LPS priming before plaque deposition impedes microglial activation and restrains Aβ pathology in the 5xFAD mouse model of Alzheimer’s disease

Microglia have an innate immunity memory (IIM) with divergent functions in different animal models of neurodegenerative diseases, including Alzheimer's disease (AD). AD is characterized by chronic neuroinflammation, neurodegeneration, tau tangles and β-amyloid (Aβ) deposition. Systemic inflammation has been implicated in contributing to the progression of AD. Multiple reports have demonstrated unique microglial signatures in AD mouse models and patients. However, the proteomic profiles of microglia modified by IIM have not been well-documented in an AD model. Therefore, in the present study, we investigate whether lipopolysaccharide (LPS)-induced IIM in the pre-clinical stage of AD alters the microglial responses and shapes the neuropathology. We accomplished this by priming 5xFAD and wild-type (WT) mice with an LPS injection at 6 weeks (before the robust development of plaques). 140 days later, we evaluated microglial morphology, activation, the microglial barrier around Aβ, and Aβ deposition in both 5xFAD primed and unprimed mice. Priming induced decreased soma size of microglia and reduced colocalization of PSD95 and Synaptophysin in the retrosplenial cortex. Priming appeared to increase phagocytosis of Aβ, resulting in fewer Thioflavin S+ Aβ fibrils in the dentate gyrus. RIPA-soluble Aβ 40 and 42 were significantly reduced in Primed-5xFAD mice leading to a smaller size of MOAB2+ Aβ plaques in the prefrontal cortex. We also found that Aβ-associated microglia in the Primed-5xFAD mice were less activated and fewer in number. After priming, we also observed improved memory performance in 5xFAD. To further elucidate the molecular mechanism underlying these changes, we performed quantitative proteomic analysis of microglia and bone marrow monocytes. A specific pattern in the microglial proteome was revealed in primed 5xFAD mice. These results suggest that the imprint signatures of primed microglia display a distinctive phenotype and highlight the potential for a beneficial adaption of microglia when intervention occurs in the pre-clinical stage of AD.


Introduction
Microglia are the resident innate immune cells in the central nervous system (CNS). They are derived from the yolk sac and migrate to the brain parenchyma during embryogenesis. Microglia play essential roles in both brain homeostasis, such as environmental surveillance, synaptic pruning, and immediate inflammatory responses in diseases, including phagocytosis and the release of various inflammatory mediators.
Microglia have a relatively long lifespan, with a median life expectancy of over 15 months in adult mouse cortex (Füger et al., 2017;Askew et al., 2017) and 4.2 years in most human brain regions (Réu et al., 2017). This gives microglia the potential to "remember" their inflammatory history, thereby inducing long-lasting consequences depending on this memory which may contribute to the progression of neurodegenerative diseases. This memory has been proposed as microglial innate immunity memory (IIM) and has been demonstrated in different animal models of neurodegenerative diseases, such as Alzheimer's disease (AD) (Candore et al., 2004;Crous-Bou et al., 2017;Silva et al., 2019).
Increasing evidence has shown a crucial role of innate immunity in AD pathogenesis and progression (Candore et al., 2004;Crous-Bou et al., 2017;Silva et al., 2019). Another risk factor contributing to AD progression is systemic infection (Perry et al., 2003(Perry et al., , 2007Kamer et al., 2010). Microglia primed with systemic inflammation have been shown to manifest contradictory outcomes in AD animal models and AD patients Henry et al., 2008;Rakic et al., 2018;Wendeln et al., 2018;Tejera et al., 2019;Asby et al., 2021). These outcomes demonstrated different proteomic/transcriptomic profiles in microglia and contrasting effects on β-amyloid (Aβ) burden. It turns out that the consequences of microglial IIM depend specifically on the timing and strength of the stimulation. Thus, IIM can suppress or enhance the immune responses to a secondary inflammatory stimulus (Wendeln et al., 2018;Neher and Cunningham, 2019).
To better understand microglial reactions, transcriptional and proteomic approaches have been used to identify a unique subpopulation of microglia, disease-associated microglia (DAM). DAM highly express ApoE, Trem2, Lpl, and CD11c and are suggested to mediate Aβ clearance (Keren-Shaul et al., 2017;Krasemann et al., 2017). However, transcriptional alterations correlate poorly with protein changes (Lundberg et al., 2010;Sharma et al., 2015). Thus, a comparative proteomic study would provide direct information about phenotypical and functional changes in microglia. Moreover, peripheral immune cells also participate in the pathology by infiltrating the CNS due to blood-brain barrier (BBB) damage (Khoury et al., 2007;Marsh et al., 2016;Bettcher et al., 2021). Though, monocytes recruited from blood are suggested not to be associated with Aβ deposition (Reed-Geaghan et al., 2020). Therefore, we would like to assess changes in monocytes from their original niche, the bone marrow (BM). Despite a growing body of evidence demonstrating that systemic inflammation could modify microglial activation and peripheral immunity, alter synapses, and worsen Aβ loads in a short period Go et al., 2016;Tejera et al., 2019;Knopp et al., 2022), a comprehensive study to elicit the long-term influences of systemic inflammation on microglial responses, peripheral immunity, and Aβ pathology has not been reported.
Given the failures of clinical trials by modulating pathology at a late stage, the importance of events that occurred at the early pathological stage has drawn increasing attention. After Aducanumab (Aduhelm TM ), Lecanemab (Leqembi TM ) has recently been approved as the second antibody against Aβ for AD by U.S. Food and Drug Administration (FDA, 2023). Research studies and clinical trials of Lecanemab highlight the importance of microglial removing protofibril Aβ in the early stage of AD (Englund et al., 2007;Johannesson et al., 2021;van Dyck et al., 2022). Thus, we focused on the alterations of Aβ accumulation with systemic inflammation introduced in the pre-clinical stage. Following the results from our previous study, where microglia presented proinflammatory activation before Aβ accumulation as early as 6 weeks in the 5xFAD mouse model (Boza-Serrano et al., 2018b), we wanted to investigate whether priming on microglia at this age will have a long-term effect on AD-related pathology. Thus, we primed 5xFAD mice (Primed-5xFAD) and age-matched wild-type mice (Primed-WT) with an intraperitoneal (i.p.) injection of LPS (1 mg/kg) at 6 weeks. We first examined whether IIM occurred 140 days after a single immune challenge and how IIM affected microglial responses towards Aβ deposition at the prefrontal cortex (PFC), dentate gyrus (DG) and retrosplenial cortex (RSC). These regions have been shown to suffer early pathological changes in AD progression and are associated with learning and memory formation (Miller, 2000;Vann et al., 2009;Mu and Gage, 2011). Microglial morphology, activation, and phagocytic capability were analyzed through immunostaining with a pan microglial marker IBA1, Galectin-3 (Gal3) (Boza-Serrano et al., 2019), and CD68, respectively. Alteration of synaptic protein colocalization was assessed using the pre-synaptic marker Synaptophysin and the post-synaptic marker PSD95 (Hong et al., 2016). To study changes related to BBB disruption, astrogliosis and infiltration of T cells were evaluated in the DG. Proinflammatory cytokines were measured in the central nervous system (CNS) and periphery. Cognitive performance was also evaluated with the novel object recognition test. Finally, we applied quantitative mass spectrometry to reveal possible signaling alterations after priming in microglia and BMderived monocytes.
Our data suggest specific imprint signatures of primed microglia when challenged in the pre-clinical AD stage. This can lead to a beneficial adaption of microglia and other cell types.

Animals
Transgenic 5xFAD mice were purchased from Jackson laboratory on a C57/BL6-SJL background. These mice overexpress human amyloid precursor protein (APP) with three mutations (the Swedish mutation, K670N/M671L; the Florida mutation, I716V; and the London mutation, V717I) and human PSEN1 with two mutations (M146L/L286V). The mutations are related to familial AD forms, and the transgenes are expressed under the neuron-specific Thy-1 promoter. Age-matched wildtype (WT) littermates were used as controls. Male mice were used and analyzed for this study at 6 months of age. Animal housing, handling and experiments were performed under international guidelines approved by the Malmö/Lund animal ethics committee (M30-16, Dnr5.8.18-01107/2018). Mice were housed in groups (5 animals/cage) and maintained in a 12 h light/dark cycle, with free access to food and water.

Intraperitoneal injection of LPS
At 6 weeks of age, mice were subjected to intraperitoneal injection of either lipopolysaccharide (1 mg/kg, L4516-from Escherichia coli O127: B8, Sigma-Aldrich) or the same amount of vehicle (saline) as control. All mice in this study were injected with the same batch of LPS. No mortality or sickness behavior was observed due to this concentration of LPS. The outline of the experiment is shown in Fig. 3F and Fig. 4A. Briefly, mice (n = 58) were first divided into two groups based on their genotype (WT and 5xFAD). To study the long-term effect of peripheral inflammation on monocytes and microglia under conditions of AD pathology, mice were divided into two groups: treated or untreated with LPS (Ctrl-WT, n = 15; Primed-WT, n = 16; Ctrl-5xFAD, n = 10; Primed-5xFAD, n = 17). At 6 months, mice underwent object recognition memory tests described in 2.7 before sacrificing for sample collection (Fig. 3F).

Quantification of microglial density and morphology
Automated thresholds were made separately for DAPI and IBA1 stainings. The mask was made by selecting the DAPI + area and then overlaying it with the IBA1 + area to quantify the microglial density in different brain regions. The number of individual microglia was counted using the 'Analyze Particle' function in Fiji ImageJ version 2.1.0 (Rasband, W.S., U.S. National Institute of Health) with a range of > 10 μm 2 .
Microglia morphology was analyzed based on their soma area (Hadar et al., 2017). The soma area was calculated under the 'Analyze Particle' function within an area of > 20 μm 2 in Fiji ImageJ version 2.1.0 (Rasband, W.S., U.S. National Institute of Health).

3D reconstruction of Aβ phagocyted by microglia
To analyze the percentage of Aβ plaque associated with microglia, images were examined as previously described (Zhong et al., 2019). Briefly, the images were taken with a digital zoom of 3 using a 63×/1.4 oil immersion objective. The constant gain and laser power were applied with a Leica TCS SP8 confocal microscope (Leica Micro systems). The whole section was scanned with a Z-stack step of 0.297 μm under 776pixel by 776-pixel resolution. To quantify the colocalization of Aβ plaque and CD68 within IBA1 + cells, 3D reconstruction was conducted using the Imaris version 9.8.2 (Bitplane).

Synaptic puncta quantification
Analysis of synaptic density and colocalization of pre-synaptic and post-synaptic puncta were made according to Manabe et al. study (Manabe et al., 2021) with modifications. Briefly, the images were taken with a digital zoom of 2 using a 63×/1.4 oil immersion objective. The images were captured from 3 to 4 sections from the same mouse from 3 to 4 randomly selected views. The constant gain and laser power were applied with Nikon Confocal A1RHD microscope. The sections were scanned with a Z-stack step of 0.5 μm and the total optical thickness of 8 μm under 1024-pixel by 1024-pixel resolution. The density of synaptic puncta and the colocalization of pre-synaptic and post-synaptic puncta were quantified using a Fiji ComDet 0.3.6.1 plugin (https://github.co m/UU-cellbiology/ComDet/wiki). The representative figures were made by Imaris version 9.8.2 (Bitplane). The density and the mean fluorescence of NeuN + cells were also quantified in the same images with Fiji ImageJ version 2.1.0 (Rasband, W.S., U.S. National Institute of Health).

Brain homogenization
The PFC and hippocampi were dissected from mouse brains and homogenized using Lysis Matrix D 2 ml tubes (MP Biomedicals). Two fractions were contained in these tissues. Briefly, the first fraction was extracted with RIPA buffer (Sigma-Aldrich) containing protease and phosphatase inhibitors using FastPrep TM -24 Classic instrument (MP Biomedicals) for 2 × 10 s at a speed of 6 m/s. The homogenates were incubated on ice for 30 min before 30 min of 14 000 × g centrifugation at 4 • C. The supernatants were collected, including soluble and membrane-bound proteins. The second fraction was obtained using the pellets from the previous step. The pellets were resuspended using 70% formic acid and sonicated for 2 min with 6 rounds (10-sec pulse and 10sec pause per round) at 60% amplitude. Next, the suspensions were incubated on ice for 30 min and spun down at 14 000 × g for 30 min at 4 • C. Following, the supernatants were neutralized with 1 M Tris-base buffer in 1:17 dilution. Protein concentration was measured with the Pierce BCA Protein Assay kit (23227, ThermoFisher Scientific) for the first fraction and the Coomassie Plus (Bradford) Assay kit (MAN0011203, ThermoFisher Scientific) for the second fraction.

Novel object recognition memory test
Memory performance was assessed by the novel object recognition memory test as previously described (Bachiller et al., 2022). First, mice (n = 9-17) were placed in an arena (55×40×40 cm) and allowed to explore it for 5 min freely. Following this, two identical objects were placed in the center of the area for 15 min (training session, T). Memory test was carried out 24 h after T by returning mice to the arena for a 10minute session with the replacement of one original object with a novel one. The time spent exploring each object was recorded using Anymaze software (Stoelting Co, Wood Dale, IL), and the relative exploration of the novel object was calculated as a discrimination index [DI = (t novelt familiar )/(t novel + t familiar )]. The criterium for exploration was based strictly on active exploration, defined as directing the nose toward the object at a distance of ± 1.5 cm and/or touching the object with the nose or vibrissae as previously described (Bachiller et al., 2017). Circling or sitting on the object were not considered exploratory behaviors. All trials were performed by the same experimenter blind to the treatments and/or the genotype. All the sessions were conducted during the light phase (08:00-14:00) in the behavioral room with dim lighting and background noise. The arenas and objects were thoroughly cleaned with 70% ethanol at the beginning of each session and between sessions to remove olfactory cues.

Multiplex cytokine ELISA
The concentrations of different cytokines in serum and soluble fractions from brain homogenization were measured with the Mesoscale Discovery platform system (MSD, Rockville, USA). The MSD U-Plex Metabolic Group 1(mouse) Kit (brain-derived neurotrophic factor (BDNF), granulocyte-macrophage colony-stimulating factor (GM-CSF), interferon (IFN)-γ, interleukin (IL)-1β, tumor necrosis factor (TNF)-α, monocyte chemoattractant protein (MCP)-1, macrophage inflammatory protein (MIP)-1α, MIP-1β, MIP-2, MMP-9; K152ACM-2) was selected to evaluate the cytokine levels. The MSD V-Plex 6E10 Kit (Aβ38, Aβ40, and Aβ42, K1200E-2) was used to measure Aβ load in the soluble and insoluble fractions. The plates were read by a QuickPlex Q120 according to the manufacturer's instructions. The data were analyzed with MSD Discovery Workbench software. In total, 5-8 mice for each condition were analyzed. However, several samples were under the lowest detection limit and were excluded from the final statistical analysis.

Cell sorting
Brain suspension was prepared using an established method from Miltenyi Biotec with their Neural Tissue Dissociation Kit (130-092-628, Miltenyi Biotec). Bone marrow was collected from the femur and tibia. Microglia and monocytes were enriched using magnetic-activated cell sorting (MACS) with CD11b microbeads (130-093-634, Miltenyi Biotec) and the Monocyte Isolation Kit (130-100-629, Miltenyi Biotec), respectively. Next, cells were sorted by a FACSArial III cytometer (BD Biosciences) with FACS Diva software (BD Biosciences). A viability staining, propidium iodide (PI) (421301, 1:1000, BioLegend), was used to exclude dead cells. PI-negative microglia were sorted for proteomic analysis. The purity of PI-negative cells was increased significantly after MACS enrichment, about 95%, previously confirmed by flow cytometry

Protein digestion and tandem mass tag (TMT) labeling
Each sample was digested with trypsin using the filter-aided sample preparation (FASP) method (Wiśniewski et al., 2009). Briefly, 100 µl of 4% sodium dodecyl sulfate (SDS) in 50 mM triethylammonium bicarbonate (TEAB) solution was added to each sample, which was then shaken and sonicated repeatedly to dissolve the proteins. The samples were reduced with 100 mM dithiothreitol at 60 • C for 30 min. The reduced samples were transferred to 30 kDa MWCO Pall Nanosep centrifugation filters (Pall Corporation), washed several times with 8 M urea and once with digestion buffer (DB, 0.5% sodium deoxycholate in 50 mM TEAB) prior to alkylation with 10 mM methyl methanethiosulfonate in digestion buffer for 20 min at room temperature. Digestions were performed by the addition of 0.3 µg Pierce MS grade Trypsin (ThermoFisher Scientific) in DB and incubated overnight at 37 • C. An additional portion of 0.3 µg trypsin was added and left to incubate for another three hours. Peptides were collected by centrifugation and labeled using TMTpro 16-plex isobaric mass tagging reagents (Ther-moFisher Scientific) according to the manufacturer's instructions. Labeled samples were combined into 2 sets, and sodium deoxycholate was removed by acidification with 10% TFA. The combined TMTlabeled samples were desalted using Pierce Peptide Desalting Spin Columns (Thermo Scientific) following the manufacturer's instructions.
Each fraction was analyzed on Orbitrap Fusion Tribrid mass spectrometer interfaced with Easy-nLC 1200 nanoflow liquid chromatography system (ThermoFisher Scientific). Peptides were trapped on the Acclaim Pepmap 100 C18 trap column (100 μm × 2 cm, particle size 5 μm, Thermo Fischer Scientific) and separated on the in-house packed C18 analytical column (75 μm X 35 cm, particle size 3 μm) using the gradient from 5% to 12% B in 5 min, from 12% to 35% B in 72 min, from 35% to 100% B in 3 min, and 100% B for 10 min at a flow of 300 nL/min. Solvent A was 0.2% formic acid and solvent B was 80% acetonitrile, 0.2% formic acid. MS scans were performed at 120 000 resolution, m/z range 375-1500. MS/MS analysis was performed in a data-dependent, with a top speed cycle of 3 s for the most intense doubly or multiply charged precursor ions. Precursor ions were isolated in the quadrupole with a 0.7 m/z isolation window, with dynamic exclusion set to 10 ppm and a duration of 45 s. Isolated precursor ions were subjected to collision induced dissociation (CID) at 30 collision energy with a maximum injection time of 60 ms. Produced MS2 fragment ions were detected in the ion trap followed by multinotch (simultaneous) isolation of the top 10 most abundant fragment ions for further fragmentation (MS3) by higherenergy collision dissociation (HCD) at 55% and detection in the Orbitrap at 50 000 resolutions, m/z range 100-500 (Wiśniewski et al., 2009).

Database search and quantification
Identification and relative quantification were performed of the combined injections using Proteome Discoverer version 2.4 (Thermo-Fisher Scientific). The database search was performed using the Mascot search engine v. 2.5.1 (Matrix Science, London, UK) against the Swiss-Prot Mouse database (July 2020). The APP fragments were matched against the Swiss-Prot Human database (Nov 2022) using sequestHT. Trypsin was used as a cleavage rule with no missed cleavages allowed; methylthiolation on cysteine residues, TMTpro at peptide N-termini and on lysine side chains were set as static modifications, and oxidation on methionine was set as a dynamic modification. Precursor mass tolerance was set at 5 ppm and fragment ion tolerance at 0.6 Da. Percolator was used for the peptide-spectrum match (PSM) validation with the strict false discovery rate (FDR) threshold of 1%. Quantification was performed in Proteome Discoverer 2.4. The TMT reporter ions were identified with 3 mmu mass tolerance in the MS3 HCD spectra. The TMT reporter S/N values for each sample were normalized within Proteome Discoverer 2.4 on the total peptide amount. Only the quantitative results for the unique peptide sequences with the minimum SPS match percentage of 65 and the average S/N above 10 were considered for protein quantification.

Bioinformatic and statistical analysis
The MS/MS data were normalized with the R package NormalyzerDE (version 1.14.0) at the condition of VSN (Willforss et al., 2019). Most of the keratin proteins were removed as they were possibly contaminations from the experiment steps. Before analysis, proteins without values appearing in all biological replicates from one experimental group were filtered. And the proteins with missing values were imputed based on 'MinProb' methods from the DEP (1.18.0) R package (Zhang et al., 2018). All differentially expressed proteins (DEPs) were decided as 23% above and below a fold change (FC) of 1 (log2 FC > 0.3 and log2 FC < -0.3) using the DEP (1.18.0) R package with adjusted p-values < 0.05. The enrichment analysis of Gene ontology (GO) of all DEPs with all genes was based on the clusterProfiler (version 4.4.4) R package Wu et al., 2021). GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) terms were selected as significantly altered using the Benjamini-Hochberg method with adjusted p-values < 0.05. Comparisons between experimental groups were analyzed with unpaired t-tests or two-way ANOVA followed by Tukey's or Holm-Šídák's multiple comparisons. P < 0.05 was considered statistically significant. All statistical analysis was done using the GraphPad Prism 7.0 software for macOS Catalina (GraphPad Software, San Diego, CA, USA). Data are presented as mean ± SD. A confidence interval of 95% was set as significant. The exact P values are given in the figure legends. The graphic abstract and experimental schemes were created with BioRender.com.

Priming induced decreased IBA1 coverage and reduction of microglial soma size at RSC and a potential increase of Aβ phagocytosis at DG in 5xFAD
We first investigated whether microglia morphology and density changed at 6 months of age in 5xFAD due to i.p. injected at 6 weeks of age with saline (Ctrl-5xFAD) or LPS (1 mg/kg, Primed-5xFAD) in three brain regions (PFC, DG, and RSC) strongly connected with memory formation and cognition (Miller, 2000;Vann et al., 2009;Mu and Gage, 2011). To analyze the morphological changes, the area of a pan microglial marker, IBA1, was first quantified at PFC, DG, and RSC ( Fig. 1A and B). Interestingly, the percentage of IBA1 + area was significantly reduced at RSC in 6 months old 5xFAD mice primed with LPS at 6 weeks of age (Fig. 1B). Next, we wanted to know whether this change was due to a lower microglial density or morphological alterations in the RSC region. By comparing the microglial density in Ctrl-5xFAD and Primed-5xFAD, we found a trend of fewer microglia at RSC after priming ( Fig. 1C, P = 0.0661). Thus, we further evaluated the morphology changes by looking at the soma size of microglia in these three regions ( Fig. 1A and D). As in the initial state of microglial activation, the cell bodies of microglia enlarged with retraction of processes (Kreutzberg, 1996). The morphological transformation from ramified to amoeboid implies potential changes in the states of microglia (Paolicelli et al., 2022). The cell bodies of microglia remarkably declined in RSC from Primed-5xFAD compared to Ctrl-5xFAD (Fig. 1D).
Following, we asked whether the morphological alteration implicated the functionality differences in different regions. Therefore, we performed CD68 staining together with MOAB2 and IBA1 to examine the priming impact on phagocytic capability in Aβ-associated microglia (Fig. 1E). The percentage of MOAB + CD68 + area colocalized with IBA1 + cells was quantified accordingly. Surprisingly, at the region of DG, it was a tendency of an increased amount of Aβ in the lysosomal compartment marked with CD68 in microglia after priming (Fig. 1F, P = 0.0516). However, no difference was seen in PFC and RSC. Taken together, a single LPS challenge given in the periphery has a long-lasting effect on microglial coverage and microglia morphology in RSC, but the morphological changes do not directly relate to microglial phagocytosis.

Priming led to smaller MOAB2 + Aβ plaques at PFC and fewer ThioS + Aβ plaques at DG in 5xFAD
As priming reduced microglial soma size and potentially increased Aβ phagocytosis in 5xFAD mice indicated in 3.1, we asked whether priming also impacted the size, number, and composition of Aβ plaque. Here, we analyzed the Aβ deposition with either MOAB2 or Thioflavin S (ThioS) at PFC, DG, and RSC regions. MOAB2 is an antibody targeting all forms of human Aβ42 and unaggregated human Aβ40 (Youmans et al., 2012), while ThioS only labels Aβ fibrils (Malmos et al., 2017). Intriguingly, the average size of MOAB2 + Aβ deposition was selectively decreased in Primed-5xFAD at the region of PFC ( Fig. 2A and B). But no difference was observed in Aβ plaque number (Fig. 2B). In contrast, significantly fewer ThioS + Aβ plaques were seen at DG in Primed-5xFAD mice ( Fig. 2C and D). And there appeared to be a trend of reduced ThioS + Aβ size at DG due to priming (Fig. 2C, P = 0.0533). Afterwards, we wanted to know whether these changes in Aβ deposition also caused alterations in neuritic dystrophy by using LAMP1 staining. No changes were observed in the coverage and intensity of LAMP1 between Ctrl-5xFAD and Primed-5xFAD mice ( Fig. 2E and F). Next, we wondered whether the LPS-induced alteration in plaque formation was due to a higher degradation of Aβ in microglia. Insulin-degrading enzyme (IDE) and Apolipoprotein E (ApoE) have been found to contribute to the degradation and clearance of Aβ (Farris et al., 2003;Jiang et al., 2008). However, no difference was observed in the expression of IDE and ApoE after priming ( Supplementary Fig. 1A, B and D). The findings indicate that the impacts of priming on different forms of Aβ depend on the specific brain region. Moreover, potentially increased phagocytosis of Aβ through the lysosomal compartment, as demonstrated in 3.1, appears to be more effective in reducing the number of ThioS + Aβ fibrils than MOAB2 + plaques at DG.

Priming improved long-term memory in WT and 5xFAD mice
To investigate how acute systemic inflammation modulated cognition in the long term, we performed novel object recognition tasks on the mice 140 days after the LPS injection (Fig. 2I). There was no increased mortality or weight loss in the mice after one dose of LPS at 6 months of age ( Supplementary Fig. 4D). Preference of positions was also evaluated in the training session. Mice used in the test showed no side preference ( Supplementary Fig. 4E). At 6 months of age, 5xFAD mice have pronounced Aβ deposition and memory deficits (Oakley et al., 2006). As expected, 5xFAD mice performed worse than WT ( Fig. 2G and H). Intriguingly, LPS priming triggered better discrimination in the novel object recognition test in both genotypes (Fig. 2G). This finding suggests that mice exhibited longer exploration times, specifically on the novel object after priming (Fig. 2H). Taken together, priming with one dose of LPS administration at 6 weeks improved the longstanding memory in WT and 5xFAD transgenic mice.

Priming led to alterations in the microglial barrier around Aβ plaques
Since priming improved the cognitive performance, and reduced the size of MOAB2 + Aβ plaques at PFC and the number of ThioS + Aβ at DG, we further investigated how priming affected the microglial barrier around Аβ plaques at PFC, DG, and RSC. Surprisingly, IBA1 + microglia surrounding Aβ depositions (MOAB-2 + staining) were selectively reduced in the 6 months old 5xFAD mice primed with LPS at 6 weeks of age in all three regions ( Fig. 3A and B). We further characterized activated microglia around Aβ depositions using Gal3, an activation marker from Ctrl-5xFAD and Primed-5xFAD groups. Colocalization of CD68 and Aβ was shown in purple in the last column. Scale bar: 7 μm. Unpaired t test, n = 4-5 per group. (F). Percentage of colocalization of CD68 and MOAB2 in IBA1 + cells at the PFC, DG, and RSC regions from the two 5xFAD transgenic groups. Unpaired t test, n = 4-5 per group. Data are presented as means ± SD. Unpaired t test, *p < 0.05, ***p < 0.001. PFC: prefrontal cortex, DG: dentate gyrus, RSC: retrosplenial cortex.
(caption on next page) Y. Yang et al. of microglia (Boza-Serrano et al., 2019). Interestingly, fewer Gal3 + microglia were seen around the plaques at DG and RSC in the Primed-5xFAD group (Fig. 3A and C). Altogether, a single LPS challenge given in the periphery has a long-lasting effect on microglia, leading to a decrease in the number of microglia surrounding Aβ plaques and fewer Gal3 + microglia near the plaques.

Priming altered the reactivity of GFAP in astrocytes
Astrocytes play an essential role in brain homeostasis including the BBB function (Mrak et al., 2001;Kuchibhotla et al., 2009), and are responsive to inflammatory stimuli. AD pathology worsens the BBB integrity and disrupts the permeability of BBB (Weller et al., 2009;Giannoni et al., 2016;Shabestari et al., 2022). Therefore, we wanted to check how priming affected astrocytic activation. We observed that in DG, astrogliosis was exacerbated by AD pathology, and the reactivity of GFAP remained high 140 days after the LPS injection (Fig. 3 D and E). Nevertheless, the coverage of GFAP evaluated with immunostaining and expression level of GFAP revealed with western blot remained no significant changes in terms of the LPS challenge but an increase as a result of Aβ pathology in 5xFAD (Fig. 3F, Supplementary Fig. 1C and D). Increased T cell infiltration commonly occurs in AD patients and mouse models (Ferretti et al., 2016;Bettcher et al., 2021). We wonder whether priming also changed the infiltration of T cells. T cell (CD3 + ) were assessed in DG, PFC, and RSC. Two cortical regions had barely visible T cell infiltration. The number of T cells in DG was higher in 5xFAD mice than in WT mice ( Fig. 3D and G). However, priming did not alter the infiltration of T cells in DG from WT and 5xFAD mice (Fig. 3G). These findings suggest that priming has long-lasting effects on GFAP + astrocytes but does not protect DG from T cell infiltration.

Priming decreased colocalization of PSD95 and Synaptophysin in RSC
Since priming altered the microglial barrier around Aβ depositions and plaques formation in 5xFAD mice indicated in 3.2 and 3.3, we further investigated the production of APP from neurons and levels of synaptic proteins related to neuronal activity in the PFC and hippocampus. These two areas were consistently reported to be impaired in AD, resulting in cognitive impairment (Oakley et al., 2006). Firstly, a western blot was used to examine the production of APP in Ctrl-5xFAD and Primed-5xFAD mice. There was no difference in APP levels in either region among the two groups ( Supplementary Fig. 1E and H). Furthermore, we wanted to know whether LPS stimulation had long-term effects on synapses. Through an analysis of the density of the pre-synaptic marker Synaptophysin and the post-synaptic marker PSD95, we observed that priming exhibited a possibility to enhance the density of PSD95 in WT, while conversely reducing it in 5xFAD mice, specifically in the PFC (Fig. 4A and B, P = 0.059). Moreover, there was a tendency for a reduction of the density of Synaptophysin due to priming in PFC (Fig. 4C, P = 0.092). Surprisingly, the colocalization of PSD95 and Synaptophysin was significantly reduced in a priming-dependent manner among WT and 5xFAD (Fig. 4D). Western blots indicated that PSD95 was selectively increased in the PFC due to priming (Supplementary Fig. 1F and H). On the other hand, priming affected the Synaptophysin levels differently in WT and 5xFAD at PFC (Supplementary Fig. 1G and H). Nevertheless, NeuN + cells were increased in RSC after priming ( Supplementary Fig. 4A and B), suggesting increased expression of NeuN in this specific brain region. Taken together, priming induces a reduction of colocalization of PSD95 and Synaptophysin at RSC and does not affect APP production in the PFC and hippocampus.

Priming did not reverse the elevation of proinflammatory cytokines induced by Aβ pathology
To further analyze how AD pathology and priming affected inflammation in the CNS and periphery, we measured the levels of 8 different proinflammatory cytokines in the brain, including the PFC and hippocampus, as well as in the blood serum. No significant changes were found in the serum in any of the experimental groups (Supplementary Fig. 2A-E). In contrast, the levels of proinflammatory cytokines MIP-1α were strongly increased in the PFC and hippocampus in the 5xFAD mice ( Supplementary Fig. 3A). There was a trend in reduction of MIP-1α levels after priming in PFC ( Supplementary Fig. 3A). Another proinflammatory cytokine, MIP-1β, was increased in PFC in 5xFAD mice compared to WT mice ( Supplementary Fig. 3B) but was below the detection limit in the hippocampus. IFN-γ also had a trend towards upregulation in the PFC after LPS priming in WT and 5xFAD mice (Supplementary Fig. 3C). Additionally, the neurotrophic factor BDNF was found to be upregulated in both regions from 5xFAD transgenic mice ( Supplementary Fig. 3D). A compensatory repair-mechanism may explain this increase at the time point in the 5xFAD model (Durany et al., 2000;Jin et al., 2004). The BDNF level may decline at later age, as reported previously in this mouse model (de Pins et al., 2019) and in AD patients (Peng et al., 2005). Altogether, we found increased levels of proinflammatory cytokines in the regions of PFC and hippocampi from mice with AD pathology but no effect of priming.

Microglia shared a large similarity with monocytes and obtained distinct proteomic profiles
To elucidate the mechanism underlying the alterations induced by one systemic LPS injection, we acquired proteomic profiles of both microglia and monocytes. CD11b + microglia were isolated from 6 months old Ctrl-WT, Primed-WT, Ctrl-5xFAD, and Primed-5xFAD (n = 3/4 mice per group) as described in 2.9 (Fig. 5A). Our previous study showed that magnetic-activated cell sorting enriched cells were ~ 95% CD11b + microglia (Boza-Serrano et al., 2018b). In inflammatory diseases or infections, Ly6C hi monocytes migrate from the BM and infiltrate tissues as a key player in the host response to insults and pathogens (Serbina et al., 2003;Ginhoux and Jung, 2014). This recruitment is reported to be regulated by CC-chemokine receptor 2 (CCR2) signaling (Serbina and Pamer, 2006;Gordon and Taylor, 2005). CCR2 signaling has also been implicated as contributing to AD pathogenesis (Khoury in the PFC, DG, and RSC from Ctrl-5xFAD and Primed-5xFAD groups. Scale bar: 50 μm. (B). Average size and the number of MOAB2 + Aβ plaques at the PFC, DG and RSC regions from the two 5xFAD transgenic groups. Unpaired t test, n = 4-5 per group. PFC: P size = 0.0036 (C). Average size and the number of Thioflavin-S + Aβ deposition at the PFC, DG and RSC regions from the two 5xFAD transgenic groups. Unpaired t test, n = 4-5 per group. DG: P number = 0.0322. (D). Representative immunostaining images of dense-core plaques using Thioflavin-S at the PFC, DG, and RSC regions. Scale bar: 100 μm. (E). Representative immunostaining images of IBA1, MOAB2 + Aβ deposition, LAMP1 + neuritic dystrophy and Methoxy-X04 + Aβ in the PFC, DG, and RSC from Ctrl-5xFAD and Primed-5xFAD groups. Magnification is shown in the last column. Scale bar: 20 μm, 50 μm. (F). Coverage of LAMP1 + dystrophic neuritis (left panel) and mean fluorescent intensity of LAMP1 (right panel) at the PFC, DG and RSC regions from the two 5xFAD transgenic groups. Unpaired t test, n = 4-5 per group. (G). Discrimination index of the mice performed with novel object recognition. Two-way ANOVA analysis (P Genotype = 0.0006; P priming = 0.0108) followed by Tukey's multiple comparisons. n = 9-17 per group. (H). Exploration time on the novel object of the mice. Two-way ANOVA analysis (P Genotype = 0.0142; P priming = 0.046) followed by Tukey's multiple comparisons. n = 9-17 per group. (I). Experimental design of novel object recognition tasks (graphic created with BioRender.com). Data are presented as means ± SD. Two-way ANOVA results are shown: #p < 0.05, ###p < 0.001. PFC: prefrontal cortex, DG: dentate gyrus, RSC: retrosplenial cortex, Hippo: the hippocampus. Data are presented as means ± SD. Unpaired t test, *p < 0.05, **p < 0.01. PFC: prefrontal cortex, DG: dentate gyrus, RSC: retrosplenial cortex. in the region of DG from Ctrl-5xFAD and Primed-5xFAD groups. CD3 + T cells are pointed with arrows. Scale bar: 50 μm. (E). Quantification of GFAP immunofluorescent intensity at DG region from all experimental groups. Two-way ANOVA analysis (P Genotype = 0.0252, P Priming < 0.0001) followed by Tukey's multiple comparisons (WT groups: P = 0.0016; 5xFAD groups: P = 0.0144). (F). Coverage of GFAP + area at DG region from all experimental groups. Two-way ANOVA analysis (P Genotype = 0.0028) followed by Tukey's multiple comparisons (P = 0.047). (G). Number of infiltrated CD3 + T cells at DG region from all experimental groups. Twoway ANOVA analysis (P Genotype = 0.0443) followed by Tukey's multiple comparisons. Data are presented as means ± SD. Two-way ANOVA results are shown: #p < 0.05, ##p < 0.01 ####p < 0.0001. Multiple comparisons results are displayed: *p < 0.05, **p < 0.01. PFC: prefrontal cortex, DG: dentate gyrus, RSC: retrosplenial cortex. Naert and Rivest, 2011;Michaud et al., 2013). Therefore, we further sorted a particular subset of BM-monocytes (CD115 + Ly6C hi CD11b + ) from the 4 experimental groups, the same as microglia samples, to assess the impact of systemic inflammation on peripheral recruitment (Fig.  5A). The percentage of CD115 + Ly6C hi CD11b + BM-monocytes was significantly increased due to the 5xFAD genotype, not the priming ( Fig. 5B and C). The median fluorescence intensity (MFI) of CD115, Ly6C, and CD11b were not altered in the subset of Ly6C hi monocytes among the four groups (Supplementary Fig. 5), suggesting that Aβ pathology enhanced the composition of CD115 + Ly6C hi CD11b + monocytes in the BM. This indicates that BM-monocytes in 5xFAD had a higher tissue infiltration potential due to a chronic inflammation caused by overexpression of Aβ in the brain.
Quantitative mass spectrometry was performed on microglia and monocytes. The samples were digested and labeled with TMT reagents (Fig. 5A). Two sets of TMT 16-plex peptides were measured and compared for differential expression. In total, 4468 proteins from microglia and 4179 from monocytes were detected (Supplementary  Table 2). Among these proteins, 66.1% were shared in microglia and CD115 + Ly6C hi CD11b + BM-monocytes. To understand how different conditions impact these two cell types, three dimensions principal component analysis (3D-PCA) was applied to get a general view of alterations. Microglia were clustered into 4 distinct groups due to four conditions (Fig. 5D). In contrast to the monocytes that did not form distinct groups depending on condition (Fig. 5E). As expected due to the brain pathology; microglial proteomics had more pronounced changes than the monocytes in response to priming, Aβ, or both stimuli (Supplementary Fig. 6). Differentially expressed proteins in microglia are listed in Supplementary Table 3. Therefore, we focused on microglia for the following analysis.

Primed microglia in the 5xFAD model seemed to internalize more Aβ resulting in less soluble Aβ42 in the region of PFC
Although there was no difference in the amount of Aβ inside the lysosomal compartment shown in 3.2, we wondered whether the priming before the appearance of Aβ plaques affected Aβ internalization. Human Uniprot databases were used to recognize different fragments of APP in microglia. In total, 5 different АPP peptides were mapped in the datasets (Fig. 6A). 3 of them were matched at the N-terminal of the APP sequence localized at 430-438, 439-450, and 586-601, possibly corresponding to soluble APP-α and APP-β (sAPPα/β). 2 of them matched with Aβ sequence (peptide-1: 677-688 and peptide-2: 688-700) in which Aβ peptide-1 distinctive existed in human. No C-terminal residues of App were found in our dataset, suggesting the APP-related peptides identified in our datasets were likely not produced inside microglia but phagocytosed extracellularly. All 5 APP-related peptides showed an AD genotype-dependent elevation ( Fig. 6B and C). Importantly, Aβ fragments were selectively higher in the microglia primed with LPS than the Ctrl-5xFAD (Fig. 6C). In contrast, the sAPPα/β showed no differences between Ctrl-5xFAD and Primed-5xFAD (Fig. 6B). These findings go along with what we have seen in 3.1 (Fig. 1E and F). It suggests that LPSpriming before Aβ plaques increased the internalization of Aβ but probably not the ability to degrade it thoroughly by microglia as IDE and ApoE does not differ among Ctrl-5xFAD and Primed-5xFAD (Supplementary Fig. 1A and B).
Next, we wanted to know whether the higher internalization in the Primed-5xFAD microglia changed soluble and insoluble Aβ in the brain. Different brain regions were homogenized to quantify Aβ40 and Aβ42 levels using multiplex ELISA plates, including the PFC and hippocampus. Remarkably, the priming significantly reduced RIPA-soluble Aβ40 and Aβ42 levels at the PFC region, whereas soluble Aβ40 and Aβ42 showed no differences in the hippocampus between the Ctrl-5xFAD and Primed-5xFAD mice (Fig. 6D). Moreover, insoluble Aβ40 and Aβ42 extracted with formic acid (FA-fraction) showed no priming difference (Fig. 6E). It appeared that the LPS-priming at 6 weeks old had a stronger impact on the levels of soluble Aβ40 and Aβ42 than the insoluble forms.
(caption on next page) Y. Yang et al. Glycolysis/Gluconeogenesis, Citrate cycle and neurodegenerative diseases such as AD, ALS, Parkinson's disease and Huntington's disease ( Fig. 8A and B). In contrast, the mRNA surveillance pathway and Spliceosome were specifically enriched for down-regulated proteins. The lysosomal pathway was connected with both up-and down-regulated proteins ( Fig. 8A and B).
Moreover, the systemic LPS-induced priming model was used to compare the long-term effect of acute inflammatory changes with or without chronic neuroinflammation in the AD 5xFAD mouse model. In microglia from Primed-WT mice, 130 proteins were significantly increased, and 51 proteins were significantly decreased (Fig. 7C). GO analysis demonstrated that the increased proteins shared the same GO terms as the overexpressed proteins from Ctrl-5xFAD microglia, such as the ATP process, Antigen processing and presentation, and Mitochondrial organization and membrane (Fig. 7D). Interestingly, these GO terms were impeded in the study of acute effects on LPS-induced WT mice (Rangaraju et al., 2018), suggesting an opposite effect of the microglia proteome following acute vs. long-term effect of a systemic LPS challenge. In addition, the reduced proteins found in Primed-WT samples were enriched for GO terms, including Monosaccharide metabolic process, Nucleotide-sugar biosynthetic and Metabolic process following LPS-induced priming (Fig. 7D). Among these proteins, 52 upregulated proteins in Primed-WT were also increased in the 5xFAD microglia (such as H2-D1, Ndufa7 and Ctsh), and 21 proteins were . Scatter plot (left) demonstrates relative expression of shared DEPs from Ctrl-WT vs. Primed-WT and Ctrl-WT vs. Ctrl-5xFAD. A list of up-and-downregulated DEPs is shown in a heatmap (right). (F). Scatter plot (left) demonstrates relative expression of shared DEPs from Ctrl-5xFAD vs. Primed-5xFAD and Ctrl-WT vs. Ctrl-5xFAD. A list of up-and-down-regulated DEPs is shown in a heatmap (right). In all volcano plots (A)-(C), the values in the x-axis are Log2 transformed fold change, and the values in the y-axis are -Log10 transformed p-value. p < 0.05 is labeled on the y-axis. Significantly up-and down-regulated proteins are shown in red and cyan. Pearson's correlation coefficient was used in the scatter plots (E) and (F) to assess the overall concordance. The highly concordant proteins are highlighted red for the increase and blue for the decrease. The conversely related proteins are highlighted in green. down-regulated in Primed-WT and Ctrl-5xFAD compared with agematched WT mice (Hmga1, Ifi204, and Ptma). A comparison of KEGG from each analysis (Ctrl-WT vs. Ctrl-5xFAD, Ctrl-WT vs. Primed-WT) demonstrated that both LPS-priming and Aβ pathology promoted Oxidative phosphorylation, Prion disease and Reactive oxygen species GO (Fig. 8B). Surprisingly, AD term appeared in the KEGG analysis when comparing Ctrl-WT and Primed-WT (Fig. 8B). It implicated that LPS priming in the microglia lasted for several months after the induction. Moreover, differentially expressed proteins in Primed-WT went concordantly with alterations in microglia isolated from Ctrl-5xFAD (Fig. 7E, r = 0.75, p < 0.001). This suggests that priming stimulated a phenotype of microglia partly comparable to the neurodegenerationinduced status.

Priming in 5xFAD promoted lysosomal processing, and suppressed Ras signaling and GTP functions
To assess the long-term impacts of priming on microglia under Aβ pathology, we compared the proteomic profiles of microglia from Primed-5xFAD with Ctrl-5xFAD, followed by analyses of differential expressed proteins due to priming or Aβ pathology alone (Ctrl-WT vs. Primed-WT, Ctrl-WT vs. Ctrl-5xFAD). 91 proteins were significantly changed in Primed-5xFAD compared to Ctrl-5xFAD microglia. Of these proteins, 62 were identified as up-regulated (APP, Gpld1, Rab9, and Rab39), and 18 were down-regulated (Arf4, Cbr3, Plcg2, and Rac1/2) (Fig. 7A, Supplementary Table 5). The up-regulated proteins due to priming in 5xFAD microglia were enriched for GO terms in Lipid-related processes and neuron projection maintenance from BP, Lysosome and High-density lipoprotein particle from CC (Fig. 7D). In contrast, the reduced proteins were associated with Maintenance of cell polarity and Regulation of lipase activity in BP; Ribosomal subunit in CC; GTP binding and ribonucleoside binding in MF (Fig. 7D). Interestingly, most of the GO terms enriched with down-regulated proteins from Primed-5xFAD microglia were linked with increased expressions in the analysis of Ctrl-WT vs. Ctrl-5xFAD (Fig. 7D). Ras signaling pathway, Leukocyte transendothelial migration, and Neurotrophin signaling pathway were constrained in Primed-5xFAD microglia compared with that in Ctrl-5xFAD ( Fig. 8A and B). Together, this data suggests that LPSpriming in microglia modified the phenotype provoked by Aβ pathology.
To clarify in which way the priming affected the microglia in 5xFAD mice compared with Aβ pathology alone, we did a linear regression on differentially changed proteins in microglia reflecting the priming effects on AD pathology vs. only the AD pathology proteomic signatures (Fig. 7F). It demonstrated the priming modified phenotype was negatively correlated with the phenotype induced by Aβ pathology alone (Fig. 6F, r = -0.49, p < 0,001). 39 proteins in Primed-5xFAD were significantly modulated in contradictory directions as that of Aβinduced microglial profiles. Among these proteins, 9 were decreased in Primed-5xFAD microglia, including Ank1, Arf4, and Mpeg1; 30 were increased in Primed-5xFAD, including Sparc, Rab39, Gpld (Fig. 7F). In summary, peripheral LPS-administration induced somewhat an opposite effect on specific proteins upregulated in Ctrl-WT vs. Ctrl-5xFAD, suggesting a suppression or delay of the microglial overactivation caused by AD pathology. These phenomena may be regulated through enhanced lysosomal functions, damped Ras-linked cascades and GTP binding, and inhibited Leukocyte transendothelial migration. These findings also indicated that IIM on microglia caused by this extent of priming might suppress neuroinflammation in AD-related pathology.

Discussion
A large body of evidence has elucidated how systemic inflammation accelerates brain pathology and neuroinflammation in a relatively short period (from 2 h to 63 days after the LPS injection) (Tejera et al., 2019;Henry et al., 2008;Jiang et al., 2022). However, the long-term effects of systemic inflammation on microglia under the progression of AD in vivo have been poorly studied. One critical work done by Neher's group shows epigenomic alterations last even 6 months after the LPS stimulation in microglia (Wendeln et al., 2018). The IIM of microglia has been demonstrated in an adult AD mouse model with systemic inflammation using different dosages of LPS (Wendeln et al., 2018). That study showed decreased cerebral β-amyloidosis induced by repeated LPS stimulation, suggested due to epigenomic and transcriptomic alterations in microglia (Wendeln et al., 2018). However, the proteomic profiles of microglia modified by IIM have not been well-documented in an AD model. Therefore, in the present study, we provide the first exclusive proteomic signatures of microglia adapted to IIM caused by systemic inflammation before Aβ deposition in 5xFAD mice. Our study adds more aspects of IIM on microglia at the protein level and the consequences of modulated microglial response in astrogliosis and behavioral performance in the AD model. The current study also emphasizes the importance of the time window for introducing the inflammatory challenge. We chose 6 weeks for the priming to set up the IIM model because we previously found that the inflammatory response of microglia began as early as 6 weeks in the presence of APP + neurons with no extracellular Aβ deposition in the same 5xFAD mouse model (Boza-Serrano et al., 2018b). As the failure of clinical trials targeting a late stage of AD, we wonder whether IIM introduced in the pre-clinical stage of AD reverses the microglial activation later in different regions of the CNS. In our IIM model, one systemic LPS injection suppressed microglial activation in the 5xFAD mice. It reduced the amount of RIPA-soluble Aβ42 in the PFC and the number of dense-core Aβ deposits, particularly at the DG. Similar effects were found in the cortex in the APP23 AD model after repeated LPS administration, but not with a single stimulation (Wendeln et al., 2018). The more aggressive 5xFAD model manifests microglial activation at 6 weeks and the first dense-core Аβ plaques at 10 weeks (Oakley et al., 2006;Boza-Serrano et al., 2018b). The APP23 mice, on the other hand, present the first Аβ plaque with massive microglial activation at an older age, about 6 months (Sturchler-Pierrat et al., 1997). The setup of our study primed the microglia in the initial phase of mild inflammation in the hippocampus (Boza-Serrano et al., 2018b). In addition, we found that in this specific stage, LPS priming altered microglial morphology (Fig. 1A-D), and how they were recruited to Aβ plaques ( Fig. 3A and B), which has not been observed in previous work to our knowledge. We, therefore, assumed that the phenotype of microglia after priming in the 5xFAD model has distinctive features compared to the ones defined earlier, called immune tolerance microglia (Wendeln et al., 2018).
Comprehensive comparisons of the microglial and monocyte proteome from a neurodegenerative (Ctrl-5xFAD), IIM (Primed-5xFAD), and inflammatory models (Primed-WT) highlight shared and differentially expressed proteins in each state. Taking advantage of proteomic analysis, we hypothesize that peripheral LPS stimulation at 6 weeks suppresses the pro-inflammatory microglial response triggered by AD pathology. To analyze how priming shapes brain pathology, Aβ40 and Aβ42 from two cognition-and learning-related regions, including PFC and hippocampus, were measured with multiplex ELISA and immunohistochemical staining. Interestingly, fewer ThioS + Aβ plaques were quantified in DG from Primed-5xFAD than Ctrl-5xFAD due to a potential increase of Aβ phagocytosis in microglia ( Fig. 1E and F, Fig. 2C and D). Meanwhile, priming also induced a reduction of soluble Aβ40 and Aβ42 in PFC (Fig. 6D) which probably led to a smaller size of MOAB2 + Aβ plaques ( Fig. 2A and B). Importantly, and surprisingly, long-term memory tests on these mice revealed that priming could rescue the cognitive deficits in 5xFAD mice.
This work also provides evidence of the long-lasting impacts of priming on other cell types, including synaptic protein colocalization and astrocytic response. We first investigated how priming affects synaptic protein colocalization in different regions. In the RSC region, we observed a decrease in colocalization of pre-synapses (Synaptophysin) and post-synapses (PSD95) in RSC (Fig. 4D), coinciding with a reduction of soma area of microglia (Fig. 1D). However, we did not observe a significant decrease in PSD95 and Synaptophysin at regions we were interested in from 5xFAD mice at this age which might be explained by the synaptic loss started at 9 months (Oakley et al., 2006). Still, cognitive impairment was confirmed in the novel object recognition tests from this age ( Fig. 2G and H). Many studies have proved that single and multiple injections of LPS systemically can cause neuronal death and synapse loss Bodea et al., 2014;Manabe et al., 2021). But these studies examined mostly the acute effects of LPS. The current study showed a reduction of colocalization of pre-synapses and postsynapses in mice 140 days after the 1 mg/kg LPS challenge. The decrease of colocalization might implicate disruption of neurotransmission and alteration in synaptic connectivity occurring in RSC after priming. Microglia modulate early synapse loss via the complement system before Aβ plaque deposition (Hong et al., 2016). Extensive evidence demonstrates that ramified microglia, rather than amoeboid microglia, actively engage in the phagocytosis of apoptotic neurons through their terminal branches (Sierra et al., 2010;VanRyzin et al., 2022). Reduced IBA1 coverage and decreased microglial soma in RSC following priming could potentially be related to an altered microgliasynapses function (Fig. 1B and D). Moreover, the astrocytes exhibited stronger GFAP intensity not only due to AD pathology but also in an LPSdependent way (Fig. 3D and E). Astrocytes are critical components in modulating BBB permeability, neuroinflammation and AD pathology. Further work is needed to elucidate how astrocytes adapt to the priming in cerebral β-amyloidosis, which is outside the scope of this study.
From proteomic profiles of microglia from three different models, we found that signatures in the inflammatory model shared similarities with AD models. Of all proteins upregulated in Primed-WT microglia, 39.7% of proteins were also upregulated in Ctrl-5xFAD, such as H2-D1, Ndufa7, and Ctsh. Among all proteins decreased in the inflammatory model, 41.2% were downregulated with AD pathology, including Hmga1, Ifi204, and Ptma. Notably, the unique microglial markers expressed with AD pathology (Clu, Trem2, ApoE, Hltra1, and Ptn), were also identified in our study (Fig. 6B) (Grau et al., 2005;Rangaraju et al., 2018;Fernández-Calle et al., 2022). We also found an increased amount of Аβ in primed microglia ( Fig. 1E and F, Fig. 6C). Based on our observation of a decrease in levels of RIPA-soluble Аβ40 and 42 in the PFC (Fig. 6D and E) and absence of any indication of decreased Аβ production (Supplementary Fig. 1E), we surmise that primed microglia have a greater capacity to internalize Аβ but may lack the capacity to degrade it. We neither found differentially changed proteins related to phagocytosis from the proteomic profiles nor alterations in the expression of IDE and ApoE due to priming in the 5xFAD mice. Therefore, it is plausible that the time point of the peak in phagocytosis took place before the time of sacrifice and our investigation. Moreover, we found that a set of proteins in Ctrl-5xFAD microglia were altered conversely after priming at 6 weeks (Fig. 7F). And CD115 + Ly6C hi CD11b + monocytes were not influenced that much either in the AD model or the inflammatory model. Even though there was a higher percentage of this subset of monocytes in the 5xFAD mice than in WT mice (Fig. 5B), this subpopulation was well-reserved inside the bone marrow. Thus, our data suggest priming in WT promotes microglia to an AD-like phenotype while elevating the microglial internalization of Aβ in the 5xFAD model.
Previously, transcriptional profiles indicated HIF-1 and Rap1 signaling are down-regulated in the microglia from four times LPSinjected APP23 mice versus vehicle-treated APP23 (Wendeln et al., 2018). Conversely, our proteomic data do not show a reduction of the Rap1 pathway in the Primed-5xFAD versus Ctrl-5xFAD. The other significantly repressed proteins in Primed-5xFAD microglia are enriched with the Ras pathway. Nevertheless, Ras signaling and the Rasdependent MAPK pathway are highly activated in neurons and microglia around Aβ plaques at an early stage of AD (Gärtner et al., 1995, Gärtner et al., 1999. The suppression of Ras signaling by priming is likely to conduct less inflammatory response in microglia. We have also observed less inflammation by examination of Gal3 + immunoreactivity in DG and RSC regions in LPS-primed microglia in the 5xFAD model ( Fig. 3A and C). It is further implied by less microglia around Aβ plaques after priming ( Fig. 3A and B) and a trend of reduction in MIP-1α release in Primed-5xFAD versus Ctrl-5xFAD ( Supplementary Fig. 23A). Taken together, we provide evidence that priming at 6 weeks reduces the Gal3associated inflammation and alters microglial barrier surrounding Aβ deposits 140 days after the peripheral immune stimulation, possibly via inhibition of the Ras pathways. Further studies are required to elucidate the mechanism of how Ras signaling regulates microglial responses in AD.
While microglia play a critical role in the clearance of Aβ deposits, other mechanisms also likely contribute to the overall reduction of Aβ levels besides microglial clearance. These mechanisms include Aβdegrading proteases, such as neprilysin and IDE, which promote the removal of Aβ in the brain (Farris et al., 2003;Miners et al., 2009;Dorfman et al., 2010). Although western blot results of IDE expression in PFC and hippocampus did not show differences (Supplementary Fig. 1A and D) between Ctrl-5xFAD and Primed-5xFAD mice. We could not rule out the possibility of increased enzymatic activity in other cognitionrelated brain regions or at an earlier time point other than 6 months old. Additionally, the activation of other cleaning systems (e.g. the glymphatic system), responsible for the clearance of metabolites from the brain, may also facilitate the removal of Aβ as well (Iliff et al., 2012;Reeves et al., 2020). Given the evidence of enhanced GFAP immunoreactivity with LPS priming in DG, it is plausible to speculate that the priming may potentially influence the functionality of the glymphatic system. Moreover, although the priming did not show any significant impact on APP expression ( Fig. 1E and H), it is important to acknowledge that we cannot entirely exclude the possibility of priming effects on the underlying processes involved in Aβ formation. These alternative mechanisms highlight the complexity of Aβ regulation and underscore the need for a comprehensive understanding of the multiple pathways involved in amyloid reduction.
The limitation of the current work is that we only examined one timepoint for the effects of LPS priming. Therefore, we likely missed the time window to fully elucidate the underlying mechanism of increased uptake of Aβ in the microglia after priming, supposedly related to internalization/degradation. Also, we cannot rule out the effects of some compensatory or adaptive mechanisms during the progression of the disease. Further experiments are required to demonstrate the impact of astrocytes in the clearance of Aβ after priming. Another drawback is that we only used males for a more homogenous group, as females increase the variability due to estrous cycle disruption. Therefore, we cannot predict whether similar phenomena will occur in females. However, the findings of this study are still worthwhile in elucidating the IIM in microglia and demonstrating the effects of priming in AD-like pathology.

Conclusion
To summarize, our findings emphasize the potential long-term effect of peripheral inflammation/priming, altering behavior and pathology related to Aβ in adulthood. Such effects lead to an innate immunity memory (IIM) in microglia encountering Aβ as a sequential insult of LPS priming. We believe that the investigation of IIM can improve our understanding of modifying the microglial response in AD. Priming may alleviate AD pathology-like consequences, possibly by enhanced Aβ internalization, altered recruitment of microglia around Aβ, and impeded microglial activation. Moreover, the imprint signatures of IIMmicroglia display a distinctive phenotype with downregulation of Ras signaling and suppression of inflammatory response. This study implies that a single peripheral immune stimulation may give rise to immune memory in microglia and the significance of an extended time window for adapting microglial responses. The findings of our work stress the important role of systemic inflammation and its effect on microglia in AD pathogenesis.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.