Selective reduction of astrocyte apoE3 and apoE4 strongly reduces Aβ accumulation and plaque-related pathology in a mouse model of amyloidosis

Further information and requests for resources and reagents should be directed to David M. Holtzman (holtzman@wustl.​edu).

APPPS1–21 mice on a C57BL/6 N background (gift from Dr. Mathias Jucker, Department of Cellular Neurology, Hertie Institute for Clinical Brain Research, University of Tübingen, Germany) overexpress human APP bearing both the Swedish mutation and PSEN1 containing an L166P mutation, both driven by the Thy1 promoter [36]. ApoE3^flox/flox and apoE4^flox/flox (FE3 and FE4, respectively), human APOE knock-in mice on a C57BL background, were generated by replacing the mouse genomic sequence from the translation initiation codon in exon 2 to the termination codon in exon 4 with its human counterparts flanked by loxP site, driven by the endogenous APOE promoter [33]. Aldh1l1-Cre/ERT2 mice on a C57BL background were obtained from Jackson Laboratories (Stock No. 031008). To generate APPPS1-21/ apoE3^flox/flox or apoE4^flox/flox mice (APPPS1;FE3Cre- or APPPS1;FE4Cre-, respectively, and collectively referred to as Cre-), we crossed APPPS1-21 transgenic mice with FE3 or FE4 for several generations. To generate Aldh1l1-Cre/ERT2/apoE3^flox/flox or apoE4^flox/flox mice (AFE3 or AFE4, respectively), we crossed Aldh1l1-Cre/ERT2 mice to FE3 or FE4 for several generations. We then crossed APPPS1;FE3Cre- or APPPS1;FE4Cre- mice to AFE3 or AFE4 mice, respectively, to produce APPPS1-21/Aldh1l1-Cre/apoE3^flox/flox or apoE4^flox/flox mice (APPPS1;FE3Cre + or APPPS1;FE4Cre+, respectively, and collectively referred to as Cre+). Finally, we crossed APPPS1;FE3Cre + or APPPS1;FE4Cre + mice to FE3 or FE4 mice, respectively, to produce experimental mice utilized. All the experimental mice involved in the final analysis were obtained from the same generation. The sex of animals in each specific experiment can be found in the corresponding figure legends. All animal procedures and experiments were performed under guidelines approved by the Institutional Animal Care and Use Committee (IACUC) at Washington University School of Medicine. All of the phenotyping and data analysis was performed by researchers who were blind to the genotype of the mice.

Tamoxifen was dissolved in corn oil at a concentration of 20 mg/ml by shaking overnight at 37 °C. After preparation, the tamoxifen solution was wrapped by foil, and stored at 4 °C for up to a month. Tamoxifen was given at 100 mg tamoxifen/kg body weight and administered via intraperitoneal (IP) injection once every 24 h for 6 consecutive days. All experimental mice received the same series of IP tamoxifen injections over 6 days.

Mice were anesthetized with 200 mg/kg pentobarbital and subsequently perfused with cold PBS containing 3 IU/ml heparin. After brain isolation, the left hemisphere was fixed in 4% paraformaldehyde for at least 24 h and then transferred to 30% sucrose and stored at 4 °C until sectioning. The right hemisphere was dissected into various parts (posterior- and anterior-cortex, hippocampus, etc.), all of which were snap-frozen using dry ice and stored at − 80 °C until further analysis.

Hippocampal-containing sections were selected for human apoE (Cell Signaling, 13,366, 1:500), GFAP (Abcam, ab53554, 1:500; Millipore, MAB3402B, 1:2000), Iba1 (Abcam, ab5076, 1:500; Wako, 019–19,741, 1:5000), Aβ (HJ3.4, in house, mouse monoclonal, 2 μg/ml), Clec7a (InviviGen, mabg-mdect, 1:50), BACE1 (Abcam, ab108394, 1:100), and RTN-3 (a generous gift from Dr. Riqiang Yang, 1:1000) immunofluorescence staining. For reticulon-3 (RTN-3) staining, sections were pre-mounted on to Superfrost + glass microscope slides. For all other stainings, sections were free-floating. Sections were washed in Tris-buffered saline (TBS) buffer for 3 times, 5 min each. After washing, sections were incubated in TBS with 0.25% Triton X-100 (TBSX) for 30 min at room temperature to permeabilize the sections, followed by 1 time TBS washing for 5 min. Then sections were incubated with 2 μM X34 in the X-34 staining buffer (40% ethanol, 60% TBS, 1:500 vol. 10 N NaOH) for 20 min. After X-34 staining, sections were washed 3 times for 2 min each in the X-34 washing buffer (40% ethanol and 60% TBS), followed by 3 times washing in TBS for 5 min each. For RTN-3 staining, antigen retrieval was performed by heating sections to 95 °C in 50 nM Sodium Citrate buffer for 30 min followed by 3 times washing in TBS for 5 min each. After washing, sections were blocked by 10% donkey serum in TBSX for 1 h at room temperature to prevent non-specific binding. Then sections were incubated with primary antibodies at 4 °C, overnight. After overnight incubation, sections were washed 3 times in TBS for 5 min each. Then sections were incubated with corresponding secondary fluorescence antibodies (Life Technologies) for 2 h at room temperature. After 3 times washing by TBS for 10 min each, sections were mounted to glass slides. Slides were coverslipped by ProLong Glass Antifade Mountant (Invitrogen, P36980) and scanned by Nikon A1Rsi Confocal Microscope, Leica Stellaris 5, or BioTek Cytation5. Representative images in Fig. 1E were captured by the BioTek Cytation5 using a 10X objective. Representative images in Figs. 2A, 3A, 4A, 5C (top panel), 6A, S1, and S2A were captured by the BioTek Cytation5 using a 4X objective. Representative images in Figs. 3F, 6C, E, S2B, and S5 were captured by the BioTek Cytation5 using a 20X objective. Representative images in Figs. 1G, 2D, 4C, and E were captured by the Leica Stellaris 5 using a 40X oil objective. Representative images in Fig. 5B (bottom panel) were captured by the Nikon A1Rsi using a 40X objective.

Acquired images were analyzed by using Image J v1.53c (https://​imagej.​net/​Fiji), Imaris 9 (https://​imaris.​oxinst.​com/​), and BioTek Gen5 (https://​www.​biotek.​com/​products/​software-robotics/​) software. ApoE fluorescent staining intensity in Fig. 1F was determined using images captured by the BioTek Cytation5. The average pixel intensity for each image was found by setting a minimal threshold to highlight positive staining, running the “Analyze Particle” function to obtain the mean pixel intensity, and then subtracting the average background pixel intensity. The average background pixel intensity was determined by calculating the average pixel intensity of APPPS1EKO images. To determine the percent of hippocampal area or cortical area covered by X-34, HJ3.4, Iba1, GFAP, or BACE1 in Figs. 2B, C, 3B, C, 4B, 5D, and 6B, images were analyzed as previously described [37, 38]. Briefly, Image J software was used to analyze images captured with the BioTek Cytation5. Regions of interest (ROI’s) of images were traced, images were thresholded to highlight positive staining, and the “Analyze Particle” function was used to obtain the percent area covered. While intense BACE1 staining was present in the hippocampus, quantification of the hippocampus was excluded because high levels of BACE1 in the mossy fibers are physiologically normal and not indicative of any neuritic dystrophy. The average intensity of fibrillar plaques in Fig. 2E was also found using ImageJ software to analyze images of X-34 stained sections captured with the BioTek Cytation5 software. The cortex was traced, images were thresholded to highlight positive staining, and the “Analyze Particle” function was used to obtain the average pixel intensity. The X-34 to HJ3.4 ratio in 3G was determined using images of the cortex captured with the BioTek Cytation5 and analyzed using ImageJ software. Images were thresholded to highlight positive staining, and the “Analyze Particle” function was used to obtain the percent area covered. The area of X-34 was then divided by the area of HJ3.4. To determine the percent area of Clec7a and Iba1 staining around X-34 plaques in Fig. 4E, ImageJ software was used to analyze images captured with the Leica Stellaris 5 confocal microscope. The images of X-34⁺ plaques were thresholded to highlight plaques, “Analyze Particles” was run, and the thresholded plaque ROI’s were combined and then enlarged by 15 μm. The enlarged ROI’s were then transferred to the corresponding Clec7a and Iba1 images, images were thresholded, and “Analyze Particles” was run to find the area of Clec7a and Iba1. The area of Clec7a was then divided by the area of Iba1. The GFAP volume to X-34 volume in Fig. 5D was found using Imaris software to analyze images obtained by the Nikon A1Rsi confocal microscope, as previously described [24]. Briefly, Surfaces were created for X-34 and GFAP and the Dilate Xtension was used to dilate the X-34 surfaces by 15 μm. The surface-surface co-localization Xtension was run using the dilated X-34 surfaces and GFAP surfaces and the volume of GFAP within 15 μm of X-34 plaque was determined based on overall X-34 plaque volume. The BACE1 and RTN-3 area per X-34 plaque was determined using BioTek Gen5 software to analyze images obtained by the BioTek Cytation5. The Cellular Analysis function was used with the primary mask thresholded and set based on the X-34 staining. The secondary mask was set using a 15 μm expanded distance from the X-34 primary mask, BACE1 and RTN-3 staining was thresholded, and the average BACE1 area per X-34 plaque and RTN-3 area per X-34 plaque was determined. To determine the amount of apoE area per Iba1 area around plaques, images were obtained with the Leica Stellaris5 confocal microscope. The images of X-34⁺ plaques were subtracted from the corresponding apoE images to remove any apoE signal co-localized with X-34, leaving only non-plaque associated apoE. The images of X-34⁺ plaques were then thresholded to highlight plaques, “Analyze Particles” was run, and the thresholded plaque ROI’s were combined and then enlarged by 10 μm. The enlarged ROI’s were then transferred to the corresponding apoE and Iba1 images, images were thresholded, and “Analyze Particles” was run to find the area of apoE and Iba1 around the plaque. To determine what apoE was located within microglia, the Iba1 images were subtracted from the corresponding apoE images to leave apoE not co-localized with Iba1. The remaining area of apoE around plaques that was not co-localized with Iba1 was determined using the enlarged ROI, thresholding the images, and running “Analyze Particles”. These areas were then subtracted from the total apoE area around plaques to determine the amount of apoE area that was co-localized with Iba1. The apoE area co-localized with Iba1 was then divided by the total area of Iba1 around the plaques.

Mouse posterior cortical tissue samples were sequentially homogenized with cold PBS, and then 5 M guanidine buffer in the presence of 1X Complete Protease Inhibitor (Roche, 11,697,498,001) and 1X phosSTOP phosphatase Inhibitor (Roche, 04906845001). First, tissues were weighed, a half spoon of beads (Next Advance, ZrOB05) were added, and samples were homogenized for 45 s on setting 3, using a bead homogenizer (Next Advance, Bullet Blender Strom 24), in cold PBS buffer at 20 μl buffer/1 mg tissue. Homogenates were centrifuged 30 min at 15,000 rpm at 4 °C. The supernatant was saved as the PBS soluble fraction. Then the same amount of 5 M guanidine buffer was added to the pellet and homogenized on bead homogenizer for 3 min on setting 8, followed by 1 h rotation at room temperature. Finally, homogenates were centrifuged for 30 min at 15,000 rpm at 4 °C. The supernatant was saved as the 5 M guanidine insoluble fraction. All fractions were stored at − 80 °C until further analyzed.

The levels of apoE, Aβ₄₀, and Aβ₄₂ in PBS and 5 M guanidine fractions were measured by sandwich ELISA and normalized to the tissue weight. The coating antibodies for human apoE, Aβ₄₀, and Aβ₄₂ were HJ15.3 (in house, mouse monoclonal, 5 μg/ml), HJ2 (in house, mouse monoclonal, 20 μg/ml), and HJ7.4 (in house, mouse monoclonal, 10 μg/ml), respectively [33, 39]. The capture antibodies were HJ15.7-biotinylated (in house, mouse monoclonal, 150 ng/ml) for human apoE and HJ5.1-biotinylated (in house, mouse monoclonal, 90 ng/ml) for both Aβ₄₀ and Aβ₄₂ [33, 39].

Both male and female mouse anterior cortex tissues were used for the gene expression analysis. Nine samples from all APPPS1 positive groups were selected based on the mean values of the X34 plaque load. Four samples from APPPS1 negative FE3Cre-, FE4Cre-, and FE4Cre + groups, 3 samples from APPPS1 negative FE3Cre + group were also used as the APPPS1 negative controls. mRNA was exacted from frozen tissues using RNeasy Micro Kit (Qiagen, 74,004) and converted to cDNA using the high-capacity RNA-to-cDNA kit (Applied Biosystems, 4,387,406), following the manufacturer’s instructions. Gene expression was conducted using Fluidigm Biomark HD Real-Time PCR System in collaboration with Genome Technology Access Core at Washington University. Using Taqman primers (Life Technologies), the fold-changes were relative to FE3Cre- after values were normalized to the mean values of β-actin and Cyc1.

Mouse cortical tissues were lysed using RIPA buffer (4°C). Samples were randomly selected and included 3 males and 3 females in each group. Complete Protease Inhibitor and phosSTOP Phosphatase Inhibitor were added freshly to the RIPA buffer. After weighing tissues, 50 μl cold RIPA/1 mg tissue and 1 spoon of beads (ZrOB05) were added to the tissue-containing Safe-Lock microcentrifuge tubes. Tissues were homogenized using a bead homogenizer (Bullet Blender Strom 24) on setting 8 for 3 min. Homogenates were centrifuged in a microcentrifuge at 15,000 rpm for 30 min at 4°C. Supernatant was collected for western blot. Micro BCA Protein Assay Kit was utilized to measure protein concentration. Five micrograms of total protein was loaded and separated by 4–12% NuPAGE gels in MOPS buffer. Gel was transferred for 7 min on a nitrocellulose membrane using the iBlot 2 system. After washing 3 times for 5 min each with TBS buffer, including 0.05% Tween (TBST), the membrane was blocked with 5% milk in TBST for 1 h at room temperature, followed by incubation with APP C-terminal antibody (Sigma, rabbit polyclonal, A8717, 1:1000) and α-tubulin antibody (mouse monoclonal, 66,031–1-lg, 1:15,000) at 4 °C. After overnight incubation, the membrane was incubated with corresponding secondary HRP-conjugated secondary antibodies for 1 h at room temperature. Images were captured using the ChemiDoc™ MP Imaging System (BIO-RAD). Full length and C-terminal fragments of APP, as well as α-tubulin bands were captured separately and analyzed by ImageJ software.

All values were reported as mean ± SEM. All statistical analyses were conducted in Prism 8 (GraphPad). Two-way or Three-way ANOVA was used for assessing significance between more than two groups. P values less than 0.05 (p < 0.05) were considered significant for all tests. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. The significant p values and F values for each experiment can be found in Supplementary Table 1. The value of n per group and what n represents in each specific experiment can be found in the corresponding figure legends and in Supplementary Table 1.

In order to assess how the loss of astrocytic apoE impacts Aβ pathology, we used Aldh1l1-Cre/ERT2 BAC transgenic mice to selectively remove APOE from astrocytes in a tamoxifen dependent manner. APPPS1;FE3Cre + and APPPS1;FE4Cre + mice (Cre+) were created to assess Aβ plaque pathology compared to APPPS1;FE3Cre- and APPPS1;FE4Cre- mice (Cre-). Once-daily intraperitoneal injections of tamoxifen were administered to mice in each group at 4 weeks of age (2–4 weeks before the initial formation of Aβ plaques in these mice) for 6 consecutive days (Fig. 1A). All mice received the same series of tamoxifen injections, including the Cre- mice that served as the control. A cohort of Cre + and Cre- mice were collected at 6 weeks of age, 1 week after completing tamoxifen administration, and showed no signs of Aβ pathology (data not shown). To investigate the efficiency of APOE removal from astrocytes after tamoxifen administration, we first assessed apoE mRNA and protein levels in the brain at 18-weeks of age (Fig. 1A). ApoE mRNA levels were assessed by qPCR and were significantly reduced by ~ 70% in the Cre + mice compared to the Cre- mice (Fig. 1B). To assess apoE protein levels, cortex tissue samples were sequentially extracted in PBS and guanidine buffers to measure soluble and insoluble apoE, respectively. Soluble apoE levels were also significantly reduced by ~ 50–70% in all Cre + mice compared to Cre- mice (Fig. 1C). Insoluble apoE levels, which mostly reflects apoE co-deposition in Aβ plaques [40, 41], were also significantly reduced in Cre + mice compared to Cre- mice, with female mice having a ~ 75–85% reduction and male mice having a ~ 60–65% reduction (Fig. 1D).

Next, we assessed the cell-type-specificity of apoE expression after tamoxifen administration by performing immunofluorescent staining of apoE (Fig. 1D). We observed a strong overall reduction in the intensity of apoE staining in the cortex and hippocampus by ~ 60–80% for the Cre + mice compared to the Cre- mice (Fig. 1E and F). Since apoE is secreted into the extracellular space in lipoprotein particles, the overall intensity of the apoE immunostaining in both the cortex and hippocampus may be influenced by a combination of both intracellular and extracellular apoE. Additionally, plaques are primarily located in the cortex at the ages we studied and thus the intense foci of apoE staining in the cortex is likely from apoE associated with amyloid deposits. Qualitative assessment of the staining for apoE also revealed a strong decrease in the presence of apoE in GFAP+ astrocytes. In the Cre- mice nearly all GFAP⁺ astrocytes were also apoE⁺ while the Cre + mice exhibited only rare apoE staining in GFAP⁺ astrocytes (Fig. 1G). We also observed apoE to still be present in Iba1⁺ microglia, particularly in the vicinity of Aβ plaques (Fig. 1G). Therefore, the tamoxifen administration was able to effectively reduce the overall apoE levels and strongly reduce apoE expression in astrocytes, while appearing to still enable microglial to express apoE in the Cre + mice. Additionally, while the immunostaining showed that overall apoE levels were greatly reduced in the Cre + mice, there were still intense foci of apoE staining co-localized with the X-34⁺ amyloid plaques, similar to what was observed in the Cre- mice (Fig. 1E and G).

The deposition of Aβ into fibrillar amyloid plaques is influenced by the expression level and isoform of apoE. With the targeted removal of astrocytic APOE resulting in a significant reduction in overall apoE levels, we investigated what impact this reduction might have on the deposition of Aβ into X34⁺ fibrillar amyloid plaques (Fig. 2A, S1). In the cortex and hippocampus, there was a large and significant decrease in fibrillar plaque load in APPPS1;FE3Cre + females and in APPPS1;FE4Cre + males and females, while the male APPPS1;FE3Cre + mice did not show a significant difference (Fig. 2B and C). In the cortex, the female Cre + and APPPS1;FE4Cre + males had a ~ 60–70% reduction in amyloid burden and in the hippocampus the reduction was ~ 75–85%. Overall, the total amyloid burden with the loss of astrocytic apoE was qualitatively similar to that observed in APPPS1;EKO mice (Fig. 2B and C). An analysis of fibrillar plaque intensity revealed Cre + mice have a significantly reduced overall intensity compared to the Cre- mice (Fig. 2D and E) and exhibit a smaller, less dense compact core. The reduction in plaque intensity for the Cre + mice was ~ 40–55% compared to the intensity of the Cre- plaques. However, a qualitative assessment of the APPPS1;EKO mice showed their fibrillar plaques form with a core that is even less compact and less intense than the Cre + mice (Fig. 2D and E).

To assess overall Aβ deposition, staining was performed using an anti-Aβ antibody (Fig. 3A, S2A). In the cortex, male APPPS1;FE3Cre + mice showed no difference in Aβ levels, while female APPPS1;FE3Cre + mice showed a trend towards lower Aβ plaque load compared to female APPPS1;FE3Cre- mice (Fig. 3B). For the APPPS1;FE4Cre + mice, both male and female mice had a significant decrease in cortical plaque coverage of ~ 60% compared to APPPS1;FE4Cre- mice (Fig. 3B). In the hippocampus, the Cre + females had significantly lower Aβ plaque coverage with a ~ 90% reduction (Fig. 3C). Interestingly, a qualitative analysis of the APPPS1;EKO mice revealed a higher Aβ plaque load than the Cre + mice (Fig. 3B and C).

To further analyze the accumulation of Aβ in the brain, the level of insoluble Aβ was assessed in the guanidine soluble (PBS-insoluble) fractions from homogenized cortical tissue samples. Both Aβ₄₀ and Aβ₄₂ levels from the guanidine fraction were assessed by ELISA. Aβ that is insoluble in PBS and detected in the guanidine fraction serves as a measure of how much Aβ has accumulated in deposits in the brain and constitutes the majority of the Aβ pool once Aβ aggregates. The Aβ₄₀ and Aβ₄₂ in the PBS soluble fraction showed no difference between each of the groups (data not shown). The APPPS1;FE3Cre + males showed no difference in Aβ₄₀ and Aβ₄₂ levels in the guanidine fraction compared to the APPPS1;FE3Cre- male mice while Aβ₄₀ was significantly decreased in the APPPS1;FE4Cre + males by 65%. Both the Aβ₄₀ and Aβ₄₂ levels were significantly decreased by in the Cre + females compared to the Cre- females, with a ~ 70–80% decrease for Aβ₄₀ and a ~ 50–60% decrease for Aβ₄₂ (Fig. 3D and E).

An assessment of the Aβ immunostaining in relation to the formation of fibrillar Aβ plaques also revealed unique patterns of Aβ deposition in the Cre + mice compared to the Cre- mice (Fig. 3F, S2B). In the Cre + mice, the fibrillar Aβ core was smaller with a greater percent of each plaque consisting of non-fibrillar Aβ (Fig. 3F). However, the APPPS1;EKO mice had an even greater and more dispersed accumulation of non-fibrillar Aβ than the APPPS1;FE3Cre + and APPPS1;FE4Cre + mice (Fig. 3F). In order to assess how much of the total Aβ that was accumulating was forming fibrillar amyloid structures, the amount of fibrillar Aβ detected by X-34 staining was compared to the total Aβ stained by HJ3.4. Immunostaining with HJ3.4 detects all forms of Aβ, so all X-34 positive Aβ plaques will also be detected by HJ3.4 staining, but not all Aβ detected with HJ3.4 will be positive for X-34. Therefore, comparing the amount of X-34 staining to the total amount of HJ3.4 staining provides insight into differences in the structural formation of plaques. The ratio of the X-34 area to HJ3.4 area revealed that the male Cre + mice had an approximate 65% reduction in the X-34 to HJ3.4 ratio, while for the females only the APPPS1;FE4Cre + mice had a significant reduction in the ratio of about 50% (Fig. 3G).

Overall, the biochemical results, combined with the Aβ staining results, show that the loss of astrocytic human apoE leads to a decrease in overall Aβ pathology compared to mice with no loss of apoE. Additionally, the loss of astrocytic apoE leads to more diffuse, less fibrillar plaques, though not as diffuse as is seen with a complete knockout of apoE. The loss of astrocytic apoE also reduces the total amount of fibrillar amyloid plaques that form in relation to the total amount of Aβ that deposits in the parenchyma. However, the changes observed in the overall Aβ plaque load, and in the types of Aβ plaques that form, could result from changes in the overall APP levels and/or in the processing of APP that occur with the loss of astrocytic apoE. To address this concern, we performed Western blotting of tissue samples for full-length APP and APP-CTFs and saw no differences between groups (Fig. S2C, S2D), indicating there were no differences in APP levels or in the processing APP after removal of astrocytic apoE.

The activation state of microglia has been shown to be influenced by apoE isoform and the presence of Aβ plaque pathology in the brain. To better understand how the loss of astrocytic apoE was influencing microglial gene expression, we performed RT-qPCR on cortical tissue samples. The APPPS1;FE3Cre- and APPPS1;FE4Cre- mice exhibited an upregulation of DAM-associated transcripts compared to FE3 or FE4 mice, respectively, indicative of a microglial response to amyloid plaques (Fig. S3D-J). The APPPS1;FE3Cre + and APPPS1;FE4Cre + mice also showed some upregulation of DAM associated genes, however the increase in these Cre + mice was reduced in comparison to the Cre- mice for several of the DAM genes (Fig. S3D-I), including Clec7a (Fig. S3F).

To assess if the overall levels of microglia were changed in the Cre + mice, potentially accounting for differences in overall DAM gene expression, brain sections were stained for Iba1 (Fig. 4A). Overall, the % area of the cortex covered by Iba1 staining remained unchanged between the APPPS1;FE3Cre + and APPPS1;FE3Cre- mice and between the APPPS1;FE4Cre + and APPPS1;FE4Cre- mice (Fig. 4B). However, the apoE3 female mice did show lower overall Iba1 staining levels than the other groups. To investigate the state of microglia surrounding Aβ plaques, an analysis of the amount of Clec7a, a marker of DAM microglia, was assessed based upon the total amount of Iba1+ microglia present (Fig. 4E). The amount of Iba1+ microglia clustering around X-34 plaques was unchanged between the Cre- and Cre + groups, indicating the loss of astrocytic apoE did not alter the ability of microglia to migrate to the plaques (Fig. 4F). However the level of Clec7a was shown to be decreased by ~ 50–70% around fibrillar plaques in the Cre + mice compared to the Cre- mice, (Fig. 4G). Furthermore, an assessment of the presence of apoE (which is also a marker of DAM microglia) in plaque associated microglia also showed a significant reduction of ~ 60–70% in Cre + mice compared to Cre- mice (Fig. 4C, D) The reduction of both microglial apoE and Clec7a levels in Cre + mice suggests an impaired microglial response to amyloid pathology in the absence of astrocytic apoE.

ApoE has been shown to influence certain functions of astrocytes, including responses to pathogenic stimuli [42, 43], however it is not fully known if the cell-specific loss of astrocytic apoE alters astrocyte reactivity to Aβ plaques. Therefore, we assessed the expression profile of genes involved with astrocyte reactivity by RT-qPCR. Overall expression of reactive astrocytic genes showed few differences between the Cre- and Cre + mice, including S100β (Fig. 5A). However, slight decreases of ~ 40–50% in GFAP expression were seen in the APPPS1;FE4Cre + mice, but not in the APPPS1;FE3Cre + mice (Fig. 5B). To visualize and further assess astrocyte reactivity, we assessed GFAP levels by immunostaining (Fig. 5C). Overall GFAP staining was not altered in APPPS1;FE3Cre + compared to APPPS1;FE3Cre- mice. However, there was a significant reduction in GFAP staining in the cortex of female APPPS1;FE4Cre + mice by 36% compared to APPPS1;FE4Cre- mice (Fig. 5B and C). The reductions in GFAP gene expression, and in immunostaining, in APPPS1;FE4Cre + mice could result from altered astrocyte reactivity to amyloid plaques or from overall reductions in amyloid burden. Therefore, to see if astrocyte reactivity was altered specifically around Aβ plaques, we assessed GFAP staining within 15 μm of each plaque (Fig. 5C). No differences were found in the levels of GFAP staining around plaques between the groups (Fig. 5E). Since GFAP staining around plaques was unaltered, it suggests the loss of apoE in astrocytes does not alter the ability of astrocytes to respond to Aβ plaques, at least morphologically, but that overall changes to astrocyte reactivity in the APPPS1;FE4Cre + mice may be driven by the level of total Aβ plaque pathology.

The deposition of Aβ into fibrillar amyloid plaques results in damage to surrounding neuronal processes and leads to the formation of large swollen axons and dendrites around plaques (neuritic dystrophy). These damaged and swollen neurites contain accumulations of various proteins that can act as markers for the dystrophy, including BACE1 [44, 45] and RTN-3 [46, 47]. We used an anti-BACE1 antibody to stain and assess the total amount of dystrophic neurites present in the cortex of mice (Fig. 6A). The BACE1 levels for the female Cre + mice were significantly decreased by ~ 50–70% compared to the female Cre- mice (Fig. 6B). As the BACE-1 positive dystrophic neurites are only present around amyloid plaques, this is consistent with overall reductions in amyloid plaque following removal of astrocyte apoE. When we assessed the area of BACE1 within 15 μm of each plaque, we found a significant increase in neuritic dystrophy in the male Cre + mice of ~ 30% and an increase of 50% in female APPPS1;FE4Cre + mice (Fig. 6C and D). To further assess neuritic dystrophy around plaques, staining for RTN-3 was also performed (Fig. 6E, S5). The RTN-3 levels in female Cre + mice were found to be ~ 90–115% higher around plaques than in Cre- females (Fig. 6F).