Intra- and extracellular β-amyloid overexpression via adeno-associated virus-mediated gene transfer impairs memory and synaptic plasticity in the hippocampus

Alzheimer’s disease (AD), the most common age-related neurodegenerative disorder, is currently conceptualized as a disease of synaptic failure. Synaptic impairments are robust within the AD brain and better correlate with dementia severity when compared with other pathological features of the disease. Nevertheless, the series of events that promote synaptic failure still remain under debate, as potential triggers such as β-amyloid (Aβ) can vary in size, configuration and cellular location, challenging data interpretation in causation studies. Here we present data obtained using adeno-associated viral (AAV) constructs that drive the expression of oligomeric Aβ either intra or extracellularly. We observed that expression of Aβ in both cellular compartments affect learning and memory, reduce the number of synapses and the expression of synaptic-related proteins, and disrupt chemical long-term potentiation (cLTP). Together, these findings indicate that during the progression AD the early accumulation of Aβ inside neurons is sufficient to promote morphological and functional cellular toxicity, a phenomenon that can be exacerbated by the buildup of Aβ in the brain parenchyma. Moreover, our AAV constructs represent a valuable tool in the investigation of the pathological properties of Aβ oligomers both in vivo and in vitro.


Results characterization of Aβ expression following incubation and injection of AAV-BRi-Aβ42
and AAV-UBi-Aβ42 vectors. The detailed description of generating AAV constructs used here was reported previously [19][20][21] . AAV vectors encoding BRI-Aβ cDNAs, fusions between human Aβ peptides and the BRI protein (known to be associated with amyloid deposition in British familial dementia), are able to promote high-level expression of Aβ peptide in the absence of APP overexpression 19 . AAV-BRI-Aβ42 and AAV-UBI-Aβ42 were created to facilitate the expression of extracellular and intracellular Aβ, respectively 22 .
Prior to in vivo testing, we incubated hippocampal neuronal cell cultures with AAV constructs encoding BRI-Aβ42 or UBI-Aβ42 fusion proteins in order to determine the optimal concentration and efficacy for each construct. Exogenous synthetic oligomers of Aβ were used as a positive control. We detected significant levels of Aβ42 in the culture medium using three concentrations of BRI-Aβ42, in contrast to the UBI-Aβ42 and EGFP control AAV constructs (Fig. 1a). We also measured the expression of Aβ oligomers in cell media and lysate of cell cultures. There is no significant increase in the cell media after incubation of either AAV constructs, even though there is an increased trend in the AAV-BRI-Aβ42 in accordance with what has been observed in Fig. 1a. We observed an increase in the levels of oligomers in the cell lysate after incubation with AAV-BRI-Aβ42 compared to EGFP (Fig. 1c). With these results we confirmed the ability of these constructs to promote the overexpression of Aβ peptides.
Next, we tested the AAV-BRI-Aβ42 and AAV-UBI-Aβ42 expression in vivo by determining the Aβ42 relative levels in the soluble and insoluble fractions of mice hippocampus after AAVs injection. Mice were divided into 3 cohorts: AAV-BRI-Aβ42-treated, AAV-UBI-Aβ42-treated or AAV-EGFP-treated. Each subject received bilateral intrahippocampal injection of a single specific AAV construct, and the brains were collected and analyzed 3 months later. The overexpression of BRI-Aβ42 construct resulted in higher expression of both soluble and insoluble Aβ42 as compared to the UBI-Aβ42 construct, while there was no detectable Aβ following EGFP incubation (Fig. 2a,b). Immunostaining for 6E10 in the BRI-Aβ42 construct demonstrated a high amyloid deposition in the hippocampus. However, animals that received the UBI-Aβ42 presented distinct neuronal processes staining with mild intraneuronal accumulation of Aβ (Fig. 2c), without accumulation of Aβ deposits 23 . Hippocampal Aβ expression promoted by AAV constructs leads to impaired cognition. Animals treated with both Aβ AAV constructs presented significant cognitive impairment, measured by performance in the Morris Water maze test (Fig. 3). In this evaluation, both groups took more time to find the hidden platformas demonstrated by latency (Fig. 3b) -and also crossed the platform fewer times (Fig. 3c). The injection of both vectors did not affect motor skills, as demonstrated by velocity and distance evaluation (Fig. 3d,e). These results are significant and demonstrate the importance of both intra-and extracellular Aβ in the development of spatial cognitive impairments. There are no significant changes with contextual fear conditioning (Fig. 3f).
In vitro synaptic function is altered by both BRi-Aβ42 and UBI-Aβ42. As Aβ is known to cause dendritic spine density and synaptic function changes 24 , we next examined if the cognitive impairments observed in our animals were associated with functional or structural changes in dendritic spines or specific synaptic markers.
Both BRi-Aβ42 and UBI-Aβ42 drive synapse loss in vivo. Consistent with the suppression of synaptosomal cLTP by BRI-Aβ42 and UBI-Aβ42, we found that our AAV constructs reduced the total number of spines ( Fig. 5a), as well as that of stable mushroom spines (Fig. 5b). The BRI-Aβ42 construct also significantly reduced stubby (Fig. 5c) and filopodia-like spines (Fig. 5d), whereas the UBI-Aβ42 construct did not affect the stubby spines, i.e. there was a higher number of these immature spines in this group. In addition, we observed reduced protein levels of the pre-and postsynaptic markers PSD-95 and synaptophysin, respectively ( Fig. 5f,g), as well as of profilin-1 (Fig. 5h), an actin cytoskeleton protein, suggesting that impairments in synaptic function and strength could be related to intra-and extracellular Aβ.

Discussion
The findings of this study imply that intra-and extracellular Aβ accumulation mediated by AAV-gene transfer can promote deleterious effects on synaptic and cognitive functions. Our study indicates that these AAV constructs of Aβ42 induce memory impairments, alongside reductions in the number of spines and of proteins related to synaptic function and LTP. These results are significant as they demonstrate that mild intraneuronal accumulation of Aβ42 (UBI-42) is as potent as combined intraneuronal and extracellular accumulation (BRI-42), significantly impacting our understanding of the detrimental aspects of AD.
The view that insoluble Aβ fibrils are the major factor in AD pathogenesis was firmly held until prefibrillar soluble Aβ oligomers were shown to be more detrimental in some experimental settings [31][32][33][34][35] . These oligomers can interact with glutamate receptors, dysregulating calcium influx and also altering spine morphology and density 36 . Moreover, when extracted directly from AD brains and injected into rodent brains, they can inhibit LTP, enhance long-term depression (LTD), and reduce dendritic spine numbers 37 . Although the potential role of extracellular Aβ oligomers is more well-established, there is also evidence that intracellular Aβ and its deposition precede the formation of plaques in animal models and human brains 38 . We observed a significant increase in both soluble  www.nature.com/scientificreports www.nature.com/scientificreports/ and insoluble fractions of Aβ, particularly with the BRI-Aβ42 construct. Moreover, this construct also induced the formation of plaques as well as intraneuronal accumulation of Aβ in the CA1 area of the hippocampus.
The first reports describing the existence of intracellular Aβ appeared shortly after the original identification of Aβ as the main component of plaques. Since then, other studies have provided evidence for intracellular Aβ accumulation in transgenic mouse brain and in post-mortem brain samples from AD and Down syndrome patients. A rat model with a doubly mutated APP, driven by a Thy1.2 promoter highly expressed in neurons, triggered the accumulation of intraneuronal human Aβ in 2-3-month-old rats, coinciding with cognitive impairments and pre-plaque generation 39 . Further evidence linking intraneuronal Aβ accumulation to cognitive deficits and synaptic dysfunction came from transgenic mouse models 38,[40][41][42] . However, the overexpression of specific genes and the subsequent formation of plaques and/or tau pathology in such models make it hard to attribute the associated cognitive impairments they promote exclusively to the presence of intracellular Aβ. Our results are significant because, through the promotion of both intra-and extracellular Aβ, we can conclude that the accumulation of Aβ peptides, even in young mice, is sufficient to promote behavioral and synaptic impairments.
Synaptic deficits and loss are the pathological hallmarks that best correlate with the progressive cognitive decline observed in AD patients [43][44][45] . Here, we have demonstrated that synaptic function as measured by FASS-LTP is affected very early. We used FASS-LTP, an approach to evaluate chemically-induced LTP directly in isolated synaptosomes. FASS-LTP identifies the subset of synaptosomes that are double-labeled for surface GluA1 and Nrx1β (GluA1 + Nrx1β +). Importantly, the increase in surface GluA1 + Nrx1β+ levels after cLTP stimulation is sustained, and mechanistically parallels the facilitation of synaptic transmission following electrically induced LTP (e.g., dependence on NMDAR and CaMKII), as previously shown 26,27 . Specifically, the generation of cLTP was reduced significantly in the presence of either intra-and extracellular Aβ. Our results are in accordance with recent findings, whereas LTP was blocked by the intracellular injection of Aβ42 into hippocampal pyramidal cells 46 , and by extracellular Aβ at the CA3-CA1 synapses of APP-knockout mice 47 . Moreover, the Aβ42-induced impairment of glutamatergic synaptic function is dependent on its internalization and intracellular accumulation 48 .
Information storage underlying learning and memory, as well as signal transduction at excitatory synapses, develop at the postsynaptic density (PSD) 49,50 . Within PSDs, the most abundant scaffolding protein is PSD-95, which is known to play key roles in synaptic plasticity 51,52 . In the pre-synapse, the loss of synaptophysin is one of the best brain correlates of cognitive decline in AD 53 , occurring early in the development of disease and accompanied by increased APP and hyperphosphaorylated tau expression in neurons of the hippocampus and entorhinal cortex 54 . Interestingly, the reduction in synaptic density is more pronounced within immature and mature plaques 55 . Moreover, the actin cytoskeleton and its dynamics play pivotal roles in modulating synaptic function by organizing the PSD, anchoring postsynaptic receptors, facilitating the trafficking of synaptic cargoes, and localizing the translation machinery in the synapses 56,57 . Therefore, impairments in signaling pathways that regulate synaptic markers such as PSD-95 and synaptophysin, and actin dynamics, such as profilin-1, could lead to the synaptic and cognitive deficits observed early in AD.
Dendritic spines are specialized anatomical structures in neuronal cells that serve as the postsynaptic element/ component for the vast majority of CNS synapses. Structural changes at dendritic spines underlie learning and memory processes in the brain, and changes in spine structure and function might lead to cognitive impairments. Spines can be classified morphologically into four types: stubby, mushroom, thin and filopodia-like, which corresponds to their maturation and function as immature, stable and mature, transient or lacking synapses, respectively 58,59 . The structure and function of dendritic spines are dynamically regulated by cellular pathways acting on the actin cytoskeleton. In the data reported here, we observed that an increase in Aβ42 both intra-and extracellularly was coupled with a reduction in important synaptic proteins and changes in the density and morphology of dendritic spines that may constitute the primary cause for the synaptic inhibition and memory impairments.
Synaptic deficits and synapse loss occur early in AD and MCI, before the onset of plaques, being some of the first signs of the neurodegenerative process 60,61 . As Aβ aggregates it can adopt different shapes such as fibrils and non-fibrillar aggregates 62,63 , and there is a consensus that Aβ alone is not the main mediator responsible for AD. However, its precise pathogenic roles, subcellular location and state are still being discussed. Here we have provided evidence that low intraneuronal accumulation of Aβ peptides provoked by hippocampal infusion of UBI-Aβ42, as well as combined intraneuronal and extracellular accumulation of Aβ provoked by BRI-Aβ42, can disrupt cognitive behavior, synaptic plasticity and spine morphology.
Some insights on the relationship of different amyloid structures have already been discussed [64][65][66] . Further studies are required to determine the roles of these AAV vectors in promoting specific Aβ isoforms. It is of great importance to the advancement of AD research the acknowledgment of how the available tools work, so they can be appropriately employed in order to provide relevant results.

Methods infusion of vectors.
Stereotaxic injection of AAV-EGFP, AAV-BRI-Aβ42 and AAV-UBI-Aβ42 into the hippocampus was performed according to previously described surgical protocols 67, 68 . Viral preps were generated as described previously 69 . Briefly, AAV vectors expressing the Aβ peptides under the control of the cytomegalovirus enhancer/chicken beta actin (CBA) promoter, a WPRE, and the bovine growth hormone polyA were generated by plasmid transfection with helper plasmids in HEK293T cells. 48 hours after transfection cells were harvested non-specific fluorescence for each marker were set by staining with secondary antibodies only. (d) Values normalized to the basal state in each experimental group, mean ± SEM. Basal vs cLTP: EGFP, *p = 0.041 (n = 6); AAV-BRI-Aβ42, P = 0.244 (n = 7); AAV-UBI-Aβ42, P = 0.461 (n = 6). and lysed in the presence of 0.5% Sodium Deoxycholate and 50U/ml Benzonase (Sigma) by freeze thawing, and the virus isolated using a discontinuous Iodixanol gradient, and affinity purified on a HiTrap HQ column (Amersham). The genomic titer of each virus was determined by quantitative PCR.
Aβ derived diffusible ligand (ADDL) preparation. ADDLs were prepared according to previous publications 71 . Briefly, Aβ1-42 was dissolved in hexafluoro-2-propanol (HFIP) and aliquoted to microcentifuge tubes. HFIP was removed by evaporation under vacuum and an aliquot of Aβ42 was dissolved in anhydrous dimethyl sulfoxide (DMSO), which was then added to ice-cold F12 medium without phenol red. This solution was incubated at 4 °C for 24 h and then centrifuged at 14 000 g for 10 min. Centrifugation produced a small pellet and the supernatant is defined as the ADDL preparation, which comprises fibril-free solutions of oligomers as well as monomers. Cells were incubated with 10 mM ADDLs for 24 h.

Morris Water Maze.
Three months after infusion of vectors, behavioral analyses were performed. Mice were trained to swim to a circular clear Plexiglas platform submerged 1.5 cm beneath the water's surface. Four trials were performed per day, for 60 seconds each with 5 minutes between trials. Mice were trained for as many days as needed for the group to reach the training criterion of 25 seconds. The probe test was assessed 24 hours after the last trial, with the platform removed. Performance was monitored with the EthoVision XT video-tracking system (Noldus Information Technology, Leesburg, VA, USA). contextual fear conditioning. During training, mice were placed in the fear conditioning chamber and allowed to explore for 2 minutes before receiving three electric foot shocks (duration: 1 s, intensity: 0.2 mA, intershock interval: 2 minutes). Animals were returned to the home cage 30 seconds after the last foot shock. Twenty-four hours later, behavior in the conditioning chamber was video recorded for 5 minutes and subsequently analyzed for freezing behavior.
Golgi staining. Following transcardial perfusion with 0.1 M phosphate-buffered saline (PBS, pH 7.4), mice brains were removed and processed using a superGolgi Kit (Bioenno Tech LLC, Santa Ana, CA), as described previously 24,72 Dendritic and spine analysis. Stereological quantifications were performed using Neurolucida software from Microbrightfield Bioscience (MBF Bioscience, Williston, VT, USA) to determine the number of spines in the stratum radiatum (SR) and the molecular layer (ML) of the hippocampal CA1 region, respectively. Briefly, every second section was used through the entire antero-posterior extent of the hippocampus (between −1.46 mm anterior and −3.40 mm posterior to Bregma according to Franklin and Paxinos, Third Edition, 2007). The SR and ML in the CA1 region were defined using a 5x objective and spines were counted using a 100x/1.4 objective. The coefficient of error (CE) value for each animal ranged between 0.03 and 0.08. Dendritic spine length was traced using a 100x/1.4 objective and data were analysed via Neurolucida Explorer software. For dendritic morphological analysis, 5 neurons per animal (n = 6) in the CA1 hippocampal area were traced using Neurolucida software. Dendritic width was measured using Image J software in electronic microscopic images (10 images per animal for a total of 6 mice per group).
immunohistochemistry. For immunohistochemistry, sections (40μm thick) were pretreated with 3% H2O2/3% methanol in Tris-buffered saline (TBS) for 30 min, followed by a TBS wash. Sections were then incubated in TBS with 0.1% Triton X-100 (TBST) for 15 min, followed by TBST with 2% BSA (Sigma-Aldrich) for 30 min. Sections were incubated with anti-6E10 (1:1000; Biolegend, San Diego, CA, USA) in TBS + 5% normal horse serum overnight at 40 C. Sections were then incubated with the appropriate secondary biotinylated antibody (1:500) in TBS containing 2% BSA plus 5% normal serum for 1 hour at room temperature, followed by Vector ABC Kit and DAB reagents (Vector Laboratories, Burlingame, CA, USA) to visualize staining. electrochemiluminescence-linked immunoassay. Quantitative biochemical analyses of human Aβ and inflammatory cytokines in mouse tissue were performed using a commercially available electrochemiluminescence-linked immunoassay from Meso Scale Discovery (MSD, Gaithersburg, MD, USA). The V-PLEX Aβ Peptide Panel 1 (6E10) was used and plates were analyzed on the MS2400 imager (MSD). Assays were performed according to the manufacturer's instructions, and all standards and samples were measured in duplicate. enzyme-linked immunosorbent assay for Aβ42. Aβ1-42 was measured in the primary neuronal hippocampus cell culture medium using a sensitive sandwich enzyme-linked immunosorbent assay system as previously described 70 .
Statistical analysis. All data between two groups were analyzed by Student's t-test comparisons, and oneor two-way analysis of variance (ANOVA), followed by Bonferroni's test for comparisons among more than 2 groups. Mann-Whitney U Test was used for the FASS-LTP data. Graphpad Prism software (Graphpad Prism Inc., San Diego, CA, USA) was used, and the significance was set at 95% of confidence. Values are presented as mean ± SEM.

Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.