Microglial translational profiling reveals a convergent APOE pathway from aging, amyloid, and tau

Alzheimer’s disease (AD) is the leading form of dementia, marked by key pathological features of extracellular amyloid plaque deposition and intracellular neurofibrillary tangles composed of hyperphosphorylated tau (Holtzman et al., 2011). Aging is a major AD risk factor, and AD is disproportionally prevalent in females (Riedel et al., 2016). Microglia, the central nervous system (CNS)–resident innate immune population, are central cellular mediators during homeostasis and disease (Davalos et al., 2005; Nimmerjahn et al., 2005; Saijo and Glass, 2011; Nayak et al., 2014). Illuminating microglial transcriptome alterations during disease conditions is critical since genome-wide association studies have also identified several human genetic AD risk alleles that likely modulate microglial function including CR1, CD33, TREM2, and APOE (Corneveaux et al., 2010; Guerreiro et al., 2013; Liu et al., 2013; Efthymiou and Goate, 2017). Significant shifts in microglial sensome transcripts in aging and those that delineate a subpopulation of disease-associated microglia in aging and amyloidosis have been observed (Hickman et al., 2013; Keren-Shaul et al., 2017). Understanding how aging, amyloidosis, and tauopathy shape microglial transcriptomes and whether commonalities exist is critical to understanding how these factors may converge to exacerbate AD.

We hypothesized that microglial cellular isolation strategies may introduce unwanted bias, as has recently been reported for muscle stem cell populations (Davalos et al., 2005; Nayak et al., 2014; van den Brink et al., 2017). We therefore explored the use of the RiboTag translational Cre/Lox-based profiling system for microglial transcript isolation based on expression of a hemagluttanin (HA) epitope-tagged core ribosomal protein (RPL22) in a cell type of interest. A tamoxifen-inducible Cx3cr1 promoter for Cre and enhanced YFP (eYFP) expression (Cx3cr1-Cre^ERT2-IRES-eYFP; Parkhurst et al., 2013) was crossed to RiboTag mice to generate double heterozygous Cx3cr1^{CreERT2-IRES-eYFP/+}; Rpl22^HA/+ mice (referred to as RiboTag), to rapidly extract microglial transcripts without the necessity of cellular isolation (Fig. 1, A and B; Sanz et al., 2009; Parkhurst et al., 2013). In this approach, all mice are haploinsufficient for Cx3cr1, but this did not cause any general overt behavioral or cognitive phenotypes (Fig. S1, A–H).

Comparison of RiboTag versus cell enrichment with sorting (referred to generally here as “cellular isolation”) methodologies for microglial transcript isolation was conducted in tamoxifen injected RiboTag mice, which induced Cx3cr1-YFP⁺-HA⁺ cells (Fig. S1 I; Parkhurst et al., 2013). For the RiboTag method, mice were perfused with PBS, and forebrain tissue was immediately dispersed by brief sonication in cycloheximide-supplemented buffer to stabilize polysomes for microglial mRNA isolation using anti-HA antibody and magnetic bead purification (Doyle et al., 2008; Sanz et al., 2009). To control for drug and genotype effects, PBS-perfused forebrains from tamoxifen-injected RiboTag mice were used for cellular isolation that underwent enzymatic digestion, Percoll gradient myelin removal, staining, and flow cytometric sorting for the live 7AAD⁻ CD11b⁺ CD45^lo/int microglial population to >98% (Fig. S1 J). We used these markers since they, or CD11b bead-based purification, have been widely used for microglial profiling studies. We also compared the RiboTag method to cellular isolation during robust microglial activation induced by acute peripheral lipopolysaccharide (LPS) challenge (Erickson and Banks, 2011; Kang et al., 2018). Injection of 2 µg/g LPS i.p. resulted in early inflammation followed by microglial morphological changes and activation at 24 h after injection (Fig. S1, K–N). The four RNAseq microglial transcriptome groups (RiboTag-saline, RiboTag-LPS, Cellular Isolation–saline, and Cellular Isolation–LPS) were analyzed with a false discovery rate (FDR) set at q < 0.1 and absolute fold change (FC) > 1.25 as analysis cutoffs. Principle component analysis (PCA) demonstrated that principle component 1 of transcript isolation methodology dictated the majority of the variability, even more so than LPS stimulation (Fig. 1 C). Comparison of the top 25 transcripts enriched (reads per kilobase per million mapped reads [RPKM] > 10-fold higher over input) in microglia and ranked by abundance according to RiboTag transcriptomes demonstrated that RiboTag-defined transcripts were also highly enriched in the cellular isolation dataset, validating the RiboTag method (Fig. S1 O). Additionally, transcripts associated with microglia, but not other CNS cells, were robustly enriched from input samples using both methodologies (Fig. 1 D). Comparison of baseline (saline) datasets yielded many profound differences, with 289 transcripts that were >10-fold higher in the cellular isolation dataset compared with the RiboTag dataset, 61 of which were >100-fold higher, including S100a8, S100a9, Ngp, Retnlg, Ifitm6, Lcn2, and Mmp8 (Fig. 1 E). These transcripts may reflect a bias due to microglial activation and/or contaminating non-microglial CD45⁺CD11b⁺ cells such as neutrophils that could express these transcripts. However, aspects of cellular isolation procedures alone resulted in S100a8, Ngp, and Lcn2 up-regulation that was not entirely attributed to neutrophil contamination (Fig. S2, A–H). Flow cytometry revealed distinct S100A8 expression in cellular isolated TMEM119⁺ microglia (Fig. 1 F), in agreement with studies of activation-associated S100a8 and S100a9 expression in peripheral immune cells (Vogl et al., 2007; Austermann et al., 2014). Following in vivo LPS challenge, striking discrepancies in cellular isolation versus RiboTag responses were revealed by both PCA analysis (Fig. 1 C) and by the sparse overlap between the top 25 most LPS–up-regulated transcripts ranked by fold induction (Fig. S2, I and J). For some transcripts such as Lcn2 and Il1b, cellular isolation elicited tremendous basal transcript expression that dampened the overall magnitude of the LPS response (Fig. 1, G and H; Kang et al., 2018). Other transcripts such as Prok2 and Cst7, appeared to have a “two-hit” effect where cellular isolation in combination with LPS challenge resulted in an exaggerated response (Fig. 1, I and J). Therefore, to avoid potential skewing of the microglial transcriptomes, we used the RiboTag method to study microglial changes in mouse models of AD and normal aging.

To this end, we generated three datasets using mice on a RiboTag background that included: aging, amyloidosis (APP/PS1), and tauopathy (AAV-Tau^P301L; see Materials and methods for further model system details). Although the APP/PS1 and the AAV-Tau^P301L are overexpression systems, both provide insights regarding the two key pathological features of human AD: amyloidosis and tauopathy. Additionally, both rely on the expression of mutated forms of either the APP or MAPT gene, which were identified in humans, to initiate pathological disease. We set an FDR of q < 0.1, and absolute FC > 1.25 for all RNAseq studies. Distinct microglial transcriptomes shifts were observed by PCA during aging, amyloidosis, and tauopathy (Fig. 2, A–C). Additionally, two examples of acute inflammation, induced by either systemic injection of LPS or polyinosinic:polycytidylic acid (poly(I:C)), also showed distinct shifts in the microglial transcriptomes relative to saline controls (Fig. 2, D and E). The top 100 up-regulated transcripts sorted by aging (Fig. 2 F), amyloidosis (Fig. 2 G), and tauopathy (Fig. 2 H), revealed a striking pattern where aging had a >75% overlap in the top 100 amyloidosis and tau induced genes. Although significant shifts in the microglial transcriptome were observed with acute insults induced by LPS and poly(I:C) challenge, there was virtually no overlap with amyloidosis and tauopathy in the top 100 transcripts induced in these conditions (Fig. 2, I and J); however, within these acute inflammatory insult paradigms, LPS and poly(I:C) shared several transcripts and pathways as well as induction of unique genes (Fig. S3). Surprisingly, we found a significant amount of overlap between aging effects and LPS and poly(I:C) conditions (Fig. 2, I and J). Overall, these data indicated that aging appears to encompass aspects of both chronic and acute insults, suggesting that it could exacerbate inflammation induced with both types of neuroinflammation.

Further examination of shared transcripts suggests that tauopathy shares a common signature with aging and amyloidosis such as up-regulation of Cst7, Itgax, Gpnmb, Clec7a, Lpl, and Lgals3 (Fig. 3, A–C), along with Apoe and Spp1. In models of amyloidosis, the protein encoded by targets, such as Cst7, Lpl, and Itgax, have been observed in amyloid associated microglia by immunofluorescence, suggesting that specific subpopulations of microglia may express these transcripts (Keren-Shaul et al., 2017; Ofengeim et al., 2017). Examination of shared genes demonstrated that Cst7, the most up-regulated gene during amyloidosis, was also significantly induced across all conditions. Primary microglial cultures with siRNA knockdown of Cst7 demonstrated a significant increase in phagocytic activity (Fig. 3, E and F), suggesting that Cst7 induction in microglial cells may contribute to the decreased phagocytic activity observed during amyloidosis, tauopathy, or aging (Krabbe et al., 2013; Keren-Shaul et al., 2017; Koellhoffer et al., 2017; Salter and Stevens, 2017).

Surprisingly, Ccl3 and Ccl4 emerged as significant hits within the top transcripts shared across all three conditions (Fig. 3, A–D). Since we demonstrated that Ccl3 is prone to induction by cellular isolation (Fig. 1), this may have masked its relative importance in other studies. When examining the abundance of several shared transcripts, we found that Ccl3 is highly abundant and falls within the top 79.7, 94.8, and 88.8% of microglial transcripts expressed in aging, amyloidosis, and tauopathy, respectively, but was low in young, WT mice. Further, of the eight significant biological pathways shared between amyloid, tau, and aging, Ccl3 and Ccl4 were present in most of these, including inflammatory response (C3ar1, Cybb, Ccl3, Crl2, Axl, Tlr2, Clec7a, Ccl4, Cd14, and Spp1), positive regulation of tumor necrosis factor production (Ccl3, Tlr2, Ccl4, and Cd14), and neutrophil chemotaxis pathways (Ccl3, Lgals3, Ccl4, and Spp1; Fig. 3 G). Interestingly, CNS neutrophil recruitment/extravasation has been shown to exacerbate pathology and cognitive impairment in AD models (Baik et al., 2014; Zenaro et al., 2015). Additionally, sepsis, which can result in CNS neutrophil recruitment, has also been associated with cognitive decline (Iwashyna et al., 2010; He et al., 2016). These data suggest that Ccl3 and Ccl4 may be integral in microglial function in aging and disease, potentially through CNS recruitment of neutrophil or other peripheral immune cells.

Though not a microglia-specific gene, the absolute abundance of Apoe even under basal conditions places it in the top 99% of all transcripts expressed in microglia, even higher than known microglial marker genes such as Tmem119, Trem2, Cx3cr1, or Aif1, the gene encoding IBA1 (Fig. 4, A–D). Examination of Apoe revealed that it was significantly up-regulated in aging, amyloidosis, and tauopathy (Fig. 4, E–G). Strikingly, Ingenuity pathway analysis of shared microglial transcripts observed in these three conditions revealed a strong APOE-driven network that converged on CCL3 and CCL4 (Fig. 4 H). Recent work examining Apoe^−/− mice in an amyloidosis model demonstrated decreased Ccl3 and Ccl4 transcripts and hippocampal CCL3 protein levels (Ulrich et al., 2018). Additionally, examination of TMEM119⁺CD11b⁺ microglia from Apoe^−/− mice revealed a significant decrease in SPP1, a gene up-regulated across all conditions and present in the APOE network (Fig. 4, H and I). Collectively, these data suggest a significant in vivo effect of APOE expression on this network. Although APOE is known to contribute to disease through many mechanisms and can exacerbate both amyloidosis and tauopathy (Liu et al., 2013; Shi et al., 2017), the role of microglial APOE and whether it serves a unique or redundant role with astrocyte APOE is unknown.

In addition to aging, sex is another AD risk factor. Previous microarray analysis revealed age-related sexually divergent neuroinflammatory and complement genes in bulk hippocampal tissue (Mangold et al., 2017). To determine how sex affects the aged microglial transcriptome, we compared RiboTag RNAseq microglial transcriptomes from 3-, 12-, and 24-mo-old male and female mice. PCA analysis revealed discrete populations that separated most by the principle component 1 aspect of aging (Fig. 5 A) with the majority of sex differences occurring at 24-mo of age (Fig. 5, B–D). A potential caveat is that the sample sizes for RNAseq was n of four to five samples per group; therefore, it is possible that additional transcripts would be significant if we had even higher numbers to provide additional statistical power. This may account for the low number of significant differences between sexes, especially at earlier ages. Notably, in both sexes the middle-age samples appeared to have an intermediate transcript up-regulation (Fig. 5 E), suggesting a gradual shift in the transcriptome. Surprisingly, of the 37 transcripts that were significantly altered between 24-mo-old females and males (q < 0.1, absolute FC > 1.25), seven (18.9%) of the transcripts were found in the APOE driven network (Fig. 4 H) with significantly higher expression of Spp1, Gpnmb, Lgals3, Apoe, Ccl3, Clec7a, and Ccl4 in females (Fig. 5 F). This suggests that this network may be critical for driving the increased susceptibility of aged females at basal states and in the context of neurodegeneration.

Collectively, our results suggest that the risk factors of aging and sex, along with the pathological features of amyloidosis and tauopathy, serve to induce a core set of microglial transcripts that drive an APOE-mediated response resulting in CCL3 and CCL4 production. Although astrocyte-derived APOE has remained a large focus of AD related APOE research, our data suggest that microglial-derived APOE may also play a significant role in disease. Additionally, the overwhelming convergence of this core set of transcripts indicates that these multiple “hits” together may significantly exacerbate microglial dysfunction, resulting in increased pathology through these key mediators that could be critical targets for future therapeutics.

C57BL/6J, APPswe/PS1ΔE9 (APP/PS1)²⁹, and Cx3cr1-Cre^ERT2-IRES-eYFP^+/−; Rpl22^HA/+ mice (The Jackson Laboratory) were housed under standard laboratory conditions in ventilated cages on 12-h light:dark cycles in a specific pathogen-free environment. To generate the Cx3Cr1creER; Rpl22 mice, two independent strains both on a C57BL/6J background derived from The Jackson Laboratory, the Cx3Cr1^{tm2.1(Cre/ERT2)} and Rpl22^tm1.1PSam were bred and used as heterozygous mice for all studies (i.e., Cx3Cr1^creER/+; Rpl22^HA/+). For LPS and poly(I:C) studies, animals were injected with 2 µg/g LPS (0111:B4; Sigma) or 12 µg/g poly(I:C) (HMW; Invivogen) i.p., respectively. To induce tagged RPL22 expression, 3–4-mo-old male Cx3Cr1^cre/+; Rpl22^HA/+ mice were injected for three consecutive days with 100 µg/g tamoxifen (Sigma). 9–10-mo-old male APPswe/PS1ΔE9 (APP/PS1) and littermate controls were used for RNAseq studies. Animal protocols were reviewed and approved by Mayo Clinic Institutional Animal Care and Use Committee.

Cx3Cr1^creER/+; Rpl22^HA/+ pups were injected intracerebroventricularly with 2.7E + 10 viral particles/ventricle and 2 µl/ventricle of rAAV1-Tau^P301L and rAAV1-GFP on postnatal day 0. Viral production and injections were performed as described (Cook et al., 2015). In brief, cryoanesthetized newborn mice were placed on a cold metal plate, and a 30-G needle was used to pierce the skull just posterior to bregma and 2 mm lateral to the midline. For each lateral ventricle, 2 µl of AAV was injected, before rewarming the pups. Male animals were aged to 9–10 mo and used for RNAseq studies.

Animals were deeply anesthetized with pentobarbital before cardiac perfusion with PBS to expunge blood from the cerebrovasculature. For biochemical analysis, forebrain tissues were quickly dissected, frozen on dry ice, and stored at −80°C until further processing. Tissues set aside for immunohistochemistry were prepared as indicated below.

For RNA extraction of tissues, frozen forebrains were briefly sonicated in Tris buffered saline with EDTA (TBSE; 50 mM Tris, pH 7.5, 150 mM NaCl, and 1 mM EDTA) with 1× protease and phosphatase inhibitors (Thermo-Fisher Scientific). An aliquot of the tissue suspension was processed for RNA. For RNA extraction of cells isolated using myelin beads or Percoll methodologies, cells were pelleted at the final step and lysed in RLT Lysis Plus buffer supplemented with β-mercaptoethanol. Total RNA was isolated from tissues using an RNeasy Plus micro isolation kit according to manufacturer’s instructions (Qiagen).

Random-primed reverse transcription was performed according to manufacturer protocols (Invitrogen; Life Technologies). cDNA was added to a reaction mix (10-µl final volume) containing 300 nM gene-specific primers and Universal SYBR green supermix (Bio-Rad). All samples were run in triplicate and were analyzed on a Quant Studio 7 Flex Real Time PCR instrument (Applied Biosystems; Life Technologies). Relative gene expression was normalized to GAPDH controls and assessed using the 2^-ΔΔCT method. Primer sequences are as follows (5′ to 3′): Gapdh F: CTGCACCACCAACTGCTTAG, Gapdh R: ACAGTCTTCTGGGTGGCAGT; Aif1(Iba1) F: GGATTTGCAGGGAGGAAAAG, Aif1(Iba1), R: TGGGATCATCGAGGAATTG; Il1b F: CCTGCAGCTGGAGAGTGTGGAT, IL1b R: TGTGCTCTGCTTGTGAGGTGCT; Tnfa F: AGCCCACGTCGTAGCAAACCAC, Tnfa R: AGGTACAACCCATCGGCTGGCA; S100a8 F: CCTTTGTCAGCTCCGTCTTC, S100a8 R: CAAGGCCTTCTCCAGTTCAG; Ngp F: CTCTTCCGCCTGCTAAGTG, Ngp R: TGAACTGGAGTGGGATATTGG; Lcn2 F: ATGTCACCTCCATCCTGGTC, Lcn2 R: CCTGGAGCTTGGAACAAATG.

PBS-perfused hemi-brains were drop fixed into 10% neutral-buffered formalin (Thermo-Fisher Scientific) overnight at 4°C. Tissue was then placed in 30% sucrose (Sigma) dissolved in PBS overnight at 4°C. 50-µm coronal brain sections were cut on a freezing-sliding microtome and stored in cryoprotectant at −20°C until staining. Sections were blocked for endogenous peroxidase activity and permeabilized with 0.6% H₂0₂ and 0.1% NaN₃ in PBS-X (1× PBS containing 0.3% Triton-X) for 30 min at room temperature. Sections were blocked with 1% milk in PBS-X followed by incubation with rabbit anti-IBA1 at 1:8,000 (019-9741; Wako) in 0.5% milk PBS-X for 2 d at 4°C. Sections were then incubated with the Vectastain kit anti-rabbit IgG (Vector Labs) overnight at 4°C followed by ABC component for 4 h and developed using the DAB kit (Vector Labs) according to manufacturer’s instructions. Images were acquired using an XT Scanner (Aperio) at a 20× magnification.

PBS-perfused brains from tamoxifen injected Cx3Cr1^creER/+; Rpl22^HA/+ mice challenged with either saline or 2 µg/g LPS i.p. were incubated with collagenase D for 20 min at 37°C. Brains were immediately disrupted into a single cell suspension, filtered through a 100-µm filter, and washed with HBSS. Samples were centrifuged for 5 min at 300 g. Supernatants were aspirated and the cell pellet was resuspended in 90% Percoll/1× HBSS. A Percoll gradient was made with additional layers of 60% Percoll/1× HBSS, 40% Percoll/1× HBSS, and HBSS. Gradients were centrifuged for 18 min at 500 g with no brake, and cells were isolated from the 60/40 interface. Cells were washed twice with HBSS, incubated with Fc block for 10 min on ice in 1% BSA/HBSS buffer, and then stained for 30 min with anti-CD11b PE/Cy7 and anti-CD45 APC (Biolegend). Cells were sorted using 7AAD⁻ CD11b⁺ CD45^lo/int as the gate to identify microglia, and sorted directly into a 1.5-mL microcentrifuge tube containing Buffer RLT using a FACS Aria II SORP (BD Bioscience). RNA was extracted using a Qiagen RNeasy Microkit (Qiagen). This process from brain extraction to sorting into a microcentrifuge tube was ∼4 h.

Cx3Cr1^cre/+; Rpl22^HA/+ mice were injected for three consecutive days with 100 µg/g tamoxifen (Sigma) resuspended in corn oil to induce nuclear localization of CreER and allow for HA-tagged Rpl22 expression. 1 wk following the final injection, animals were challenged i.p. either with saline or 2 µg/g LPS. At 24 h after injection, animals were anesthetized with pentobarbital and PBS-perfused before forebrain harvest and immediate tissue sonication in homogenization buffer (50 mM Tris, pH 7.4, 100 mM KCl, 12 mM MgCl₂, and 1% NP-40) supplemented with 1 mM DTT, 1× protease/phosphatase inhibitors (Thermo-Fisher Scientific), 200 U/ml RNAsin, 100 µg/ml cycloheximide, and 1 mg/ml heparin. The cycloheximide serves to stabilize mRNA onto ribosomal complexes(Doyle et al., 2008; Sanz et al., 2009). A portion of this was immediately added to buffer RLT and RNA was purified to serve as “input” for each sample. Lysates were centrifuged at 4°C at 10,000 g for 10 min. Supernatants were incubated with anti-HA antibodies (Biolegend) and rotated at 4°C for 4 h before addition of protein G magnetic beads (Bio-Rad) and overnight incubation at 4°C while rotating. Beads were magnetized and non-HA containing supernatants were removed before washing with a high salt buffer (50 mM Tris, 300 mM KCl, 12 mM MgCl₂, 10% NP-40, 1 mM DTT, and 100 µg/ml cycloheximide). Three washes were conducted before magnetizing the beads, removing all supernatants and then releasing HA bound transcripts from the beads using Buffer RLT supplemented with 2-β-mercaptoethanol from a (Qiagen) RNeasy microkit followed by in-column DNase I treatment and isolation of RNA according to manufacturer’s instructions. Samples were amplified for RNAseq by cDNA library preparation using a NuGen Ovation RNA v2 kit (NuGen).

All samples were sequenced at Mayo Clinic Genome Facility using an Illumina HiSeq 4000. Reads were mapped to the mouse genome mm10 and RPKM were generated for each transcript. RPKM values were log2 transformed, and low abundance transcripts were removed from further analysis (log2 RPKM > −2 across all experimental groups were considered too lowly expressed). Differential expression, PCA, and statistical analysis were performed using Partek Genomics Suite (Partek Inc.) with ANOVA followed by multiple test comparison adjustment of P values using Bonferroni correction with FDR set at q < 0.1. Pathway analyses were performed using DAVID (National Institute of Allergy and Infectious Diseases) to examine significantly enriched pathways.

Following PBS perfusion, forebrains were removed, minced, and immediately incubated in StemPro Accutase (Thermo-Fisher Scientific) for 30 min at 4°C while rocking. Samples were centrifuged for at 4°C at 300 g for 10 min. The tissue pellet was resuspended in ice-cold HBSS and triturated sequentially using a 10-ml pipette, 5-ml pipette, and a P1000 pipet tip. Cells were filtered through a 100-µm strainer and then a 70-µm strainer before centrifugation at 4°C at 300 g for 10 min. The pellet was resuspended in 1.8 ml MACS buffer (0.5% BSA in PBS) and 200 µl myelin depletion beads were added, and depletion was conducted according to manufacturer’s instructions (Miltenyi Biotec). Tissues were taken before Accutase digestion, following Accutase digestion, and after myelin removal to generate RNA using a Qiagen RNeasy micro kit (Qiagen) for RT-qPCR.

Cells isolated from PBS-perfused brains following collagenase digestion and myelin removal using a Percoll gradient were washed in HBSS. Cells were resuspended in PBS and incubated for 15 min with Zombie violet (Biolegend) at room temperature. Cells were washed with PBS and then resuspended in staining buffer (1% BSA and PBS) and blocked with Fc block with 1:100 anti-CD16/32 (Biolegend) for 15 min on ice, followed by surface staining with anti-CD11b PE/Cy7, anti-Ly6G BV711 (Biolegend), and rabbit anti-TMEM119 (Abcam) for 15 min. Cells were washed with PBS, centrifuged for 5 min at 300 g, and then stained Alexa Fluor 488 conjugated with donkey anti-rabbit antibody (Jackson ImmunoResearch) for 15 min on ice. Following a PBS wash and 5 min centrifugation at 300 g, cells were resuspended in fixation buffer (Biolegend) for 15 min at room temperature. Cells were washed with PBS and then washed with permeabilization buffer (Biolegend) three times. Goat anti-S100A8 (R&D Systems) intracellular stain was conducted for 15 min on ice. Samples were washed with permeabilization buffer and then stained with donkey anti-goat Cy5 for 15 min on ice. Following two washes, cells were resuspended in 1% BSA/PBS buffer for analysis. Samples were read on an Attune NxT (Thermo-Fisher Scientific) and analyzed using FlowJo v10 software (FlowJo).

RNAseq data are deposited in Gene Expression Omnibus database (GSE117646).

Fig. S1 shows system setup including behavioral and cognitive assessment of Cx3cr1^+/+, Cx3cr1^gfp/+, and Cx3cr1^gfp/gfp animals; confocal analysis of HA expression in tamoxifen injected heterozygous Cx3cr1^{CreERT2-IRES-eYFP/+}; Rpl22^HA/+ mice (referred to as RiboTag in the manuscript); and cellular sorting strategies. A timecourse of LPS induced CNS inflammation to determine the time point for LPS studies is shown. The top 25 transcripts in microglia determined by RNAseq in RiboTag versus Cellular (RPKM > 10-fold higher over input) is also shown. Fig. S2 details the impact of different enzymatic conditions, myelin removal methodologies, and neutrophil depletion on activation/contamination markers. A comparison of the top LPS induced microglial transcripts determined using RNAseq data ranked according to both the cellular isolation dataset, and the RiboTag dataset 24 h after LPS challenge is also shown. Fig. S3 displays the Venn diagram of LPS and poly(I:C) induced microglial transcripts as well as the top 10 biological pathways 24 h following systemic challenge with either stimulant. The top induced transcripts based on RNAseq data ranked either using the LPS or poly(I:C) data are also shown. Supplemental material for this paper can be found at https://www.fryerlab.com/ribotag website. This content has not been peer reviewed.

This work was supported by grants from Mayo Foundation, GHR Foundation, Mayo Clinic Center for Individualized Medicine, Mayo Clinic Gerstner Family Career Development Award, Ed and Ethel Moore Alzheimer’s Disease Research Program of Florida Department of Health (grant 6AZ06 to J.D. Fryer), and National Institutes of Health grants NS094137, AG047327, and AG049992 (to J.D. Fryer); MH103632 (to S.S. Kang); AG057997 and AG016574 (J.D. Fryer and S.S. Kang); NS100693, NS097273, NS084974, and NS100693 (L. Petrucelli).

The authors declare no competing financial interests.

Author contributions: S.S. Kang and J. Fryer designed all the experiments. S.S. Kang, J.P. Sens, and C. Cook performed the research. X. Wang and J.-P. Kocher performed RNAseq sequence processing. M.T.W. Ebbert assisted in RNAseq analysis and website construction. K.E. Baker provided colony breeding and maintenance for mouse studies and additional technical support. S. Kang and J.D. Fryer wrote the manuscript with input from L. Petrucelli.