Apolipoprotein E4 expressed by microglia impairs microglial functions and enhances neurotoxicity

Background Microglia, the major cell type that mediates active immune defence in the central nervous system (CNS), constantly survey the brain parenchyma through highly motile processes. Mounting evidence has implicated both benecial and toxic roles of microglia when over-activated upon neuronal injury. Understanding the function of microglia in the brain may uncover the regulatory mechanisms for neuroinammation and facilitate the development of a novel therapeutic strategy for Alzheimer's disease (AD). The ε4 allele of apolipoprotein E (APOE) is a major genetic risk factor for the late onset AD. ApoE, as the major cholesterol carrier in the brain, has been implicated in AD pathogenesis. However, how APOE and APOE isoforms directly regulate microglial functions remains largely unknown. Methods Using primary culture of microglia from Apoe knockout (KO) mice, APOE3 and APOE4 targeted replacement (TR) mouse, we investigated the characteristics of microglial secreted apoE particles and the biological effects of apoE isoforms on microglial inammatory response, migratory ability, cell viability and proliferation. Meanwhile, microglia-neuron co-culture system was utilized to study the effects of apoE isoforms on neurite outgrowth. Results Herein, we found that microglia secret abundant lipidated apoE. Interestingly, apoE4 particles from primary microglia exhibited a higher lipidation status compared to apoE3 particles. Furthermore, apoE4 microglia exhibited a reduced migratory ability as well as enhanced inammatory responses and neurotoxicity, indicating microglial apoE4 is involved in unfavourable functions. Conclusions Our ndings revealed the critical roles for apoE and apoE isoforms in regulating microglial functions. Our results also indicate that targeting apoE-mediated microglial inammatory responses may serve as a potential therapeutic strategy for AD. pro-inammatory responses and reduced migration in response to stimuli compared with apoE3 microglia. Most importantly, apoE4 microglia exhibited strong inhibitory effects on neurite outgrowth. Our ndings support a proposed model that apoE isoforms affect microglial functions, leading to subsequent neurotoxicity and dysfunction of microglial clearance (see details of schematic model in Figure 8). Together, our studies show that the characteristics of apoE from microglia have a critical role in regulating microglial functions, and demonstrate the unfavourable role of the APOE4 allele in microglia function. These results highlight the potential for targeting apoE-mediated inammatory responses as a therapeutic strategy for AD.

Primary astrocytes were prepared as described by a previous protocol [41] with modi cation. Simply, mixed glial cells from newborn (postnatal 1 to 3 day old) pups were cultured in astrocyte culture media (DMEM, high glucose + 10% heat-inactivated fetal bovine serum + 1% penicillin/streptomycin). The medium was changed 2 days after plating of the mixed cortical cells and every 3 days thereafter. At day 9 or day 10, when astrocytes were con uent, mixed cells were shaken at 220 rpm for 30 min to remove the upper microglial cells. Trypsin (Sigma, T2601) was used to split attached astrocytes for further culture or use.
Primary cortical neurons were obtained from 11 to 17 day old embryos of wild-type C57BL/6 mice and cultured in neurobasal medium (GIBCO) supplemented with 0.5 mM GlutaMAX (GIBCO), 2% B27 (GIBCO), and 1% penicillin-streptomycin (Invitrogen) on cover glasses pre-coated with poly-D-lysine solution (50 µg/mL). At day 5 of the in vitro study (DIV5), the neurons were treated with 10 M cytosine arabinofuranoside (Sigma Aldrich) for 2 days to remove glial cells. At DIV7 the culture medium was then replaced with fresh neurobasal medium containing B27 and penicillin-streptomycin. For neuron-microglia co-cultures, microglia were re-suspended in neuronal culture medium and were seeded on top of primary neurons at DIV8 to a nal ratio of 1:2 (microglia:neuron).

| Apoe knockdown by siRNA
Two different Apoe speci c siRNAs (chemically synthesized by Dharmacon) were used to knockdown Apoe in microglia by electroporation using an Amaxa Nucleofector and a glial speci c Nucleofector kit (LONZA), according to the manufacturer's instructions. Each electroporation reaction contained 4 × 10 6 cells and 300 nM siRNA. Transfected cells were seeded and used for subsequent experiments.

| Assessment of microglial migration
Microglial cells (5 × 10 4 ) in culture medium were added to the upper well of each Transwell insert (Corning), each of which bears an uncoated lter with 8 μm diameter holes. After 48 hr, the medium in the Transwell insert was replaced with 0.2% FBS/DMEM, and DMEM containing 10% FBS, adenosine triphosphate (ATP; 300 nM), or lipopolysaccharide (LPS; 1 µg/mL) (Sigma Aldrich) was added to the lower well to induce migration. After 24 hr, the cell-bearing lters were xed in 4% paraformaldehyde for 10 min, rinsed with PBS, and the microglial cells remaining on the upper side of each lter removed with a cotton swab. The lters were stained with 0.1% crystal violet (Sigma Aldrich) for 30 min and then rinsed with PBS and water. The number of cells that had migrated to the underside was counted (8 random elds/ lter) at 20x or 10x magni cation using an Olympus DP71 microscope (Olympus, Tokyo, Japan).

| BrdU incorporation assay
BrdU incorporation assay was performed using BrdU cell proliferation ELISA kit (Abcam) following the manufacturer's protocol. Brie y, microglial cells were seeded at a density of 1.0 x 10 4 cells/well and BrdU was added to the cells for 48 hr for incorporation. Incorporated BrdU was examined through anti-BrdU antibodies after the cells were xed, permeabilized and denatured. The cells were then incubated with horseradish peroxidase-conjugated secondary antibodies and the colored reaction product quanti ed using a microplate spectrophotometer (Varioan Flash; Thermo Fisher Scienti c).

| Western blot analysis
All cells were lysed in RIPA buffer and total protein concentrations were determined via a BCA Protein Assay Kit (Thermo Scienti c). Total protein (40 µg) was loaded into 10% SDS-PAGE. Western blot was performed as has been previously described [42]. Brie y, gels were rst transferred onto PVDF membranes (Millipore). After which, primary antibodies were incubated overnight at 4°C, followed by secondary antibody incubation. The following antibodies were used in this study: Conditioned media were generated by culturing cells in serum-free media for 24 hr in T75 asks and then collected. Conditioned media were then concentrated 20 fold using a 10-kDa cut-off lter (Millipore) and centrifuged to remove cellular debris before storage at 4°C prior to fractionation. Samples were run through an AKTA FPLC system through a Heparin column (heparin a nity chromatography) or a Superose TM 6 column (size exclusion chromatography; GE Healthcare). Fractions were collected and stored at 4°C until further analysis.

| Cholesterol assay
Cholesterol levels were analyzed using the Amplex Red cholesterol assay kit (Life Sciences) according to the manufacturer's instructions. Brie y, samples were added into an opaque 96-well plate. Standards and samples were incubated with Amplex Red reagent (300 µM Amplex Red, 2 units/mL cholesterol oxidase, 2 units/mL HRP, and 0.2 units/mL cholesterol esterase) at 37°C for 30 min and uorescence was measured using excitation in the range of 550 nm and emission detection at 590 nm.

| ApoE ELISA
A 96-well plate (Fisher Scienti c) was coated overnight with WUE4 antibody [43]. After blocking in 1% nonfat milk in PBS, samples of the appropriate dilution were incubated with the detection antibody (K74180B), followed by incubation with streptavidin-poly-HRP antibodies (Fitzgerald). Finally, tetramethylbenzidine (TMB) (Sigma Aldrich) was added to each well and the substrate-peroxidase reaction was stopped by sulfuric acid stop solution. The absorbance was read at 450 nm using a BioTek 600 plate reader. ApoE concentration of samples was calculated against a standard curve derived from serial dilutions of recombinant human apoE3 or apoE4 protein (Fitzgerald).

| Analysis of apoE/lipoprotein particles sizes by Native PAGE
48 hr post shaking and seeding of primary cells, the medium was replaced with serum-free medium and the conditioned medium harvested and concentrated. ApoE particles in the concentrated medium were quanti ed and normalized to apoE3 in the conditioned medium. Then, equal amounts of apoE3 and apoE4 proteins were separated by Native PAGE Novex 4-20% Tris-Glysine gels (Thermo Fisher) under native conditions following the manufacturer's instructions and transferred to a PVDF membrane (Millipore) at 300 mA for 1.5 hr. Ponceau S staining solution (0.1% (w/v) Ponceau S in 5% (v/v) acetic acid) was used in blotting to visualize the molecular mass markers. Particle sizes were estimated using the Native Mark Unstained Protein Standard (Invitrogen). After washing, the membrane was incubated with goat anti-apoE antibody (K74180B, Meridian Life Science) overnight at 4°C, followed by avidinlabeled donkey anti-goat antibody. Western blot bands were quanti ed by Image J software.

| ApoE associated cholesterol assay
Avidin-agarose beads (Pierce) were pre-coupled with biotinylated polyclonal anti-apoE antibody (K74180B, Meridian Life Science) and then incubated overnight with concentrated conditioned media at 4°C. Complexes of bead-antibody-apoE were washed with TBS buffer three times, then apoE associated cholesterol was eluted by 0.1% Triton X-100/TBS. 0.1 M glycine (pH 2.5) was used to elute immunoprecipitated apoE and then neutralized with 1 M Tris (pH 8.5).
2.15 | Quanti cation of neurite outgrowth and cytokines level in co-culture system 15-25 cells from 5-6 images were analysed and quanti ed following a method reported previously [44].
Neurite number (neurite initiation sites) and length were counted and measured using Image J. The levels of IL-1β and IL-6 in the co-culture system were examined using Quantikine ELISA kits according to manufacturer's instructions (R&D Systems, Minneapolis, MN). The absorbance was measured at 450 nm with a 590 nm correction and concentrations were calculated.

| Statistical analysis
All quanti ed data is represented as mean ± SEM. Statistical signi cance was determined with an unpaired t test, one-way ANOVA test, or two-way ANOVA and Tukey's post hoc test (GraphPad Prism 6). All experiments were performed in at least triplicate and p <0.05 was considered signi cant.

| Primary microglia express abundant lipidated apoE
To examine the role of apoE in microglia, we rst compared the protein levels of apoE between primary microglia and astrocytes from C57BL/6 WT mice by Western blotting. Interestingly, we found that compared to astrocytes, microglia expressed higher levels of apoE (~ 12 fold) in cell lysate (Figure 1a,b).
To evaluate the stability of apoE protein in these two cell types, the cells were treated with cycloheximide (CHX) to inhibit protein synthesis and harvested at indicated time points (0, 1, 2, 4 hr). Our results showed that microglial apoE was signi cantly more stable than astrocytic apoE (~ 70% v.s. ~ 10%) (Figure 1c,d), which may contribute to the above observation that primary microglia exhibited higher apoE levels than primary astrocytes.
Since APOE is a major lipid carrier in the brain, we compared the lipidation status of apoE derived from primary microglia to that of astrocytes. ApoE in conditioned medium from each cell type was puri ed by heparin a nity chromatography and the levels of apoE-associated cholesterol were examined. We found that microglial apoE carries a similar (p=0.2125) amount of cholesterol compared to apoE derived from primary astrocytes (Figure 2a). Notably, size exclusion chromatography (SEC) showed that apoE can be detected earlier in microglia-derived conditioned medium, indicating that microglial apoE complexes exhibit increased size over astrocyte apoE complexes ( Figure 2b). As ATP-Binding Cassette transporter A1 (ABCA1) is a key regulator of apoE lipidation [45][46][47], we next examined its expression in primary microglia. ABCA1 expression was abundant in primary microglia and signi cantly upregulated compared to astrocyte levels (~ 3 fold) (Figure 2c,d). Taken together, these results indicate that microglia express abundant levels lipidated apoE.

| Microglial apoE particles exhibit different lipidation status depending on isoforms
To investigate whether apoE particles secreted from microglia display different lipidation status, we examined the levels of ABCA1 in primary microglia derived from APOE3-TR and APOE4-TR mice, which express the human APOE isoform driven by the endogenous murine Apoe promoter. We found that apoE4 primary microglia exhibited higher ABCA1 levels (~1.6-fold) compared to apoE3 microglia (Figure 3a,b). Conversely, both the protein and mRNA levels of APOE were lower (protein, ~40% less; mRNA, 60% less) in apoE4 primary microglia compared to those in apoE3 microglia (Figure 3a,c,d). These results indicate that microglial apoE may be more lipidated. To directly assess whether the lipidation status of microglial apoE is isoform-dependent, we detected the level of apoE-associated cholesterol in particles isolated from apoE3 and apoE4 primary microglia. ApoE-containing particles in the conditioned medium were immunoprecipitated with a biotinylated apoE antibody and the quantities of cholesterol coimmunoprecipitated with apoE were quanti ed as previously described [48]. The particles from apoE4 primary astrocytes carried less cholesterol (~ 20% less) than that from apoE3 astrocytes (Sup Figure 1a) [35]. Intriguingly, apoE-containing particles from apoE4 primary microglia exhibited higher cholesterol levels (~ 1.7 fold) than those with apoE3 (Sup Figure 1b), a result consistent with the higher ABCA1 levels found in apoE4 primary microglia. Furthermore, we evaluated the sizes of apoE3 and apoE4-containing particles secreted from microglia by non-denaturing gel electrophoresis followed by Western blotting as described [48]. The sizes of apoE/lipoprotein particles were catergorized as large particles (> 690 kDa), medium particles (232-690 kDa), or small particles (< 232 kDa). Compared with apoE3 primary microglia, apoE4 microglia secreted more large particles (12% more, ratio to total apoE/lipoprotein) and less small particles (~ 10% less, ratio to total apoE/lipoprotein) (Sup Figure 1c,d). Conversely, APOE4 primary astrocytes secreted more small particles and less large particles than their APOE3 counterparts (Sup Figure 1e,f). Collectively, our results indicate that in contrast to astrocytes, particles from microglia exhibit distinct biochemical properties depending on the APOE isoforms, which may play a critical role in determining their functions.

| APOE4 microglia exhibits proin ammatory phenotype
It has been proposed that microglia release cytotoxic mediators and mediate in ammatory response in the AD brain [49]. To investigate the roles of Apoe in microglial functions, we utilized primary microglia from Apoe-knockout (KO) mice and established technique to knock down (KD) Apoe in microglia from WT mice. As expected, apoE was undetectable in Apoe-KO primary microglia (Sup Figure 2a,b). We also successfully knocked down Apoe in WT microglia using two siRNAs targeting distinct regions of Apoe by electroporation (Sup Figure 2c,d). To assess whether apoE in microglia mediates immune responses, we rst examined the levels of in ammatory cytokines induced by LPS in wide-type (WT) or Apoe-KO primary microglia. Our results demonstrated that the mRNA levels of Il-1β and Il-6 were dramatically increased in Apoe-KO microglia treated with LPS, compared to WT microglia. Whereas, there was no difference between these cells in the absence of LPS (Figure 4a,b). Similar results were also observed in Apoeknockdown (KD) primary microglia (Figure 4c,d). These results illustrate that defects in Apoe expression and/or function may exaggerate neuroin ammation, which could contribute to AD pathogenesis.
To further examine whether the microglial in ammatory response is APOE isoform-dependent, we compared the pro-in ammatory cytokine levels upon LPS stimulation in apoE3 and apoE4 primary microglia. We found that primary microglia from apoE4-TR mice were hypersensitive (Il-1b mRNA: apoE4, 40-fold v.s. apoE3, ~28-fold; Il-6 mRNA: apoE4, 17000-fold v.s. apoE3, 10000-fold) to LPS (Figure 4e,f), suggesting that apoE4 microglia exhibits a reduced ability to suppress in ammation compared with apoE3 microglia. Therefore, our studies indicate that microglial apoE plays an anti-in ammatory role upon immune stimulation, however, apoE4 microglia displays a pro-in ammatory phenotype.

| APOE4 microglia exhibits reduced migratory ability
Microglia are highly motile cells that act as the main form of active immune defense in the CNS [50]. To investigate whether apoE is required for microglial migration, we rst compared the migratory property of WT and Apoe-KO microglia using a two-dimensional "wound healing" assay [51]. We found that Apoe-KO primary microglia exhibited reduced migration ability (40~50% less migrated cells) compared with WT cells (Figure 5a). We next performed a three-dimensional cell migration assay using Transwell chambers (Figure 5b) with 10% FBS as a chemoattractant. This study revealed that the number of Apoe-KO microglia which migrated from the upper chamber to the lower chamber was signi cantly reduced (50% less) compared to WT microglia (Figure 5c), suggesting that the expression of Apoe is required for microglial mobility. To investigate whether APOE isoforms affect microglial migration, we performed a cell migration assay using primary microglia isolated from apoE3-TR and apoE4-TR mice (Figure 6a). It has been reported that LPS and ATP can e ciently induce microglial migration [52][53][54]. Thus, we examined how microglia with different apoE isoforms migrated upon the stimulation of these chemoattractants, in addition to the above used 10% FBS. Interestingly, apoE4 microglia exhibited reduced migratory activity (40-50% less) in response to the three inducers compared to apoE3 microglia (Figure 6b), indicating that apoE4 microglial migration is impaired. This impaired migration of apoE4 microglia may result in a reduced capacity for immune defense in the brain and contribute to the pathogenesis of AD.
3.5 | APOE4 microglia secretes pro-in ammatory cytokines that lead to severe neurotoxicity Previous studies suggest that microglia may exert distinct actions in response to stimuli which can be either harmful or bene cial to neuronal growth [55]. To investigate how microglia with different apoE isoforms affect neural outgrowth, we established a co-culture system (Figure 7a1) of primary microglia and neurons to examine the effects of microglia secretions on neurite outgrowth. Neurite outgrowth was assessed by the numbers of neurite initiation sites and lengths of neurites ( Figure  7a2). Co-culture assay revealed that the neurite outgrowth was suppressed when neurons were cocultured with apoE3 or apoE4 primary microglia (Figure 7b-d), when compared with the no microglia on the top insert group (control). However, the neurites exhibited a much greater reduction outgrowth within the apoE4 microglial co-culture system that the apoE3 co-culture (Figure 7b-d). As pro-in ammatory cytokines secreted by activated microglia has been shown to inhibit neurite outgrowth [55,56], we next examined the cytokine levels in the co-culture system. As expected, the levels of pro-in ammatory cytokines, Il-1β and Il-6, were increased in the system when neurons were co-cultured with either apoE3 (IL-1β, ~ 4.5 fold; IL6, ~ 7.5 fold) or apoE4 (Il-1β, ~ 11 fold; Il6, ~ 10 fold) primary microglia (Figure 7e,f). Furthermore, even both apoE3 and apoE4 microglia exibit proin ammatory response in condition of coculture system, apoE4 microglia elicited stronger in ammatory responses with higher cytokines levels compared to apoE3 microglia in the co-culture system (Figure 7e,f). Our results indicate that apoE4 microglia result in greater neurotoxic effects by promoting neuroin ammation and this could exacerbate AD pathogenesis.

| APOE isoforms do not affect microglial viability and proliferation
Microglial proliferation and viability play important roles in neuronal protection and tissue repair [57,58]. We thus used staurosporine [59] or GM-CSF [60,61] as inducers to assess the effects of Apoe de ciency on microglial viability and proliferation. We found that Apoe de ciency had no effect on the viability or proliferation of microglia (Sup Figure 3a,b). Furthermore, the viability and proliferation in microglia from apoE3-TR mice were not signi cantly changed when compared to those in microglia from apoE4-TR mice (Sup Figure 3c,d). These results suggest that apoE may not be involved in microglial viability or proliferation phenotypes.

Discussion
Resident microglia are a specialized population of immune cells in the brain, acting as the rst defense of the CNS [62]. Given that in ammation-related genes --including CD33 and TREM2, which are primarily expressed in microglia [63][64][65][66][67] --have been found to be risk factors for AD, the signi cance of microglia in AD development is more prominent [24,68,69]. ApoE has been shown to be involved in AD pathogenesis, including modulation of Aβ clearance, tau pathogenesis, lipid transport, and synaptic functions [28,30,70,71]. However, the roles of apoE in microglial functions have remained elusive.
Neuroin ammation induced by microglial activation is an early event and an important pathological feature in the pathogenesis of AD [69,72,73]. Our results showed that Apoe-KO primary microglia displayed exacerbated in ammatory responses, indicating the anti-in ammatory role of apoE. Thus, increasing apoE expression in primary microglia could facilitate a reduction of in ammatory responses during AD development. Additionally, apoE isoforms have been shown to differently regulate neuroin ammatory responses. In the brain hippocampus, apoE4-TR mice showed increased glial activation in response to intracerebroventricular LPS stimulation compared to apoE2-TR and apoE3-TR mice [74]. Additionally, glial cells (~95% astrocytes, 5% microglia) from apoE4-TR mice have been found to secrete more robust pro-in ammatory cytokines than apoE3-TR mice [75]. In our study, we demonstrated that apoE4 primary microglia was more hypersensitive to LPS stimulation than apoE3 microglia, providing a mechanistic link betweenapoE4-mediated neuroin ammation and AD development. Our results are consistent with previous reports that different apoE isoforms differentially regulate the microglial in ammatory response [76]. Together, these ndings indicate that apoE isoforms might impact in ammatory responses through diverse mechanisms. However, whether apoE4 from different cell types regulates neuroin ammation through its gain-of-functions in the pro-in ammatory response or loss-of-functions in the anti-in ammatory requires further investigation.\ The ability of microglia migrating to the injury sites in the brain in response to various stressors is critical to their physiological and pathophysiological actions [77]. Therefore, migration is a key component for microglia in the clearance of cell debris in tissue repair [52,78]. Our study demonstrated thatapoE is required for microglial migration and in an apoE isoform-dependent manner, in which apoE4 microglia exhibit reduced migration toward distinct chemotaxis. This is consistent with previous evaluations, where the migration of apoE2 and apoE4 microglia were suppressed in response to complement C5a and ATP [79]. Several studies have suggested that matrix metalloproteinases (MMPs) play important roles in regulating microglial migration. ATP, speci cally, induces microglial migration by regulating MMP2 and MMP9 [53,80]. Furthermore, MMP2, MMP9, MMP12, and MMP14 have all been found to be increased in LPS-mediated microglial migration [78,81]. It has been reported that the blood brain barrier recover from traumatic brain injury faster in apoE3-TR mice than those in apoE4-TR mice, a feature which may due to the differing levels of MMP-9 expression [82]. Whether differing apoE isoforms affect microglial migration through differentially mediating the activation of MMPs warrants further investigation. Glial cells are critical regulators of synaptic connectivity both in healthy and diseased brains [83].
Connections between neurons have also been shown to be modi ed by microglia [84]. Additionally, microglia have shown the capacity to medicate synapse pruning [85][86][87][88]. One study showed that when neurons are co-cultured with glial cells (~95% astrocytes), apoE4 glia cells display a gain-of-toxic function in mediating neuronal growth [89]. However, how microglia with different APOE isoforms affect neurite outgrowth remains unclear. Here, we found that microglia co-cultured with neurons suppressed neurite outgrowth. Interestingly, apoE4 primary microglia secreted higher levels of pro-in ammatory cytokines in the co-culture system than their apoE3 counterparts. These ndings indicate that apoE4 microglia may induce higher neurotoxicity through promoting the neuroin ammatory responses. Microglial cells have been reported to express one of the highest levels of APOE transcript in mice brain tissues, second only to astrocytes [90]. We also found that microglia expressed abundant APOE, which was lipidated and more stable than that from astrocytes. These results are consistent with previous studies in murine BV2 cells [91]. Intriguingly, microglial apoE4 particles were larger and exhibited a higher lipidated status than microglial apoE3 particles, which is the opposite of what was observed in primary astrocytes (i.e. apoE3> apoE4). Enhanced lipid synthesis has been reported in macrophages during in ammation [92] and high-cholesterol diets have been shown to induce microglial activation in the hippocampus of rabbits [93], implying that higher lipid levels may increase in ammatory responses. It is possible that higher cholesterol levels carried by the APOE4 microglia particles may contribute to their pro-in ammatory aspects, leading to enhanced neurotoxicity in AD pathogenesis.
In Tarja Malm [94], Li-Huei Tsai [95] and Julia TCW (preprint, bioRxiv, doi: https://doi.org/10.1101/713362)'s studies using human iPSC-derived microglia models, they observed a dysfunctional or impaired phenotype of APOE4 iPSC-derived microglia when compared with that of APOE3, indicating that APOE4 might contribute to AD risk via dysregulating microglia. However, the opposite conclusion has been reported in another study performed with an murine microglial cell line [96]. Our results on this contentious issue are consistent with those studies using human iPSC-derived microglia cells, although the microglia used in this work were isolated from mouse models.
There is an urgent and desperate need for and effective AD treatment [97,98] and the targeting APOE has become one of the most attractive potential strategies [30,99]. Understanding the characteristics of microglial apoE in CNS will facilitate the successful development of APOE-targeting strategies in AD treatment. Our results demonstrate the critical roles of apoE and different roles of apoE isoforms in microglial functions, yet these results need to be further veri ed in additional animal models. Further understanding of the molecular mechanisms and intracellular pathways underlying apoE-mediated microglial functions may shed light on how apoE and apoE isoforms contribute to AD pathogenesis and can be targeted for AD therapy.

Conclusions
In the current study, we found that microglia-derived apoE crucially modulates microglia functions in an isoform-dependent manner. ApoE in microglia plays an anti-in ammatory role and regulates migratory behaviors of microglia without signi cant effects on microglial proliferation or viability. Furthermore, apoE4 primary microglia showed enhanced pro-in ammatory responses and reduced migration in response to stimuli compared with apoE3 microglia. Most importantly, apoE4 microglia exhibited strong inhibitory effects on neurite outgrowth. Our ndings support a proposed model that apoE isoforms affect microglial functions, leading to subsequent neurotoxicity and dysfunction of microglial clearance (see details of schematic model in Figure 8). Together, our studies show that the characteristics of apoE from microglia have a critical role in regulating microglial functions, and demonstrate the unfavourable role of the APOE4 allele in microglia function. These results highlight the potential for targeting apoE-mediated in ammatory responses as a therapeutic strategy for AD. The datasets used during the current study are available from the corresponding author on reasonable request.