplane, South Windsor, CT). The total CTβ signal in each superior colliculus was calculated using the formula: total CTβ area = ΣCTβ section area/ssf x asf x tsf [59]. The section sampling fraction (ssf) was the number of sections analyzed over the total number of sections obtained from each superior colliculus (7/36), the area sampling fraction (asf) was the area sampled divided by the total area (1), and the thickness sampling fraction (tsf) was the section thickness sampled divided by the total section thickness (1). This analysis provided a representative value of CTβ-positive area which was then multiplied by the width of the entire superior colliculus to yield the total CTβ volume. To confirm CTβ uptake by RGCs, whole retinas were incubated overnight at 4 °C with goat IgG against the RGC-specific marker brain-specific homeobox/POU domain protein 3a (Brn3a, 0.27 μg/ml, Santa Cruz Biotechnology, Santa Cruz, CA) followed by secondary Alexa Fluor 594 anti-goat IgG (1 μg/ml; Jackson ImmunoResearch Laboratories, West Grove, PA). Retinas were rinsed, mounted, and the total number of CTβ-positive and Brn3a-positive neurons was quantified by independent random stereological sampling.

The following siRNA sequences against Tau (siTau) were purchased: i) 5′-CCUAGAAAUUCC AUGACGA-3′, ii) 5′-GGACAGGAAAUGACGAGAA-3′ iii) 5′-GCUGAUAGGCAGUUUAC AA-3′, iv) 5′-UGGGUGGGCUAGAUAGAUA-3′ (ON-TARGETplus mouse Mapt siRNA-SMART pool, GE Dharmacon, Lafayette, CO). The following control siRNA (siCtl) sequence was used: 5′-UGGUUUACAUGUUGUGUGA-3′ (ON-TARGETplus, non-targeting siRNA #2, GE Dharmacon). Each siRNA was injected into the vitreous chamber using a custom-made sharpened glass microneedle as described above (3.8 μg/μl, total volume: 2 μl).

Mice were subjected to transcardial perfusion with 4% paraformaldehyde and retinas were dissected out and fixed for an additional 15 min. Free-floating retinas were blocked overnight at 4 °C in 10% normal goat serum, 2% bovine serum albumin, 0.5% Triton X-100 in PBS, and incubated with the RGC-specific marker RBPMS (1:1000, PhosphoSolutions) for 5 days at 4 °C. Retinas were then incubated with Alexa 594-coupled secondary antibody (2 μg/ml, Life Technologies) for 4 h at room temperature, washed, mounted with the nerve fiber layer side up, and visualized with a Zeiss Axio Observer (Carl Zeiss). RBPMS-labeled RGCs were counted within three square areas at distances of 0.25, 0.625 and 1 mm from the optic disc in each of the four retinal quadrants for a total of twelve retinal areas as described by us [4].

Visual deficits and pathological changes have been described in the retinas of AD patients [21, 31]. Therefore, we first asked whether the level of endogenous retinal tau was altered in 3xTg mice and, if so, whether it correlated with tau changes in the brain. For this purpose, retinal and brain protein samples from 3- and 6-month-old 3xTg mice were analyzed and compared to those from age-matched wild-type controls. These time points were selected because they precede the appearance of reported behavioral and cognitive defects in this model [50, 60]. Western blots of soluble retinal homogenates using an antibody against total tau (K9JA), which binds the microtubule binding domain of the protein irrespective of its phosphorylation state [61, 62], revealed the presence of four predominant tau isoforms of 37-kDa, 50-kDa, 55-kDa and 100-kDa (Fig. 1a). The 100-kDa band most likely corresponded to big tau, a high molecular weight tau isoform detected only in retinal and peripheral neurons [63‐65]. Densitometry analysis showed that all four tau isoforms increased in 3-month-old 3xTg mouse retinas relative to age-matched wild-type controls (Fig. 1c). Analysis of cortical and hippocampal homogenates revealed the presence of four tau isoforms of 37-kDa 50-kDa, 55-kDa, and 60-kDa (Fig. 1b). In contrast to retina, however, the levels of all brain tau isoforms in 3-month-old 3xTg mice were similar to those in age-matched controls (Fig. 1d). No signal was detected in retinal or brain samples from tau null mice thus confirming the specificity of the K9JA antibody in these tissues (Fig. 1a, b). Western blot and densitometry analyses revealed an increase of the 55-kDa tau isoform in both retina and brain samples from of 6-month-old 3xTg mice, while all other isoforms remained unchanged (Fig. 1e-h). These results demonstrate that tau protein increases in transgenic retinas early in the disease process and precedes tau accumulation in the brain.

Alterations in tau phosphorylation contribute to tau dysfunction in pathological conditions [14, 66, 67]. Therefore, we analyzed tau phosphorylation in retinal samples from 3xTg mice using the phospho-tau specific antibodies AT8 (S202, T205) [68, 69], PS199 (S199) [70], and PHF1 (S396, S404) [71, 72]. These antibodies were selected because they recognize phospho-specific epitopes that correlate with AD severity [73]. Visualization of western blots probed with AT8 or PHF1 revealed an increase in tau phosphorylation at these epitopes in 3-month-old 3xTg retinas (Fig. 2a, b). Densitometry analysis of phospho-tau signals with respect to total tau, which markedly increased at this age, confirmed greater tau phosphorylation on residues S202 and T205 of all isoforms (AT8, Fig. 2a, d), while a relative decrease at residue S199 was observed (PS199: 55-kDa, 100-kDa) (Fig. 2b, e). Conversely, 6-month-old 3xTg retinas displayed reduced phosphorylation with AT8 on all isoforms (Fig. 2g, j), and increased phosphorylation with PS199 (50-kDa, 55-kDa, 100-kDa) (Fig. 2h, k). No significant changes were detected with PHF1 in 3- or 6-month-old 3xTg retinas relative to controls (Fig. 2c–l).

Phosphorylation can lead to changes in the conformation of tau protein during the course of AD, which can play a key role in pathological tau accumulation and cleavage [74, 75]. The conformation of tau in 3xTg retinas was investigated using the antibodies MC-1 and ALZ-50, which recognize the early folding back of the N-terminus on the microtubule domain in a hairclip configuration linked with tau aggregation [74, 76, 77]. No changes in conformation-dependent epitopes recognized by MC-1 or ALZ-50 were detected in 3- or 6-month-old 3xTg retinas relative to age-matched controls (Fig. 3). Together, these data demonstrate increased tau phosphorylation at AT8 epitopes in the early presymptomatic stages, whereas increased phosphorylation at PS199 appears at a later phase of the disease, in the absence of conformational changes. We conclude that retinal tau undergoes complex age-related and epitope-specific changes in AD.

The cellular distribution of tau in the retina was investigated by immunohistochemistry using the K9JA antibody against total tau. In agreement with previous reports [4, 78], a low basal level of tau expression was found in all retinal layers except the outer nuclear layer (Fig. 4a, c). Consistent with our biochemical findings, retinal tau increased in 3-and 6-month-old 3xTg mice and its accumulation was more pronounced in the inner plexiform layer (IPL), where RGC dendrites are located (Fig. 4b, d). Labeling with an antibody against tubulin isoform βIII (TUJ1), an RGC-specific marker that strongly labels the soma and dendrites of these neurons [79, 80], confirmed localization of tau within the somatodendritic compartment (Fig. 4e-g). Further analysis of tau distribution on flat-mounted retinas using RBPMS, which selectively labels RGC soma [81, 82], confirmed tau accumulation in RGC bodies in 3xTg retinas relative to controls (Fig. 4h-m). A similar distribution of tau was observed in 6-month-old transgenic mice, with more robust tau build up at this age (Fig. 4n-s).

Further analysis of tau expression in flat-mounted retinas, specifically changes in the intraretinal RGC axons, was investigated by co-localization of tau with the axonal marker neurofilament-H (NF-H). In wild-type retinas, low basal levels of tau protein were detected in RGC intraretinal axons, visualized with NF-H (Fig. 5a-c, g-i). In contrast, a pronounced increase of tau signal in NF-H-positive axons was detected in 3xTg retinas at both 3 and 6 months of age (Fig. 5d-f, j-l). To establish whether tau accumulation in the retina resulted from increased gene expression, real-time qPCR was performed using primers that recognized all tau isoforms (pan-tau) or big tau [63‐65]. No significant changes in tau mRNA levels were detected in AD retinas compared to controls (Fig. 5m). These data indicate that tau accrues early in the AD retina, predominantly in RGC dendrites and intraretinal axons, and that this accumulation is not the result of increased gene expression.

In normal physiological conditions, tau is enriched in axons with low levels found in dendrites and soma [83]. In AD and other tauopathies, tau detaches from axonal microtubules and accumulates in the somatodendritic compartment of affected neurons [84]. To assess whether the distribution of tau in RGC axons within the optic nerve was altered in 3xTg mice, we carried out confocal imaging of nerve cross sections colabeled with tau and the axonal marker NF-H. In control optic nerves, tau was enriched in RGC axons visualized with NF-H (Fig. 6a-c, m- o). In contrast, 3xTg optic nerves from both 3- and 6-month-old mice displayed a striking reduction in axonal tau protein (Fig. 6g-i, s-u). High magnification confocal images of individual optic nerve fascicles revealed that many NF-H positive axons were still detected in transgenic animals in spite of marked reduction of tau protein levels (Fig. 6j-l, v-x), indicating that a decrease in tau was not the result of axonal degeneration. Western blot analysis of optic nerve homogenates confirmed a visible reduction of tau protein in 3- and 6-month-old 3xTg mice (Fig. 6y, Y’, z, Z’). Taken together, our results demonstrate that tau is markedly reduced in RGC axons within the optic nerve early in the course of the disease.

RGCs, like other long projecting neurons, rely heavily on axonal transport for proper function. Anterograde transport impairment is recognized as an early sign of RGC damage and dysfunction [55]. Therefore, we examined whether the ability of RGCs to transport the anterograde tracer CTβ to terminals in the superior colliculus, the primary target of RGCs in the rodent brain [85‐87], was altered in 3xTg mice. CTβ is an excellent reagent to study axonal transport because of its high sensitivity, ability to effectively move anterogradely or retrogradely from the injection site, dependency on intact microtubules thus serving as readout of active transport, capacity to label the entire neuron including extremely fine terminals, restriction from labeling fibers of passage, and demonstrated efficacy to label axonal tracts in the visual system following ocular administration [54‐56, 88‐96]. Alexa Fluor 488-conjugated CTβ was injected intravitreally and its accumulation in the contralateral superior colliculus was quantified using unbiased stereological sampling [54]. Mutant APP, PS1 and tau proteins are expressed throughout development in 3xTg mice, thus we first aimed to establish whether there were developmental alterations in axonal transport in young mice (21 days), a week after eye opening. Our data demonstrate that there was no difference in the amount of brain CTβ between wild-type and 3xTg at 21 days of age (Fig. 7a, b). In contrast, a substantial reduction in the CTβ-labeled volume was observed in 3- and 6-month-old 3xTg mice relative to age-matched controls (Fig. 7c-f). Quantification of total CTβ volume confirmed a 57% reduction in the superior colliculi of 3xTg mice suggesting major deficits in the ability of RGCs to transport cargos to their targets (Fig. 7g). These findings indicate that deficits in anterograde axonal transport in 3xTg mice are not of developmental origin, but rather reflect pathological changes detected early in the course of the disease.

To rule out that anterograde transport deficits were caused by the inability of 3xTg RGCs to uptake CTβ, we examined CTβ-injected retinas with Brn3a, a selective marker of RGC nuclei [97]. Abundant cytoplasmic CTβ within RGCs was observed at 24 h after tracer injection in both wild-type and transgenic retinas (Fig. 7h, i). Quantification of Brn3a-positive RGCs containing CTβ revealed a similar number of neurons in both 3xTg and control mice (Fig. 7j), thus confirming effective tracer uptake. To determine whether anterograde transport deficits reflected neuronal death, RGC density was quantified in 3- and 6-month-old transgenic retinas using the cell-specific marker RBPMS. A similar RGC density was found in 3xTg and wild-type mice at 3 months of age, confirming the absence of significant cell death at a time when major transport deficits are already apparent (Fig. 7k–o). In 6-month-old transgenic retinas, only a modest reduction in RGC density was detected relative to controls (~15%, Fig. 7m–o), therefore the substantial axonal transport loss at this age cannot be ascribed solely to retinal neuron death. Collectively, these results demonstrate that major deficits in axonal transport along RGC axons are a relatively early feature of neuronal dysfunction in AD pathology that precedes cell death.

To investigate whether tau accumulation underlies the axonal transport deficits in 3xTg RGCs, we sought to decrease tau protein levels using siRNA followed by analysis of CTβ transport. First, we assessed the ability of a targeted siRNA against tau (siTau) to reduce retinal tau protein levels. We previously demonstrated that siRNA delivered by intravitreal injection is rapidly taken up by RGCs [98]. Western blot analysis of retinal extracts from 3-month-old 3xTg eyes that received siTau showed a significant reduction in tau protein relative to age-matched transgenic mice that received a control non-targeting siRNA (siCtl) (Fig. 8a). Quantitative analysis confirmed that siTau induced a 27%, 42% and 50% reduction of the 50-kDa, 55-kDa and 100-kDa tau isoforms, respectively, relative to control siRNA-treated eyes (Fig. 8b). Next, we investigated whether siRNA-mediated tau knockdown improved RGC axonal transport. For this purpose, siTau was injected intraocularly once a week for a total of three weeks. The multiple injection regimen was selected based on our previous findings that siRNA-mediated protein knockdown in the retina is transient [98]. Four days after the last siTau injection, CTβ was administered in the eye and the amount of the tracer in the superior colliculus quantified three days later. Visualization of caudal-to-rostral sections from the superior colliculus of siTau-treated 3xTg mice showed a marked increase in CTβ relative to transgenic mice that received control siCtl (Fig. 8c, d). Quantitative analysis confirmed that tau knockdown promoted a significant increase in anterograde axonal transport compared to control animals (20%, Fig. 8e). Our results demonstrate that attenuation of retinal tau levels improves axonal transport, suggesting that early tau accumulation in the retina impairs axonal function in AD.