Gene therapy for Alzheimer's disease targeting CD33 reduces amyloid beta accumulation and neuroinflammation

Abstract

Neuroinflammation is a key contributor to the pathology of Alzheimer's disease (AD). CD33 (Siglec-3) is a transmembrane sialic acid-binding receptor on the surface of microglial cells. CD33 is upregulated on microglial cells from post-mortem AD patient brains, and high levels of CD33 inhibit uptake and clearance of amyloid beta (Aβ) in microglial cell cultures. Furthermore, knockout of CD33 reduces amyloid plaque burden in mouse models of AD. Here, we tested whether a gene therapy strategy to reduce CD33 on microglia in AD could decrease Aβ plaque load. Intracerebroventricular injection of an adeno-associated virus (AAV) vector-based system encoding an artificial microRNA targeting CD33 (miRCD33) into APP/PS1 mice reduced CD33 mRNA and TBS-soluble Aβ40 and Aβ42 levels in brain extracts. Treatment of APP/PS1 mice with miRCD33 vector at an early age (2 months) was more effective at reducing Aβ plaque burden than intervening at later times (8 months). Furthermore, early intervention downregulated several microglial receptor transcripts (e.g. CD11c, CD47 and CD36) and pro-inflammatory activation genes (e.g. Tlr4 and Il1b). Marked reductions in the chemokine Ccl2 and the pro-inflammatory cytokine Tnfα were observed at the protein level in the brain of APP/PS1 mice treated with miRCD33 vector. Overall, our data indicate that CD33 is a viable target for AAV-based knockdown strategies to reduce AD pathology. One Sentence Summary: A gene therapy approach for Alzheimer's disease using adeno-associated virus vector-based knockdown of CD33 reduced amyloid beta accumulation and neuroinflammation.

Figure 1

Schematic representation of miR control and miR^CD33 vector injections in APP/PS1 mice. (A) AAV vector encoding miR targeting CD33 (miR^CD33). The construct also encodes GFP allowing visualization of transduction efficiency in the brain. CBA, CMV IE enhancer/chicken beta actin hybrid promoter; WPRE, woodchuck hepatitis virus posttranscriptional regulatory element; pA, polyA signal; ITR, inverted terminal repeat. (B) RNA was extracted from a single cell clone HEK293T line stably expressing mouse CD33 that was transfected with miR control or miR^CD33 AAV expression plasmids. RNA was used to measure CD33 expression levels normalized to GAPDH transcripts. CD33 mRNA levels were reduced by 32.11% (****P < 0.0001, unpaired t test with Welch’s correction) in cells transfected with miR^CD33 compared to miR control. Data are represented as mean ± SEM. (C) Schematic representation of treatment: vectors were injected into both lateral ventricles of APP/PS1 mice. (D) Representative images of the GFP signal detected in brain sections from mice injected with control vector or vector encoding miR^CD33. Scale bar represents 1000 μm. See also Supplementary Material, Figure S1.

Figure 2

CD33 mRNA expression and TBS-soluble Aβ levels, but not amyloid plaque burden, are decreased in the brain of APP/PS1 mice by AAV-miR^CD33 (late intervention). (A–D) Eight-month-old mice were injected into the lateral ventricles with AAV-miR control or AAV-miR^CD33 and sacrificed at 11 months of age. (A) RNA was extracted from brain around the injection site and was used to measure CD33 expression levels normalized to GAPDH transcripts. CD33 mRNA levels were decreased by 36.5% (albeit non-significant, P = 0.167) in miR^CD33-treated APP/PS1 mice compared to control mice. (B) and (C). ELISA (B) and Meso Scale Discovery multi-spot assay (C) of Aβ40 and Aβ42 peptides in TBS-soluble (B) or formic acid (FA)-soluble (C) fractions isolated from mouse brain tissue spanning the injection site. (B) Treatment with miR^CD33 compared to control vector leads to decreased levels of TBS-soluble Aβ40 (30.1% decrease, P = 0.1117) and Aβ42 (35.1% decrease, *P = 0.0279) in APP/PS1 mice. (C) Modest but non-significant reductions in FA-soluble levels of Aβ40 and Aβ42 were observed in APP/PS1 mice treated with miR^CD33 versus control vector. (D) Aβ plaque load quantitated stereologically in brain sections that were labeled with an antibody directed against Aβ. Plaque area and plaque density were assessed in both mouse groups, and data were normalized to the miR control group. Symbols in graphs represent individual animals (n = 10 for miR control, n = 9 for miR^CD33). Data are represented as mean ± SEM.

Figure 3

Early intervention therapy decreases CD33 mRNA levels and TBS-soluble Aβ levels in APP/PS1 mice. Two-month-old APP/PS1 mice were injected into the lateral ventricles with vector encoding miR control or miR^CD33 and sacrificed at 10 months of age for analysis. (A) Levels of CD33 mRNA were significantly reduced with miR^CD33 versus miR control (30.1% decrease, *P = 0.0107, unpaired t test with Welch’s correction) in APP/PS1 mice. (B) and (C) Meso Scale Discovery multi-spot assay of Aβ38, Aβ40 and Aβ42 peptides in TBS-soluble (B) or formic acid (FA)-soluble (C) fractions isolated from mouse brain tissue around the injection site. (B) TBS-soluble Aβ38 levels were comparable in both mouse groups. Early injection with miR^CD33 versus miR control led to decreased levels of TBS-soluble Aβ40 (25.1% decrease, *P = 0.0274, Mann–Whitney test) and Aβ42 (30.8% decrease, P = 0.1996) in APP/PS1 mice. (C) Levels of FA-soluble Aβ38 were similar in both mouse groups. Modest but non-significant reductions in FA-soluble Aβ40 and Aβ42 were noticed in miR^CD33-treated mice compared to control mice. Symbols in graphs represent individual animals (n = 8 for miR control, n = 9 for miR^CD33). Data are represented as mean ± SEM. See also Supplementary Material, Figure S2.

Figure 4

Early intervention therapy decreases amyloid plaque burden in APP/PS1 mice. Two-month-old APP/PS1 mice were injected into the lateral ventricles with vector encoding miR control or miR^CD33 and sacrificed at 10 months of age. (A) Photomicrographs of selected coronal brain sections labeled with an antibody targeting Aβ (red, rabbit polyclonal antibody specific for the N-terminal fragment of Aβ) to reveal plaques and an anti-GFP antibody (green, chicken polyclonal antibody specific to GFP) to visualize the transduction efficiency and expression of miR control and miR^CD33 vectors in the brain of APP/PS1 mice. Brain sections were also counterstained with DAPI (blue). Scale bar represents 1000 μm. (B and C) Quantification of amyloid plaque area and plaque density in the cortex (B) and hippocampus (C) of APP/PS1 mice treated with miR control or miR^CD33 vector. Treatment with miR^CD33 versus control vector decreased amyloid plaque burden as measured by area (24.4% decrease, ^*p = 0.035, Unpaired t test with Welch's correction) and density (22.2% decrease, p = 0.066) in the cortex of APP/PS1 mice. Treatment with miR^CD33 versus miR control also reduced amyloid plaque burden in the hippocampus of APP/PS1 mice. Symbols in graphs represent individual animals (n = 8 for miR control, n = 9 for miR^CD33). Data are represented as mean ± SEM. See also Supplementary Material, Figures S3 and S4.

Figure 5

PCR array analysis reveals decreased levels of microglial receptor transcripts and pro-inflammatory activation genes in miR^CD33-treated APP/PS1 mice compared to control mice. RNA was extracted from the early intervention mice that were treated with miR control or miR^CD33 vector. RNA was reverse transcribed and subjected to PCR array that consists of 14 transcripts. Graphs with green bars depict transcripts that were decreased in the miR^CD33 group compared to control, while graphs with orange bars depict transcripts that were increased in the miR^CD33 group versus control. CD33 expression levels were reduced by 33.41% (**P = 0.0078) in miR^CD33-treated mice compared to control mice. Several microglial receptor transcripts (e.g. Itgax/CD11c, CD47 and CD36) and pro-inflammatory activation genes (e.g. Tlr4 and Il1b) were downregulated after miR^CD33 treatment versus miR control in APP/PS1 mice. Each symbol in graphs represents an individual animal (n = 8 for miR control, n = 9 for miR^CD33). Data are represented as mean ± SEM.

Figure 6

Treatment with miR^CD33 compared to miR control leads to decreased protein levels of pro-inflammatory cytokines in APP/PS1 mice. Brain tissue homogenates from the early intervention treated mice were used in a CCL2-specific ELISA and Meso Scale Discovery cytokine assay to detect expression of the indicated proteins. The pro-inflammatory cytokine TNFα was significantly decreased (38.48%, ***P = 0.0005, Mann–Whitney test) in miR^CD33-treated mice compared to control mice. Trends towards reduced levels of chemokines, CCL2 (44.58%, P = 0.059) and CXCL1 (31.3%, P = 0.075), were observed in miR^CD33-treated mice versus control mice. Graphs with blue bars depict transcripts that were decreased in the miR^CD33 group, while graphs with pink bars depict transcripts that were increased in this group. Each symbol in graphs represents an individual animal (n = 8 for miR control, n = 9 for miR^CD33). Data are represented as mean ± SEM.