Attenuating astrocyte activation accelerates plaque pathogenesis in APP/PS1 mice

Transgenic mice

APPswe/PS1ΔE9 transgenic mice (APP/PS1; ref. 43; Jackson Laboratories, Bar Harbor, ME, USA) were crossed with Gfap^−/−Vim^−/− mice (22). Because the two knockout genes and transgene segregated independently, this cross required 3 generations of crosses to produce APP/PS1 Gfap^+/+Vim^+/+ and APP/PS1 Gfap^−/−Vim^−/− mice: 1) APP/PS1^+/− Gfap^+/+ Vim^+/+ × APP/PS1^−/− Gfap^−/− Vim^−/−; 2) APP/PS1^+/− Gfap^+/− Vim^+/− × APP/PS1^−/− Gfap^+/− Vim^+/−; and 3) APP/PS1^+/− Gfap^−/− Vim^−/− × APP/PS1^−/− Gfap^−/− Vim^−/− or APP/PS1^+/− Gfap^+/+ Vim^+/+ × APP/PS1^−/− Gfap^+/+ Vim^+/+. Mice of both genotypes were maintained on the same mixed backgrounds (B6C3 × C57BL 129SV 129Ola).

Tissue preparation

Mice were deeply anesthetized with isoflurane and transcardially perfused with 0.01 M PBS. The brains were removed, and the hemispheres were separated. Left hemispheres were immediately dissected, snap-frozen on dry ice, and stored at −80°C for biochemical analysis. Right hemispheres were fixed in 4% paraformaldehyde for 24 h and transferred to 30% sucrose in 0.2 M PBS. After brains were saturated with fixative, they were snap-frozen on dry ice and stored at −80°C. Coronal sections, 50 μm thick, were made with a sliding microtome and stored in 0.2 M PBS, 30% sucrose, and 30% ethylene glycol at −20°C.

Immunohistochemistry

Sections were permeabilized with 0.3% Tween-20 in Tris-buffered saline (TBS-T20) for 10 min, and endogenous peroxidase activity was quenched by a 10-min treatment of 0.3% H₂O₂ solution in TBS. Tissue was washed with TBS, blocked with 3% dry milk in TBS-T20 for 1 h, and incubated with primary antibody overnight. A fresh solution of streptavidin and horseradish peroxidase-conjugated biotin (1:400, Vector Laboratories, Burlingame, CA, USA) was applied to tissue for 90 min, followed by 0.025% 3-3′-diaminobenzadine tetrachloride in 0.25% NiCl and 0.05% H₂O₂ for 15 min. The slices were placed on glass slides, dried overnight, dehydrated, and mounted. Primary antibodies were mouse anti-Aβ antibody (HJ3.4, 1:1000; ref. 44), rabbit anti-reticulon-3 (RTN-3) antibody (1:4000; a gift from Dr. Riqiang Yan, Lerner Research Institute, Cleveland Clinic, Cleveland, OH, USA; ref. 45), and rabbit anti-synaptophysin antibody (1:200; Cell Signaling, Danvers, MA, USA).

Immunofluorescence

After perfusion fixation (as above), brains were postfixed for 4 h in the same fixation solution. Free-floating sections (50 μm) were cut and collected. Rabbit anti-Iba-1 antibody (1:1000; Wako Pure Chemicals Industries, Tokyo, Japan), rabbit anti-glutamine synthetase (GS; 1:4000; Abcam, Cambridge, MA, USA), or mouse anti-neuronal nuclei (NeuN, 1:400; Millipore, Billerica, MA, USA) was applied to the sections overnight at 4°C. Sections were incubated with Alexa Fluor 488-conjugated donkey anti-mouse IgG antibody (Alexa Fluor 488; 1:400; Molecular Probes, Grand Island, NY, USA). Sections were washed, mounted, and examined by confocal microscopy (Carl Zeis, Thornwood, NY, USA).

Plaque staining

Brain slices were mounted on glass slides. Tissue was permeablized with 0.25% Triton for 30 min and stained with X-34 (gift from Robert Mach, Mallinckrodt Institute of Radiology, Washington University School of Medicine, St. Louis, MO, USA) dissolved in a solution of 40% ethanol and 60% water, pH 10, for 20 min. Tissue was then rinsed in distilled water and mounted. For FD Neurosilver staining (MTR Scientific, Ijamsville, MD, USA), fixed sections were stained according to the manufacturer's protocol.

Aβ ELISA

Dissected cortices were homogenized in PBS and then in 5 mM guanidine HCL in TBS (pH 8). After each homogenization step, samples were centrifuged at 12,000 rpm for 20 min, and supernantants were collected for ELISA. Pellets were used for subsequent extraction steps. All ELISA samples were diluted in a final buffer consisting of 0.25% BSA, 500 mM guanidine, and 200 mM Tris-PBS (pH 7.4), with protease inhibitors. Antibodies HJ2 (anti-Aβ_35–40) and HJ7.4 (anti-Aβ_37–42) were used to capture Aβ_x-40 and Aβ_x-42, respectively; biotinylated HJ5.1 (anti-Aβ13–28) was used to detect Aβ (44). Aβ_1-x was captured with m266 antibody (anti-Aβ_13–28) and detected with biotinylated 3D6 antibody (anti-Aβ_1–5) (44). Streptavidin-poly-horseradish peroxidase-40 (Research Diagnostics, Flanders, NJ, USA) was added, and slow ELISA TMB (Sigma, St. Louis, MO, USA) was used to develop the peroxidase. A BioTek 600FL microtiter plate reader (BioTek, Winooski, VT, USA) 600FL was used for colorimetric detection and analysis. Standard curves were generated from synthetic Aβ peptide (American Peptide, Sunnyvale, CA, USA).

Real-time quantitative PCR (qPCR)

Messenger RNA was extracted from frozen cortical tissue using TRIzol, and reverse transcribed with the High Capacity cDNA Reverse Transcription kit (Applied Biosystems, Carlsbad, CA, USA). qPCR was performed using the ABI 7500 in the default thermal cycling mode with Power SYBR (Applied Biosystems, Foster City, CA, USA) using the primers shown in Supplemental Table S1. Mouse β-actin was used as a normalization reference. To confirm the specificity of qPCR reactions, dissociation curves were analyzed at the end of qPCR assay. Relative mRNA levels were calculated using the comparative C_t method, and expressed as a percentage of control (APP/PS1 Gfap^+/+ Vim^+/+).

Western blotting

Dissected cortices were homogenized in 1% Triton X-100 in PBS and centrifuged at 12,000 rpm for 20 min, and supernantants were collected. Protein concentration was determined with the BCA kit (Pierce, Rockford, IL, USA). Equal amounts of protein from each sample were run on 3–8% Tris acetate or 4–12% Bis-Tris XT gels (Bio-Rad, Hercules, CA, USA) and transferred to PVDF membranes with the iBlot dry blotting system (Invitrogen, Grand Island, NY, USA). Membranes were blocked with 5% dry milk in 1% Triton X-100 in TBS for 1 h. Membranes were probed with the following antibodies diluted in 2.5% dry milk in 1% Triton X-100 in TBS overnight: APP and its C-terminal fragment (1:1000; Invitrogen, Grand Island, NY, USA), matrix metalloproteinase 2 (MMP-2; 1:200; Abcam, Cambridge, MA, USA), matrix metalloproteinase 9 (MMP-9; 1:200; Abcam), apoE (1:30000; Calbiochem, Darmstadt, Germany), apolipoprotein J (apoJ; 1:4000; Covance, Princeton, NJ, USA), and S100B (1:200; Dako, Houston, TX, USA). Membranes were washed and probed with biotinylated secondary antibodies (1:5000; Santa Cruz Biotechnology, Santa Cruz, CA, USA), and visualized with Pico West (Pierce) or Amersham ECL Advance (GE Healthcare, Piscataway, NJ, USA) kits. Normalized band intensity was quantified with ImageJ software (U.S. National Institutes of Health, Bethesda, MD, USA; ref. 44).

In vivo microdialysis

In vivo microdialysis to assess brain interstitial fluid (ISF) Aβ in the hippocampus of awake, freely moving mice was performed as described previously (46), to sample soluble molecules (<30 kDa) in the ISF. Aβ capable of entering the probe has been dubbed “exchangeable Aβ” (eAβ). Under isoflurane anesthesia, guide cannulas (BR-style; Bioanalytical Systems, West Lafayette, IN, USA) were cemented into the left hippocampus (3.1 mm behind the bregma, 2.5 mm lateral to the midline, and 1.2 mm below the dura at a 12° angle). Microdialysis probes (2 mm) were inserted through the guides so the membrane was contained entirely within the hippocampus (BR-2, 38-kDa molecular mass cutoff membrane; Bioanalytical Systems). Microdialysis perfusion buffer was artificial cerebrospinal fluid containing 4% BSA (Sigma) that was filtered through a 0.1-μm membrane. Flow rate was a constant 1.0 μl/min. Samples were collected every 60 min with a refrigerated fraction collector into polypropylene tubes and assessed for Aβ_1-x by ELISA (as described above) after each experiment. ISF Aβ half-life was determined as described previously (46). After steady-state ISF Aβ levels were established, mice received an intraperitoneal injection of compound E (20 mg/kg), a potent, blood-brain permeable γ-secretase inhibitor (ref. 47; synthesized by AsisChem, Waltham, MA, USA) to rapidly block Aβ production. Microdialysis samples were collected every 60 min for 6 h and assayed for Aβ_x-40 by ELISA. The half-life of ISF Aβ was calculated from the slope of the semilog plot of percentage change in Aβ vs. time (46). Only Aβ values that were continually decreasing were included in half-life analysis (the first 3 h after compound E administration).

Viral vectors

An adeno-associated virus 2 (AAV2) construct (Stratagene, Santa Clara, CA, USA) was modified as follows. The 1.2-kb cytomegalovirus promoter and β-globin intron were removed from the plasmid using MluI and SalI restriction sites and replaced with the GFAP promoter and green fluorescent protein (GFP). The construct was transformed into DH5α bacterial cells, which were grown on ampicillin-containing medium overnight. The Hope Center Viral Core (Washington University, St. Louis, MO, USA) generated AAV2 virions containing our construct at 6 × 10¹¹ viral genomes/ml.

Stereotactic injection

AAV-GFAP-GFP (2 μl) was injected into the cortex of 12-mo-old APP/PS1 Gfap^+/+Vim^+/+ and APP/PS1 Gfap^−/−Vim^−/− mice (n=3/group) at the following stereotactic coordinates: 2.0, ±1.7, −0.5. The virus was slowly infused through a 30-gauge needle (0.2 μl/min) over 10 min; the syringe was left in place for an additional 10 min before it was slowly withdrawn. In each mouse, the AAV was allowed to incubate for 4 wk.

Plaque quantification

Brain sections (50 μm) were collected every 600 μm from the rostral anterior commissure to the caudal hippocampus. The sections (n=9–10/group) were stained with X-34 or immunostained with HJ3.4 antibodies and imaged with a ×20 0.75 N.A. objective on a NanoZoomer digital scanner (Hamamatsu, Bridgewater, NJ, USA) to create high-resolution digital images (TIFF). Individual images were exported to ImageJ. Cortex and hippocampus were manually delineated, and brightness thresholding was used to delineate plaques. For each slice, the total area of plaque coverage was measured in the cortex and hippocampus, expressed as percentage of total area in each mouse and averaged for each group as described previously (44). For individual plaque analysis of RTN-3 burden and corresponding plaque size, images of randomly selected plaques from stained brain sections were captured using a ×60 1.4 N.A. oil immersion lens on a Nikon Eclipse E800 (Nikon, Tokyo, Japan) with attached Optronics camera (Optronics, Goleta, CA, USA), exported to ImageJ, and quantified by using brightness-thresholding to delineate dystrophic neurites and manual tracing to delineate plaque profiles. Cross-sectional areas of both dystrophic neurites and plaque size were measured and graphed on a scatterplot by genotype. All imaging in this study was performed at room temperature.

Stereology

RTN-3-stained slices were collected every 300 μm (n=5/group). Dystrophic neurites in the cortex were quantified using unbiased stereology (Cavalieri-point counting method) with a ×60 1.4 N.A. oil immersion lens on a Nikon Eclipse E800 with attached Optronics camera as described previously (48).

Astrocyte, microglia, and neuron quantification

Brain sections labeled with GS, Iba-1, or NeuN for immunofluorescence were systematically collected every 300 μm, while AAV-GFAP-GFP-transfected tissue was collected 100 and 200 μm on either side of the injection needle track (4 slices/mouse, n=3/group). Labeled sections were imaged by confocal microscopy (Zeiss Axiovert 200 M with LSM 5 Pa) with a ×40 1.2 N.A water immersion or a ×67 1.4 N.A. oil immersion objective (Zeiss, Thornwood, NY, USA) to image several XY planes along a Z-axis. Images were exported to ImageJ for quantification of cell count and plaque invasion (overlap between cell and plaque staining). Some confocal Z stacks were used to reconstruct 3-dimensional images for rotation and captured on video.

Statistical analysis

Various measures in APP/PS1 Gfap^+/+Vim^+/+ and APP/PS1 Gfap^−/−Vim^−/− mice were compared using two-tailed Student's t test when data were normally distributed. Normality was tested with the Kolmogorov-Smirnov test (SigmaStat 2.0 statistical software; Systat Software, Chicago, IL, USA). For data with non-normal distributions or insufficient sample sizes to test normality, the Mann-Whitney U test was performed. The level of significance was set at P ≤ 0.05.

Gfap/Vim deletion exacerbates amyloid plaque pathogenesis

APP/PS1 mice (APPswe PS1ΔE9; ref. 43) were crossed with Gfap^−/−Vim^−/− mice (20, 22) to produce APP/PS1 Gfap^+/+Vim^+/+ and APP/PS1 Gfap^−/−Vim^−/− mice. Mice of both genotypes were grossly normal and fertile, without obvious behavioral or anatomical differences. Histological examination of neurons using cresyl violet staining revealed no qualitative differences in neuronal layers of the cortex or hippocampus (Supplemental Fig. S1). Quantitative analysis using NeuN immunofluorescence in regions of the cortex did not reveal differences in neuronal density (Supplemental Fig. S1I–K). Brain sections from these mice were immunostained with anti-Aβ antibody (HJ3.4) to quantify Aβ plaque load or stained with X-34 to quantify amyloid load. Aβ load and amyloid load were significantly increased in the cortex and hippocampus of APP/PS1 Gfap^−/−Vim^−/− mice at 8 and 12 but not 4 mo of age (Fig. 1A–F). In APP/PS1 mice, plaque accumulation begins at ∼4 mo of age and increases thereafter. These data suggest that Gfap and Vim deletions do not influence the initial deposition of plaques but exacerbate subsequent plaque accumulation, consistent with the contention that Gfap/Vim deletion selectively affects astrocyte activation—an event that occurs after initial plaque formation. By examining the frequency distribution of plaque sizes in 12-mo-old mice, it is apparent that the increase in plaque load is a result of increased plaque numbers of all sizes (Fig. 1G).

Plaque load was corroborated by Aβ quantification with ELISA after serial extractions of brain tissue with PBS and guanidine. Consistent with the plaque load data, no difference in Aβ levels was detected between genotypes at 4 mo of age. At 8 and 12 mo, APP/PS1 Gfap^−/− Vim^−/− mice had significantly higher Aβ levels in the guanidine extract but not in the PBS fraction (Fig. 2). These data suggest that Gfap and Vim deletion selectively alters insoluble (plaque-associated) Aβ but not the soluble pools of Aβ in APP/PS1 mice.

To further examine the effects of Gfap and Vim deletion on Aβ levels in ISF, in vivo microdialysis was performed in 3-mo-old mice. Microdialysis probes were implanted unilaterally in the hippocampus of APP/PS1 Gfap^+/+Vim^+/+ and APP/PS1 Gfap^−/−Vim^−/− mice under isoflurane anesthesia. After probe implantation, mice remained awake with freedom of movement throughout the experiment. Steady-state measurements of ISF Aβ levels revealed no difference in mean exchangeable Aβ_1-x levels between the APP/PS1 Gfap^−/− Vim^−/− mice and APP/PS1 Gfap^+/+Vim^+/+ mice (Fig. 3A). To determine whether the gene deletions affect Aβ turnover, mice were treated with the γ-secretase inhibitor compound E, and Aβ levels in ISF were measured to create a decay curve from which the ISF Aβ half-life was calculated. The ISF Aβ half-life was unchanged by deletion of Gfap and Vim (Fig. 3B). Thus, Gfap and Vim deletion does not alter ISF Aβ levels or metabolism but selectively alters amyloid plaque load.

Gfap/Vim deletion exacerbates neuritic dystrophy

In general, APP transgenic mice do not show significant neuronal cell death or brain atrophy, in contrast to brains of patients with AD (49). However, APP mice develop dystrophic neurites immediately around plaques (50), suggesting that this neuronal pathology is independent of cell death. We examined the impact of Gfap and Vim deletion on neuritic dystrophy using unbiased stereological counts of brains immunostained with RTN-3, a protein that selectively accumulates in dystrophic neurites (45). At 12 mo of age, dystrophic neurites were significantly more abundant in APP/PS1 Gfap^−/−Vim^−/− mice than in APP/PS1 Gfap^+/+Vim^+/+ mice (Fig. 4A–E). Because APP/PS1 Gfap^−/−Vim^−/− mice have greater plaque load than APP/PS1 Gfap^+/+Vim^+/+ mice, we normalized dystrophic neurite counts to amyloid load and still found a significant increase (Fig. 4F). Furthermore, using a random sampling of plaques from mice of both genotypes, we plotted plaque size vs. dystrophic neurite burden (cross-sectional area). Across all sizes of plaques, dystrophic neurites were greater in APP/PS1 Gfap^−/−Vim^−/− (Fig. 4G). To verify labeling of dystrophic neurites with RTN-3, we performed colabeling studies using RTN-3 immunostaining and silver staining in adjacent 15-μm brain sections. RTN-3 immunostaining consistently delineated dystrophic neurites demonstrated by silver staining in APP/PS1 mice (Supplemental Fig. S2). These findings suggest that activated astrocytes either protect neurons from developing dystrophic neurites or are involved in the clearance of dystrophyic neurites.

Although Gfap/Vim gene deletion exacerbates neuritic dystrophy, we observed no difference in the pattern of synaptophysin immunoreactivity surrounding plaques and no difference in the amount of synaptophysin or PSD95 mRNA between the cortices of each genotype (Supplemental Fig. S3).

Gfap/Vim deletion does not alter expression or processing of APP or Aβ-degrading proteases

To further understand the molecular mechanisms of the effects of Gfap and Vim deletion on plaque pathogenesis, we examined the expression of genes and proteins involved in Aβ synthesis and degradation. RNA extracted from cortex of 12-mo-old mice was subjected to real-time PCR to compare transcript levels for APP processing secretases [ADAM10, ADAM17, PS1, memapsin-2, nicastrin, γ-secretase subunit anterior pharynx defective 1 homolog B (APH-1b), PEN2], lipoprotein mediators (apoE, apoJ, LRP1, ABCA1), and Aβ-degrading enzymes (neprilysin, endothelin-converting enzyme, insulin-degrading enzyme, MMP-9, MMP-2). Only APH-1b and MMP-9 showed significant, but modest, decreases (20-30% decrease) in mRNA levels in the APP/PS1 Gfap^−/−Vim^−/− mice (Fig. 5A); transcript levels for all other genes were unchanged. Western blot analysis showed no differences between genotypes in APP expression and C-terminal fragments (Fig. 5B, E), suggesting that Gfap and Vim gene deletion did not alter APP expression or processing. ApoE and apoJ (clusterin) showed no significant differences (Fig. 5C, E). Similarly, MMP-2 and MMP-9 showed no significant changes in APP/PS1 Gfap^−/−Vim^−/− mice (Fig. 5D, E), although pro-MMP9 and MMP9 showed a decreasing trend.

Astrocytic invasion of plaques is attenuated in APP/PS1 Gfap^−/−Vim^−/− mice

To examine astrocyte morphology in vivo, mice were injected with AAV2 constructs carrying GFP driven by a Gfap promoter (AAV-GFAP-GFP). Because immunostaining with individual antibodies cannot reveal the full extent of astrocyte morphology (16), we expressed GFP in astrocytes in vivo to observe the complete cellular profile of the cells. Mice received a cortical injection of viral particles at 12 mo of age. After 1 mo, the mice were euthanized and processed for histology.

In regions surrounding plaques, astrocytes in APP/PS1 Gfap^+/+Vim^+/+ mice were hypertrophic with thick proximal processes, characteristic of reactive astrocytes (Fig. 6A, C). Processes were also evident deep within X-34-stained plaques, suggesting process invasion within plaques (Fig. 6C). In contrast, astrocytes from APP/PS1 Gfap^−/−Vim^−/− mice did not exhibit process hypertrophy, and distal processes appeared as fine strands. Moreover, little overlap between astrocyte and plaque was evident (Fig. 6B, D). In fact, regions immediately adjacent to the plaques were devoid of astrocyte processes, forming a “halo” around plaques (Fig. 6D). This difference in astrocyte interaction with plaques was even more striking in 3-dimensional reconstructions of the confocal images in Fig. 6C, D (Supplemental Videos S1 and S2).

Quantitative assessment of astrocyte numbers (GFP-expressing astrocytes) did not reveal a significant difference in the density of astrocytes (n/mm²) between genotypes (Fig. 6E). This finding was confirmed by quantification of glutamine synthetase-immunostained astrocytes (Supplemental Fig. S4), suggesting that deletion of Gfap and Vim alters the activation and morphology of astrocytes without altering their viability. However, in APP/PS1 Gfap^−/−Vim^−/− mice, there was far less overlap between astrocyte processes and plaque staining than in APP/PS1 Gfap^+/+Vim^+/+ mice (Fig. 6F). Astrocytes distant from amyloid plaques were indistinguishable between the two genotypes (Fig. 6G, H). These findings suggest that Gfap/Vim deletion results in astrocytes less capable of process hypertrophy and invasion into amyloid plaques. Previous work has shown BDNF expression is increased in amyloid-associated astrocytes (51) and that this may play a role in amyloid associated pathology (52). Quantitative PCR analysis revealed no difference in BDNF message levels (Supplemental Fig. S3G).

Microglial morphology was also examined by Iba-1 immunostaining. Gross morphological differences between microglia of the two genotypes were not found (Fig. 7A–D). The density of Iba-1-positive microglia (number of cells/area) was slightly higher in APP/PS1 Gfap^−/−Vim^−/− than in APP/PS1 Gfap^+/+Vim^+/+ mice (Fig. 7E). In addition, the overlap between microglia and plaques was two-fold greater in APP/PS1 Gfap^−/−Vim^−/− mice than in APP/PS1 Gfap^+/+Vim^+/+ mice (Fig. 7F). Thus, microglia seem to be more abundant around plaques in the absence of activated astrocytes. Real-time PCR from the cortex of 12-mo-old mice confirmed an increased gene expression of CD11b and Iba-1 in APP/PS1 Gfap^−/−Vim^−/− mice; however, expression of astrocyte-specific genes (glutamine synthase and S100b), as well as cytokines and chemokines (IL-1b, IL-6, IL-10, TNF-α, TGF-β, iNOS), was unchanged (Fig. 7G). Despite this increase in microglial density in APP/PS1 Gfap^−/−Vim^−/− mice, amyloid plaque load was increased.