The adaptive immune system restrains Alzheimer’s disease pathogenesis by modulating microglial function

To examine the impact of the adaptive immune system on AD pathogenesis, we backcrossed a well-established AD transgenic line, 5xfAD mice (19), onto a Rag2^−/−/Il2rγ^−/− double-knockout background, creating mice that lacked T cells, B cells, and natural killer (NK) cells. Although NK cells are typically considered part of the innate immune system, recent studies suggest they also play important roles in adaptive immunity (20). In the process of generating these immune-deficient Rag-5xfAD and “Rag-WT” littermates, we also produced strain-matched equivalent immune-competent AD transgenic and wild-type mice termed “WT-5xfAD” and “WT-WT,” respectively (Fig. S1). Although the parental 5xfAD mice are maintained on a purebred C57Bl6 strain, the Rag2^−/−/Il2rγ^−/− line is maintained as a C57Bl6/Bl10 hybrid. To verify that the resulting lines exhibited equivalent genetic backgrounds, we therefore examined tail DNA using a 384 mouse SNP array (Charles River Laboratories), which revealed equivalent mixtures of Bl6 and Bl10 alleles in each group.

To confirm that the resulting Rag-5xfAD and Rag-WT mice lacked B, T, and NK cells, flow cytometry was performed on splenocytes extracted from 6-mo-old mice. Analysis confirmed that all immune-deficient animals, regardless of AD transgene expression, lacked B cells, CD4 and CD8 T cells, as well as NK cells and NK T cells (Fig. S1 B–E). Given the large role of inflammation in the pathogenesis of AD, we also examined whether immune-competent animals exhibited altered proportions of these immune cells. Analysis of splenocytes by flow cytometry found that percentages of B, CD8 T, NK, and NK-T cells were all unchanged with 5xfAD transgene expression. However, there was a small but significant difference in the percentages of CD4 T cells between the WT-WT and WT-5xfAD animals (Fig. S1 B–E).

To determine whether deletion of these peripheral immune cell populations influences AD pathogenesis, we performed a highly sensitive multiplex ELISA to quantify soluble and insoluble levels of Aβ within the brain. Surprisingly, and despite the use of an aggressive model of amyloidosis, we found that all Aβ species examined (Aβ38, 40, 42) were elevated nearly twofold in Rag-5xfAD vs. WT-5xfAD half-brains (Fig. 1 A and B). Additionally, immunohistochemistry was performed and plaque volume calculated from confocal z-stack images using IMARIS image analysis software (Fig. 1 C and D). Analysis of the dentate gyrus of the hippocampus, one of the most plaque-dense areas in the 5xfAD model, revealed that Rag-5xfAD mice exhibited a highly significant (P < 0.001), nearly fourfold increase in total plaque volume in this region (Fig. 1E). Quantification of small-, medium-, and large-sized plaques also revealed that Rag-5xfAD animals exhibit 2.5- to 3-fold increases in the number of plaques of each size compared with WT-5xfAD (P < 0.05; Fig. 1F).

Given the substantial increase in Aβ observed in Rag-5xfAD mice, we next sought to determine whether these findings arose from increased Aβ production or decreased clearance. Although autosomal-dominant AD is characterized primarily by mutations that increase production of Aβ or Aβ42/40 ratio (21, 22), recent studies demonstrate that sporadic AD patients primarily accumulate Aβ as a result of impaired clearance (23–25). However, the 5xfAD model, as with most AD transgenic animals, includes familial AD mutations, and thus increased Aβ production could potentially underlie the observed changes in amyloid load. We therefore examined the protein levels of human amyloid precursor protein (APP) and Presenilin-1 (PS-1) by Western blot. Although Rag-5xfAD and WT-5xfAD mice exhibited the expected transgene-mediated increases in APP and PS-1 vs. wild-type controls, no differences in APP and PS-1 expression were detected between Rag-5xfAD and WT-5xfAD groups (Fig. S2 A–C). Furthermore, quantitative PCR (qPCR) analysis of APP processing genes found no significant changes in expression of hAPP, hPSEN1, Adam10, Adam17, Bace1, and Bace2 between WT-5xfAD and Rag-5xfAD (Fig. S2D). Thus, it appears that the observed elevations in plaque load in Rag-5xfAD are not due to increased APP production and/or processing, but rather likely mediated via altered Aβ clearance.

Next, we used microarrays and functional genomic analysis to provide an unbiased assessment of the pathways that are altered in immune-deficient vs. immune-intact AD and wild-type mice. Whole-brain mRNA from all four groups was compared using Affymetrix Mouse Transcriptome 1.0 array, revealing 2,552 significantly altered genes between Rag-5xfAD and WT-5xfAD groups (Fig. 2 A and B). Interestingly, the great majority of these changes reflected decreased expression (green) of many T-cell– and B-cell–associated transcripts in Rag-5xfAD vs. WT-5xfAD mice (Fig. 2B), supporting the notion that these peripheral populations can be found within the brain. In contrast, many of the up-regulated genes (red) in Rag-5xfAD mice consisted of chemokine and cytokine signaling factors involved in recruitment of peripheral cells as well as altered microglial-enriched transcripts. Next, GO analysis was used to identify potential biological or signaling mechanisms that were significantly altered between Rag-5xfAD and WT-5xfAD brains. This analysis further implicated both adaptive and innate immunity as changes in GO terms such as cytokine activity, Ig binding, and antigen-receptor–mediated signaling were highly significant (Fig.2C). To determine whether changes affected microglia function and whether microglial-associated transcripts were specifically altered, we also performed hierarchical cluster analysis of microglial-enriched genes and compared these results to neuronally enriched mRNAs (Fig. 2 D and E), as identified from a recently published online RNA-seq database (26). As shown, Rag-5xfAD microglial gene expression clusters very well together but completely separate from WT-5xfAD mice. In addition, Rag-5xfAD mice exhibit profound increases (red) in the great majority of microglial transcripts (Fig. 2D), far beyond that observed in WT-5xfAD mice compared with the WT-WT group. Interestingly, no significant changes are detected between WT-WT and Rag-WT microglial genes (Dataset S1) and these groups fail to separate by hierarchical clustering (Fig. 2D), suggesting that the impact of the adaptive immune system on microglial gene expression occurs primarily in response to AD pathology rather than an inherent developmental difference. In contrast to the interactions between AD pathology and immune deficiency on microglia gene expression, neuronal genes showed far greater variability as Rag5xfAD and WT-5xfAD groups failed to cluster together (Fig. 2E). Thus, changes in neuronal expression are far more variable and less systemic than the observed microglial alterations. Together, these analyses prompted us to further examine the role of microglia dysfunction in Rag-induced exacerbation of AD pathology.

Given that our genomic analysis strongly implicated microglial dysfunction, we next sought to determine whether alterations in microglial number or morphology could explain the increased Aβ. To quantify microglial number and to model the branching of microglial processes, immunohistochemical and IMARIS-based 3D analysis of Iba1+ microglia was performed (Fig. 3 A–D). Upon activation by reactive stimuli, microglial typically become more amoeboid with shorter and less complex processes that can be accurately quantified via unbiased IMARIS 3D rendering (7). Confocal Z stacks were captured from the dentate gyrus by a blinded observer, and then microglial number, process length, and branching were assessed. As expected, we found that WT-5xfAD mice exhibited substantial increases in microglial cell number vs. WT-WT mice and a corresponding reduction in microglial branching and process length, indicative of activated phenotype (Fig. 3 E–G). However, we also found that Rag-5xfAD mice exhibited an additional significant increase in microglial number and decreases in branching and process length vs. WT-5xfAD mice, suggesting that Rag-5xfAD microglia may be further or differentially activated relative to WT-5xfAD mice (Fig. 3 E–G). Interestingly, we also found that Rag-WT microglia had on average slightly longer processes and a small increase in branching vs. WT-WT microglia (Fig. 3 F and G). Thus, it appears that the loss of B, T, and NK cells can subtly modulate the phenotype of microglia within the brain in the absence of pathology.

Given the observed changes in microglial number and morphology in Rag-5xfAD mice, we reasoned that other alterations in microglial function likely occur. We therefore used a multiplex ELISA-based assay [Meso Scale Discovery (MSD) proinflammatory panel 1] to measured protein levels of several important cytokines within the brain. For each of these cytokines, we found no differences between WT-WT and Rag-WT mice (Fig. 3 H–L). In contrast, several key proinflammatory cytokines produced by microglia including interleukin 1β (IL-1β), interleukin 6 (IL-6), and tumor necrosis factor α (TNFα) were significantly elevated in WT-5xfAD animals compared with WT-WT mice. Furthermore, these same cytokines were even further elevated as much as twofold in Rag-5xfAD mice compared with WT-5xfAD (Fig. 3 H–J). Although levels of interferon-gamma (IFNγ) were significantly increased in WT-5xfAD animals compared with WT-WT, no change was observed in Rag-5xfAD animals from basal Rag-WT levels, consistent with a lack of T-cell–derived IFNγ that would be expected in Rag mice (Fig. 3K). In addition to traditional proinflammatory cytokines, two recent studies demonstrated that the antiinflammatory cytokine interleukin 10 (IL-10) negatively regulates amyloid pathology (27, 28). We therefore also examined levels of IL-10 in all four groups. However, our analysis revealed no differences in IL-10 between Rag-5xfAD and WT-5xfAD (Fig. 3L), and thus, IL-10 is likely not responsible for the observed differences in Aβ in this model.

As microglial gene expression, morphology, and cytokine production were different in Rag-5xfAD mice, we wondered whether microglial phagocytosis of Aβ might also be altered. We therefore used high-power confocal z-stack imaging with IMARIS software colocalization, surface reconstruction, and volumetric quantification to analyze Aβ phagocytosis (28–30). Using this approach, we quantified the proportion of Aβ localized to microglial phagolysosomes and detected a nearly 39.6% decrease in phagocytic efficiency (P = 0.016) in Rag-5xfAD microglia vs. WT-5xfAD microglia (Fig. 4).

While examining immunolabeling with various mouse monoclonal antibodies in WT-5xfAD mice, we observed an unexpected but consistent pattern of microglial labeling. We hypothesized the labeling could be endogenous mouse Ig within the brain. We therefore used an anti-mouse f(ab′)2 fluorescently labeled secondary antibody to determine whether endogenous mouse IgG might be present within the WT-5xfAD brain. Indeed, we found significant immunolabeling of mouse IgG in association with Iba1+ microglia in WT-5xfAD brains (Fig. 5 A–C). In contrast, IgG labeling was greatly diminished in WT-WT mice (Fig. 5 E–G) and absent in Rag-5xfAD and Rag-WT mice (Fig. 5 E–O). Further examination of endogenous IgG labeling also revealed a high degree of labeling in the choroid plexus of both WT-WT and WT-5xfAD mice (Fig. 5 D and H). Although the choroid plexus can often exhibit nonspecific immunoreactivity, the specificity of this IgG labeling was confirmed by the absence of staining in the Rag-WT and Rag-5xfAD mice (Fig. 5 L and P). To further examine the differences in IgG levels between WT-5xfAD and WT-WT brains, we measured total mouse IgG within the brain and plasma by ELISA. Confirming our histological findings, we detected a significant 50% increase in IgG levels in the brains of WT-5xfAD vs. WT-WT mice (Fig. 6A). However, this difference was not due to an increase in peripheral IgG levels, as plasma IgG measurements revealed no differences between WT-WT and WT-5xfAD mice (Fig. 6B).

The observed increase in WT-5xfAD brain IgGs could be mediated by a breakdown in the blood–brain barrier (BBB), as recent studies have shown that the BBB can become disrupted in AD (31, 32). To determine whether this was the case, we used an Evans Blue permeability assay to assess BBB integrity (33) in all four genotypes. Interestingly, this experiment revealed no alterations in BBB permeability (Fig. S3 A and B); thus, IgG accumulates in the brains of WT-5xfAD by an alternative mechanism that currently remains unclear.

To determine whether these endogenous IgGs were directed against Aβ, we collected sera from WT-WT, WT-5xfAD, and Rag-5xfAD mice and measured anti-Aβ antibody titers using a well-validated ELISA (34, 35). Despite testing extremely undiluted samples of sera, we found no differences between WT-WT and WT-5xfAD in levels of anti-Aβ IgG antibodies (Fig. 6E). In contrast, highly diluted samples taken from positive control mice that had been actively immunized against Aβ (red bar) exhibited a robust signal in this assay (Fig. 6E). Thus, neither WT-WT nor WT-5xfAD mice exhibit induction of anti-Aβ antibodies. To further confirm these results, we used a sensitive ELISpot assay to quantify numbers of anti-Aβ IgG producing B cells within the spleen. Using Rag-5xfAD mice that lack B cells as a negative control and mice actively immunized with Aβ peptide as a positive control, we detected no differences in anti-Aβ IgG producing B cells between WT-WT and WT-5xfAD splenocytes (Fig. 6C). As the immune response to a brain-localized antigen would be expected to be increased within the cervical lymph nodes, we next repeated this assay using cells extracted from the deep and superficial cervical lymph nodes. Analysis again revealed no differences between WT-WT and WT-5xfAD mice (Fig. 6D). We also performed an assay to determine whether WT-5xfAD mice exhibited a specific T-cell response to Aβ as indicated by IFNγ production following Aβ exposure. Similar to the B-cell ELISpot, this assay again revealed no differences in the reactivity of T cells to Aβ between genotypes (Fig. 6F). Last, further examination of brain IgG localization by high-magnification confocal microscopy confirmed that WT-5xfAD IgGs do not directly associate with amyloid plaques (Fig. 6 G–J) and instead colocalize with Iba1+ microglia, again suggesting that these brain-localized antibodies are not Aβ specific.

To further confirm the importance of brain-localized IgG in restraining Aβ pathology, we used Image Streamer flow cytometry to accurately quantify the effects of preimmune IgGs, purified from the sera of wild-type unimmunized mice, on microglial Aβ phagocytosis. Unlike standard flow cytometry, the Amnis Image Streamer system captures 256 fluorescent and phase-contrast images of cells as they pass through the system. On-board computer algorithms then calculate whether a given fluorescent signal is localized within the cell as opposed to being stuck to the outside membrane (Fig. 7A). Using this system, we examined whether the presence of preimmune mouse IgG could induce the phagocytosis of fibrillar fluorescently labeled Aβ by BV2 microglial cells. Cells were cultured in serum-free media and treated overnight with or without IL-1β to further mimic the proinflammatory state of the AD brain. Ten hours later, cells were exposed to fluorescently labeled fibrillized Aβ with or without preimmune IgGs for 1 h and then examined using the Image Streamer system. Interestingly, whereas relatively few cells phagocytosed Aβ following treatment with PBS/PBS, treatment with preimmune IgG induced a sixfold increase in the number of cells with internalized Aβ (Fig. 7 A–C). Furthermore, of those cells that did actively phagocytose Aβ, ones treated with IgG showed a twofold increase in the amount of Aβ phagocytosed as measured by mean fluorescence intensity within the cell (Fig. 7D). IL-1β/PBS pretreatment did not significantly change the amount of phagocytosis compared with PBS/PBS treatment. However, combined treatment of IL-1β/IgG significantly reduced the number of cells that phagocytosed Aβ vs. IgG alone, but slightly increased the amount of Aβ within those cells (Fig. 7 C–E), consistent with the known effects of IL-1β on microglial activation state and phagocytosis (36).

Next, we performed an additional in vitro experiment to determine the mechanism by which preimmune IgG activates phagocytic machinery to enhance Aβ clearance. IgG binding to Fc receptors can activate a signal transduction cascade that leads to membrane remodeling, actin reorganization, and the formation of a phagasome (reviewed in ref. 37). In macrophages and monocytes, this signaling pathway is initiated via phosphorylation of immunoreceptor tyrosine-based activation motifs (ITAMs) by enzymes of the Src tyrosine-kinase family and the subsequent docking and activation of spleen tyrosine kinase (Syk). This in turn activates Phosphatidylinositol 3-kinase (PI3K) and other downstream signaling pathways to induce phagosome formation. To determine whether IgG treatment of BV2 microglial cells activates this same canonical pathway, we pretreated BV2 cells with inhibitors of Src/Syk or PI3K and then exposed them to fluorescent Aβ and IgG for 30 min. Cells were then washed and fixed, and fluorescent Aβ internalization was imaged and quantified by a blinded observer. As before, we again found that treatment with preimmune IgGs dramatically increases Aβ phagocytosis (Fig. S4). In contrast, pretreatment with a selective inhibitor of Src and Syk tyrosine kinases [3,4-methylenedioxy-β-nitrostyrene (MNS)] or the PI3K inhibitor wortmannin, completely abrogated Aβ phagocytosis (Fig. S4). Thus, it appears that IgG-induced uptake of Aβ by microglial cells involves activation of Src/Syk/PI3K phagocytosis signal transduction pathways.

Consistent with our in vitro analysis, previous studies have demonstrated that injection of preimmune IgG into the brain of a different transgenic AD model can promote Aβ clearance (38). To determine whether a similar approach can reduce Aβ in a mouse model that lacks B, T, and NK cells, we replicated this design by stereotactically injecting preimmune IgG (2 μL of 1 mg/mL) into one hippocampus and PBS into the contralateral hippocampus of Rag-5xfAD mice. Seven days later, mice were killed, and amyloid burden was assessed, revealing a significant reduction in Aβ plaque volume by 50.9% (P = 0.01) on the IgG-injected side (Fig. S5 A–C). Further subanalysis demonstrated that IgG delivery also significantly reduced the number of large plaques (P = 0.01; Fig. S5D).

To determine whether replacing the missing adaptive immune cell populations in the periphery could also reverse the effects of immune deficiency on AD pathology, adoptive transfer of whole bone marrow was performed. At 2 mo of age, Rag-5xfAD mice received an injection of either 500,000 whole bone marrow cells from age- and sex-matched immune-intact WT-5xfAD mice or an equivalent injection of vehicle into the retroorbital venous sinus. Mice were allowed to age normally for 4 mo, and then brains were processed for analysis. Successful engraftment was confirmed by flow-cytometric analysis of splenocytes at the time of killing (Fig. S6 A and B). Next, using both biochemical and histological endpoints, we assessed the impact of bone marrow transplantation on Aβ pathology. Immunohistochemical analysis of hippocampal amyloid burden revealed a significant 47.4% reduction in plaque volume in mice that received bone marrow transplants (P < 0.05; Fig. 8 A–C). Subanalysis of plaque load demonstrated that numbers of both medium and large plaques were significantly reduced with bone marrow transplantation (P < 0.05) and numbers of small plaques were almost significantly decreased (P = 0.056) (Fig. 8D). To confirm our immunohistochemical analysis, we followed up with ELISA measurement of Aβ and detected a significant reduction in Aβ40 and nearly a significant reduction in Aβ38 (P = 0.058) (Fig. 8E). To determine whether endogenous mouse IgG was present in the brain following bone marrow engraftment, we also examined mouse IgG levels by ELISA and immunohistochemistry. These analyses confirmed that replacement of peripheral immune cells could elevate levels of IgG within the brain (Fig. 8F) and likewise led to the characteristic colabeling of microglia with endogenous mouse IgG (Fig. 8 G–N). Taken together, these data strongly implicate adaptive immunity, the production of IgGs, and cross talk between peripheral and central immune systems as being critical for regulating and restraining the development of AD pathology.

All animal procedures were performed in strict accordance to the guidelines of the National Institutes of Health and University of California Institutional Animal Care and Use Committee. The Rag-5xfAD immune-deficient AD mouse model was created by backcrossing 5xFAD transgenic mice onto a Rag2/il2rγ double-knockout background. Briefly, 5xfAD mice (MMRRC strain: 034848-JAX) express two cointegrated and coinherited transgenes (APP and PS-1). The APP transgene includes three familial AD mutations (Swedish, Florida, London), and the PS-1 transgene includes two mutations (M146L and L286V) (19). Purebred C57Bl6 5xfAD mice were crossed with Rag2/il2rγ double-knockout mice (Taconic; 4111) (59), followed by repeated littermate crosses from each generation (Fig. S1A), to create mice that are heterozygous for the APP/PS-1 transgenes and lack both copies of the Rag2 and Il2rγ genes (Il2rγ −/y in males). We also simultaneously generated strain-matched immune-deficient mice wild type for the AD transgenes hereafter referred to as Rag-WT, as well as strain-matched immune-competent AD and WT mice referred to as WT-5xfAD and WT-WT, respectively. Sex- and age-matched mice were used, and all animals were group housed on a 12-h/12-h light/dark cycle with access to food and water ad libitum.

Fluorescent immunohistochemical analysis followed previously described established protocols (60). Primary antibodies used for immunohistochemical analysis included the following: Iba1 (Wako), Iba1 (Abcam), CD68 (Abcam), Aβ40 (Life Technologies), Aβ42 (Life Technologies), 82E1 (IBL America), GFAP (Millipore), f(ab′)2 fragment of goat anti-mouse IgG directly conjugated to Alexa Fluor 488 or 555 (Life Technologies). Sections were incubated in primary antibodies overnight followed by detection with appropriate Alexa Fluor-conjugated secondary antibody (Life Technologies) and coverslipped using Fluoromount-G with or without DAPI (Southern Biotech). Fibrillar amyloid was visualized using Amylo-Glo (Biosensis) diluted 1:100 in 0.01 M PBS (61).

Immunofluorescent sections were visualized, and images were captured using an Olympus FX1200 confocal microscope. Aβ plaques were quantified using the Surfaces function followed by volume analysis in IMARIS software (Bitplane). IMARIS-based quantification of microglial number, morphology, and Aβ phagocytosis followed previously described methods using the Colocalization and Surfaces functions (28–30, 33). To calculate Aβ internalization ratio, the volume of Aβ within CD68⁺ phagolysosomes was normalized to microglia number and total Aβ volume within the field. All images were captured and analyzed by a blind observer using coded slides.

mRNA was extracted from frozen half-brains using RNA Plus Universal Mini Kit (Qiagen), and sample purity and concentration were verified by Bioanalyzer (Agilent). Affymetrix GeneChip Mouse Transcriptome 1.0 arrays were then used to collect global transcriptional profiles in 16 samples (4 per group) by the University of California, Irvine (UCI) Genomics High-Throughput Facility. Background correction and normalization of raw data were conducted via robust multiarray analysis algorithm implemented in Bioconductor package “oligo 1.32.” All samples passed quality control analysis. Additionally, array features were filtered with Bioconductor packages “genefilter 1.50” and “mta10sttranscriptcluster.db 8.3.1” for Entrez IDs resulting in ∼24,600 genes for further analysis.

Unsupervised hierarchical clustering (R packages flashClust 1.01-2 and gplots 2.17) were performed on subsets of genes that have previously been shown to be enriched in microglia or neurons (26). Differentially expressed genes (DEGs) were determined between all contrasts of genotypes, and multiple testing was done with Benjamini–Hochberg method to obtain the corresponding |logFC| and P value implemented in Bioconductor package “limma 3.24.25.” Genes with adjusted P value of <0.01 were used for constructing Venn diagram (JMP Pro-11.2 software) and GO analyses. Bioconductor packages “topGO 20.2,” “GO.db 3.1.2,” and “gage 2.18” were used to perform GO enrichment for the DEGs found significant in Rag5xfAD vs. WT5xfAD contrast. P < 0.05 was defined as the cutoff to identify the statistical significance of enrichment analyses.

Blood was collected from retroorbital sinus, and titers of anti-Aβ antibodies in mouse sera were determined by ELISA as previously described (34, 35). Spleens, deep cervical lymph nodes (DCLNs), and superficial cervical lymph nodes (SCLNs) were collected, and antibody-forming B cells specific to Aβ were detected in splenocytes and pooled DCLN/SCLN lymphocytes by ELISPOT (Mabtech). Splenocytes or lymphocytes were incubated for 24 h in 96-well plates coated with Aβ peptide, and the assay was performed following manufacturer’s protocol (Mabtech). Sera and splenocytes were also collected from mice immunized with Aβ as a positive control (34, 35). Analysis of IFNγ production by T cells was performed in DCLN/SCLN lymphocytes by ELISpot assay (BD Biosciences). Cultures of lymph node cells were restimulated in vitro with soluble Aβ (10 µg/mL) for 20 h. The numbers of spot-forming cells per 10⁶ lymph node cells stimulated with Aβ were then counted.

Phagocytosis was assessed following previous methods with some modifications (36, 62). First, BV2 cells (63) (3.0 × 10⁵ cells per well) were exposed to either proinflammatory cytokine IL-1β (20 ng/mL; Thermo Fisher Scientific) or 0.1% BSA in PBS for 10–12 h. After IL-1β or PBS preexposure, the cells were changed to serum-free DMEM and then exposed to fAβ_1–42 (10 μg/mL) with murine IgG (0.5 mg/mL; Jackson ImmunoResearch Laboratories) or PBS for 1 h at 37 °C. Next, the cells were washed three times with prewarmed PBS to remove unassociated fAβ_1–42. Cold PBS was added to wells, and the cells were collected, centrifuged (5 min, 600 × g), washed in PBS, and suspended in flow cytometer buffer (PBS with 0.1% BSA and 0.5 mM EDTA) and placed on ice. BV2 cells were incubated with rat anti-CD16/32 Fc receptor block (2 mg/mL; BD Biosciences) for 5 min at 4 °C. Cells were then stained with anti–CD45-APC clone 30-F11 (Tonbo Biosciences) at 1:200 in flow cytometer buffer. Samples were then analyzed using Amnis Imagestreamer^x Mark II Imaging Flow Cytometer (Millipore). fAβ_1–42 phagocytosis was analyzed using the IDEAS software onboard Internalization Wizard algorithm. Figures represent single cells that internalized fAβ_1–42 from each group (n = 6) with mean and SE reported.

Age- and sex-matched immune intact 5xfAD mice served as donors for bone marrow adoptive transfer. Donor mice were killed by CO₂ asphyxiation, femurs were removed, and whole bone marrow was harvested by flushing the marrow contents with PBS. Marrow was then treated with ammonium chloride–potassium buffer to lyse red blood cells, filtered through a 70-μm nylon mesh, and cell numbers were counted by hemocytometer. Each recipient (n = 4–5 females/group) was anesthetized with isoflurane before receiving 500,000 live cells in 100 μL or equivalent volume of PBS via retroorbital injection. To confirm engraftment, at the time of killing, splenocytes were collected and analyzed for the presence of B, T, and NK cells by flow cytometry as described in SI Materials and Methods.

Detailed descriptions of all other standard experimental procedures, including tissue processing, flow cytometry, Western blot, ELISA, Evans Blue assay, qPCR, cell culture, Aβ preparation, phagocytosis inhibition, intracranial IgG injection, and statistical analysis, can be found in SI Materials and Methods.

At 7 mo of age, mice were anesthetized with Euthasol before intracardial perfusion with 0.01 M PBS. Immediately following perfusion, brains were quickly removed, and hemispheres were separated. The cerebellum was removed from the right hemisphere and then immediately flash-frozen in dry ice for subsequent biochemical analysis. The left hemisphere was drop-fixed in 4% (wt/vol) paraformaldehyde for 48 h at 4 °C. After 48-h fixation, the left hemispheres were transferred to 0.01 M PBS and 0.02% NaN₃ for storage until sectioning. Brains were sectioned coronally on a freezing microtome at 40-μm thickness.

Right hemispheres, previously frozen on dry ice and stored at −80 °C, were crushed on dry ice using mortar and pestle. Approximately one-quarter of homogenate was separated for mRNA analyses, and the remaining portion was subsequently homogenized in solution of T-PER (Pierce) and phosphatase and protease inhibitor mixtures (Thermo Scientific and Roche) and processed as previously described (60).

Lymphocytes isolated from the spleen were immunophenotyped with fluorescent antibodies for the following cell surface markers for pan T cells: PE-conjugated CD3e (500A2; BD Biosciences), PerCP-eFluor710–conjugated CD3e (17A2; eBioscience); CD4 and CD8 T cells: PE-conjugated CD4 (GK1.5; BD Biosciences), Pacific Blue-conjugated CD4 (RM4-5; Biolegend), APC-conjugated CD8 (Ly-2; BD Biosciences), PE/Cy7-conjugated CD8 (53-6.7; eBioscience); NK cells: APC-conjugated CD49b (DX5; BD Biosciences), PE-conjugated NK1.1 (PK136; Biolegend); B cells: FITC-conjugated CD45R/B220 (RA3-6B2; BD Biosciences), Alexa Fluor 710-conjugated CD19 (ebio1D3; eBioscience). All cells for flow cytometry were FC blocked with anti-CD16/32 (1:200; BD Biosciences) for 10 min at 4 °C before staining with surface marker antibodies. Cells were run on an LSRII flow cytometer (BD Biosciences) or BD FACSAria II (BD Biosciences) and analyzed with FlowJo software (FlowJo).

SDS/PAGE Western blot and dot blot were performed according to previously established protocols (60). Primary antibodies used for Western blot analysis included the following: 6E10 (BioLegend) and PS1 (Novus Biologicals). Following initial labeling, blot membranes were stripped in 0.2 M NaOH for 20 min followed by block and labeling with anti-GAPDH (Santa Cruz) as loading control.

Direct injection of IgG into hippocampus was adapted from previous published reports (38). Stereotaxic injection of whole-molecule IgG (Jackson ImmunoResearch) was performed according to previously described surgical protocols (38, 60). Briefly, 4-mo-old mice were anesthetized and placed in stereotaxic frame (Kopf) under continuous isoflurane anesthesia. Using 10-μL Hamilton syringe and 30-gauge needle, mice received 2-μL injection of 1 mg/mL IgG in the right hemisphere and 2 μL of 0.01 M PBS in the left hemisphere at the following coordinates relative to bregma: anteroposterior, −2.06 mm; dorsoventral, −2.25 mm; mediolateral, ±1.5 mm. Needle was cleaned with consecutive washes of PBS, 70% (vol/vol) ethanol, and PBS in between hemispheres and animals. Animals were allowed to recover on heating pad before being placed back in home cages. One week following surgery, animals were perfused with 0.01 M PBS and entire brain removed for immunohistochemistry, confocal microscopy, and IMARIS analysis as described above.

Evans Blue dye injection and BBB permeability analysis were performed according to a previously published protocol (33). Mice were injected i.p. (4 mL/kg) with 2% (wt/vol) Evans Blue (Sigma-Aldrich) in PBS. After 6 h, mice were perfused, and liver, kidney, and brain were removed and placed on ice. Organs were homogenized in PBS at 150 mg/mL (vol/wt) and centrifuged at 10,000 × g for 10 min, and then supernatant was loaded onto a clear 96-well plate in duplicate. Intensity of Evans Blue dye was then measured by spectroscopy at 595 nm. An uninjected mouse acted as a blank control, and its tissue absorbance values were subtracted from experimental animal values.

Total mouse IgG levels were measured by ELISA according to manufacturer’s protocol (eBioscience). Mouse brain homogenate was loaded neat and normalized to the total protein content of each sample. Mouse plasma samples were diluted 1:10,000 as directed by manufacturer before being loaded onto the plate.

Quantitative biochemical analysis of mouse cytokine and human Aβ were conducted using commercially available electrochemiluminescent multiplex assay system [Meso Scale Discovery (MSD)]. Human Aβ triplex (6E10 capture antibody) was used for simultaneous measurement of Aβ38, Aβ40, and Aβ42 in both soluble and insoluble protein fractions. Mouse proinflammatory panel 1 was used for simultaneous measurement of proinflammatory cytokines in soluble brain protein fraction.

cDNA was produced from sample mRNA using Applied Biosystems High Capacity RNA-to-cDNA Kit (Thermo Fisher Scientific). All samples were then run in triplicate per gene on 384 qPCR plates using Applied Biosystems Gene Expression Master Mix (Thermo Fisher Scientific) according to manufacturer’s protocol using the following TaqMan primers: Human APP (Hs00169098_m1), Human PSEN1 (Hs00997789_m1), Adam10 (Mm00545742_m1), Adam17 (Mm00456428_m1), Bace1 (Mm00478664_m1), and Bace2 (Mm00517138_m1). Both probe only and probe plus RNA were run in triplicate per gene as negative controls. Plates were run immediately after loading on a Viia 7 Real-Time PCR System (Thermo Fisher Scientific). CT values were exported to Microsoft Excel and ΔΔct values and fold overexpression were calculated. Results are reported as fold overexpression relative to WT-5xfAD animals ± SEM.

The BV2 immortalized microglial cell line (63) [generously provided by Dr. Andrea Tenner (University of California, Irvine)] was cultured in complete media and incubated at 37 °C with 5% CO₂. Complete media was composed of DMEM high glucose (25 mM) supplemented with 10% (vol/vol) heat-inactivated FBS, sodium pyruvate (1 mM), Glutamax dipeptide (2 mM), and 1% penicillin/streptomycin.

Fibrillar fluorescent Aβ (fAβ_1–42) was generated as described previously (36). Briefly, fluorescently labeled Aβ_1–42 peptide (Anaspec) was first dissolved in 0.1% NH₄OH to 1 mg⋅mL⁻¹, and then further diluted in sterile endotoxin-free water and incubated for 7 d at 37 °C. fAβ was thoroughly mixed before cell exposure.

BV2 microglial cells were cultured as described (see BV-2 Microglial Cell Line Culture). Twenty-four hours before phagocytosis assay, cells were passaged and plated on chamber slides (Thermo Fisher Scientific) at density of 25,000 cells per chamber. Twelve hours before assay, cells were switched to serum-free media mirroring previous phagocytosis experiment. Thirty minutes before experiment, cells were treated with vehicle (DMSO), PI3K inhibitor, 20 nM wortmannin (Cell Signaling), or 50 μM Src/Syk kinase inhibitor, 3,4-methylenedioxy-β-nitrostyrene (MNS) (Tocris Bioscience). Following pretreatment, PBS or IgG (0.5 mg/mL) in addition to Aβ (5μg/mL) were added to the wells. After 30 min, cells were washed four times in prewarmed PBS and placed in 4% PFA for 30 min, followed by two subsequent PBS washes. Bright-field phase and fluorescent images were taken using Olympus IX71 microscope and 4× objective. Pictures from three independent wells per treatment were used for analysis by blind observer. Optical density of fluorescent images was measured following background subtraction using ImageJ and total cells per field were counted using the bright-field image and manual cell counter plugin of ImageJ.

Statistical analysis was performed using Statview 5 statistics and GraphPad Prism 6 software. Comparisons involving more than two groups used one-way ANOVA followed by Fisher’s post hoc test. Comparisons of two groups used two-tailed Student’s t test. All differences were considered significantly different when P < 0.05.