AAV-mediated neuronal expression of an scFv antibody selective for Aβ oligomers protects synapses and rescues memory in Alzheimer models

Abstract

The accumulation of soluble oligomers of the amyloid-β peptide (AβOs) in the brain has been implicated in synapse failure and memory impairment in Alzheimer’s disease. Here, we initially show that treatment with NUsc1, a single-chain variable-fragment antibody (scFv) that selectively targets a subpopulation of AβOs and shows minimal reactivity to Aβ monomers and fibrils, prevents the inhibition of long-term potentiation in hippocampal slices and memory impairment induced by AβOs in mice. As a therapeutic approach for intracerebral antibody delivery, we developed an adeno-associated virus vector to drive neuronal expression of NUsc1 (AAV-NUsc1) within the brain. Transduction by AAV-NUsc1 induced NUsc1 expression and secretion in adult human brain slices and inhibited AβO binding to neurons and AβO-induced loss of dendritic spines in primary rat hippocampal cultures. Treatment of mice with AAV-NUsc1 prevented memory impairment induced by AβOs and, remarkably, reversed memory deficits in aged APPswe/PS1ΔE9 Alzheimer’s disease model mice. These results support the feasibility of immunotherapy using viral vector-mediated gene delivery of NUsc1 or other AβO-specific single-chain antibodies as a potential therapeutic approach in Alzheimer’s disease.

Graphical abstract

In this work, Ferreira and colleagues show that targeting Aβ oligomers (AβOs) with an oligomer-specific single-chain variable-fragment antibody (NUsc1) delivered to the brain using an adeno-associated viral vector (AAV-NUsc1) rescues memory in AβO-infused WT mice and in aged APPswe/PS1ΔE9 Alzheimer’s disease model mice, supporting the feasibility of gene-mediated immunotherapy.

Introduction

Dementia affects 55 million people worldwide, and Alzheimer’s disease (AD) accounts for 60%–70% of dementia cases.¹ Remarkable research efforts in the past few decades have elucidated many of the mechanisms thought to underlie AD brain dysfunction, cognitive impairment, and behavioral alterations in AD. However, despite significant progress in our understanding of AD pathogenesis, effective disease-modifying drugs are still lacking.²

A distinct therapeutic target in AD has emerged in recent years, comprising neurotoxic soluble oligomers of the amyloid-β peptide (AβOs). AβOs are structurally distinct from amyloid fibrils, which are insoluble and deposit in the brain parenchyma as the amyloid plaques that are histopathological hallmarks of AD. The oligomer hypothesis for AD has been proposed as an alternative to the so-called amyloid cascade hypothesis,^3,4 based on a large body of evidence implicating AβOs as causal agents of synapse damage and cognitive impairment in AD.^3,5,6 AβOs accumulate in AD brain and cerebrospinal fluid (CSF)^7,8,9,10 and are prominent in various transgenic animal models of AD, including those with little or no amyloid plaque burden.^11,12 Experimental exposure to AβOs causes cognitive deficits in animal models and induces major features of AD pathology, including tau hyperphosphorylation, brain inflammation, synapse elimination, and selective nerve cell death.³

Specific targeting of pathogenic oligomers is expected to provide improved target engagement and efficacy of Aβ-directed therapies.¹³ Immunotherapeutic approaches targeting a broad spectrum of A species (including monomers, oligomers, amyloid fibrils, and plaques) may have failed due to a reduction in the effective mass of antibodies that remain available to bind and neutralize the most neurotoxic Aβ species (i.e., oligomers), as a large portion of such antibodies may be lost in off-target interactions (little toxic or non-toxic monomers, fibrils, and plaques).¹⁴

We here demonstrate the preclinical efficacy of targeting AβOs with NUsc1, an AβO-specific single-chain variable-fragment (scFv) antibody selected by phage display from a library of human-derived scFv antibodies. NUsc1 shows little or no reactivity to non-toxic Aβ monomers or to insoluble amyloid fibrils and targets a specific subpopulation of AβOs of apparent molecular mass >50 kDa, which binds to neurons and is highly toxic to synapses.^15,16 An important feature of scFv antibodies is the lack of the Fc domain, which reduces their capacity to activate cellular immune/inflammatory responses.^17,18,19 From a therapeutic standpoint, this is appealing, as brain inflammation has been a significant adverse effect in clinical trials of immunoglobulin-based AD immunotherapies.^20,21,22,23 In addition, the shorter sequence of scFvs compared with intact immunoglobulins facilitates their gene delivery by adeno-associated virus (AAV)-based vectors.

Gene-mediated expression of therapeutic antibodies in the brain obviates the need for regimens of repeated injections in patients^18,19 and allows for local production of antibodies, thus reducing losses associated with inefficient blood-brain barrier crossing and systemic clearance mechanisms. AAV vectors are available in various serotypes, are commercially available for use in gene therapy,²⁴ and are currently being tested in numerous clinical trials. For example, Rafii et al.^25,26 showed that administration of an AAV2 vector harboring the nucleotide sequence for human nerve growth factor induces sustained transgene expression in the brains of control individuals and AD patients without major complications in a 2-year follow-up period.^25,26 We now describe the development of an AAV9 vector that drives neuron-specific expression of NUsc1 within the brain and rescues memory in AD models.

Results

NUsc1 prevents AβO-induced impairment in hippocampal long-term potentiation and memory in mice

We first tested whether exogenous recombinant NUsc1 could block the inhibition of hippocampal long-term potentiation (LTP) and memory loss induced by AβOs. We exposed mouse hippocampal slices to 200 nM AβOs and measured LTP responses elicited by high-frequency stimulation at Schaffer collaterals. Consistent with previous reports,^27,28 LTP was inhibited in hippocampal slices exposed to AβOs. In contrast, slices treated with NUsc1 before application of AβOs exhibited normal LTP (Figures 1A and 1B). Control measurements showed that NUsc1 had no effect on LTP in control hippocampal slices (i.e., in the absence of AβOs).

Figure 1.

Recombinant NUsc1 prevents AβO-induced inhibition of long-term potentiation (LTP) in hippocampal slices and memory impairment in mice

(A) LTP was measured in mouse hippocampal slices. Baseline responses were recorded for 20 min, after which slices were perfused for 20 min (black horizontal line) with vehicle, purified recombinant NUsc1 (200 pM), AβOs (200 nM), or NUsc1 + AβOs. LTP was elicited by high-frequency stimulation (black arrowhead) at CA3 (Schaffer collaterals) and recording was performed at CA1 for 2 h after stimulus. The main plot shows field excitatory post-synaptic potential (fEPSP) slopes measured as a function of time under different experimental conditions. Values at each time point are represented as the mean ± SE. Representative fEPSP traces before (black lines) and after high-frequency stimulus (colored lines) are illustrated on the right for each experimental condition. (B) Plot of mean fEPSP measured 2 h after stimulus. n = 7–9 slices from 5 to 7 individual mice per experimental condition. Values represent the mean ± SE, and symbols represent individual slices. The p value for each comparison is shown; two-way ANOVA followed by Dunnett’s post hoc test. (C and D) NUsc1 prevents AβO-induced memory impairment in mice. Three-month-old Swiss mice received an i.c.v. infusion of NUsc1 (10 fmol) 30 min prior to i.c.v. infusion of AβOs (10 pmol). Animals were tested in the novel object recognition (NOR) task 24 h (C) and 7 days (D) after AβO infusion. Percentages of time spent exploring familiar (F) and novel (N) objects are represented by white and colored bars, respectively. Values represent the mean ± SE and symbols represent individual mice. Color coding is the same in all panels. n = 17–19 total animals per group, tested in two independent experiments. The p value for each experimental condition is shown in the graphs; two-tailed one-sample Student’s t test comparing the percentage of novel object exploration time to the chance value of 50%.

To determine whether NUsc1 could prevent memory impairment induced by AβOs, we treated mice with an intracerebroventricular (i.c.v.) infusion of NUsc1 30 min before i.c.v. administration of AβOs and assessed their memory using the novel object recognition (NOR) task. Consistent with our previous reports,^28,29,30 AβO-infused mice exhibited memory impairment both 24 h and 7 days following i.c.v. infusion of AβOs (Figures 1C and 1D). In contrast, mice treated with NUsc1 exhibited normal performance in the NOR test both 24 h and 7 days post-infusion of AβOs.

Construction of AAV-NUsc1

We next sought to determine whether neuronal transduction by an AAV vector constructed to drive expression of NUsc1 could be protective in AD models. We constructed an AAV9 vector harboring the nucleotide sequence for NUsc1 containing a single amino acid substitution for expression in eukaryotes (see materials and methods) downstream of a signal peptide (SP) for secretory pathway export and under the control of the synapsin I promoter for selective neuronal expression (Figure 2A). AAV9 was used in the current study because it is an efficient serotype for transduction in the CNS and one of the most frequently used via both intracerebral and systemic routes of administration.³² AAV vectors are generally considered safe, although, in some instances, the use of AAV vectors in gene therapy trials has raised safety issues, notably related to the induction of detrimental immune responses against the capsid, hepatotoxicity, or thrombotic microangiopathy.^33,34 However, such adverse events occurred when using high vector doses delivered by systemic or local injection, whereas much lower doses of vectors are administered in direct intracerebral injections. To our knowledge, no major adverse events have been reported to date after intracerebral administration of AAV vectors.

Figure 2.

Transduction by AAV-NUsc1 drives NUsc1 expression and secretion in adult human brain slices, decreases AβO binding, and prevents dendritic spine loss in rat hippocampal neurons

(A) Map of the pAAV-NUsc1 plasmid used for AAV-NUsc1 vector production. The heavy and light chains of NUsc1 are linked by a flexible linker (VH/linker/VK). NUsc1 has a signal peptide (SP) for secretion and two C-terminal (His and Myc) epitope tags. Expression is under the control of neuron-specific promoter synapsin I, and the WPRE domain was used to enhance NUsc1 expression. (B) AAV-NUsc1 drives the expression of NUsc1 in adult human brain slices in culture. Human cortical slices were infected at 3 days in vitro (DIV) with increasing doses of AAV-NUsc1 vector. At 7 DIV, slices were collected and NUsc1 expression was quantified by qPCR, normalized by 18S ribosomal RNA, and plotted as a function of viral particle load per milligram of tissue. Symbols represent individually treated slices. (C) Dot immunoblot (anti-His antibody) analysis of culture media from vehicle-treated or AAV-NUsc1-infected (10⁸ vp/mg) human brain slices. Representative results (top) and quantification (bottom) of independent cultures from two donors are shown. Values represent means ± SE. (D) Hippocampal cultures were transduced (or not) with AAV-NUsc1 (MOI = 10⁴) at 14 DIV and exposed to AβOs (500 nM) at 20 DIV. AβO binding to neurons was detected using oligomer-specific NU4 monoclonal antibody³¹ following 3 h of exposure to AβOs. Representative images are shown for cultures exposed to AβOs (left) or to AβOs following transduction by AAV-NUsc1 (right). Graph shows integrated fluorescence of bound AβOs (NU4 immunoreactivity, puncta along dendrites). Bars represent means ± SE of five experiments (normalized for cultures exposed to AβOs alone) with independent cultures and AβO preparations. Symbols correspond to individual cultures. The p value is shown in the graph; two-tailed paired Student’s t test. Scale bars correspond to 10 μm. (E) Hippocampal cultures were transduced (or not) with AAV-NUsc1 (MOI = 10⁴) at 14 DIV and exposed to AβOs (500 nM) at 20 DIV. Dendritic spine density was assessed by labeling with Alexa-conjugated phalloidin after 24 h of exposure to AβOs. Representative images are shown for vehicle- or AβO-exposed cultures previously transduced (or not) by AAV-NUsc1, as indicated. Scale bars correspond to 15 μm. Insets show optical zoom images of isolated dendrite segments. Scale bars correspond to 1 μm. The number of dendritic spines along 20 μm dendrite segments was quantified. Three dendrite segments from five neurons were quantified in triplicate experiments in each experimental condition. Bars represent means ± SE, n = 3 experiments with independent neuronal cultures and AβO preparations; symbols represent independent cultures; p values for each comparison are shown in the graph; repeated measures one-way ANOVA followed by Dunnett’s multiple comparisons test.

By selectively targeting neurons and employing a single dose of 3 × 10⁹ viral particles (vp) of AAV-NUsc1, we aimed to prevent overexpression of NUsc1 by glial and other cell types in the brain, thus minimizing the potential for concentration-dependent antibody self-aggregation and for induction of aberrant immune or inflammatory responses.

AAV-NUsc1 induces neuronal expression of NUsc1 in human adult cortical slices

To examine the translational potential of AAV-NUsc1 in AD, we first tested the capacity of AAV-NUsc1 to drive scFv antibody production and secretion in adult human brain tissue.³⁵ Because NUsc1 expression in our AAV-NUsc1 construct is driven by the synapsin promoter and its secretion is induced by the presence of an SP, we evaluated the potential of adult human neurons to express and secrete NUsc1. Initial tests with an AAV9-mCherry control vector confirmed diffuse transduction in human adult cortical slices in culture (Figure S1). Human cortical slices exposed to increasing titers of AAV-NUsc1 in culture showed dose-dependent expression of NUsc1 (Figure 2B). NUsc1 protein expression and secretion to the medium were confirmed by dot immunoblot analysis (Figure 2C). Results indicate that AAV-NUsc1 transduces adult human neurons and drives the expression and secretion of NUsc1.

AAV-NUsc1 reduces AβO binding to neurons and prevents AβO-induced loss of dendritic spines

To determine whether AAV-NUsc1 could protect neurons from the toxic impact of AβOs, we transduced mature rat hippocampal cultures (14 days in vitro [DIV]) with AAV-NUsc1. Secretion of NUsc1 into the culture medium was verified by western blotting at 21 DIV (Figure S2). Based on our previous report that treatment with exogenous recombinant NUsc1 reduces AβO binding to neurons,¹⁶ we asked whether NUsc1 produced and secreted by AAV-NUsc1-transduced neurons could similarly block neuronal binding of AβOs. Indeed, transduction by AAV-NUsc1 caused ∼50% decrease in AβO binding to dendrites in hippocampal neurons (Figure 2D).

We further investigated whether the reduction in AβO binding by NUsc1 could protect neurons from AβO-induced loss of dendritic spines. Whereas neurons exposed to AβOs showed reduced dendritic spine density compared with control cultures, neurons transduced by AAV-NUsc1 prior to exposure to AβOs exhibited normal dendritic spine density (Figure 2E).

AAV-NUsc1 rescues memory in mouse models of AD

As a control experiment prior to assessing the in vivo efficacy of AAV-NUsc1, we examined brain distribution and neuronal expression of an mCherry transgene following i.c.v. infusion of an AAV9-mCherry control vector in mice. Results revealed that AAV9-mCherry induced expression of mCherry over the entire mouse brain, with prominent expression in the hippocampus and striatum after 8 weeks (Figure 3A). Having verified that the i.c.v.-infused AAV9 vector successfully transduced neurons in AD-relevant brain areas, we next infused AAV-NUsc1 i.c.v. into mice and verified NUsc1 expression in their brains by qPCR and immunoblotting (Figure 3B).

Figure 3.

Transduction by AAV-NUsc1 prevents AβO-induced memory loss and reverses memory deficits in aged APPswe/PS1ΔE9 mice

(A) AAV-mCherry (3 × 10⁹ viral particles) was infused via i.c.v. in 3-month-old Swiss mice. Brain distribution and expression of mCherry were evaluated by immunohistochemistry 8 weeks after infusion. The main image shows a photomontage of coronal sections (−1.82 mm from bregma) from a control, uninfected mouse (left hemisphere) and an AAV-mCherry-transduced mouse (right hemisphere). Sections were stained with DAPI (blue). Scale bar corresponds to 1 mm. Insets show higher magnification images of areas contained within the dashed white rectangles. Scale bar corresponds to 100 μm. (B and C) Three-month-old Swiss mice received an i.c.v. infusion of 3 × 10⁹ viral particles of AAV-NUsc1 8 weeks prior to i.c.v. infusion of AβOs (10 pmol). (B) NUsc1 expression in the brain was determined by (left) qPCR (means ± SE, n = 13 animals per group; two-tailed Mann-Whitney test) and (right) western blot (arrow points to band corresponding to NUsc1; asterisks mark non-specific bands). (C) Animals were tested in the NOR task 24 h after infusion of AβOs. Percentage of time spent exploring the novel object is represented by colored bars. Symbols correspond to individual mice. n = 12–17 animals tested in three independent experiments. (D–G) APPswe/PS1ΔE9 mice (9–18 month-old male and female mice) received an i.c.v. infusion of 3 × 10⁹ AAV-NUsc1 particles. (D) Sandwich ELISA employing 6×His and Myc tags present in NUsc1 shows NUsc1 levels in the hippocampus (black, WT mice; red, APP/PS1 mice; blue, AAV-NUsc1-treated WT mice; green, AAV-NUsc1-treated APP/PS1 mice); Student’s t test. (E–G) Eight weeks after infection, mice were tested in the NOR task and in the three-chambered social recognition test. (E) Percentage of time spent exploring the novel object in the NOR test is represented by colored bars. Symbols represent data for individual mice. (F) In the three-chambered social recognition test, animals were first habituated in the middle chamber of the apparatus and were then given an option between exploring an empty chamber (E) or exploring a chamber containing a stranger mouse (S1) of the same sex and similar age (white bars labeled “E” versus colored bars labeled “S1” in the graph). Symbols represent individual mice. (G) In the social novelty part of the task, mice were given the option to explore the already familiar mouse (S1) or a novel mouse (S2). Time spent exploring novel (dark bars) and familiar (light bars, color coded as in F) mice was quantified. Values represent means ± SE and symbols represent individual mice. n = 10–15 animals tested in three independent experiments. The p value for each experimental condition is shown in the graph; two-tailed one-sample Student’s t test comparing percentage of exploration time of the novel mouse (S2) to the chance value of 50%. (E–G) Blue symbols represent male mice and pink symbols represent females.

Two months after i.c.v. infusion of AAV-NUsc1 or AAV-mCherry (control vector), animals received an i.c.v. infusion of AβOs. When tested in the NOR memory test, both AβO-infused mice (Figure 3C) and AAV-mCherry-treated/AβO-infused mice (Figure S3) failed the task. In contrast, AβO-infused mice that had been transduced with AAV-NUsc1 exhibited normal performance in the NOR task (Figure 3C). Control experiments showed that transduction by AAV-NUsc1 had no impact on the performance of control, vehicle-infused mice in the NOR test (Figure 3C).

Finally, we tested the beneficial action of AAV-NUsc1 on memory impairment exhibited by aged male and female APPswe/PS1ΔE9 AD model mice. Memory tests on APPswe/PS1ΔE9 mice (or wild-type [WT] littermates) were performed 2 months after i.c.v. infusion of AAV-NUsc1. NUsc1 brain levels were evaluated by ELISA in WT and APP/PS1 mice (Figure 3D) and confirmed that AAV-NUsc1-transduced mice showed clear production of NUsc1 (2–5 ng NUsc1/μg hippocampal protein). APPswe/PS1ΔE9 mice failed both the NOR test and the social memory phase of the three-chambered social recognition test (Figures 3E and 3G), but not the sociability phase of the three-chambered test (Figure 3F). Remarkably, APPswe/PS1ΔE9 mice transduced with AAV-NUsc1 exhibited normal performances in both NOR and social memory tasks (Figures 3E and 3G), indicating reversal of memory impairments in male and female APP/PS1ΔE9 mice.

Discussion

NUsc1 shows little or no reactivity to non-toxic Aβ monomers or to insoluble amyloid fibrils^15,16 and targets a specific subpopulation of AβOs of apparent molecular mass >50 kDa that bind to neurons and are highly toxic to synapses. Recent unpublished observations from our group using NUsc1-functionalized beads to immunoprecipitate AβOs indicate that <15% of a typical AβO preparation (containing both low- and high-molecular-mass oligomers²⁹) are recovered after elution from such beads. These recent data (A. Sebollela et al., unpublished data) reinforce the notion that NUsc1 selectively targets a relatively minor subpopulation of toxic oligomeric species.

Here, we first found that exogenous recombinant NUsc1 prevented the inhibition of LTP in hippocampal slices and memory impairment induced by AβOs in mice. From a mechanistic perspective, AAV-mediated neuronal expression of NUsc1 reduced AβO binding to neurons and blocked AβO-induced loss of dendritic spines in cultured hippocampal neurons. Transduction by AAV-NUsc1 prevented memory deficits in AβO-infused mice and, notably, reversed memory impairments in aged APPswe/PS1ΔE9 mice. Of significant relevance from a potential translational perspective, AAV-NUsc1 effectively transduced and drove the expression and secretion of NUsc1 in cultured adult human brain slices.

A number of clinical trials involving Aβ immunotherapies in AD are currently under way, and considerable attention has been given to the recent FDA approval of an antibody targeting multiple Aβ assemblies (aducanumab).³⁶ However, considerable controversy remains regarding the benefits of aducanumab and of other IgGs targeting heterogeneous Aβ assemblies (monomers, fibrils, and amyloid plaques, in addition to oligomers). Significantly, Liu et al. have shown that targeting of amyloid fibrils by N-terminal antibodies against Aβ can lead to a “dust-raising effect,” favoring the formation of AβOs from fibrils and consequently enhancing neurotoxicity,³⁷ an effect that was also observed in humans.³⁸ Specific targeting of oligomers by NUsc1 may thus result in substantial improvement in target engagement and efficacy.

Importantly, none of the approaches developed thus far have taken advantage of the genetic tools available to induce sustained brain expression of an scFv to specifically block the most toxic Aβ species. To our knowledge, this is the first study to utilize a gene therapy approach to achieve sustained neuronal expression of a human-derived scFv antibody that selectively targets a highly toxic subpopulation of AβOs and exhibits minimal reactivity to Aβ monomers and fibrils. We found that this strategy led to protection against AβO-induced memory decline in WT mice and, significantly, to reversal of age-dependent memory impairment in APPswe/PS1ΔE9 mice.

Gene therapy has recently attracted considerable interest in neurology stemming from its successful introduction into clinical practice for spinal muscular atrophy.³⁹ Moreover, gene therapy has reached clinical trials in a number of neurological disorders, including lysosomal storage diseases, aromatic L-amino acid decarboxylase deficiency disorders, and Parkinson’s disease.⁴⁰ Two gene therapy clinical trials have been registered in ClinicalTrials.gov for AD (NCT00087789 and NCT03634007). While the first approach using an AAV2 vector to induce brain NGF expression failed to prevent cognitive impairment in AD patients,²⁶ it showed no adverse effects. The second approach using an AAVrh.10hAPOE2 to treat APOE4 homozygotes is currently recruiting patients for phase 1. Preclinical gene therapy studies targeting α-secretase activator/PKC modulator or sAPPα⁴¹ have not yet reached clinical trials.

In the current proof-of-principle study, we performed i.c.v. administration of AAV-NUsc1 to allow for diffusion of the AAV vector and expression of NUsc1 in multiple brain regions. However, AAV-NUsc1 could also be directly infused into specific brain regions if needed. We further note that capsid modification could be introduced to facilitate AAV-NUsc1 delivery via intravenous or even intranasal administration.

Potential advantages of the current approach include (1) the need for a single infusion of AAV-NUsc1 to achieve sustained brain expression of NUsc1; (2) local brain expression and secretion of a single-chain antibody devoid of the Fc domain and, thus, less likely to induce aberrant immune/inflammatory responses; (3) specific targeting of a particular subpopulation of AβOs that are highly toxic to synapses and neurons,^15,16 (4) reduced off-target interactions with Aβ monomers and plaques; and (5) being amenable to further development, including systemic administration of modified AAV vectors capable of effectively crossing the blood-brain barrier.⁴² Based on our current findings, we propose that development of AAV-NUsc1 represents a step toward reaching immuno-gene therapy for AD.

Materials and methods

Expression and purification of NUsc1

NUsc1 was expressed in the HB2151 E. coli strain with isopropyl β-D-1-thiogalactopyranoside (IPTG) induction, as described.¹⁵ Soluble NUsc1 was purified from both the supernatant and the lysate of HB2151 cells using a protein A affinity column (GE Healthcare). The solution containing eluted antibody was buffer-exchanged into phosphate-buffered saline (PBS) (Dulbecco’s PBS without calcium or magnesium; Corning) (pH 7.4) and concentrated using 10-kDa cutoff centricons (Merck) before storage at −80°C. The final concentration of purified protein was determined using Bradford reagent (Bio-Rad). Purified NUsc1 was routinely checked by SDS-PAGE and size-exclusion chromatography (SEC).

AAV-NUsc1 and AAV-mCherry vectors

NUsc1 scFv was subcloned from the original plasmid for expression in bacteria (piT2-NUsc1) by PCR with the following alternative pairs of primers: PCR1, FWD1 (5′–3′) TATTACTCGCGGCCCAGC/REV1 (5′–3′) CTATGCGGCCCCATTCAG; PCR2, FWD2 (5′–3′) GCTAGCTATTACTCGCGGCCCAGC (with a restriction site for Nhe1)/REV2 (5′–3′) TCGCGACTATGCGGCCCCATTCAG (with a restriction site for Nru1). The PBSKII plasmid was digested with SmaI and used to link the product of each PCR, yielding intermediate plasmids PBSKII-NUsc1 PCR1 and PBSKII-NUsc1 PCR2. The PBSKII-NUsc1 PCR2 plasmid was digested with Nru1, blunt-ended with Klenow polymerase, and digested with Nhe1, thus releasing the fragment coding for the open reading frame of NUsc1. This fragment was then cloned into plasmid pA-EAU2⁴³ to generate intermediate plasmid pA2-NUsc1, in which NUsc1 is driven by the human cytomegalovirus (HCMV) promoter. Sequencing of this plasmid revealed a STOP codon between heavy- and light-chain sequences, which is methylated in bacteria. Since this STOP codon would lead to translation interruption in mammalian cells, site-directed mutagenesis was conducted to replace the STOP codon with a glutamic acid (Glu) codon, resulting in the pA2-NUsc1 plasmid for NUsc1 expression in eukaryotic cells. The NUsc1 protein was then fused to two N-terminal epitope tags (His and Myc tags). In addition, an export signal (SP) was added to promote secretion into the extracellular medium. Finally, the NUsc1 open reading frame (ORF) was extracted from pA2-NUsc1 and subcloned into the pA2-SGWA plasmid to obtain plasmid pAAV-NUsc1, in which the NUsc1 ORF is driven by the synapsin I promoter. This plasmid also contains a WPRE sequence to enhance NUsc1 expression (Figure 2A).

AAV9 vector production was achieved by co-transfection of 293 cells with pAAV-NUsc1, helper plasmid, and rep-cap plasmid. Cells were collected 48 h post-transfection and lysed to harvest the AAV particles, which were purified by CsCl gradient. The vector was titrated by qPCR. The rAAV particles purified by iodixanol gradients were then quantified; the number of genome-containing particles was assessed by qPCR. Following transduction of mammalian cells, the AAV-NUsc1 vector drives expression of a 30-kDa protein, as expected for the NUsc1 transgene (Figure S2).

The AAV-mCherry control vector was purchased from Virovek (Hayward, CA) and consists in an AAV9 vector harboring the nucleotide sequence for the fluorescent protein mCherry under the control of the synapsin I promoter.

Preparation and characterization of AβOs

AβOs were prepared from synthetic Aβ_1–42 (California Peptide) as previously described.^44,45 Oligomer preparations were routinely characterized by size-exclusion high-performance liquid chromatography (HPLC) and, occasionally, by Western blots using oligomer-sensitive NU4 monoclonal antibody⁴⁶ and comprised a mixture of Aβ dimers, trimers, tetramers, and higher-molecular-weight oligomers.^27,29,31 Protein concentration was determined using the BCA assay (Thermo Scientific-Pierce).

Neuronal cultures

Hippocampal cultures were prepared from E18 Wistar rat embryos and were maintained in neurobasal medium supplemented with B27 (Invitrogen) for 3 weeks as described.⁴⁷ Cultures were treated with vehicle or AAV-NUsc1 vector using an MOI of 10⁴ at 14 DIV and were exposed to 500 nM AβOs (or vehicle) at 20 DIV. The supernatant was collected at 21 DIV for determination of the presence of NUsc1 by western blotting.

Immunocytochemistry and phalloidin labeling

Cells were fixed and blocked as described,⁴⁸ incubated with AβO-selective NU4 mouse monoclonal antibody (1 μg/mL)⁴⁶ overnight at 4°C, and incubated for 3 h at 23°C with Alexa conjugated secondary antibody. Spines were labeled with Alexa-conjugated phalloidin (which binds to spine-localized dense bundles of F-actin) for 20 min at 23°C, according to the manufacturer’s instructions (Invitrogen). Coverslips mounted with Prolong containing DAPI were imaged on a Zeiss Axio Observer Z1 microscope using an EC Plan-Neofluar 63×/1.25 oil M27 objective. Dendritic spines were quantified manually using NIH ImageJ.

Electrophysiological recordings

Electrophysiological recordings were performed as described.⁴⁹ Briefly, transverse hippocampal slices (400 μm) were prepared and transferred to a recording chamber where they were maintained at 29°C and perfused with artificial cerebrospinal fluid (aCSF, 2 mL/min flow rate) continuously bubbled with 95% O₂ and 5% CO₂. Field extracellular recordings were performed by stimulating the Schaffer collateral fibers through a bipolar tungsten electrode and recording in CA1 stratum radiatum with a glass pipette filled with aCSF. After evaluation of basal synaptic transmission, a 20-min baseline was recorded every minute at an intensity eliciting a response approximately 35% of the maximum evoked response. Slices were then perfused for 20 min with vehicle, 200 nM AβOs (expressed as Aβ monomer concentration), 200 pM NUsc1, or AβOs + NUsc1. This dose was based on previous findings by Puzzo et al.⁵⁰ After treatment, LTP was induced by theta-burst stimulation (four pulses at 100 Hz, with bursts repeated at 5 Hz, and three tetanic 10-burst trains at 15-s intervals). Responses were recorded for 2 h after tetanization and were measured as field excitatory post-synaptic potentials (fEPSP) slopes expressed as a percentage of baseline.

Human cortical slice culture

Cortical tissue was obtained from adult patients submitted to amygdalohippocampectomy for the treatment of refractory temporal lobe epilepsy at the University Hospital of the Federal University of Rio de Janeiro. Collection of this tissue for research purposes was under informed consent of the donors and approved by the Institutional IRB of the Federal University of the State of Rio de Janeiro under no. CAAE: 69409617.9.0000.5258. A fragment of temporal cortex (surgical access tissue) was collected at the operating room and was immediately processed and cultured as described.⁴⁶ Briefly, tissue was sliced at 400 μm using a McIlwain tissue chopper. Slices were plated in 24-well plates (one slice/well) containing 400 μL Neurobasal A (Gibco) supplemented with 1% Glutamax (Gibco), 1% penicillin/streptomycin (Gibco), 2% B27 (Gibco), and 0.25 μg/mL amphotericin B (Gibco) supplemented with 50 ng/mL BDNF (Sigma Aldrich). Cultures were maintained at 37°C and 5% CO₂. These experiments were performed paired, i.e., AAV-NUsc1-treated slices were utilized in parallel with control slices obtained from the same donor. Cultures from three different donors (M, age 61; F, age 39; F, age 45) were used in this study.

Dot immunoblots

Frozen samples of culture medium from human cortical slices were thawed and concentrated using 10-kDa-cutoff Amicon Ultra-0.5 mL centrifugal filters. Samples (20 μg total protein in 200 μL) were spotted onto a nitrocellulose membrane using a vacuum-assisted dot-blot apparatus (Bio-Dot Apparatus 1706545, Bio-Rad). Blots were blocked with 5% BSA in Tween-TBS at room temperature for 2 h and incubated at 4°C overnight with anti-His antibody (1:200; Sigma Aldrich) in blocking buffer. Membranes were then incubated with anti-mouse secondary antibody conjugated to IRDye 800CW (Licor, Lincoln, NE; 1:10,000) at room temperature for 2 h, imaged on an Odyssey Imaging System (Licor), and analyzed using NIH ImageJ. The integrated density of the dots corresponding to culture medium from AAV-NUsc1-treated samples was normalized by the corresponding control.

Animals and intracerebroventricular infusions

Three-month-old male Swiss or 9- to 18-month-old APPswe/PS1ΔE9 (or WT C57BL/6 littermates) male and female mice were used. Animals were housed in groups of five per cage with free access to food and water, under a 12-h light/dark cycle with controlled room temperature and humidity. All procedures followed the principles of laboratory animal care of the National Institutes of Health and were approved by the Institutional Animal Care and Use Committee of the Federal University of Rio de Janeiro (protocol IBqM 136/15) and the University of Western Ontario (protocols 2016–104 and 2016–103).

For i.c.v. infusion of AAV-NUsc1, AAV-mCherry, AβOs, or vehicle, animals were anesthetized for 7 min with 2.5% isoflurane (Cristália, São Paulo, Brazil) using a vaporizer system and were gently restrained only during the injection procedure. The i.c.v. injection was performed using a free-hand technique that has been implemented and well characterized in several previous studies by our group.^28,29,30 A 2.5-mm-long needle was unilaterally inserted 1 mm to the right of the midline point equidistant from each eye and 1 mm posterior to a line drawn through the anterior base of the eyes. Swiss mice received 3 × 10⁹ viral particles of AAV-NUsc1 (in a final volume of 3 μL) 8 weeks before the infusion of 10 pmol AβOs (or an equivalent volume of vehicle). An AAV-mCherry vector was used in control experiments following the same protocol. When indicated, 1.5 μL of purified recombinant NUsc1 (0.01 pmol) was administered 30 min before AβOs via the same i.c.v. injection site.

Transgenic APPswe/PS1ΔE9 mice (and WT littermate controls) received 3 × 10⁹ viral particles of AAV-NUsc1 via i.c.v. 2 months before behavioral tests. The injection volume was 3 μL, as this is the largest volume that can be injected safely into the lateral ventricle of mice. Because this dose of viral particles caused no signs of toxicity, brain degeneration, or behavioral changes in mice, we maintained this dose in our subsequent studies. Mice were closely monitored for the duration of the experiment and showed no signs of toxicity following AAV infusion. Three independent APP/PS1 cohorts were used in the current study. Cohort 1 comprised 20-month-old mice (21 females), cohort 2 comprised 11- to 16-month-old mice (8 males, 4 females), and cohort 3 comprised 14- to 16-month-old mice (10 males, 12 females). Within each independent experiment (carried out with independent cohorts), ages at behavioral testing (2 months after AVV-NUsc1 injection) were not different among experimental groups.

Immunohistochemistry

Animals were anesthetized and perfused with saline, followed by 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). Fixed brains were removed, cryoprotected in increasing concentrations of sucrose, frozen in dry ice, and stored at −80°C. Coronal sections (40 μm) were obtained on a cryostat (Leica Microsystems) and stored in PBS (pH 7.4). Immunohistochemistry was performed after washing the sections extensively with PBS. Groups of six sections per animal were immersed in 0.1% Sudan Black B for 30 min, washed three times in PBS, and blocked for 2 h with 0.3% Triton X-100 and 5% BSA in PBS at room temperature. Sections were then incubated overnight with anti-mCherry primary antibody (1:100; Thermo Fisher) diluted in PBS, washed, and incubated for 2 h with Alexa Fluor 594-conjugated secondary antibody (1:1,000; Life Technologies). After a final washing step, the sections were briefly stained with DAPI and mounted with Prolong (Thermo Fisher). Images were acquired on a Zeiss Axio Observer Z1 microscope.

Image processing

Representative images shown in Figures 2D, 3A, and S1 were processed using Zeiss ZEN software. Contrast was enhanced by adjusting the tonal range of the histograms (i.e., brightness/contrast adjustment) equally across all images being compared. Care was taken to avoid clipping. DAPI was used as a counterstain and thus adjusted more freely, with the goal of not obscuring other channels. Nonetheless, processing of DAPI images was quite similar across all images. No gamma corrections or any other type of non-linear corrections were made in any channel on any of the images shown. All analyses and quantifications were made on raw images prior to any processing.

Western immunoblots

Forty-eight hours after i.c.v. infusion of AβOs, hippocampi and cortex were dissected and immediately frozen in liquid nitrogen. For total protein extraction, samples were thawed and homogenized in PBS containing a phosphatase and protease inhibitor cocktail (Thermo Scientific-Pierce). Protein concentrations were determined using the BCA kit. Samples containing 30 μg protein were resolved in 15% polyacrylamide Tris-glycine gels (Invitrogen) and were electrotransferred to nitrocellulose membranes at 350 mA for 1 h. Blots were incubated with 5% BSA in Tween-TBS at room temperature for 2 h and incubated at 4°C overnight with anti-His tag primary antibody diluted in blocking buffer. Membranes were then incubated with anti-mouse secondary antibody conjugated to IRDye 800CW (1:10,000) at room temperature for 2 h, imaged using the Odyssey imaging system (LiCor), and analyzed using NIH ImageJ.

RNA extraction and quantitative real-time PCR

Samples were homogenized and RNA was extracted using the SV total RNA isolation kit (Promega). RNA purity was determined by the 260/280 nm absorbance ratio. One microgram of RNA was used for cDNA synthesis using the High-Capacity cDNA reverse transcription kit (Applied Biosystems). qPCR was performed on an Applied Biosystems 7500 RT-PCR system using the Power SYBR kit (Applied Biosystems). Cycle threshold (Ct) values were used to calculate fold changes in gene expression using the 2^−ΔΔCt method.⁵¹ The following primers were used to detect the expression of NUsc1: forward, TCAGCAGAAACCAGGGAAAG; reverse, CTGCTGATGGTGAGAGTGAAA. Actin-β was used as a housekeeping gene for mouse samples and the primers used were forward, 5′TGTGACGTTGACATCCGTAAA-3′; reverse, 5′GTACTTGCGCTCAGGAGGAG-3′. Human slice samples were normalized by 18S ribosomal RNA and the primers used were forward, ATCCCTGAAAAGTTCCAGCA; reverse, CCCTCTTGGTGAGGTCAATG.

NUsc1 ELISA

NUsc1 was quantified by ELISA in hippocampal extracts from WT or APP/PS1 mice. Immediately prior to the assay, aliquots of extracts (prepared in TBS) were diluted in PBS + 2% BSA (Sigma) to a final concentration of 0.6 mg/mL total protein. Samples were added to plate wells coated with rabbit anti-Myc antibody (Sigma; 1:1,000 in PBS) and incubated for 14 h at 4°C. After washes with PBS + 0.1% Tween 20 (Sigma), mouse anti-His antibody (Sigma; 1:5,000 in PBS + 2% BSA) was added and incubated for 1 h. Detection was carried out using anti-mouse IgG-HRP antibody (GE; 1:5,000 in PBS + 2% BSA) followed by the addition of TMB substrate (Sigma). Reaction was stopped by adding 0.5 M H₂SO₄. A NUsc1 standard curve was built using purified recombinant NUsc1 at 0, 1, and 16 nM (R² = 0.99).¹⁵ Optical density values from control (vehicle-infused) mice were averaged, and the resulting mean value was subtracted from signals obtained from AAV-NUsc1-treated mice to obtain the mass of NUsc1 in each extract.

Novel object recognition task

The task was performed in an open field arena measuring 30 × 30 × 45 cm (W × L × H). The floor of the arena was divided by lines into nine equal rectangles. Test objects were made of glass or plastic and had different shapes, colors, sizes, and textures. During sessions, objects were fixed to the box to prevent displacement caused by exploratory activity of the animals. Previous tests showed that none of the objects used evoked an innate preference. Before training, each animal was submitted to a 5-min habituation session to freely explore the empty arena. Training consisted of a 5-min session during which the animals were placed at the center of the arena in the presence of two identical objects. The amount of time spent exploring each object was recorded. Sniffing and touching the object were considered exploratory behavior. The arena and objects were cleaned thoroughly between trials with 40% ethanol to eliminate olfactory cues. In the test session, performed 2 h after training, one of the two objects used in the training session was replaced by a new one. Time spent exploring familiar and novel objects was measured. Results were expressed as a percentage of time exploring each object during the test session and were analyzed using a one-sample Student t test, comparing the mean exploration time for each object against the fixed (chance) value of 50%. An animal that recognizes the familiar object (i.e., that learns the task) explores the novel object >50% of the total time.

Three-chambered social recognition task

Each animal was positioned in an apparatus divided into three equally sized chambers and tested along three sessions lasting 5 min each. The first session consisted in free exploration of the middle chamber. The second session evaluated social interaction by quantifying the time spent exploring each of the side chambers, one containing a small empty wire cage and the other containing an identical cage containing another mouse of the same sex and similar age compared with the test animal inside. Finally, the third session consisted in placing a novel mouse in the empty cage on the opposite chamber and quantifying the amount of time spent exploring the familiar and the novel mice, to evaluate social memory.

Statistical analysis

Results were tested using the D’Agostino and Pearson omnibus normality test and are represented as means ± SE. All analyses were performed using GraphPad Prizm software. Outliers were identified and were removed when applicable. Sample sizes and statistical tests used to analyze the results are specified in the corresponding figure legends.