Do the retina and brain share similar molecular and cellular pathological footprints? Insights from an Alzheimer´s disease animal model

Background: The concept 'the retina as a window to the brain' in Alzheimer´s disease (AD) has been explored in recent years since patients sometimes present visual alterations before the first symptoms of dementia. The retina is an extension of the brain and can be assessed by non-invasive methods. However, assessing the retina for AD diagnosis is still a matter of debate. Using the triple transgenic mouse model of AD (3xTg-AD), this study was undertaken to investigate whether the retina and brain undergo similar molecular and cellular changes during the early stages of AD pathology, and if the retina could anticipate the pathological alterations occurring in the brain. Methods: We used the 3xTg-AD and wild-type mice (C57BL6/129S), at 4 and 8 months of age, and assessed several parameters in the retina and brain (hippocampus and cortex): amyloid-beta (Aβ) and hyperphosphorylated tau (p-tau) levels, barrier permeability, cell death, neurotransmitter levels and glial changes. Results: We detected increased Aβ levels in the hippocampus and cortex and increased p-tau in the hippocampus, retina and cortex of 3xTg-AD mice. The brain and retinal barriers were unaffected. At 4 months, the content of some synaptic proteins increased in the brain but not in the retina. No cell death, including retinal ganglion cells loss, was detected in 3xTg-AD mice. Overall, no changes were observed in glutamate and GABA levels in all regions. There was an increase in astrogliosis in the hippocampus at 4 months and a decrease in the retina at 8 months. No changes were detected in Müller cells reactivity. Furthermore, we did not find changes in the number of microglia in 3xTg-AD mice, but we detected a different profile in microglia branching in the hippocampus and retina, at 4 months, where the number and length of the processes increased in the hippocampus and decreased in the retina. Conclusions: At the early stages of pathology, the retina, hippocampus and cortex of 3xTg-4 time

AD are not significantly affected, but already present some molecular and cellular alterations. The retina did not mirror the changes detected in the brain in this animal model of familial AD, and these observations should be taking into account when using the retina as a potential diagnostic tool for AD.

Background
The retina and the brain have the same embryonic origin. The retina is part of the central nervous system (CNS) and is considered an extension of the brain being connected to the brain by the optic nerve. The concept of 'the retina as a window to the brain' has been increasingly explored in several brain neurodegenerative diseases, including Alzheimer's disease (AD). The potential use of the retina as a diagnostic tool has some advantages.
We can assess the retina by non-invasive and inexpensive methods. Moreover, the retinal layers present an unique organization that easily enables the detection of structural changes (1). Visual problems have been detected in some AD patients prior to the appearance of the first signs of dementia (2). However, the precise molecular, cellular and functional changes in the retina of AD patients are still unclear and not completely elucidated yet (reviewed in (3)). The retinas of AD patients present increased glial reactivity (4-6) and cell death (7-10) (post-mortem analysis). However, the presence of AD hallmarks in the retina of AD patients is controversial, with some claiming their presence and others demonstrating their absence (11)(12)(13)(14). Concordantly, in AD animal models there are also controversial findings and missing gaps that need to be filled in, such as those related with glial reactivity, cell death and blood brain barrier (BBB) and blood retinal barrier (BRB) permeability. There is a lack of studies assessing simultaneously the brain and retina at the early stages of the disease, in order to understand whether the pathology impacts the brain and retina with the same extension and also to clarify which region is firstly affected.
We recently described structural and functional changes in the brain and retina of the triple transgenic mouse model of AD (3xTg-AD) (15,16), a well characterized model that develops both senile plaques and neurofibrillary tangles in a progressive and agedependent manner (17). We reported that 3xTg-AD mice, as early as 4 and 8 months of age, already present impaired locomotion activity and recognition memory as well as a decrease of hippocampal volume and taurine levels (15). In the retina, at these early time points, 3xTg-AD mice present retinal thinning accompanied by an atrophy of the visual cortex (16). Thus, in the present study we aimed to disentangle the molecular and cellular changes that might occur at the early stages of the pathology, which may underlie the structural and functional changes described previously, both in the retina and brain of the 3xTg-AD mouse model (16), as well as to evaluate whether the molecular and cellular changes are similar in the retina and brain and if the retina can anticipate the changes observed in the AD brain. To achieve our goals, we assessed several molecular and cellular parameters, simultaneously in the retina and brain of 3xTg-AD mice, at the early stages of the pathology (4 and 8 months of age).

Animals
Experiments were performed in 3xTg-AD mice harboring three human mutated genes (PS1M146V, APPSWE and TauP301L) (17) and in age-matched Wild Type (WT) animals (with C57BL6/129S genetic background), at 4 and 8 months of age. All animals were maintained under controlled light (12 h day/night cycle), relative humidity (68%), and temperature (21.8 ± 2 °C), with access to food and water ad libitum. All experiments using animals were approved by the Animal Welfare (Órgão Responsável pelo Bem-Estar Animal -ORBEA ii) Specific homogenization protocol for Aβ detection in the retina For Aβ detection, the retinas were lysed as previously described (18). Briefly, retinas were lysed in tissue homogenization buffer (THB: 2 mM Tris, 250 mM sucrose, 0.5 mM EDTA, 0.5 mM Ethylene Glycol Tetraacetic Acid (EGTA)) supplemented with cOmplete™ Mini and sonicated with 10 pulses of 1 s each. Then, 0.4% diethylamine (DEA) solution (0.4% DEA in 100 mM NaCl) was added to the homogenate in the proportion 1:1. Samples were centrifuged at 128 000 x g during 1 h, at 4 °C, and the supernatant was collected and neutralized with 0.5 Tris-HCl (pH 6.8). After protein quantification (BCA assay), samples were denatured by adding 6x concentrated sample buffer and heating for 5 min, at 95 °C.

ii) Retina cryosections
Retina cryosections were fixed in ice-cold acetone at -20 °C for 10 min, followed by rehydration in PBS. The slices were permeabilized in 0.25% Triton X-100 in PBS for 30 min, and blocked in 10% normal goat serum and 1% BSA in PBS, for 30 min at RT. The slices were then incubated overnight at 4 °C with the primary antibodies ( Table 2) diluted in 1% BSA in PBS. After washing, the slices were incubated with the secondary antibodies (Table 2) and DAPI (1:5000), for 1 h at RT, and then mounted with mounting medium and kept at 4 °C until being visualized. iii) Retina whole mounts Retina whole mounts were blocked in 0.1% Triton X-100, 3% BSA and 10% goat serum in PBS for 1 h at RT, and then incubated with the primary antibody (anti-Iba-1; 1:500; (GABA) and glutamate were measured by separation in a reverse-phase high-performance liquid chromatography (HPLC). Briefly, amino acids were separated by a reverse-phase GilsonASTED HPLC system, composed of a Spherisorb ODS column (particle size: 5 lm; excitation and emission wavelengths of 340 nm and 410 nm, respectively. A linear gradient elution program was applied for amino acid elution: eluent A (30 mmol/l sodium acetate buffer; pH 6.8) from 100-50%, and eluent B (methanol) from 0-50%, with a flow rate of 2.5 ml/min. The integration of the amino acid peak area and further calculations were carried out by the Gilson system software. The amino acid quantification was carried out using standard amino acids solutions under the same conditions, and normalized for protein levels.

3-D Neurolucida Software
Tridimensional morphometric analysis of microglial cells in retina whole mounts and brain slices was performed by manual reconstruction using the Neurolucida software (MBF Bioscience, Williston, VT, USA), as previously described (21)

Results
Since the retina and brain share the same embryonic origin, the concept 'The retina as a window to the brain' in neurodegenerative diseases has been explored. Here, we investigated which molecular and cellular changes can be detected in the retina and brain of the 3xTg-AD mouse model. Moreover, we assessed if those changes show a similar profile in the retina and brain and if some changes are firstly detected in the retina. Since we are particularly interested in identifying the initial changes that might occur in the retina and brain, we focused only in the early stages of the pathology.
Assessment of the molecular hallmarks of AD in the retina and brain of 3xTg-AD mice Amyloid precursor protein (APP), Aβ, p-tau and beta-secretase (BACE) protein expression levels were assessed by Western blotting in the hippocampus, cortex and retina homogenates of 3xTg-AD and age-matched WT mice, at 4 and 8 months of age (Fig. 1).
The APP levels in 3xTg-AD mice were significantly higher in the hippocampus, at 8 months Assessment of blood-retinal and brain barrier integrity of 3xTg-AD mice It has been reported an impairment of BBB integrity associated with AD (22). The control of permeability of BBB and BRB integrity is assured by the correct assembly of tight junctions (TJ) protein complexes. We investigated the brain and retinal barrier integrity by assessing the content of the tight junction proteins claudin-5, occludin and zonula occludens-1 (ZO-1), in the hippocampus, cortex and retina homogenates of 3xTg-AD and age-matched WT mice, at 4 and 8 months of age (Fig. 2).
The Western blotting (Fig. 3A). assay. At these early stages, we did not find any cell death by apoptosis (Fig. 3D) in the three regions studied.
Assessment of glutamate and GABA levels in the retina and brain of 3xTg-

AD mice
Changes in the content of neurotransmitters were already reported in the brain of AD animal model and patients (25,26). However, there is lack of information about this topic in the retina. We evaluated the levels of gamma-aminobutyric acid (GABA) and glutamate in the hippocampus, cortex and retina homogenates of 3xTg-AD and age-matched WT mice, at 4 and 8 months of age, by HPLC (Fig. 4).
The levels of GABA and glutamate did not change in the hippocampus of 3xTg-AD mice comparing with age-matched WT, at 4 and 8 months (Fig. 4A). In the cortex, at 4 months, 3xTg-AD mice presented a significant decrease in the GABA levels (WT: 0.006 ± 0.0002 nmol/mg protein; AD: 0.004 ± 0.0006 nmol/mg protein; p < 0.05) and a tendency for a decrease in the levels of glutamate. At 8 months no changes were detected in the cortex (Fig. 4B). In the retina, no statistically significant changes were detected in the levels of GABA and glutamate, but at 8 months there was a tendency for increased levels of glutamate (Fig. 4C).
Evaluation of glial protein immunoreactivity in the retina and brain of 3xTg-AD mice The contribution of glial cells for AD pathophysiology has been widely explored (reviewed in (27) Fig. 6K-M). In the retina, microglia presented a hyper-ramification in the more proximal branch orders and an atrophy in the more distal branches, comparing with the agematched WT (Fig. 6N-P).

Discussion
The concept 'the retina as an additional tool for AD diagnosis' has been increasingly explored, since some AD patients present visual alterations before the appearance of the first cognitive symptoms. Moreover, the retina and the brain share the same embryonic origin.
We recently reported the existence of structural and functional changes in the brain and retina of 3xTg-AD mice (15,16). These alterations occur as early as 4 months of age.
Because of the detection of these early changes in this animal model of AD, in the present work we investigated the presence of potential early molecular and cellular alterations in the retina and brain that may underlie the structural and functional changes detected in this model. Moreover, we intended to check whether the potential changes in the retina and brain were similar, as well as to check whether some changes could eventually occur in the retina before they appear in the brain.
Increased Aβ levels were detected in the hippocampus and cortex of 3xTg-AD mice at 4 and 8 months, despite no changes detected in APP and BACE at both time points, with the exception of increased levels of APP in the hippocampus at 8 months. These results suggest that the increased levels of Aβ detected in the brain may result from an increase of BACE activity or an impairment of Aβ removal. Several evidences point out mechanisms of Aβ clearance as a potential therapeutic strategy to modulate AD (29)(30)(31). In the retina, no changes were detected in Aβ levels. Actually, the presence of Aβ in the retina of animal models of AD is still a matter of controversy (reviewed in (3). In fact, we tried different approaches and techniques to detect Aβ in the retina and we were only successful by using a specific protocol of homogenization abovementioned. It was recently reported Aβ staining in retinal slices of 3xTg-AD mice, at 5 post-natal weeks (32). We also found a similar result, however, when we incubated both retinal slices and whole mounts only with the secondary antibody, we observed the same staining pattern. Thus, the Aβ staining previously reported may actually result from precipitation of the secondary antibody or unspecific binding. In fact, 3xTg-AD mice are known to be one of the animal models of AD with lower Aβ expression in the retina (33).
We also checked the hyperphosphorylated tau protein in the hippocampus, cortex and retina of 3xTg-AD mice. For that, we used an antibody that recognizes tau protein phosphorylated at serine 396, a phosphorylation site associated with early AD (34). We detected increased p-tau levels in the hippocampus and retina of 3xTg-AD mice, at both time points, while in the cortex this increase was only observed at 4 months. Similarly, others reported increased p-tau in the retinas of 3xTg-AD mice already at presymptomatic stages (32). However, it is important to highlight that these authors used a different antibody that recognizes the protein tau phosphorylated at serine 202 and threonine 205, which is more associated with the later stages of AD.
Regarding barriers integrity, our results indicate that claudin-5, occludin and ZO-1 levels are not affected in the retina and brain of 3xTg-AD mice at early stages (4 and 8 months of age). Additionally, we did not detect albumin extravasation from the blood to brain or retinal parenchyma. In summary, our results show that at early time points the BRB and BBB are not compromised in 3xTg-AD mice. From the best of our knowledge, this is a pioneering study assessing simultaneously the integrity of BBB and BRB in an AD animal model. Previous studies were mainly focused in BBB, and still no consistent results were reported (35)(36)(37)(38). In the retina, attenuated and disorganized ZO-1 and occludin staining was reported in 5xFAD mice, an AD animal model with a more aggressive phenotype, at 8 months (39).
We also assessed cell loss, and synaptic and neurotransmitters changes. Cholinergic Additionally, others also reported a decrease in the number of ChAT + cells in the retina of APP/ PS1 mice, associated with an increase of cell death by apoptosis, but this was observed at 13-16 months (41). In fact, we cannot discard the possibility that the 3xTg-AD mice can also present changes in the number of ChAT + cells at an older age.
We assessed cell death in the hippocampus, cortex and retina to clarify whether cell death could occurs simultaneously in the retina and brain of 3xTg-AD mice at the early stages.
We did not detect any evidence of cell death by apoptosis in all regions analyzed.
Formerly, it was described an increase in the number of cells stained for cleaved caspase-3 both in the retina and hippocampus of 3xTg-AD mice at 5 weeks of age that disappear at 30-40 weeks (7-10 months) (32). However, caspase-3 has been also implicated in the regulation of synaptic plasticity (42,43) and in mitochondria function (44). Grimaldi and colleagues claimed that increased staining for caspase-3 indicates neuronal apoptosis (32), but they did not show any evidence of cell death in the retina and brain of 3xTg-AD mice.
Loss of RGCs in AD patients has also been described (reviewed in (3). We did not find a significant decrease in the number of RGCs, at 4 and 8 months. Additionally, we did not find changes in the function of RGCs in 3xTg-AD mice (16), therefore suggesting that RGCs are not affected in this animal model at the early stages of pathology.
Regarding synaptic changes, our results suggest that the retina and brain are not similarly affected, at 4 months, in this animal model. At 8 months, we did not detect any changes in the levels of the synaptic proteins evaluated in the three regions analyzed. Although others detected decreased synaptophysin and PSD95 levels in the hippocampus and cortex of 3xTg-AD mice at 4 and 7 months (45), most studies did not observe changes in synaptic proteins at 6, 8 and 9 months in this AD animal model (46)(47)(48). At 4 months we detected an increase in the content of several synaptic proteins, which has been corroborated by other studies, at early stages (57). This increase might be due to a compensatory mechanism. Also, in postmortem brains from AD patients it was described an increase in several synaptic proteins at early stages of the pathology (Braak 3 and 4) that did not persist at later stages (Braak 5 and 6) (49). These studies corroborate our findings, suggesting that a biphasic synaptic protein response during AD progression is likely to occur. These evidences lead us to hypothesize that compensatory mechanisms (increase of synaptic proteins) might occur to keep homeostasis and cell survival. However, with the disease progression the synaptic deficit reaches a certain threshold where the brain cannot compensate anymore.
Several reports have demonstrated that AD patients present a disturbance in glutamate and GABA levels, leading to a neuronal circuitry disruption, and consequently affecting memory (50)(51)(52)(53)(54)(55). Changes in neurotransmission occur prior to cognitive decline (25). We detected a decrease of GABA in the cortex of 3xTg-AD mice at 4 months, but no more significant changes were observed in the GABA and glutamate levels in all regions analyzed. Others reported a decrease in vesicular glutamate transporter type I (vGluT1) density, without neuronal damage, in the hippocampus of 3xTg-AD mice, at 4 months (56).
At the early stages the total glutamate levels do not change, but with the alterations in VGluT1 the glutamate transport could be compromised and therefore the glutamate levels at the synaptic cleft.
Evidences point out the glial cells as key elements in the AD pathophysiology (reviewed in (57). Regarding GFAP, our results suggest the occurrence of differential glial fluctuations over time in the areas analyzed. It was recently reported an increase of astrocytic reactivity in the retina of 3xTg-AD mice at 5-10 and 30-40 weeks (32), which is not in agreement with our findings. However, others also reported inconsistent glial changes in the brain and retina in APP/PS1 mice (58). Moreover, we did not find significant changes in vimentin, suggesting no significant changes in Müller cells, at the early stages. Others also reported no changes in Müller cells in 3xTg-AD (59) and in APP/PS1 mice (60), at early stages. Conversely, evidence of an increase of GFAP staining in the Müller cells end-feet in the retina of 3xTg-AD mice at 9 months of age (32) suggests an increase of Müller cells reactivity. Taking into account our results and previous findings, it appears that Müller cells reactivity may occur in AD, but at later stages.
During an insult microglial cells increase their proliferative rate, change their morphology, and acquire a reactive state in order to reestablish the homeostasis (61). No changes were detected in the number of microglial cells neither in the hippocampus nor in the retina of 3xTg-AD mice at early stages. Former studies also reported no changes in the number of microglial cells in the retina (32) and brain of 3xTg-AD mice, at early stages (62). We also assessed microglial morphology, i.e., the number and length of processes. In the hippocampus of 3xTg-AD mice we observed a hypertrophy of microglial cells (only at 4 months), while in the retina microglial cells display an atrophic morphology. Moreover, at 8 months, retinal microglial cells featured a biphasic morphological structure since presented atrophy in the last orders and a hypertrophy in the initial ones. These results suggest a complex microglial remodeling in AD and also indicate there is a differential microglia remodeling in the retina and hippocampus during AD pathology. To our knowledge, this study is the first one that simultaneously analyzed the microglia structure in the brain and retina in the context of AD. It was reported an increase of microglial cells complexity in the retina of 3xTg-AD mice at early stages (5-10 weeks) (32). This apparent contradictory results could be explained because they used a different approach for 3D microglia reconstruction (imageJ plug-in) and the time point assessed was different (5-10 weeks versus 4 months in our work). We recently reported changes in retinal microglia of 3xTg-AD mice at older ages. Microglia acquired a different morphology and orientation along the retina, and their localization changed from a parallel to a perpendicular position relative to the retinal surface (63). In the brain of AD patients, it was found a reduction in the length of processes and arborized area of microglial cells, indicative of microglia reactivity (64). This group of findings led us to hypothesize that in the early stages microglial cells adopt a hyper-ramified morphology trying to maintain the appropriate brain surveillance. However, with disease progression, microglia acquire a less complex morphology accompanied with a senescent state.

Conclusion
Our results show that at the early stages there are no significant and consistent molecular and cellular changes in 3xTg-AD mice in the regions analyzed. Moreover, some of the few alterations detected in the brain and retina of 3xTg-AD mice do not match, suggesting that at molecular and cellular level the retina does not seem to mirror the brain changes associated with AD phenotype.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.  Evaluation of the tight junction protein levels in the retina and brain of 3xTg-AD mice. Claudin-5, occludin and zonula occludens-1 (ZO-1) protein expression levels were evaluated by Western blotting in the hippocampus, cortex and retina homogenates of 3xTg-AD and age-matched WT mice, at 4 and 8 months of age.
Values were normalized to β-actin (internal control), and are shown as percentage of age-matched WT mice ± SEM (n=5-12). Representative Western blots of claudin-5, occludin and ZO-1 and β-actin are presented below the graphs. Glutamate and GABA levels in the retina and brain of 3xTg-AD mice. The glutamate and GABA levels were evaluated by HPLC in the hippocampus, cortex and retina homogenates of 3xTg-AD and age-matched WT mice, at 4 and 8 months of age. Values were normalized to the amount of protein, and are shown as mean ± SEM (n=4-9). *p < 0.05, 3xTg-AD mice compared to age-matched WT mice.