AAV1 vectors encoding APPsw and BRI-Aβ42 or BRI-Aβ40 fusion proteins were generated and injected bilaterally into the hippocampus of adult rats, alone and in combination. Brain tissue was examined for presence of transgene expression three weeks later. As seen in Figure 1, infusion of vectors resulted in widespread expression of the transgene in neurons throughout the hippocampus, detectable by immunohistochemistry using an antibody to human Aβ, residues 1–17 (6E10). AAV-BRI-Aβ42, AAV-BRI-Aβ40 and combined AAV-BRI-Aβ40+42 injected brains had extensive transgene expression in the rostral-caudal axis, extending ~2 mm on either side of the injection site, and throughout the layers of the hippocampus. In these brains the granule cell and molecular layers of the dentate gyrus and pyramidal cells of CA1 were transduced (Figure 1). More variable transduction levels were seen in CA2-CA4 subfields and hilar interneurons. In contrast, no transgene expression was observed in naïve (non-injected) rat brain. Given the lack of specific antibodies for the furin-cleaved BRI protein, it is not possible to distinguish whether the staining reflects Aβ or the BRI-Aβ fusion proteins. However, based on ELISA data (see below) and previous studies in transgenic mice, it is clear that the BRI-fusion proteins result in enhanced production of the encoded Aβ peptide 4.

In AAV-APPsw brains, Aβ immunostaining extended as far in the rostral-caudal direction as in BRI-Aβ fusion animals, but staining was not as widespread within the layers of hippocampus. These brains contained much lower expression levels in the granule cells of dentate gyrus and CA1 (Figure 1) than that observed in BRI-Aβ fusion brains, with no positive immunostaining in CA2-4 or hilar interneurons. Distinctive sub-cellular expression patterns were observed with the different vectors – transgene detected following transduction with the BRI-Aβ40 and BRI-Aβ42 vectors filled the cell bodies; transduction obtained with the APPsw vector resulted in sparse transgene expression within the cell soma but expression extended out into the processes.

Once hippocampal transgene expression had been confirmed in a small number of test animals, further animals were injected with AAV vectors and tested for development of cognitive deficits at 3 months post-infusion. In the open field, naive animals crossed an average of ~130 lines in a 5 min period; BRI-Aβ42, BRI-Aβ40 and APPsw animals crossed a comparable number of lines with mean line crosses in these groups of ~140 (Figure 2A). However BRI-Aβ40+42 animals crossed significantly more lines than naïve controls (203 vs 130 lines, p < 0.01) indicating higher baseline locomotor activity than other groups. Naïve rats spent ~13% of their time in the open (center of field) as did the BRI-Aβ40 animals, with a small reduction in time spent in center field in both the BRI-Aβ42 and APPsw groups relative to naive animals. The most marked reduction in time spent in the open however, was in the BRI-Aβ40+42 animals which spent on average only half the amount of time in the open as naive controls (~6% vs 13%) (Figure 2B).

In the Morris water maze rats were required to learn the location of the hidden platform over 5 training blocks, with the start point varied between each trial so that the task was difficult enough that differences in performance would be measurable. As seen in Figure 3A, naïve animals learned the location of the platform faster than AAV-treatment groups, with naïve animals reaching a minimum latency of ~18s by day 3 and not improving further beyond that time. All Aβ treatment groups took longer to reach the minimum latency than naïve controls, with the most notable difference between groups on day 3 when BRI-Aβ40+42 animals displayed an increased latency to find the platform. Over all 20 acquisition trials, naive rats had a mean latency of ~26 sec (Figure 3B), with all Aβ treatment groups having longer latencies than this, particularly BRI-Aβ40+42 animals which performed the worst overall with a mean latency over 20 trials of ~37s (ANOVA P = 0.07). Repeated measures ANOVA on pathlength taken to find the platform during the five days of acquisition training confirms the results obtained with latency data. Overall the difference between treatment groups was not significant (P = 0.061) (Figure 3C) but BRI-Aβ40+42 animals took a significantly longer mean pathlength to find the platform over all 20 trials, Figure 3D (ANOVA P = 0.02; BRI-Aβ40+42 > naïve, ** p < 0.01).

Animals were put through a probe trial 24 h after the last training trial to check retention of spatial learning (Figure 3E, F). Naive animals spend ~30% of their time in the target quadrant, as did the BRI-Aβ42, BRI-Aβ40 and APPsw animals. BRI-Aβ40+42 animals spent the least amount of time and distance in the target quadrant of any treatment group (~20%) indicating less retention of platform location than in other groups. Results from the visible platform test demonstrated that visibility was not impaired in treated animals compared to naïve controls, with no difference in latencies to swim to a visible target (data not shown).

In the passive avoidance test there was no difference in baseline latencies to enter the dark chamber between the treatment groups – on being placed in the light chamber for the first time, animals took an average of ~18s to move to explore the dark chamber, indicating no baseline differences in anxiety between treatment groups in this behavioral test (Figure 4A). AAV-BRI-Aβ40 animals had a significantly reduced latency to re-enter the dark chamber 24 hr after receiving a shock as compared to naïve controls (~200s vs 500s, p < 0.05), indicating these animals had an associational learning deficit. BRI-Aβ40+42 animals also had a reduced latency (~240s vs 500s) to enter the dark chamber but high variability between animals meant this did not reach statistical significance. BRI-Aβ42 and APPsw groups had reduced latencies (down by ~20% and 40% respectively) but were not significantly different to controls.

In the novel object recognition test, rats were habituated to the test environment and then given two identical objects to explore (O1, O2). In this first trial all groups spent equal amounts of time exploring both objects, with a ratio O2/O1 of ~1.2 for all groups. Rats were returned to the test chamber 15 min later and presented with one familiar (O3) and one novel (N) object. Animals in all groups recognized and explored the novel object more than the familiar object (with a ratio N/O3 of ~1.9–2.4 for all groups) (Figure 4C). However it is noticeable that BRI-Aβ40+42, BRI-Aβ42, and APPsw animals all spent less total time (O1+O2+O3+N) exploring objects than naïve controls (BRI-Aβ42 < naive, p < 0.05; APPsw < naïve, p < 0.01; Figure 4B) suggesting these animals are less inclined to explore but can still recognize a novel object when encountered. In this behavioral assay BRI-Aβ40 animals were comparable to naïve controls, with similar exploration times and similar ability to discriminate between novel and familiar objects.

Overall, expressing APPsw or either one of the BRI-Aβ fusions alone resulted in distinct cognitive deficits, but interestingly animals expressing both Aβ species exhibited more robust behavioral deficits.

Animals were killed and the dissected hippocampus examined for levels of RIPA-soluble and RIPA insoluble FA-soluble Aβ peptides (either Aβ40 or Aβ42) by ELISA to assess the levels of Aβ present at the time of behavioral testing (Figure 5A). Naïve rats had negligible levels of Aβ40 or Aβ42 in the hippocampus (all < 2 pmol/gm tissue). AAV-BRI-Aβ40 injected animals had an increase in both RIPA- and FA-soluble Aβ40 but no increase in RIPA- or FA-soluble Aβ42. In AAV-BRI-Aβ42 animals there was a marked increase in FA-soluble Aβ42 to ~72 pmol/gm but only low levels of RIPA-soluble Aβ42 were detected in these brains, suggesting that Aβ42 is accumulating in insoluble deposits. Aβ40 levels (RIPA- or FA-soluble) were not increased above baseline in AAV-BRI-Aβ42 animals suggesting an increase in Aβ42 production did not influence endogenous Aβ40 levels. Animals injected with the combination of BRI-Aβ40+42 vectors had higher levels of RIPA-soluble and FA-soluble Aβ40 than naïve animals and an increase in FA-soluble Aβ42, but no increase in RIPA soluble Aβ42. Interestingly, this level of insoluble Aβ40 is higher than that observed in rats injected with BRIAβ40 alone. AAV-APPsw animals had an increase in the level of FA-soluble Aβ40 (14.3 pmol/gm vs none detected in naïve) but levels of RIPA-soluble Aβ40 and RIPA-soluble and insoluble Aβ42 were comparable to naïve controls.

The other hemisphere from each brain was immunostained with anti-Aβ to detect any deposits of Aβ. Only BRI-Aβ42 animals showed "amorphous" plaque-like structures within the hippocampus following immunostaining with pan anti-Aβ 1–16 antibody (Figure 5B). These deposits appeared to be diffuse, confirmed when they did not stain with either thioflavin S or Congo Red (data not shown). Even though there was evidence for significant accumulation of RIPA-insoluble Aβ in the combined BRI-Aβ40+42 expressing rats, these plaques were not seen in the combined treatment group, any other treatment group or the naïve controls. Immunostaining performed on adjacent sections to those with plaque deposition in the BRI-Aβ42 rats showed no evidence of phosphorylated tau, astrogliosis or microgliosis in the vicinity of these diffuse plaques.

AAV1 vectors encoding APPsw, BRI-Aβ42 and BRI-Aβ40 were generated and injected into the hippocampus of adult rats. Transgene expression was confirmed immunohistochemically in a group of test animals (n = 3 per vector). The effects of long-term over-expression of BRI-Aβ42, BRI-Aβ40, combined BRI-Aβ42 + BRI-Aβ40, or APPsw in the rat hippocampus were assessed, with animals tested 3 months post-infusion for development of cognitive deficits and brain tissue examined for Aβ levels and presence of AD pathology (n = 12 per group).

cDNAs for APPsw (gift of D Selkoe) and BRI-Aβ fusions 29 were sub-cloned into an AAV expression plasmid under the control of a CBA (chicken beta-actin) promoter and containing a SAR (scaffold attachment region) element, WPRE (woodchuck hepatitis virus post-transcriptional-regulatory element), and bovine growth hormone polyadenylation signal flanked by AAV2 inverted terminal repeats (ITRs).

HEK293 cells were co-transfected with AAV and helper plasmids using standard CaPO₄ transfection. Cells were harvested 60 hr following transfection, and AAV1 vectors purified from the cell lysate by ultracentrifugation through an iodixanol density gradient followed by Q column purification, then concentrated and dialysed against PBS 6465. Vectors were titered using real-time PCR (ABI Prism 7700) and all vector stocks diluted to 1 × 10¹³ genomes/ml.

Animal studies were approved by The University of Auckland Animal Ethics Committee, and rats supplied by the Animal Resources Unit, The University of Auckland. 250–300 g male Wistar rats were used for all studies. Animals were anaesthetized with sodium pentobarbitone (Nembutal, Virbac Laboratories, 90 mg/kg, i.p.) and placed in a Kopf stereotaxic frame. AAV1 vectors were infused bilaterally into the hippocampus at the following stereotaxic co-ordinates: flat skull – anterior-posterior (AP) -4.0 mm, lateral (L) 2.1 mm, and ventral (V) 4.5 mm from skull surface, bregma = zero). 3 μl vector (3 × 10¹⁰ genomes) was infused into each side at 70 nl/min. Animals receiving both BRI-Aβ vectors still received a total of 3 × 10¹⁰ genomes in 3 μl (1.5 × 10¹⁰ genomes of each vector) into each side.

All animals were pre-handled for 7 days prior to testing and tests were carried out in the order described. All tests were conducted by an experimenter blinded to the treatments.

Behavioral data was analyzed using ANOVA (repeated measures where appropriate) with Dunnett's post-hoc analysis.

Test rats (used to confirm transgene expression) were euthanised with pentobarbitone and perfused transcardially with 60 ml saline followed by 60 ml 4% paraformaldehyde in 0.1 M phosphate buffer (PFA). Brain tissue was post-fixed for 24 h in 4% PFA, cryoprotected in increasing concentrations (10, 20, 30%) of sucrose in PBS and cut into 40 μm free-floating sections using a cryostat. All other rats were sacrificed one week after the conclusion of behavioral testing – animals were euthanised with pentobarbitone and brains removed and dissected longitudinally. One hemibrain was post-fixed in 4% PFA and used for immunohistochemical and histological analyses; the other half was frozen immediately on dry ice and used for Aβ ELISAs.

Sections from transgene expression brains were immunostained with anti-Aβ 1–17 (Chemicon MAB1560, 6E10, specific to human Aβ 1–17) according to the following protocol. Sections were washed in 1 × PBS containing 0.2% Triton (PBS-T), and incubated in 1% H₂O₂ in 50% methanol for 30 min to bind endogenous peroxidase present in the tissue. Sections were washed extensively in 1 × PBS-T, then treated with 70% formic acid for 30 min and again washed in PBS-T. 200 μl of primary antibody diluted 1:1500 in immunobuffer (1 × PBS-T containing 1% normal goat serum, 0.4 mg/ml methiolate) was applied overnight at room temperature on a rocking table. The following day sections were washed in 1 × PBS-T and incubated in 200 μl of biotinylated anti-mouse (Sigma; diluted 1:250 in immunobuffer) for 3 hr at room temperature. Following further washes in 1 × PBS-T, sections were incubated in 200 μl ExtrAvidin Peroxidase (Sigma; diluted 1:250 in immunobuffer) for 2 hr at room temperature. Sections were washed in PBS-T and antibody binding was visualised using 3', 3-diaminobenzidine (DAB; Sigma, St. Louis, MO) at 0.5 mg/ml DAB in 0.1 M phosphate buffer with 0.01% H₂O₂.

Hemibrains from long-term animals were post-fixed in 4% PFA and stained for Aβ plaques as described previously 66. Paraffin sections (5 μm) were pretreated with 80% formic acid (FA) for 5 minutes, boiled in water using a rice steam cooker, washed, and immersed in 0.3% H₂O₂ for 30 minutes to block intrinsic peroxidase activity. Sections were then incubated with 2% normal goat serum in PBS for 1 hour, with 33.1.1 (Pan Aβ 1–16 mAb) at 1 μg/ml dilution overnight, and then with HRP-conjugated goat anti-mouse secondary mAb (1:500 dilution; Amersham Biosciences) for 1 hour. Sections were washed in PBS, and immunoreactivity was visualized by DAB according to the manufacturer's specifications (ABC system; Vector Laboratories). Adjacent sections were stained with 4% thioflavin-S for 10 minutes. Additional sections were immunostained for activated microglia using anti-Iba1 (1:3000; Wako Chemicals); astrocytes using anti-GFAP (1:1000, Chemicon); phosphorylated tau with CP13 (1:100) and PHF-1 (1:100; kindly provided by Dr Peter Davies, Albert Einstein School of Medicine, Bronx, NY).

Rat hippocampal and cerebellar tissue was homogenized in radio-immunoprecipitation assay (RIPA) buffer and ultracentrifuged to separate RIPA-soluble from insoluble fractions. RIPA-insoluble proteins were extracted using 70% FA, neutralized with Tris base buffer and samples diluted (from 10–5000-fold). Extracted Aβ was then measured using a sandwich ELISA system as described previously 66 – Aβ42 capture with mAb 2.1.3 (mAb40.2,) and detection with HRP-conjugated mAb Ab9 (human Aβ 1–16 specific); Aβ40 capture with mAb Ab9 and detection with HRP-conjugated mAb 13.1.1 (mAβ40.1).