Gene therapy using Aβ variants for amyloid reduction

Abstract

Numerous aggregation inhibitors have been developed with the goal of blocking or reversing toxic amyloid formation in vivo. Previous studies have used short peptide inhibitors targeting different amyloid β (Aβ) amyloidogenic regions to prevent aggregation. Despite the specificity that can be achieved by peptide inhibitors, translation of these strategies has been thwarted by two key obstacles: rapid proteolytic degradation in the bloodstream and poor transfer across the blood-brain barrier. To circumvent these problems, we have created a minigene to express full-length Aβ variants in the mouse brain. We identify two variants, F20P and F19D/L34P, that display four key properties required for therapeutic use: neither peptide aggregates on its own, both inhibit aggregation of wild-type Aβ in vitro, promote disassembly of pre-formed fibrils, and diminish toxicity of Aβ oligomers. We used intraventricular injection of adeno-associated virus (AAV) to express each variant in APP/PS1 transgenic mice. Lifelong expression of F20P, but not F19D/L34P, diminished Aβ levels, plaque burden, and plaque-associated neuroinflammation. Our findings suggest that AAV delivery of Aβ variants may offer a novel therapeutic strategy for Alzheimer’s disease. More broadly our work offers a framework for identifying and delivering peptide inhibitors tailored to other protein-misfolding diseases.

Graphical abstract

Park and colleagues describe the identification and initial testing of Aβ sequence variants for amyloid reduction in Alzheimer’s disease. They develop an AAV vector to deliver peptide variants into the brain, showing this approach slows aggregation of human Aβ and lowers plaque load in a mouse model of the disease.

Introduction

The recent announcement of cognitive benefit from extended analysis of the EMERGE aducanumab trial has given new hope to the idea that targeting amyloid β (Aβ) may be clinically viable.¹ Although the findings must now be confirmed with additional clinical data before regulatory approval would be advised,² this glimmer of hope from an Aβ-lowering strategy suggests that the enormous effort invested in Aβ may not have been wasted. Although it seems unlikely that optimal Alzheimer’s disease treatment will target Aβ alone, it may well comprise one part of future combination treatments tailored to disease stage. Antibody therapies may be the first Aβ-lowering strategy to reach the clinic; however, their side-effect profile and need for repeated intravenous delivery may be obstacles to widespread use. Additional strategies for Aβ reduction should be developed alongside additional targets; here, we propose an alternate approach that attacks Aβ using its physical properties to therapeutic advantage.

The discovery that Aβ fibril formation is controlled by central domain amino acids 17–21 (LVFFA) has given rise to the idea that structural variants might be used to interfere with Aβ aggregation.³ Small peptides, like KLVFF (Aβ 16−20) or KLVFF-derived compounds, interact with Aβ to modify aggregation kinetics and reduce Aβ toxicity in vitro.^4,5 Further support for a peptide-based strategy comes from the identification of two familial mutations, Aβ^A2T and Aβ^A2V, which reduce aggregation of wild-type (WT) Aβ in a dominant-negative manner.6, 7, 8, 9, 10, 11 Although A2T is protective in both heterozygous and homozygous carriers, homozygous A2V is pathogenic and causes early-onset Alzheimer’s disease.^8,9 This distinction is critically important for the design of synthetic peptide inhibitors, and many past candidates have been eliminated due to spontaneous aggregation.^5,12 To overcome this fatal limitation, we sought Aβ sequence variants that (1) did not self-assemble, (2) prevented aggregation of WT Aβ, and (3) diminished cytotoxicity of the remaining aggregate.

Despite their promise in vitro, two major drawbacks of most peptide therapies are their short half-life due to rapid proteolytic degradation in the bloodstream and the challenge of delivery across the blood-brain barrier.^13,14 Viral gene therapy has the potential to overcome both of these obstacles. However, viral transduction is more straightforward for the expression of intracellular proteins or natively secreted proteins than for small peptides that must be delivered extracellularly. We tackled this extracellular delivery problem for Aβ peptide inhibitors by creating a minigene vector to express variant peptide at the cell membrane where its release into the extracellular space would be regulated by endogenous proteases. We then used adeno-associated virus (AAV) to deliver our vector into the neonatal (P0) mouse brain for widespread, lifelong neuronal expression to test efficacy for Abeta reduction in a mouse model of Alzheimer's amyloidosis expressing famial mutations in both amyloid precursor protein (APP) and presenilin 1 (PS1). Our findings provide initial preclinical support for novel gene therapy to prevent and potentially reverse Aβ aggregation in Alzheimer’s disease.

Results

F20P and F19D/L34P variants inhibit aggregation of WT Aβ42

The central hydrophobic region of Aβ (17Leu-Val-Phe-Phe-Ala21) governs the rate of monomeric assembly and forms the β-sheet hairpin in mature fibrils.15, 16, 17, 18 Given its importance in fibril formation, we asked whether targeted amino acid substitution in this region would yield peptides that could prevent aggregation of the wild-type Aβ42 peptide. Based on past studies of this domain, we examined 5 different Aβ42 variant peptides: V18P, F19D, F19P, F19D/L34P, and F20P.19, 20, 21 We first confirmed, using the Thioflavin-T (ThT) assay, that none of our five variants self-aggregated into fibrils, which would be a fatal flaw in any interventional peptide (Figure 1A). We next tested whether any of the five could competitively inhibit the fibrillization of wild-type (WT) Aβ42 when both peptides were mixed at equimolar ratios prior to starting the fibrillization reaction (Figure 1B). Four of the five variant peptides reduced fibrillization of WT Aβ, whereas one (V18P) exacerbated aggregation. Two peptide variants selected for further study, F20P and F19D/L34P, appeared to completely prevent aggregation. We next tested these two peptides for the potential to disassemble WT Aβ42 when added after fibril formation. Both variants disaggregated WT fibrils in a concentration-dependent manner; however, F19D/L34P was considerably more effective at fibril disassembly than F20P, attaining >80% loss of the ThT signal at the highest concentration tested (Figures 1C and 1D). Finally, we examined the potential for variant Aβ peptides to abate cytotoxicity caused by WT Aβ oligomers. For this experiment, WT Aβ42 was used alone or mixed at an equimolar ratio with variant peptide prior to oligomer assembly. 24 h later, peptide solutions were added to the media of the murine neuroblastoma 2a cell line (N2a) cell cultures. Oligomers assembled from WT Aβ42 caused nearly 50% cell loss compared to cultures treated with oligomerization buffer alone (Figure 1E). In contrast, neither F20P nor F19D/L34P significantly altered cell viability on its own. Notably, cell survival increased from 50% of control with WT Aβ42 alone to approximately 70% when WT peptide was co-incubated with either variant during oligomer assembly. Our results demonstrated that F20P and F19D/L34P variants meet the minimum in vitro criteria for candidate aggregation inhibitors.

Figure 1.

Variant Aβ peptides diminish fibrillization and cytotoxicity of wild-type (WT) Aβ

(A) Thioflavin-T (ThT) assay for self-aggregation of variant peptides compared to wild-type (WT) Aβ42. None of the 5 variants displayed self-aggregation during the reaction; for comparison, WT Aβ42 (green) reached plateau ThT binding within 1 h. (B) Competition assays testing inhibition of WT Aβ42 aggregation. Each variant was mixed 1:1 with WT Aβ42 and incubated with ThT. V18P exacerbated aggregation of the WT peptide; the other four variants abated fibrillization. F20P (red) and F19D/L34P (blue) prevented aggregation and were selected for further study. (C) Fibril disassembly assay for F19D/L34P. WT Aβ42 fibrils were exposed to varying concentrations of monomeric F19D/L34P peptide. Compared with fibrils alone, incubation with F19D/L34P produced a concentration-dependent reduction of ThT fluorescence, consistent with disassembly of Aβ fibrils. (D) Fibril disassembly for F20P, as described for F19D/L34P. F20P also produced a concentration-dependent reduction of ThT fluorescence but was considerably less effective than F19D/L34P. (E) MTS assay shows that unlike WT Aβ42 oligomers (green), neither F20P (red) nor F19D/L34P (blue) caused toxicity of N2a cells on its own; both were similar to untreated controls (black). In contrast, equimolar co-incubation of WT Aβ42 with either variant during oligomeric Aβ formation diminished subsequent N2a cytotoxicity. x axis values represent the volume (in microliters) of the 100 μL oligomeric Aβ reaction that was added to N2a media for testing. This equates to 0.1 or 10 μM of monomeric WT Aβ42 starting material, plus an equivalent amount of variant peptide where indicated. ANOVA, ∗p < 0.05, ∗∗p < 0.01 Bonferroni post-test for comparison with WT Aβ42 alone. All data are presented as mean ± SEM.

A vector design to secrete Aβ variants from mammalian cells

Although our peptide variants reduced Aβ aggregation in the test tube, our goal was to test them in the brain at levels high enough to be therapeutically effective. Our first step toward this goal was to design an expression strategy to stably express variant Aβ at the cell membrane where its release into the extracellular space would be regulated by endogenous γ-secretase. Previous studies achieved this goal by fusing the signal peptide from APP to the APP C-terminal fragment (β-CTF);²² however, whereas this construct did produce extracellular Aβ, β-CTF overexpression also unexpectedly caused Aβ-independent lysosomal-autophagic pathology.23, 24, 25 We therefore sought to avoid potential lysosomal complications by using the smallest β-CTF fragment needed for γ-secretase cleavage. We used the Gaussia luciferase signal peptide (GLSP) to target the membrane and fused this to a series of β-CTF fragments.²⁶ The longest fragment was 54 amino acids in length and included Aβ42 and the entire APP transmembrane (TM) domain plus two intracellular lysine (KK) residues (Figure 2A). Four additional constructs were each truncated by 2 amino acids into the TM domain; the shortest construct contained just 45 residues (Aβ42 + 3 additional). We transfected the different constructs into N2a cells and measured secreted Aβ in the conditioned cell culture medium using immunoprecipitation (IP) followed by immunoblot. We found that the complete APP TM domain plus KK was both necessary and sufficient to achieve Aβ secretion at levels equivalent to those observed using a construct encoding the full-length CTF (Figure 2A). Addition of the γ-secretase inhibitor (GSI) LY411575 blocked release of variant Aβ into the media in a concentration-dependent manner, confirming that Aβ secretion was dependent on γ-secretase activity (Figure 2A). We then cloned this minimal expression construct into an AAV delivery vector for subsequent in vivo expression under control of the ubiquitous CAG promoter (Figure 2B). The woodchuck poliovirus response element was included at the 3′ end to stabilize the mRNA and was followed by a bovine growth hormone poly(A) sequence. Finally, we used mass spectrometry to confirm that this vector would truly produce Aβ40/42 when expressed in cells. We transfected N2a cells with the viral Aβ F20P vector and then immunoprecipitated human Aβ from the media as we had for the WT peptide. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) was performed without enzymatic digestion to determine whether a full-length peptide was made. This analysis confirmed the presence of peptides with the appropriate mass for Aβ40 F20P (Figure 2C) and Aβ42 F20P (data not shown), along with the correct sequence for each peptide (Figure 2D).

Figure 2.

An expression construct optimized for efficient secretion of variant Aβ

(A) A sequential deletion strategy was used to identify the shortest C-terminal fragment (CTF) sufficient for γ-secretase cleavage. Constructs were transfected into N2a cells, and Aβ was harvested from the media to measure secretion. The optimal construct contained the full transmembrane (TM) domain plus two intracellular lysines (KK). This minigene achieved Aβ secretion equivalent to full-length CTF (bottom left blot; arrows indicate secreted Aβ versus uncleaved Aβ + residual TM domain for each construct, indicated by its ending residues). Secretion of WT Aβ42 using the “KK” minimal construct was blocked in a dose-dependent manner by γ-secretase inhibitor (GSI) LY411575 (bottom right blot; arrows again indicate secreted Aβ versus uncleaved Aβ + TM domain; GSI concentrations in nanomolar). (B) Design of the AAV vector for delivery of variant Aβ peptides in vivo. The expression cassette contains the Gaussia luciferase signal peptide at the N terminus of Aβ, followed by the Aβ42 variant, and the minimal APP C-terminal TM sequence, all controlled by the CAG promoter. (C) Mass spectrometry analysis of immunoprecipitated Aβ confirms extracellular release of full-length variant Aβ. N2a cells expressing Aβ F20P were used to isolate secreted peptide from the media by 6E10 immunoprecipitation. MS1 spectra of the eluted peptides display the expected mass for Aβ40 F20P (shown here) and Aβ42 F20P (data not shown). The peak for intact Aβ40 F20P at m/z = 856.4441 with a +5 charge state indicates a monoisotopic mass of 4,277.1841 Da, whereas the most abundant isotopomer indicates a mass of 4,279.1851 Da and is within 10 ppm error of the expected mass. (D) MS2 fragmentation of the peptide matching the mass of Aβ40 F20P confirms its sequence identity, with fragment ion tolerance less than 20 ppm error.

Using AAV to express Aβ variants in vivo

Our previous studies demonstrated that intraventricular injection of AAV into the P0 mouse brain yields widespread and persistent neuronal transgene expression compared to other methods.²⁷ We began by injecting WT mice to confirm that the peptides would be expressed and secreted in vivo. Both Aβ F19D/L34P and F20P expression constructs were packaged into AAV8 and injected into the lateral ventricles of P0 mice. Mice were euthanized 3−4 weeks after injection to examine expression levels and viral distribution. Viral spread was assessed by immunostaining using human-specific APP antibody (6E10) to detect variant Aβ. Viral expression was biased to the frontal cortex, and so subsequent analyses focused on this region (Figure 3A). Co-immunostaining with 6E10 to detect variant Aβ and Y188 to detect endogenous mouse APP confirmed that the virally delivered precursor peptide was located at the cell membrane alongside endogenous APP (Figure 3B). We used IP followed by MS in homogenates from WT mice transduced with Aβ F20P to test for expression of full-length peptide in vivo. Both Aβ40 and -42 F20P from brain homogenate displayed elution times identical to Aβ isolated from cell media; both peptides were again identified at the expected mass and sequence confirmed (Figures 3C and 3D; data shown for Aβ42 F20P). Notably, none of the partial cleavage products, which could have been immunoprecipitated with 6E10 (such as the intact TM domain without signal peptide or γ-secretase-cleaved Aβ still attached to the signal peptide), appeared in the spectra. We next measured the amount of Aβ produced by each variant using ELISA. The human-specific capture antibody and end-specific detection antibodies of this assay ensured that the peptides we measured were (1) human and therefore virally delivered variant Aβ and (2) mature Aβ that had undergone γ-secretase. F20P expressed well in vivo. Similar to native APP, this construct produced Aβx-40 peptide at levels several-fold greater than Aβx-42 (Figure 3E). In contrast, F19D/L34P produced high levels of Aβx-42 but no detectable Aβx-40. This anomaly may have resulted either from the Aβ40 end-specific antibody failing to bind the F19D/L34P peptide or from a shift in γ-secretase processing due to the L34P substitution. Had processing been affected, we might have observed an increase in Aβx-42 production, but Aβx-42 levels from F19D/L34P were similar to those produced by F20P. This led us to suspect that binding of the end-specific Aβ40 antibody was disrupted by the L34P substitution. These results suggest that viral transduction with our minimal expression vector achieved cell membrane localization and γ-secretase cleavage, releasing variant Aβ peptide into the brain.

Figure 3.

Neonatal (P0) AAV injection produces neuronal expression of variant Aβ

AAV encoding either F19D/L34P or F20P Aβ was injected into the lateral ventricles of WT P0 mice. 3 weeks or 7.5 months later, mice were harvested for immunostaining and/or ELISA analysis. (A) Anti-human Aβ immunostaining (6E10, green) demonstrates widespread viral expression in the cortex of this sagittal section harvested 3 weeks after P0 injection of AAV-F20P. (B) Co-staining for virally delivered variant Aβ (F20P; 6E10, green) and endogenous mouse APP (Y188, red) demonstrates good concordance, suggesting membrane delivery of the variant peptide in cortical neurons. (C) Mass spectrometry of immunoprecipitated Aβ confirms the production of full-length variant Aβ in vivo. Mice expressing Aβ F20P were used to isolate peptide from brain homogenate by 6E10 immunoprecipitation. MS1 spectra of the eluted peptides display the expected mass for Aβ42 F20P (shown here) and Aβ40 F20P (data not shown). The peak for intact Aβ42 F20P at m/z = 893.2675 with a +5 charge state (δ = 0.2) indicates a monoisotopic mass of 4,461.3011 Da, whereas the most abundant isotope configuration was 4,463.3136 Da and is within 15 ppm error of the expected mass. (D) MS2 fragmentation of the peptide shown on the left confirms the sequence identity of Aβ42 F20P, with a fragment ion tolerance of 20 ppm error. (E) Human Aβ was detected by ELISA in the soluble fraction of frontal cortex homogenates from WT mice euthanized at 3 weeks of age. Both variants produced human Aβ42; however, Aβ40 was only detected in mice transduced with F20P. (F) Expression of variant human Aβ could still be detected by ELISA 7.5 months after P0 viral injection. Values are for the soluble fraction of frontal cortex homogenates from non-transgenic (NTG) animals. Note that the absolute values of (C) and (D) cannot be directly compared, as the assays were performed at different times using kits from different manufacturing lots. (G) The ratio of Aβ40:42 produced by each variant remained constant between 3 weeks (upper panel) and 7.5 months (lower panel). 3 weeks: n = 9−10 per condition; 7.5 months: n = 2 uninjected; n = 4−5 F19D/L34P; and n = 4−6 F20P. ANOVA, ∗p < 0.05, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Data are presented as mean ± SEM.

Virally delivered variant peptides reduce Aβ load in APP/PS1 transgenic mice

We then asked whether variant Aβ peptides might be used to alleviate Aβ pathology in a mouse model of Alzheimer's amyloidosis. We injected virus encoding Aβ F20P or F19D/L34P into P0 APP/PS1 mice and non-transgenic (NTG) littermates. Animals were harvested at 7.5 months of age, shortly after the onset of amyloid deposits in this model.²⁸ Cortical tissue from NTG siblings was used to confirm that virally expressed Aβ was still produced at this age (Figure 3F). AAV-injected NTG mice showed the same relative levels of Aβ40:42 as at 3 weeks, with elevated x-42 for both variants and elevated x-40 only detected from F20P (Figure 3G).

In APP/PS1 animals, plaque burden was assessed by Aβ immunostaining in the frontal cortex of one hemisphere and by Aβ ELISA in the other. Compared with uninjected APP/PS1 control mice, mice treated with F20P showed an almost 75% decrease in Aβ deposition (Figures 4A and 4B). In agreement with immunostaining, levels of insoluble Aβx-40 and Aβx-42 in mice treated with F20P were also markedly decreased (Figure 4C). The effects of F19D/L34P expression in APP/PS1 mice were more variable than F20P. Insoluble Aβx-40 levels were significantly reduced in APP/PS1 mice expressing F19D/L34P; however, neither Aβx-42 nor plaque load was significantly diminished (Figures 4B and 4C). Thus, although both F20P and F19D/L34P significantly diminished the total accumulation of Aβ, F20P was clearly more effective at inhibiting aggregation in APP/PS1 mice. Importantly, neither F20P- nor F19D/L34P-treated NTG animals accumulated detectable levels of human Aβ over 7.5 months of expression (data not shown; total Aβ40 + 42, soluble + insoluble, uninjected 100.7 ± 45.5 [n = 6]; F19D/34P 72.3 ± 23.6 [n = 11]; F20P 158.2 ± 35.2 pg/mL [n = 10]; ANOVA F(2, 24) = 2.02, p = 0.15).

Figure 4.

Lifelong expression of variant Aβ reduces plaque load and Aβ accumulation in APP/PS1 mice

APP/PS1 mice were injected at P0 with AAV encoding Aβ F19D/L34P or F20P and harvested 7.5 months later. (A) Aβ immunostain reveals decreased plaque accumulation in mice treated with variant Aβ peptide. (B) Cortical plaque load measured as percent Aβ area confirms that F20P mice harbored less amyloid than untreated mice. n = 5 uninjected, n = 4 F19D/L34P, n = 8 F20P. (C) Meso Scale Discovery (MSD) ELISA for human Aβ peptide in guanidine extracts of cortical tissue echoes the plaque histology. Aβ40 levels were reduced by both variants, whereas F20P also reduced Aβ42 levels. n = 8 uninjected, n = 5 F19D/L34P, n = 12 F20P. ANOVA, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Data are presented as mean ± SEM.

Variant Aβ expression decreases overall neuroinflammation in APP/PS1 mice but may induce mild astrocytosis

Amyloid plaques elicit a pronounced neuroimmune response in which hypertrophic astrocytes and microglia migrate or divide to surround cored deposits. Under most conditions, the extent of glial induction parallels the severity of amyloid load. Generally speaking, treatments that slow plaque accumulation also temper these changes in glial morphology and localization. We therefore examined whether the clustering of hypertrophic astrocytes and microglia that normally delimits cored Aβ deposits would decrease alongside plaque load in mice treated with variant Aβ. We focused our analysis on the F20P variant since it had a more pronounced effect on plaque load than F19D/L34P. Both glial fibrillary protein (GFAP)-positive astrocytes and Iba1-positive microglia prominently outline amyloid plaques in the cortex of untreated APP/PS1 mice (Figures 5A and 5B). Indeed, the surface area of both GFAP and Iba1 exceeds that of Aβ immunostaining in untreated transgenic mice (Figure 5C). Lifelong treatment with virally delivered Aβ F20P significantly diminished staining for both of these markers, suggesting that neuroinflammation decreased with amyloid load. However, we also noted an unexpected increase of cortical GFAP staining in virally injected NTG mice compared with uninjected controls (Figures 5A and 5C). The GFAP-positive astrocytes were largely restricted to a band neighboring the corpus callosum, and the effect was specific for astrocytes: Iba1 levels in NTG mice were unchanged by viral exposure (Figures 5B and 5C). The same pattern of peri-callosal GFAP staining can be discerned in F20P-treated APP/PS1 mice; however, this area also contained a plaque-associated astrocytosis that confounded quantitation (Figure 5A). We suspected that this reaction might arise from non-specific transduction of astrocytes nearest to the ventricle, as our past studies have shown a similar pattern of astrocytic transduction by AAV8 following delayed P0 viral injection.²⁷ To test this prediction, we co-immunostained for human Aβ (6E10) and GFAP in 7.5-month-old NTG mice. Although we detected a few astrocytes that expressed human Aβ, these co-labeled cells comprised only a small fraction of the GFAP+ population. These data suggest that whereas viral expression of the variant Aβ peptide may cause mild astrogliosis on its own, the net effect in a model of amyloidosis is to abate the severity of chronic neuroinflammation commensurate with the reduction in plaque load.

Figure 5.

Variant Aβ peptide diminishes reactive gliosis commensurate with plaque load

Both APP/PS1 and WT (NTG) mice were harvested 7.5 months after intracerebroventricular (i.c.v.) P0 viral injection to deliver the F20P variant Aβ. Uninjected siblings were used for comparison. (A and B) GFAP immunostaining was used to detect astrocytes; Iba1 to detect microglia. Fluorescent immunostaining, shown in the bottom row, was counterstained with Thioflavin-S to detect amyloid plaques. Both the size of glial foci and the number of surrounding cells were decreased by F20P treatment in APP/PS1 mice compared with uninjected animals (upper and lower rows). In contrast, viral injection had no impact on microglial morphology or density in NTG mice but increased the number of GFAP+ cells along the corpus callosum (middle row). (C) Quantification of the colorimetric immunostains for GFAP and Iba1 confirmed the qualitative findings that F20P treatment diminished the area of glial staining commensurate in APP/PS1 mice but elevated GFAP levels in NTG mice compared with uninjected controls. (D) Viral expression of human Aβ was detected in a subset of forebrain astrocytes. Co-immunofluorescence for human Aβ (6E10, green) and GFAP (red) detected sparse co-labeled cells (arrows) in NTG mice injected with F20P. APP/PS1, n = 5 uninjected; n = 7−8 F20P; NTG, n = 5−6 uninjected; n = 5 F20P. A small portion of GFAP+ astrocytes (red) co-labeled with anti-human Aβ antibody 6E10 (green), suggesting they had been transduced at P0 by AAV injection. Two-way ANOVA, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001. Data are presented as mean ± SEM.

Discussion

The past few decades have seen considerable progress in designing β-sheet breakers and blockers to prevent or reverse the aggregation of Aβ. Synthetic chemicals, short peptides, and peptide mimetics have been tested against nearly every portion of the Aβ peptide.^3,12,29,30 Yet, compared with the vast number of inhibitors described in the literature, relatively few have been carefully examined in vivo.^4,10,12,31, 32, 33, 34, 35 We set out to bridge this gap. Our study began in vitro to identify peptide inhibitors that met four therapeutic criteria and ended in vivo by vetting our top two peptide candidates in an animal model of Aβ amyloidosis. Our work yielded the counterintuitive finding that peptides differing from pathogenic Aβ by just one or two amino acids effectively diminished Aβ fibrilization, destabilized existing fibrils, and lowered oligomer toxicity over short periods in vitro, while reducing amyloid aggregation and neuroinflammation when used chronically in vivo.

Our initial peptide candidates were chosen from past work testing specific amino acid substitutions on Aβ aggregation. Of particular importance to our thinking were studies examining proline-scanning mutations, which identified amino acids 18−20 within the central hydrophobic domain as influencing Aβ self-recognition, oligomerization, and cytotoxicity.19, 20, 21^,36, 37, 38 This work prompted three of our five peptide candidates: V18P, F19P, and F20P. In addition to these proline structural variants, we also tested how introduction of a negatively charged residue into the hydrophobic domain would affect polymerization in the F19D variant. Finally, we examined the potential for multiplexing both charge and structural variants through a combination of F19D with the proline mutation L34P. The L34P variant had previously been shown to impair fibrilization both on its own and in combination with the F19S substitution.39, 40, 41 Each of the residues that we targeted resides just inside of or immediately adjacent to β-sheet stretches of fibrillar Aβ, within hydrophobic clusters thought to stabilize β-sheet conformation.^16,42

Based on prior studies, we expected and found that none of our variants self-aggregated, even when incubated for more than 2.5× longer than it took for WT Aβ42 to fully aggregate. However, two outcomes did surprise us. First, whereas four of the five variant peptides diminished aggregation of WT Aβ42 in competition assays, V18P unexpectedly exacerbated aggregation, suggesting that it incorporated into the growing fibrils alongside the WT peptide. Past work had shown that this peptide was highly resistant to self-aggregation but had not examined co-incubation with WT Aβ.^20,37 This outcome obviously ruled out further development of V18P as a therapeutic candidate. For the purposes of therapeutic development, our findings with F19D/L34P were even more revealing. This peptide appeared to be our best candidate in vitro. It thwarted fibrillization of WT Aβ and efficiently disassembled pre-formed Aβ fibrils in a dose-dependent manner. We were therefore surprised to find that this variant was considerably less effective than F20P in vivo, which had displayed less robust fibril disassembly than F19D/L34P in vitro. At least two factors may have contributed to this discrepancy between in vitro and in vivo efficacy. The first and most obvious difference between test tube and brain is the reductive nature of the ThT assays. Conditions in vitro poorly reflect the range of Aβ species and cellular environment found in vivo and here, may not have been an accurate predictor of in vivo efficacy. An alternative, but not mutually exclusive explanation, is that the behavior of F19D/L34P may have been affected by the unexpected absence of Aβ40 in vivo. We do not know if the F19D/L34P Aβ40 peptide was produced but not detected by our end-specific ELISA antibody or instead, if the L34P mutation altered β-CTF processing to prevent Aβ40 cleavage. Taken at face value, the absence of F19D/L34P Aβ40 suggests that the ratio of monomeric Aβ species generated from the variant expression constructs may influence their efficacy in vivo (i.e., Aβ40:42, Aβ38:42, etc.). We instead suspect that the total amount of variant peptide produced in vivo is critical to efficacy, regardless of length, and that the absence of Aβ40 production by F19D/L34P significantly diminished the amount of variant Aβ available to inhibit aggregation. This discrepancy between in vitro promise and in vivo performance underlines the importance of animal testing in preclinical assessment of candidate aggregation inhibitors.

Animal model testing of peptide drugs for Alzheimer’s disease has been severely hampered by two key weaknesses: short half-life in vivo and low brain penetrance.^14,43,44 We overcame these limitations using AAV to deliver and continuously express our variant Aβ constructs directly within the central nervous system (CNS). AAV vectors are being widely developed for gene therapy in peripheral tissues, with two AAV therapies now approved by the US Food and Drug Administration for CNS disorders.^45,46 Nevertheless, AAV is not a panacea for gene delivery in all cases. Manufacturing for AAV is currently not scaled for common disorders, such as Alzheimer’s disease; approved AAV treatments target only rare genetic disorders and come with a staggering cost.⁴⁷ Preexisting immune response to AAV can hamper therapeutic efficacy if not identified and managed properly.^48,49 Perhaps most critically for Alzheimer’s disease, targeting AAV to the brain is relatively inefficient, even with newly evolved serotypes.^50,51 We averted this issue by direct injection into the P0 brain; achieving similar delivery in the adult brain will be more challenging. Although we anticipate future experiments in mice using brain-penetrant serotypes, such as PHP.eB, to test the efficacy of variant Aβ peptides introduced after amyloid onset, these vectors may be less effective for neuronal transduction in humans.51, 52, 53, 54 Given the promise of gene therapy for CNS disorders, we anticipate that even this obstacle will eventually be overcome either through new serotypes or new delivery methods.

Ensuring the absence of immune reaction due to AAV exposure is particularly critical for neurological therapies. Although treatment with F20P Aβ significantly diminished both astro- and microgliosis associated with amyloid pathology in APP/PS1 mice, we unexpectedly found that WT mice expressing variant peptides showed a mild astrocytic reaction. This reaction could also be appreciated in APP/PS1 mice but could not be as clearly discerned due to plaque-associated astrocytosis in the same area. AAV8 has the ability to transduce several cell types in the brain, including astrocytes.^55,56 A fraction of this peri-callosal astrocytosis was due to unintended transduction of astrocytes by P0 AAV injection and may be alleviated in future studies by using a neuron-specific promoter in place of the non-selective CAG element.^57,58 However, this explanation accounts for only a fraction of the observed reaction, as GFAP+ astrocytes, which lacked Aβ expression, far outnumbered those that co-labeled for 6E10. An alternative explanation may be that astrocytes in this region react to variant Aβ secreted from neighboring neurons. If so, the response was not specific to F20P and was also seen in NTG animals treated with F19D/L34P (data not shown). Future studies will be needed to test whether neuronal expression of any virally delivered protein elicits this glial reaction (i.e., using an “inert” label, such as GFP or lacZ), or if the injection procedure itself without virus causes a similar response. Although pericallosal astrocytosis has been observed in mouse models of hydrocephalus,59, 60, 61 it is hard to envision how a transient increase in ventricular pressure that may occur following viral injection at P0 could elicit such a persistent phenotype.

Another issue that we feel must be addressed in future studies is the long-term consequence of introducing excess Aβ into the brain. Here, we demonstrate that variant Aβ expression did not cause any sign of peptide accumulation in WT mice over 7.5 months’ time. Similarly, expression of variant Aβ in APP/PS1 mice diminished rather than exacerbated human Aβ levels with age. Although 7.5 months is considerably longer than past in vivo studies of peptide aggregation inhibitors,^4,10,31, 32, 33, 34, 35 it is far shorter than would be needed for treatment of human disease where a potential intervention must be safe over decades of exposure. Obviously, mice will not support such long-term studies, but future experiments could test longer periods of viral expression, up to 2 or 2.5 years in rodent models or even longer in non-human primates. A related concern is defining what small-n Aβ aggregates are formed by variant Aβ peptides in a WT Aβ environment. We do not know at a mechanistic level how F20P and F19D/L34P prevent aggregation of WT Aβ nor how they promote disassembly of pre-formed fibrils. We show here that F20P and F19D/L34P both diminish the cytotoxicity of oligomers formed by WT Aβ42, but we do not know what species they form instead. Future studies must now define the biophysical properties of Aβ aggregates formed by mixtures of our variant and WT peptides to examine their stability and propensity for alternative aggregation pathways before we can make informed decisions about potential safety for chronic use. Finally, we must consider whether variant peptides may be differentially effective at distinct stages of disease. For example, might the efficient in vitro fibril dissociation displayed by F19D/L34P make it better suited for use after plaque onset than in the preventative setting tested here?

The work presented here takes the first steps in applying viral-mediated gene therapy to overcome past limitations on advancing peptide Aβ inhibitors from the bench to the brain. The broader impact of our work lies in the potential for applying this strategy to other protein misfolding diseases where peptide treatments have been eschewed for technical reasons that can now be overcome through expression engineering and viral technology. Many neurodegenerative diseases involve the aggregation of misfolded proteins into toxic species that provoke neurodegeneration. Although each disease is canonically associated with distinct protein aggregates, the molecular mechanisms of misfolding, oligomerization, and fibrilization can be strikingly similar.⁶² This may reveal a common vulnerability that can be similarly exploited with targeted peptide inhibitor therapies to prevent aggregation or promote dissolution of toxic protein aggregates. The experiments described here provide a roadmap from in vitro characterization to expression constructs for preclinical testing in animal models that we hope will support development of additional inhibitors for other neurodegenerative proteinopathies.

Materials and methods

Preparation of WT Aβ42 or variant peptide stocks

Synthetic Aβ42 WT or variant peptides were purchased from Biomatik (Wilmington, DE, USA). To prepare stock solutions of aggregate-free Aβ42 WT or variant peptides, powdered peptide was dissolved in 50% acetonitrile, frozen, and lyophilized overnight to remove any residual trifluoroacetic acid. Lyophilized Aβ42 peptides were dissolved in hexafluoro-2-propanol (HFIP; #105228; Sigma-Aldrich, St. Louis, MO, USA). The Aβ-HFIP solution was incubated at room temperature (RT) for 30 min and then divided into aliquots. HFIP was evaporated overnight in a fume hood and then transferred to SpeedVac for 1 h to remove any remaining traces of HFIP. Tubes containing the peptide film were kept over desiccant at −20°C until used. Immediately prior to experimental use, lyophilized peptides were dissolved in DMSO to a final concentration of 5 mM and sonicated for 10 min in a bath sonicator.

Preparation of oligomeric and fibrillar WT Aβ

Oligomeric Aβ was generated by dissolving peptide in Ham’s F-12 media (#30611040-1; Fisher Scientific, Pittsburgh, PA, USA) to a final concentration of 100 μM WT Aβ42 peptide or 100 μM WT + 100 μM variant and then incubating at 4°C for 24 h without shaking. Fibrillar Aβ was generated by dissolving WT peptide in PBS to a final concentration of 100 μM and incubating at 37°C for 24 h without shaking.

ThT assay to test kinetics of Aβ self-aggregation, competition with WT peptide, and fibril disassembly

Self-aggregation of Aβ42 WT or variant monomers was tested at a starting concentration of 10 μM in PBS containing 5 μM ThT. ThT fluorescence was measured using an Infinite M1000 Pro Plate Reader (Tecan, Mannedorf, Switzerland) at an excitation wavelength of 440 nm and emission wavelength of 485 nm. Reactions were incubated without shaking at 37°C and then shaken for 5 s prior to reading fluorescence. Competition between variant and WT Aβ was performed similarly but using a starting concentration of 10 μM for each peptide in the mixture (1:1, WT:variant), with the exception of WT alone, which was tested at a final concentration of 10 μM. Fibril disassembly was assessed by mixing 10 μL of WT Aβ42 fibrils (described above) in a 1:1 ratio with 5, 10, or 20 μM monomeric F19D/L34P or F20P peptides in PBS containing 5 μM ThT. Fluorescence was measured without shaking every 24 h for 48 h of incubation at 37°C.

Oligomeric Aβ toxicity assay

N2a cells were grown in Eagle’s minimal essential medium (EMEM) (#112-018-101; VWR, Radnor, PA, USA), supplemented with 1 × 10⁴ U/mL penicillin/streptomycin (#15140-122; Life Technologies, Carlsbad, CA, USA) and 10% fetal bovine serum (FBS; #MT35010CV; Fisher Scientific) at 37°C in 5% CO₂. Confluent cells were trypsinized, diluted in EMEM containing 1% N2 supplement to minimize cell growth, and then plated 5,000 cells/well in transparent flat-bottom, 96-well plates (#07-200-89; Fisher Scientific). 10 μL of each oligomer preparation (WT Aβ alone or WT + variant; described above) or 10 μL of a 1:10 dilution was added to 90 μL of culture medium. Cell viability was determined 24 h after treatment using an MTS assay (#G3582; Promega, Madison, WI, USA), according to the manufacturer’s directions. Briefly, assays were performed by adding 20 μL of AQueous One Solution Reagent directly to culture wells, incubating for 2 h at 37°C in a 5% CO₂ atmosphere, and then recording absorbance at 490 nm using an Epoch 2 spectrophotometer (BioTek, Winooski, VT, USA).

Plasmid constructs

PCR was used to add the GLSP (amino acids MGVKVLFALICIAVAEA) onto the N terminus of Aβ-CTF. First, a synthetic DNA for GLSP with a partial Aβ sequence was made using oligo GLSP-1 5′-ATGGGCGTGAAGGTCCTGTTCGCCCTGATTTGCATCGCCGTCGCAGAGGCAGATGCAGA-3′ and oligo GLSP-2 5′-TCTGCATCTGCCTCTGCGACGGCGATGCAAATCAGGGCGAACAGGACCTTCACGCCCAT-3′. These oligonucleotides were annealed by incubation in a thermocycler programmed to start at 95°C for 2 min and then gradually cool to 25°C. Second, Aβ-CTF carrying a partial GLSP sequence was generated by amplifying a plasmid containing the human APP WT sequence (pBS-hAPPWT-IRES-GFP) with forward primer (F) 5′-GTCGCAGAGGCAGATGCAGAATTCCGACATGAC-3′ and reverse primer (R) 5′-GCGCGGATATCCTAGTTCTGCATCTGCTCAAAG-3′. GLSP-Aβ-CTF was then generated by Gibson assembly from the two templates, joining the GLSP + partial Aβ to the Aβ-CTF + partial GLSP, using a forward primer to add a Kozak sequence 5′-GCGCGAAGCTTGCCACCATGGGCGTGAAGGTCCTGTT-3′ and reverse primer 5′-GCGCGGATATCCTAGTTCTGCATCTGCTCAAAG-3′. The resulting GLSP-Aβ-CTF fragment was digested with HindIII and EcoRV and subcloned into pAAV containing the CAG promoter, Woodchuck hepatitis virus post-transcriptional regulatory element (WPRE), and AAV2 inverted terminal repeat (ITR) sequences to create pAAV-GLSP-Aβ-CTF.

The GLSP-Aβ-CTF deletion series was constructed by cloning various GLSP-Aβ-CTF deletions into pAAV. The GLSP-Aβ-CTF deletion series was amplified by PCR from the full-length pAAV-GLSP-Aβ-CTF using a common forward primer with a series of reverse primers and then digested with HindIII and EcoRV to ligate into pAAV.

Forward primer

• The forward primer used was Aβ-GLSP-forward (Kozak): 5′-GCGCGAAGCTTGCCACCATGGGCGTGAAGGTCCTGTT-3.

Reverse primers

The reverse primers used were Aβ-reverse (KK): 5′-GCGCGGATATCTTACTACTTCTTCAGCATCACCAAGGTG-3; Aβ-reverse (ML): 5′-GCGCGGATATCTTACTACAGCATCACCAAGGTGATGA-3; Aβ-reverse (LV): 5′-GCGCGGATATCTTACTACACCAAGGTGATGACGATCA-3; Aβ-reverse (IT): 5′-GCGCGGATATCTTACTAGGTGATGACGATCACTGTCG-3; Aβ-reverse (IV): 5′-GCGCGGATATCTTACTAGACGATCACTGTCGCTATGA-3; and Aβ-reverse (IA): 5′-GCGCGGATATCTTACTACGCTATGACAACACCGCCCA-3.

To build pAAV-GLSP-Aβ(F19D/L34P)-KK containing the F19D/L34P substitution, two PCR reactions were performed. A fragment of GLSP-Aβ(F19D/L34P)-KK was amplified from pAAV-GLSP-Aβ-KK using forward primer 5′-GCGCGAAGCTTGCCACCATGGGCGTGAAGGTCCTGTT-3′ and reverse primer 5′-ACCATGGGTCCAATGATTGCACCTTTGTTTGAACCCACATCTTCTGCAAAGTCCACCAA-3′. A second fragment of GLSP-Aβ(F19D/L34P)-KK was amplified using forward primer 5′-TTGGTGGACTTTGCAGAAGATGTGGGTTCAAACAAAGGTGCAATCATTGGACCCATGGT-3′ and reverse primer 5′-GCGCGGATATCTTACTACTTCTTCAGCATCACCAAGGTG-3′. The resulting fragments were joined by Gibson assembly using forward primer 5′-GCGCGAAGCTTGCCACCATGGGCGTGAAGGTCCTGTT-3′ and reverse primer 5′-GCGCGGATATCTTACTACTTCTTCAGCATCACCAAGGTG-3′. The resulting insert was digested with HindIII and EcoRV and subcloned into pAAV to create pAAV-GLSP-Aβ(F19D/L34P)-KK.

For pAAV- GLSP-Aβ(F20P)-KK containing the F20P substitution, two PCR reactions were performed. A fragment of GLSP-Aβ(F20P)-KK was amplified from pAAV-GLSP-Aβ-KK using forward primer 5′-GCGCGAAGCTTGCCACCATGGGCGTGAAGGTCCTGTT-3′ and reverse primer 5′-TCTTCTGCAGGGAACACCAATTTTTG-3′. A second fragment of GLSP-Aβ(F20P)-KK was amplified using forward primer 5′-CAAAAATTGGTGTTCCCTGCAGAAGA-3′ and reverse primer 5′-GCGCGGATATCTTACTACTTCTTCAGCATCACCAAGGTG-3′. The resulting fragments were joined by Gibson assembly using forward primer 5′-GCGCGAAGCTTGCCACCATGGGCGTGAAGGTCCTGTT-3′ and reverse primer 5′-GCGCGGATATCTTACTACTTCTTCAGCATCACCAAGGTG-3′. The resulting insert was digested with HindIII and EcoRV and subcloned into pAAV to create pAAV-GLSP-Aβ(F20P)-KK. All restriction enzymes were purchased from New England Biolabs (Ipswich, MA, USA).

Collection of Aβ peptides from N2a conditioned media

N2a cells were grown in 6-well plates (for Western) or 10 cm dishes (for MS) using Dulbecco’s modified Eagle’s medium (DMEM) (#12-604F; VWR), supplemented with 1 × 10⁴ U/mL penicillin/streptomycin and 10% FBS until approximately 90% confluent. Cells were then transfected with 2.5 μg/well (6-well) of sequentially deleted APP-CTF sequences (e.g., pAAV-GLSP-AβWT-KK) using Lipofectamine LTX (#15338030; Fisher Scientific) or with 14 μg/dish (10 cm) of pAAV-GLSP-Aβ(F20P)-KK. 24 h later, cells were washed with 1× Dulbecco’s PBS, and a serum-reduced medium consisting of DMEM and 0.2% FBS was added before the cells were returned to 5% CO₂ at 37°C. Conditioned media were harvested 24 (Western) or 48 h later (MS) and centrifuged at 10,000 rpm for 5 min at 4°C. To test the dependence of Aβ release on γ-secretase activity, cells were transfected as above, with the subsequent media replacement including 0.1 or 1 nM of LY411575 (#SML0506; Sigma). Supernatants were collected and supplemented with protease inhibitors (#5892970001; Sigma) and 0.01% NaN₃ (#S2002; Sigma) to prevent proteolytic degradation.

Aβ IP from conditioned media for immunoblot

Protein G Dynabeads (50 μL; #10003D; Fisher Scientific) were loaded with 1 μg of mouse anti-Aβ antibody 6E10 (#SIG-39320; BioLegend, San Diego, CA, USA) for 1 h at RT. Conditioned media from transfected N2a cells (6 mL) were incubated with the Dynabead-6E10 complex in a rotator overnight at 4°C. The following day, beads were collected using magnetic separation to remove the supernatants. Beads were washed 3× with PBS containing 0.02% Tween-20. Immunoprecipitated complexes were eluted with 50 mM glycine (pH 2.8). Samples were then denatured with an equal volume of 2× Laemmli sample buffer for 10 min at 70°C and electrophoresed on 16.5% Criterion Tris-Tricine Gels (#3450063; Bio-Rad, Hercules, CA, USA). Proteins were transferred to nitrocellulose using a Trans-Blot Turbo Transfer System (#170-4159; Bio-Rad). Membranes were blocked in PBS containing 0.1% Tween-20 and 5% non-fat dry milk for 1 h at RT and probed overnight at 4°C with 6E10 diluted 1:5,000 in blocking solution. Antibody binding was detected using mouse anti-immunoglobulin G (IgG) secondary antibody conjugated with IRDye, diluted 1:20,000 in block (#26-32210; LI-COR Biosciences, Lincoln, NE, USA). Blots were imaged with an Odyssey Fc Imager and analyzed with Image Studio software (LI-COR Biosciences).

Aβ IP from conditioned media for MS

Approximately 60 mL of N2a conditioned media containing Aβ F20P was incubated with 10 μg mouse anti-Aβ antibody 6E10 (#SIG-39320; BioLegend, San Diego, CA, USA), divided into two batches of 30 mL each, on a rotator overnight at 4°C. The following day, Protein G Dynabeads (100 μL; #10003D; Fisher Scientific) were added to the media and incubated for 2 h at 4°C. Beads were collected with magnetic separation to remove the supernatant, then washed 3× with PBS containing 0.02% Tween-20, followed by two washes in PBS without detergent. Finally, immunoprecipitated complexes were eluted in 80 μL of 10% formic acid solution for MS analysis.

MS

The eluted immunoprecipitates from brain homogenate and culture media expressing Aβ F20P were dried with a SpeedVac and resuspended in 50 μL of 20% formic acid. 1 μL of IP reaction was loaded onto a 10-cm, 100-μm inner-diameter C3 column (ZORBAX 300SB-C3; 300 Å 5 μm), self-packed into fused silica, pulled to form a nanoelectrospray emitter. Online high-performance LC (HPLC) was performed on a Thermo Scientific U3000 RSLCnano ProFlow system. A 50-min linear gradient from 0% buffer B to 35% buffer B, using buffer A: 2% acetonitrile, 0.1% formic acid, and buffer B: 98% acetonitrile and 0.1% formic acid. The column eluant was introduced into a Thermo Scientific Orbitrap Fusion Lumos by nanoelectrospray ionization. A static spray voltage of 2,200 V and an ion transfer tube temperature of 320°C were set for the source.

MS1 was performed by the Orbitrap at a 60-k resolution setting, in positive mode with quadrupole isolation. An automatic gain control (AGC) target of 5.0e6 with 200 ms maximum injection time, two microscans, and a scan range of 350−2,000 mass-to-charge ratio (m/z) were used. Target precursor m/z selected for MS2 fragmentation included monoisotopic and most abundant masses for +4,+5, and +6 Aβ F20P 38, 40, and 42 amino acid unmodified ions. Higher-energy collisional dissociation (HCD) fragmentation with a normalized collision energy of 45% was used. MS2 acquisition was performed using the Orbitrap with the 15-k resolution setting, an AGC target of 1e6, a max injection time of 100 ms, a scan range of 150−2,000 m/z, and three microscans. Next, we specifically targeted the masses of α-secretase-cleaved Aβ 1−16 along with an uncleaved and partially cleaved precursor (with/without signal peptide and residual TM domain) for MS analysis to enhance sensitivity for these targets. The parameters were duplicated with a targeted mass list containing these alternatively cleaved peptides.

Data processing was performed by a custom analysis suite, as described previously.^63,64 The Aβ F20P 38, 40, and 42 peptides, α-secretase-cleaved Aβ 1−16, along with an uncleaved and partially cleaved precursor peptide (with/without signal peptide and residual TM domain) were used as template sequences for the search. Representative spectra for unmodified Aβ F20P are shown.

Viral packaging

All AAVs were prepared by the Gene Vector Core at Baylor College of Medicine using a method similar to one previously described.⁶⁵ HEK293T subclone 1F11 cells were grown in DMEM (#CM002-050; GenDEPOT, Barker, TX, USA), supplemented with 10% FBS (#97068-085; VWR) and 1× antibiotic/antimycotic (#CA002-010; GenDEPOT). Serotype 8 AAV was prepared by co-transfection of three plasmids (expression vector [1.14 μg/15 cm plate], p5E18-VD2/8 Rep-Cap plasmid [4.57 μg/plate], and pAdDF6 helper plasmid [2.29 μg/plate]) using 24 μL/plate of iMFectin Poly DNA Transfection Reagent (#I7200-101; GenDEPOT). AAV purification was performed using a protocol based on Ayuso et al.⁶⁶ but with 15% iodixanol containing 0.75 M NaCl. 3 days after transfection, cells were collected while the media were retained for subsequent polyethylene glycol (PEG) precipitation. The cell pellet was re-suspended in 1 mL per plate of 50 mM Tris (pH 8.0) containing 5 mM MgCl₂ and 0.15 M NaCl, lysed by adding 0.1 vol of 5% sodium deoxycholate for 30 min at RT, and then incubated with 10 μg/mL of DNase I and RNase A for 1 h at 37°C. Cell lysates were clarified by centrifugation at 5,000 × g for 10 min at 4°C. The culture media were incubated with 10 μg/mL of DNase I and RNase A for 1 h at 37°C and then incubated overnight at 4°C in 8% PEG (stock 40% PEG 8,000 plus 2.5 M NaCl). AAV was collected from this mixture by centrifugation at 2,500 × g for 30 min. The pellet containing AAV was resuspended in a minimal volume of HEPES-buffered saline (HBS; 50 mM HEPES, 0.15 M NaCl, 1% sarcosyl, and 20 mM EDTA, pH 8.0). Cell-associated and secreted AAV preparations were combined for iodixanol density centrifugation. The resulting AAV particles were dialyzed against Mg- and Ca-free PBS using an Amicon Ultra-15 centrifugal filter (100,000 kDa nominal limit; Millipore, Burlington, MA, USA) and the titer determined by real-time PCR.

Mice

WT ICR animals, purchased from the Center for Comparative Medicine at Baylor College of Medicine, were used generate offspring for P0 viral injection. WT pups from these litters were virally injected at P0, as described below, and harvested to assess neocortical viral expression at 1 month of age. Male APPswe/PS1dE9 bigenic mice were obtained from the Mutant Mouse Resource and Research Center at Jackson Laboratory (stock #34832-JAX, B6.Cg-Tg(APPswe,PSEN1dE9)85Dbo/Mmjax). These were mated with WT C57BL/6J mice to establish a backcross colony or with FVB/NJ females to generate first filial generation (F1) offspring for study. Both male and female, NTG (WT) and APP/PS1-positive offspring were used for P0 viral injection to evaluate variant Aβ expression at 7.5 months of age. All animal experiments were reviewed and approved by the Baylor College of Medicine Institutional Animal Care and Use Committee and conform to relevant regulatory standards.

P0 intraventricular injections

Stereotaxic injection of AAV into the lateral ventricles of P0 mouse pups was performed as described previously.^67,68 Within 6 h after birth, neonates were collected from the cage and prepared for injection by cryoanesthesia. Following cessation of movement, viral solutions of 4 × 10⁶ transducing unit (TU)/μL, diluted in sterile PBS containing 0.05% trypan blue, were injected into the lateral ventricles using a 10-μL syringe (Hamilton, Reno, NV, USA; #7653-01) fitted with a 32-gauge needle (Hamilton; #7803-04, RN 6PK PT4). Two sites per hemisphere were injected with 1 μL of viral solution per site using a P0 stereotaxic device (X, Y, Z) = (±0.8, ±1.5, −1.5 mm) and (±1.35, ±2.0, −1.7 mm) from lambda. Injected pups were placed on a warming pad to regain normal color and movement before being returned to their biological mother for care.

Tissue harvest

WT ICR mice were studied for viral spread and expression at 3 weeks of age; transgenic APP/PS1 animals and their NTG siblings were studied for Aβ level, plaque load, and gliosis at 7.5 months of age. Mice were killed by sodium pentobarbital overdose and transcardially perfused with PBS and heparin. Brains were removed and dissected along the midline. The rostral half of the left cortex was snap frozen on dry ice for biochemistry. The right hemisphere was immersion fixed in 4% paraformaldehyde for 48 h at 4°C, cryoprotected in 30% sucrose at 4°C, and sectioned at 35 μm for histology.

Tissue homogenization

Frozen frontal cortex was sonicated in PBS containing 5 mM EDTA, 1× protease inhibitor (#05892970001; Roche, Basel, Switzerland), and 1× PhosSTOP (#04906845001; Roche) and centrifuged at 100,000 × g for 30 min at 4°C. The pellet was resuspended in an equal volume of PBS containing 1% Triton X-100 (PBS-X) and mixed by gentle rotation for 30 min at 4°C. Samples were centrifuged at 100,000 × g for 30 min at 4°C and the supernatant saved as the PBS-X soluble fraction. The pellet was resuspended in an equal volume of 5 M guanidine hydrochloride in 50 mM Tris (pH 6.8) and mixed by gentle rotation overnight at RT. Samples were centrifuged at 16,000 × g for 30 min at RT and the supernatant saved as the guanidine soluble fraction.

Aβ IP from brain extract for MS

Brain extract from WT ICR mice injected with AAV-F20P at P0 and harvested at 3 weeks was used for IP-MS. 500 μL of PBS-X soluble extract was mixed with 500 μL radioimmunoprecipitation assay (RIPA) buffer (PBS containing 5 mM EDTA, 0.5% Igepal, 0.5% sodium deoxycholate, 0.2% SDS, and protease inhibitor) and then incubated with 5 μg mouse anti-Aβ antibody 6E10 overnight at 4°C on a rotating platform. The following day, Protein G Dynabeads (50 μL) were added to the homogenate and incubated for 2 h at 4°C. Beads were collected with magnetic separation to remove the supernatant and then washed 3× with PBS containing 0.02% Tween-20, followed by two washes in PBS without detergent, before the immunoprecipitated complexes were eluted with 40 μL of 10% formic acid solution for MS.

Meso scale discovery assay

The concentration of human Aβx-40 and -42 in frontal cortex extracts prepared in PBS-X (3 weeks of age) or guanidine (7.5 months of age) was measured using Multiplex Aβ Peptide Panel 1 (#K15200E; Meso Scale Diagnostics, Rockville, MD, USA). The assay was performed essentially as instructed by the manufacturer, using the provided 6E10 antibody for capture (human Aβ 1−16) and end-specific 40/42 antibodies for detection. Samples were diluted to stay within the linear range of the assay, requiring dilution of 1:1 for PBS-X and 1:250 for guanidine. Initial dilution of the PBS-X and guanidine fraction was done in PBS containing 1% protease-free BSA (#820451; MP Biomedical, Santa Ana, CA, USA). The final working dilution was prepared with Diluent 35 included in the kit; Aβ blocking reagent was not used. Samples were read on a SECTOR Imager 6000 and concentrations calculated using Discovery Workbench software (Meso Scale Diagnostics).

Immunofluorescence

A 1/12 series of sections was rinsed with Tris-buffered saline (TBS) and blocked with TBS containing 0.1% Triton X-100 (TBS-T) and 10% normal goat (Gt) serum for 1 h at RT before overnight incubation at 4°C with rabbit (Rb) anti-GFAP (Dako; #Z0334), Rb anti-Iba1 (#019-19741; Waco, Richmond, VA, USA), mouse anti-Aβ antibody 6E10 (BioLegend), and/or Rb anti-Aβ antibody Y188 (#ab32136; Abcam, Cambridge, UK), each diluted 1:500 in blocking solution. After several washes in TBS, sections were incubated with Alexa Fluor-568 Gt anti-Rb and Alexa Fluor-488 Gt anti-mouse secondary antibodies (#A-11036 and #A-21121; Life Technologies), diluted 1:500 in block for 2 h at RT. GFAP/Iba1 sections were washed with TBS before being counterstained for 8 min at RT with 0.002% Thioflavin-S, diluted in TBS. Thioflavin-stained sections were washed twice in 50% ethanol, followed by several washes in TBS before being mounted. All sections were mounted onto Superfrost Plus slides (#12-550-15; Fisher Scientific) and coverslipped with Vectashield mounting medium (#H1400; Vector Laboratories, Burlingame, CA, USA).

Immunohistochemistry

For labeling Aβ plaques, free-floating sections were washed of cryoprotectant with TBS, incubated in 88% formic acid for 1 min, and then rinsed with TBS. Endogenous peroxidases were quenched with 0.9% H₂0₂ in TBS-T for 30 min at RT. Following washes in TBS, sections were blocked with 5% normal Gt serum in TBS-T for 1 h at RT. Sections were incubated at 4°C overnight with primary antibody (1:500, Rb anti-Aβ; Thermo Fisher Scientific/Zymed; #71-5800) diluted in blocking buffer. The following day, sections were washed in TBS and incubated with secondary antibody for 2 h at RT (1:500, biotinylated Gt anti-Rb, Vectastain Elite ABC Kit; Vector Labs; #PK-6101). 30 min before use, A + B reagent was made using 50 μL of each solution in 5 mL of TBS. Sections were washed several times in TBS and then incubated in A + B reagent for 90 min at RT. Sections were again washed with TBS and then developed with filtered DAB solution (Sigma; #D4293) and quenched with TBS washes. Sections were mounted on Superfrost Plus slides (Fisher Scientific; #12-550-15) and dried overnight. Slides were processed through an ethanol series (70%, 95%, 100%, xylene) before being coverslipped with Permount (Fisher Scientific; #SP15-100).

For astrocyte and microglial labeling, sections were processed as above with the following modifications. Endogenous peroxidases were quenched using 0.9% H₂0₂ in TBS-T + 0.05% Tween-20 (TBS-TT) for 30 min at RT. Blocking solution consisted of 5% normal Gt serum (NGS) in TBS-TT. Primary antibodies were diluted in TBS-TT (1:1,000, Rb anti-GFAP, Dako; or 1:1,000, Rb anti-Iba1, Wako).

Quantification of Aβ, GFAP, and Iba1 surface area

Tiled images were acquired using a Zeiss Axio Scan.Z1 at 10× magnification (Carl Zeiss AG, Oberkochen, Germany). Exposure time and lamp intensity were constant for all sections. Two sagittal sections between 0.24 and 2.04 mm from bregma were randomly chosen for quantification. The frontal cortex anterior to the hippocampus was outlined as the region of interest (ROI) for analysis using Fiji 1.51W (NIH, USA). Images were converted to 8-bit, and a threshold was applied using the Yen algorithm with the ROI as reference. The percent area above threshold with the ROI was used for quantitation.

Statistics

Statistical comparisons and graphing were done using GraphPad Prism 6.0h. Comparisons of two groups were done by Student’s t test; comparisons of three or more groups were done by one- or two-way ANOVA, followed by Bonferroni post-test. Grubb’s test was used to identify outlier data points (https://www.graphpad.com/quickcalcs/Grubbs1.cfm); this resulted in removal of 1 data point from Figure 3E (F19D/L34P Aβ42); two from Figure 3F (F19D/L34P Aβ40 and F20P Aβ42), one from Figure 3G (7.5 months F19D/L34P), and one from Figure 4B (F20P). Graphs display group mean ± SEM.