Lipidome and proteome of astrocyte and microglia ApoE lipoprotein reveal differences based on cell type and ApoE isoform

TL;DR

This study compares the lipid and protein content of ApoE lipoproteins from astrocytes and microglia, showing differences based on cell type and ApoE isoform, which may influence Alzheimer's disease risk.

Contribution

The study reveals distinct lipid and protein profiles of ApoE lipoproteins based on their cellular origin and ApoE isoform, offering new insights into their roles in CNS function and disease.

Findings

Microglia-derived ApoE lipoproteins are enriched in cholesteryl esters, while astrocyte-derived ones are enriched in sphingomyelin (SM).

ApoE4 microglial lipoproteins show enrichment in complement component 1q and Lpl, whereas ApoE2/3 lipoproteins are enriched in Ankk1 and apolipoprotein C1.

Astrocyte ApoE lipoproteins are associated with glucose metabolism and acute phase response proteins, while microglial ones are linked to immune and synapse-related functions.

Abstract

Apolipoprotein E (ApoE) is the primary, most abundant apolipoprotein of the CNS and plays an important role in brain metabolism and lipid homeostasis. In the CNS, ApoE is primarily secreted by astrocytes under homeostatic conditions and by microglia in certain disease-related conditions. APOE has three major alleles: APOE2, APOE3, and APOE4. APOE4 is the strongest genetic risk factor for late-onset Alzheimer’s disease (AD), and APOE2 results in decreased risk relative to APOE3. ApoE derived from astrocytes and microglia have been hypothesized to play different roles in the disease pathogenesis of AD. In this study, we profiled the lipidome and proteome of ApoE lipoproteins secreted by astrocytes or microglia and found that they differed according to the cellular source of ApoE and the ApoE isoform. Lipidomics revealed that microglia-derived ApoE lipoproteins were enriched in cholesteryl esters, whereas astrocyte ApoE lipoproteins were enriched in SM. Proteomics revealed that astrocyte ApoE lipoproteins were enriched in proteins involved in glucose metabolism and acute phase response. Microglia-secreted lipoproteins were enriched in proteins involved in complement activation, synapse pruning, proteolysis, and the innate immune response. Further comparison of ApoE lipoproteins from APOE4 microglia revealed that ApoE4 lipoproteins were enriched in complement component 1q and Lpl compared with ApoE2 and ApoE3 microglial lipoproteins, which were enriched in Ankk1 (ankyrin repeat and kinase domain containing 1) and apolipoprotein C1. These results provide the molecular foundation for better understanding of how ApoE functions as an apolipoprotein with the lipoprotein cargo being dependent on the cellular source and ApoE isoform, ultimately contributing to CNS homeostasis and disease pathogenesis.

Figures

A: Schematic showing workflow for purification of intact ApoE lipoprotein particles, followed by proteomic and lipidomic analyses. Primary glia were first isolated from the cortex of P2–P3 pups. Mixed glial cultures were then shaken to isolate primary microglia cultures and primary astrocyte cultures. Cultures were switched to serum-free media prior to collection of media. Astrocyte or microglia-conditioned media were flowed over immunoaffinity columns to isolate ApoE lipoproteins. Intact particles were then assessed using MS-based lipidomics and proteomics. B: Nonreducing SDS-PAGE gel shows t

A: Volcano plot showing enrichment of lipid species of astrocyte immunopurified lipoprotein versus microglia immunopurified lipoprotein. Microglia ApoE lipoproteins are enriched in several CE species. Astrocyte ApoE lipoproteins are enriched in several SM species and FC. For lipidomics analysis, the following groups were analyzed: ApoE2 astrocytes (n = 6), ApoE3 astrocytes (n = 6), ApoE4 astrocytes (n = 6), ApoE2 microglia (n = 4), ApoE3 microglia (n = 4), and ApoE4 microglia (n = 4). B: PCA of the lipidome of ApoE lipoproteins shows partial separation of samples based on the cellular source o

A: Histogram of the most abundant proteins measured in astrocyte ApoE lipoproteins. Abundance was determined by excluding common contaminants, showing differential expression compared to proteins measured in ApoE^−/−^ astrocyte-conditioned media, and the protein abundance being greater than the total mean + 0.5 standard deviation threshold. Error bars represent mean ± SEM. For proteomics analysis, the following groups were analyzed: ApoE2 astrocytes (n = 5), ApoE3 astrocytes (n = 5), and ApoE4 astrocytes (n = 5). B: PCA of the proteome of astrocyte ApoE lipoproteins. Circles are colored based

A: Histogram of the most abundant proteins measured in microglia ApoE lipoproteins. Abundance was determined by excluding common contaminants, showing differential expression compared to proteins measured in ApoE^−/−^ microglia-conditioned media, and the protein abundance being greater than the total mean + 0.5 standard deviation threshold. Error bars represent mean ± SEM. For proteomics analysis, the following groups were analyzed: ApoE2 microglia (n = 3), ApoE3 microglia (n = 3), and ApoE4 microglia (n = 4). B: PCA of the proteome of microglia ApoE lipoproteins. Circles are colored based on

A: Volcano plot showing differential protein abundance of proteins bound to astrocyte versus microglia immunopurified lipoprotein. Proteins involved in glucose metabolism (Pkm, HECT domain E3 ubiquitin protein ligase 4 [Hectd4], and ApoD) and acute phase response (fibronectin 1 [Fn1] and Serpina3n) are differentially abundant in astrocyte ApoE lipoproteins. Proteins involved in complement activation (C1qa, C1qb, and complement factor I [CfI]), synapse pruning (C1qa and C1qb), proteolysis (matrix metallopeptidase 12 [Mmp12], serine carboxypeptidase 2 [Cpb2], CfI, and cathepsin B [Ctsb]), and th

A: Volcano plot showing decreased CE species in ApoE lipoproteins immunopurified from T0901317-treated astrocyte-conditioned media compared to ApoE lipoproteins immunopurified from vehicle-treated astrocyte-conditioned media. For lipidomics analysis, the following groups were analyzed: ApoE lipoprotein immunopurified from T0901317-treated ApoE4 astrocyte-conditioned media (n = 3) and ApoE lipoprotein immunopurified from vehicle-treated ApoE4 astrocyte-conditioned media (n = 3). B: PCA of the lipidome of ApoE lipoproteins immunopurified from T0901317-treated astrocyte-conditioned media compared

B). Clu (ApoJ) also colocalized with astrocytes in APPPS1-APOE3 and APPPS1-APOE4 knock-in mice (Supplemental Fig. S6B). Together, these results reinforce the validity of our previous results and the cell specificity of the proteins secreted on ApoE lipoproteins from these various cell types.Fig. 7A: Purity of astrocyte cultures is demonstrated by positive staining of GFAP, an astrocyte marker, and negative staining for Iba1, a microglial marker. Purity of microglial cultures is demonstrated by negative staining of GFAP, an astrocyte marker, in astrocyte culture and positive staining for Iba1,

Full text

Apolipoprotein E (ApoE) is an apolipoprotein that is secreted by a wide variety of cell types and immune cells, including astrocytes (1, 2, 3) and microglia (4, 5). In the CNS, ApoE is lipidated by ABCA1 and secreted as a discoidal lipoprotein (1, 2, 6). APOE is the most important genetic risk factor for late-onset Alzheimer’s disease (AD), with one copy of the APOE4 allele increasing disease risk by about 4-fold and two copies increasing disease risk by about 12-fold relative to the most prevalent APOE3 allele (7, 8). Conversely, APOE2 decreases the risk of developing AD relative to APOE3. Astrocytes are the predominant producer of ApoE in the CNS, with astrocyte-derived ApoE shown to be important for supporting neuronal health (9, 10). In the context of AD pathology, such as amyloid-β (Aβ) plaques or tauopathy, microglia strongly upregulate APOE expression (4, 11). Microglia have also been implicated in the initial fibrillization of Aβ, where ApoE and Aβ aggregates colocalize in lysosomes (12). Cell type-specific production of ApoE also plays a critical role in disease pathogenesis, as evidenced by studies showing astrocyte-specific knockout of ApoE results in markedly decreased Aβ pathology and tau-mediated neurodegeneration (13, 14, 15). The mechanism linking ApoE to AD risk has been difficult to untangle due to the wide variety of cells that secrete and uptake ApoE. Despite the importance of astrocyte and microglial ApoE to disease pathogenesis, the significance of astrocyte- versus microglia-secreted ApoE is unclear, as the protein and lipid composition of these particles has not been directly assessed.

ApoE4 has also been reported to alter the lipid metabolism of human-induced pluripotent stem cell (iPSC)-derived astrocytes and microglia (16). In astrocytes, ApoE4 was found to upregulate de novo cholesterol synthesis despite higher levels of intracellular cholesterol. Another report found that APOE4 expressed in iPSC-derived astrocytes led to increased accumulation of unsaturated fatty acids and lipid droplet formation (17). This higher production of lipids in ApoE4-expressing astrocytes has been reported to affect neuronal health and function and increase Aβ production in cocultures of astrocytes and neurons (16, 18, 19). Under inflammatory conditions, astrocytes can secrete ApoE-carrying neurotoxic lipids that directly cause neuronal death (20).

In the context of demyelination, ApoE4 has been reported to cause microglia to be hypoactive in clearing myelin debris and to accumulate higher levels of lipid droplets compared to ApoE3-expressing microglia (21). Knockout of ApoE in a model of focal demyelination resulted in increased inflammation and greater lesion area due to an inability of microglia to effectively process myelin debris and resolve inflammation (22). Knockout of ApoE in the context of demyelination also drives an increase in esterified cholesterol, not only in microglia but also in astrocytes (23). Knockout of ApoE also leads to an impaired ability of microglia to cluster around Aβ plaques (24). This impaired microglial response leads to greater neuritic dystrophy around plaques, indicative of greater neuronal damage (24, 25). However, these effects may not be the result of microglia-secreted ApoE, as microglia-specific knockout of ApoE has relatively small effects on Aβ pathology relative to the effects of selectively knocking out ApoE in astrocytes, underlining the importance of understanding differences in the role of different cellular sources of ApoE (26, 27).

In this study, we used lipidomics and proteomics to characterize the lipidome and proteome of ApoE2, ApoE3, or ApoE4 lipoproteins secreted by microglia and astrocytes. ApoE is the dominant constituent of the isolated ApoE lipoproteins, as expected (1, 2). However, subtle differences in protein content and lipidome distinguish ApoE lipoproteins secreted by astrocytes and microglia. We also found that the lipidome and proteome of ApoE4 lipoproteins secreted by astrocytes and microglia differ from ApoE2 and ApoE3 lipoproteins. These findings pave the way for a better understanding of how the cellular origin and isoform of ApoE affects disease pathogenesis.

Materials and Methods

Animals

Mixed glia cultures were generated using pups from mice with the human APOE gene knocked into the endogenous mouse Apoe locus (28). Mice were housed in a specific pathogen-free facility with ad libitum access to food and water. Mice were housed under a 12-h light/dark cycle. Due to pooling of multiple pups to generate mixed glia cultures, sex was not considered as a biological factor. For each independent n used in this experiment, 3–5 pups were pooled for the initial mixed glial culture. This culture was then used to isolate astrocyte and microglia cell cultures. Each independent culture was maintained separately, and ApoE was immunopurified from the astrocyte- or microglia-conditioned media, respectively. All animal procedures and experiments were performed under guidelines approved by the Institutional Animal Care and Use Committee at Washington University School of Medicine.

Primary glial cultures and conditioned media collection

The frontal cortex from pups aged P2–P3 was dissected in dissection media (10 mM Hepes [Gibco], 0.6% glucose [Sigma], 1X penicillin/streptomycin [Gibco] in HBSS [Gibco]). Briefly, mice were decapitated, the skull and meninges were removed, and the frontal cortex was dissected and placed in fresh dissection media. Tissue was washed three times with dissection media and then incubated with 0.125% trypsin (Sigma) and 500 μg/ml DNase I (Sigma) for 10 min at 37°C. Tissue was homogenized and spun down at 1,000 g for 5 min at 4°C. Media were removed, and tissue was resuspended in 5 ml of dissection media. An additional 500 μg/ml of DNase I was added and incubated for 5min at 37°C. Tissue was spun down at 1,000 g for 5 min. Supernatant was aspirated, and tissue was resuspended in 10 ml of glia media (DMEM/F12 [Gibco], 1X GlutaMAX [Gibco], 1X sodium pyruvate [Gibco], 1X Hepes [Gibco], 1X penicillin/streptomycin, 10% fetal bovine serum [Gibco]). Cells were plated on a Geltrex (Gibco)-coated T75 flask (TPP). After 24 h, media were aspirated, cells were washed with Dulbecco’s PBS (DPBS) (Gibco), and fresh glia media were added. Once the astrocyte layer was confluent, media were switched to glia media supplemented with 100 ng/ml granulocyte-macrophage colony-stimulating factor (BioLegend). Cells were incubated an additional 7 days, then microglia were removed from culture by shaking flasks at 250 rpm at 37°C overnight. Media-containing microglia were collected and spun down at 1,000 g for 5 min at 4°C. Microglia were plated on poly-l-lysine (Sigma)-coated T75 flasks. Meanwhile, astrocytes were washed with DPBS and then suspended in 0.05% trypsin-EDTA (Gibco). Astrocytes were plated on Geltrex-coated T75 flasks. Cell cultures were maintained until cells were confluent in glia media (microglia media were supplemented with 10 ng/ml macrophage colony-stimulating factor (BioLegend). Once confluent, cells were washed three times in DPBS and switched to serum-free glia media (DMEM/F12, 1X GlutaMAX, 1X sodium pyruvate, 1X Hepes, 1X penicillin/streptomycin, 1X N2 neuronal supplement [Gibco]). Cells were incubated in 50 ml serum-free media for 24 h. Media were collected and replaced with serum-free media. This procedure was repeated three times. Astrocyte- or microglia-conditioned media was filtered through a 20 μm filter and supplemented with a final concentration of 0.02% sodium azide. In order to test the impact of the T0901317 liver X receptor (LXR) agonist on ApoE lipoprotein composition, astrocyte cultures were supplemented with a final concentration of 2 μM T0901317 dissolved in DMSO. An equivalent volume of DMSO was added as a vehicle treatment.

Immunopurification of ApoE

An immunoaffinity column was prepared by dialyzing 150 mg of anti-ApoE antibody HJ15.4 in coupling buffer (0.1 M NaHCO_3_, 0.5 M NaCl, pH 8.3) with four changes every 6 h (29, 30). Five grams of cyanogen bromide-activated Sepharose beads (Cytiva) were washed with 1 L of 1 mM HCl through a sinter glass funnel. Beads were washed with 50 ml of coupling buffer. Beads were spun down at 250 g in a 50 ml conical for 5min. Dialyzed antibody was added to the beads and incubated with rotation for 2 h at room temperature. Beads were spun down at 250 g for 5 min. Beads were washed with 100 ml of coupling buffer. Beads were then blocked with 1 M ethanolamine (pH 8.0) for 2 h at room temperature. Beads were washed with three alternating cycles of 100 mL acetate buffer (0.1 M acetate buffer, 0.5 M NaCl, pH 4.0) and Tris buffer (0.1 M Tris-HCl, 0.5 M NaCl, pH 8.0). Beads were washed in 100 ml of column buffer (20 mM phosphate buffer, 50 mM NaCl, 0.02% sodium azide, pH 7.4). A slurry of beads was prepared, and a 5 ml column was packed with beads. A column was packed with 50 ml of column buffer at 1 ml/min and an addition of 50 ml at 0.7ml/min. Once beads were prepared, conditioned media were cycled over the column at 1 ml/min overnight. The column was washed with 10 column volumes (CVs) of column buffer. The column was washed with 5 CVs of washing buffer (20 mM phosphate buffer, 0.5 M NaCl, 0.02% sodium azide, pH 7.4). The column was loaded with elution buffer (3 M sodium thiocyanate) and incubated for 1 h. Five CVs of elution buffer were run over the column and collected. The collected material containing eluted ApoE was concentrated with an Amicon Ultra 15 ml concentrator with a 10 kDa molecular weight cutoff (Sigma). The immunopurified ApoE was dialyzed in column buffer with four changes every 6 h.

ApoE ELISA

To measure the concentration of ApoE, we used an in-house ELISA. ELISA plates were coated with 10 μg/ml of anti-ApoE antibody HJ15.7 diluted in carbonate-coating buffer. The coating antibody was incubated overnight at 4°C with shaking. Plates were washed five times with 190 μl/well PBS. Plates were blocked with 2% BSA in PBS for 1 h at room temperature with shaking. Samples and standards (recombinant ApoE4; Leinco) were prepared in ApoE ELISA buffer (1% BSA in PBS with 1X protease inhibitors) and incubated overnight at 4°C with shaking. The ELISA plates were washed five times with 190 μl/well PBS and incubated with 50 ng/ml biotinylated anti-ApoE antibody HJ15.4 diluted in ApoE ELISA buffer for 1h at room temperature with shaking. Plates were then washed five times with 190 μl/well PBS and incubated with streptavidin poly-HRP (Fitzgerald) at 1:12,000 diluted in ApoE ELISA buffer for 1 h at room temperature with shaking. The plates were then washed five times with 190 μl/well PBS before developing with Super Slow 3,3′,5,5′-Tetramethylbenzidine for 30 min at room temperature with shaking. Absorbance was measured using a BioTek plate reader at a wavelength of 640 nm. The standard curve was fitted using a four-parameter fit, and values within the linear range of the curve were used to estimate ApoE concentration.

Nondenaturing gradient gel electrophoresis and SDS-PAGE analysis of ApoE

Nondenaturing gradient gel electrophoresis was performed on samples to estimate the size of the lipidated ApoE lipoprotein. A high molecular weight native ladder (Cytiva) was used to estimate the size of lipidated ApoE lipoprotein. Samples were normalized based on ApoE ELISA to 1 μg per well and diluted in 4X Native NuPAGE sample buffer. The sample (10 μl) was loaded onto a 4%–20% Tris-glycine gel and run at 50 V for 10 min, then 100 V for 16 h at 4°C. For SDS-PAGE, samples were normalized to 1 μg per well and diluted in 4X LDS buffer with 10X reducing agent. The sample (10 μl) was loaded onto a 4%–12% Bis-Tris gel with MES buffer. Gels were run at 50 V for 10 min and then 200V for 30 min. For both gels, Western blots were performed by transferring at 30 V for 1 h at 4°C onto an activated PVDF membrane using the XCell transfer system. Blots were briefly washed with TBS and then blocked in 5% milk in TBS with Tween-20 (TBST) for 1 h at room temperature with shaking. Blots were incubated with 1 μg/ml HJ15.7 diluted in blocking buffer overnight at 4°C. Blots were washed three times for 5 min each in TBST. Blots were incubated in anti-mouse HRP secondary (Invitrogen) at 1:1,000 diluted in blocking buffer for 1 h at room temperature with shaking. Blots were washed three times for 5 min in TBST and then 3X for 5 min in TBS, developed for 3 min with Ultra ECL (Sigma), and then imaged on a Bio-Rad imager.

Negative-stain imaging of ApoE lipoprotein

Immunopurified particles were diluted to a concentration of 200 nM in column buffer (20 mM phosphate buffer, 50 mM NaCl, pH 7.4). Copper carbon film grids were glow discharged for 30 s. The sample (10 μl) was incubated for 1 min and then washed five times in deionized water. Grids were incubated in 10 μl of uranyl formate for 1 min. Grids were blotted and allowed to dry prior to imaging. Grids were imaged using a JEOL TEM microscope at 120 kV and an AMT XR111 high-speed camera at 50,000 magnification (0.212 nm/pixel).

Staining and imaging on a confocal microscope

For images of astrocyte and microglia cell cultures, cells were cultured on poly-L-lysine-coated coverslips. Cells were fixed in 4% paraformaldehyde/4% sucrose for 15 min at room temperature. Cells were stained with the following antibodies: GFAP (1:500 dilution; Invitrogen, catalog no.: 3-0300), Iba1 (1:500 dilution; Abcam, catalog no.: ab178846), and clusterin (Clu) (1:250 dilution; R&D Systems, catalog no.: AF2747) or GFAP (1:1,000 dilution; Invitrogen, catalog no.: 53-9892-82), Iba1 (1:500 dilution; Abcam, catalog no.: ab178846), and complement component 1q (C1q) (1:250 dilution; Abcam, catalog no.: ab11861). For images of frontal cortical sections, brains were dissected from ApoE3, ApoE4, APPPS1-ApoE3, or APPPS1-ApoE4 knock-in mice and fixed overnight in 4% paraformaldehyde. Hemibrains were then saturated in 30% sucrose prior to sectioning. Hemibrains were sectioned at 30 μm on a sliding microtome. Sections were stained with antibodies against GFAP (1:1,000 dilution; Invitrogen, catalog no.: 13-0300), Iba1 (1:1,000 dilution; Abcam, catalog no.: ab178846), Clu (1:500 dilution; R&D Systems, catalog no.: AF2747), and Lpl (1:500 dilution; Abcam, catalog no.: ab93898). The following secondaries were used at 1:500 dilution: anti-rat AlexaFluor 488 (Invitrogen, catalog no.: A21208), anti-rat AlexaFluor 568 (Invitrogen, catalog no.: A78946), anti-goat AlexaFluor 568 (Invitrogen, catalog no.: A11057), anti-rabbit AlexaFluor 647 (Invitrogen, catalog no.: A31573), and anti-mouse AlexaFluor 594 (Invitrogen, catalog no.: A21203). Samples were coverslipped and imaged on a Leica confocal microscope using a 63X oil objective lens.

Proteomics of immunopurified astrocyte and microglia ApoE lipoprotein

Immunopurified particles from astrocyte and microglia-conditioned media were lysed and denatured in SDS lysis buffer containing 2% SDS, 5 mM DTT, 50 mM Hepes (pH 8.5), and protease inhibitors (1×), followed by boiling at 95°C for 10 min. Samples were then run on a short SDS gel and processed for in-gel digestion following an established protocol (31). Briefly, protein bands were excised from Coomassie-stained gels, cut into ∼1 mm^3^ pieces, reduced with 5 mM DTT, alkylated with 10 mM iodoacetamide, and digested overnight at 37°C. The resulting peptides were extracted, dried, and resuspended in 5% formic acid (FA).

For each sample, the digested peptides were separated using a C18 column (10 cm × 75 μm, 1.9 μm particle size; Bruker Daltonics) on a nanoElute 2 liquid chromatography system (Bruker Daltonics) following a previously optimized protocol (32). Peptides were eluted using a 5%–26% buffer B gradient over 60 min (buffer A: 0.1% FA in water; buffer B: 0.1% FA in acetonitrile) at a flow rate of 0.25 μl/min. Proteomics data acquisition was performed on a timsTOF SCP mass spectrometer (Bruker) using the data-independent acquisition parallel accumulation-serial fragmentation method. This platform provides robust quantitative performance, with a linear dynamic range of approximately 4–5 orders of magnitude for protein quantification (33, 34). The limit of quantification typically reaches the low attomole range (1 attomole = (10^−18^ mol) × (6.022 × 10^23^ molecules/mol) = 6.022 × 10^5^ molecules) to several tens or even hundreds of attomoles, depending on peptide ionization properties and sample matrix complexity (35, 36). Full MS data were acquired over an m/z range of 100–1,700 with an ion mobility range of 0.70–1.30 1/K_0_. DIA was performed with an m/z of 300–1,100 m/z range using m/z 20 isolation windows and a 166 ms ramp time. Collision energies were linearly ramped from 20 eV (1/K_0_ = 0.7 vs.·cm^−2^) to 59 eV (1/K_0_ = 1.3 vs.·cm^-2^) as a function of ion mobility.

Raw timsTOF data were processed using data-independent aquisition-neural network (version 1.8) (37), searched against an in silico spectral library generated from mouse (UniProt Proteome ID: UP000000589) and bovine (UniProt Proteome ID: UP000009136, which was added to detect contaminated bovine serum proteins). Search parameters included enzyme specificity set to trypsin/P with up to two missed cleavages; carbamidomethylation of cysteines (+57.02146) as a fixed modification; methionine oxidation (+15.99491) as a variable modification (maximum of two variable modifications per peptide); precursor charge range of 1–4; precursor m/z range of 300–1,800; fragment m/z range of 200–1,800; and a precursor false discovery rate threshold of 1%. Match-between-runs was enabled. To remove residual FBS contamination in the cell media, precursors matching bovine proteins were excluded, and only mouse-specific precursors were retained for quantification. Quantitative analysis was performed following the same approach as previously described (32).

Lipidomics of immunopurified astrocyte and microglia ApoE lipoprotein

ApoE protein (1 μg) was dialyzed into 20 mM phosphate buffer, 50 mM NaCl, pH 7.4, and frozen at −80°C. Lipidomic analysis was performed by using multidimensional MS-based shotgun lipidomics at the Functional Lipidomics Core at the Barshop Institute for Longevity and Aging Studies at the University of Texas Health Center, San Antonio, TX. Briefly, lipid extraction of immunopurified ApoE lipoprotein was prepared using a modified Bligh-Dyer extraction procedure, and a premixed solution of internal standards was added based on the protein concentration of the ApoE lipoprotein as previously described (38). For shotgun lipidomics, lipid extract was further diluted to a final concentration of ∼500 fmol total lipids per μL prior to the mass spectrometric analysis. The analysis was performed on a triple quadrupole mass spectrometer (TSQ Altis; Thermo Fisher Scientific, San Jose, CA) and a Q Exactive mass spectrometer (Thermo Scientific, San Jose, CA), both of which were equipped with an automated nanospray device (TriVersa NanoMate; Advion Bioscience Ltd, Ithaca, NY) as described (39). Identification and quantification of lipid species were performed using an automated software program (40, 41). Individual molecular species of free fatty acids were identified and quantified after derivatization as previously described (42). Total cholesterol and free cholesterol (FC) were quantified after derivatization with methoxyacetic acid as previously described (43). The mass levels of individual cholesteryl ester (CE) species were calculated from the total CE mass, and their composition was profiled using precursor-ion scanning of m/z 369. Data processing (e.g., ion peak selection, baseline correction, data transfer, peak intensity comparison, and quantitation) was performed as described (41). The internal standards used for quantification were phosphatidylcholine (PC) 14:1_14:1 (Avanti) for PC, SM 18:1;O2/12:0 (Avanti) for SM, cholesterol-d7 (Avanti) for cholesterol, and cholesteryl oleate-d7 (Avanti) for CE. Quantification included 24 species of PC, 7 species of SM, and 9 species of CE; cholesterol was analyzed as a single species. The results were normalized to the protein content (nmol lipid/mg protein) (44).

Statistics and visualization

Volcano plots are populated using unadjusted P value with the significance threshold determined by the false discovery rate-adjusted P value (45, 46). Fold change is normalized by taking the log2 fold change. Proteins are plotted based on the -log_10-unadjusted P-value and the log_2 fold change. Proteins with an adjusted P-value less than 0.05 are colored red. Dashed vertical lines indicate a log_2_ fold change of 1. Dashed horizontal lines indicate an adjusted P value less than 0.05. Lipid abundance significance was determined by taking the total nmol per 1 mg ApoE of the lipid species in each class, excluding the values determined for the ApoE^−/−^ samples. The Mann-Whitney U test was performed between groups in order to determine significance (∗<0.05, ∗∗<0.01, and ∗∗∗<0.001). For comparison of ApoE lipoprotein size immunopurified from T091317 and vehicle-treated astrocytes, the diameter of 300 ApoE lipoproteins was manually measured using ImageJ (ImageJ 1.54f, NIH). Comparison of sizes was performed using a Chi-square test between the two groups to determine significance (∗<0.05, ∗∗<0.01, and ∗∗∗<0.001). For heatmaps, the z-score is determined for the row so that each individual value is subtracted from the row mean and then divided by the standard deviation of that row (46). Code troubleshooting and optimization were assisted by Claude 3.7 Sonnet (Anthropic), an artificial intelligence language model. Gene Ontology (GO) term analysis was performed using the DAVID bioinformatics database (47, 48). The Venn diagram was created using Molbiotools. Proteomics was performed on immunopurified ApoE lipoprotein from astrocyte and microglia serum-free conditioned media. Contaminants resulting from previous exposure to serum-containing media were filtered by removing bovine-specific peptides as well as peptides shared between bovine and mouse. To remove other contaminants, only proteins differentially abundant compared to ApoE^−/−^ conditioned media controls were considered as present. Additionally, common contaminants (i.e., keratins and IgG fragments) were manually excluded (Supplemental Table S5). To exclude proteins with minimal abundance, all proteins with an abundance less than the mean plus 0.5 standard deviations were excluded from analysis (Supplemental Table S5). For astrocyte and microglia analyses, criteria were applied using only astrocyte or microglia samples, respectively. For combined analysis, astrocyte and microglia samples were pooled, and exclusion criteria were applied.

Results

Immunopurification of lipidated ApoE from primary astrocytes and microglia

To characterize intact, lipidated ApoE lipoproteins (Fig. 1A), primary astrocyte and microglia cultures were generated from mice expressing the human ApoE isoforms under the endogenous murine Apoe promoter (28). Representative samples were analyzed by nonreducing SDS-PAGE, reducing SDS-PAGE, and nondenaturing PAGE (Fig .1B-D). Nonreducing SDS-PAGE shows the presence of disulfide-linked ApoE in ApoE2 and ApoE3 lipoproteins (Fig. 1B). A faint band is seen at ∼72 kDa that represents an SDS-resistant dimer that is not reducible (Fig. 1C), suggesting that this band does not represent a disulfide-linked dimer in ApoE4-derived samples, in line with previous results (49). No corresponding protein bands were detected in samples derived from ApoE^−/−^ mice. Nondenaturing PAGE reveals that the ApoE lipoprotein remains lipidated following immunopurification in line with previous reports (1, 49). The sizes of ApoE lipoproteins secreted by primary astrocytes and microglia are similar (Fig. 1D). In line with previous reports (49), astrocyte-secreted ApoE2-secreted lipoprotein tends to form the largest lipoproteins, with ApoE4 forming the smallest particles; however, the distribution of the size range of all of the particles substantially overlaps (Fig. 1D). The integrity of the immunopurified ApoE lipoprotein was further confirmed by negative-stain transmission electron microscopy. Micrographs of the immunopurified particles show intact ApoE lipoproteins isolated from astrocytes and microglia (Fig. 1E).Fig. 1A: Schematic showing workflow for purification of intact ApoE lipoprotein particles, followed by proteomic and lipidomic analyses. Primary glia were first isolated from the cortex of P2–P3 pups. Mixed glial cultures were then shaken to isolate primary microglia cultures and primary astrocyte cultures. Cultures were switched to serum-free media prior to collection of media. Astrocyte or microglia-conditioned media were flowed over immunoaffinity columns to isolate ApoE lipoproteins. Intact particles were then assessed using MS-based lipidomics and proteomics. B: Nonreducing SDS-PAGE gel shows that ApoE2 and ApoE3 form disulfide-linked dimers in lipidated ApoE. ApoE4, which does not contain a Cys residue, does not. An SDS-resistant dimer is seen; however, this is resistant to the addition of DTT (Fig. 1C), suggesting that it is not a disulfide-linked dimer. The absence of signal in ApoE^−/−^ samples shows that no ApoE is present in samples collected from ApoE^−/−^ astrocyte or microglia-conditioned media. ApoE protein (1 μg) was loaded per well. Molecular weight markers on the left are in kilodaltons. C: Reducing SDS-PAGE shows the purification of ApoE from astrocyte and microglia-conditioned media. An SDS-resistant ApoE dimer is seen at ∼72 kDa. ApoE protein (1 μg) was loaded per well. Molecular weight markers on the left are in kilodaltons. D: Nondenaturing PAGE shows that ApoE lipoproteins are present as intact lipidated lipoproteins following immunopurification. The size distribution of ApoE2 lipoproteins is shifted to larger particle sizes compared to ApoE4 lipoproteins when secreted by both astrocytes and microglia. ApoE3 lipoproteins are of an intermediate size. Markers on the left show the radius in nanometers of the markers. E: Negative-stain transmission electron micrographs of immunopurified ApoE lipoproteins stained using uranyl formate show that ApoE lipoproteins were immunopurified as intact discoidal lipoproteins from both astrocytes and microglia.

Lipidomics of immunopurified ApoE lipoproteins reveal differential lipid abundance between astrocyte- and microglia-derived ApoE lipoproteins

Lipidomics was performed on immunopurified ApoE lipoproteins that were normalized by ApoE protein concentration. Comparing astrocyte- versus microglia-derived ApoE lipoproteins revealed enrichment of CE species in microglia and enrichment of SM species in astrocytes (Supplemental Table S1, Fig. 2A). Comparison of lipid species by ApoE isoform showed no significant differences (Supplemental Table S2). Principal component analysis (PCA) of the lipidomics samples showed partial separation of the samples by cell type but not by ApoE isoform (Fig. 2B). This suggests that the lipid composition of the ApoE lipoprotein is predominantly dependent on the cell from which it is secreted. Assessing the distribution of lipid classes by cell type supports the analysis of individual lipid species explored in the volcano plot. FC was significantly decreased in microglia-derived particles (P < 0.05, Supplemental Table S3) compared to astrocytes (Fig. 2C). Conversely, CEs were significantly increased (P < 0.001, Supplemental Table S3) in microglia-derived particles compared to astrocytes (Fig. 2C). The CE/FC ratio in astrocyte ApoE lipoprotein (0.125) is significantly less than the CE/FC ratio in microglia ApoE lipoprotein (0.820), in agreement with the relative increase of CEs in microglia ApoE lipoprotein and concordant decrease in FC in microglia ApoE lipoproteins. Assessing the changes in the lipidome profile by ApoE isoform revealed minimal differences. ApoE4 lipoprotein appeared to have slightly higher levels of total cholesterol and FC than ApoE2 and ApoE3; however, these changes are not statistically significant (Fig. 2D, E, Supplemental Table S4). Comparing the abundances of individual lipid species across ApoE isoforms revealed that the lipidome of microglia-secreted ApoE lipoprotein is largely independent of ApoE isoform (Supplemental Fig. S1A–C). Comparing the lipid abundance by ApoE isoform in astrocytes reveals that ApoE2 lipoproteins have reduced CE species compared to ApoE4 (Supplemental Fig. S1D–F).Fig. 2A: Volcano plot showing enrichment of lipid species of astrocyte immunopurified lipoprotein versus microglia immunopurified lipoprotein. Microglia ApoE lipoproteins are enriched in several CE species. Astrocyte ApoE lipoproteins are enriched in several SM species and FC. For lipidomics analysis, the following groups were analyzed: ApoE2 astrocytes (n = 6), ApoE3 astrocytes (n = 6), ApoE4 astrocytes (n = 6), ApoE2 microglia (n = 4), ApoE3 microglia (n = 4), and ApoE4 microglia (n = 4). B: PCA of the lipidome of ApoE lipoproteins shows partial separation of samples based on the cellular source of the ApoE lipoprotein but not by ApoE isoform. Circles represent the 95% confidence interval based on the cell type. C: Distribution of lipid classes by cell type shows the abundance of different lipid classes. The total amount of each lipid class is expressed as the nmol amount of protein per 1 mg of ApoE protein. The total amount of FC is significantly decreased in microglia ApoE lipoprotein (P < 0.05). This is explained by the increase in CEs in microglia ApoE lipoprotein (P < 0.001). D: Distribution of lipid classes by ApoE isoform shows the abundance of different lipid classes. No significant differences between lipid classes were observed when pooled based on ApoE isoform. This provides additional evidence that the primary determinant of the ApoE lipoprotein lipidome is the cellular source. E: Heatmap of measured lipid species in immunopurified astrocyte and microglia ApoE lipoproteins. The lipid composition of ApoE lipoproteins is different based on the cellular source of the ApoE lipoproteins, with microglial ApoE lipoproteins being enriched in CE species and astrocyte ApoE lipoproteins being enriched in SM lipid species.

Proteomics of astrocyte-secreted lipoproteins show minimal differences based on the ApoE isoform

Before filtering proteins, 1,302 proteins were detected in astrocyte- and microglia-secreted ApoE lipoproteins. In order to assess similarity to previous results from human studies of lipoproteins in cerebrospinal fluid (CSF) and plasma, we compared our findings to those in the brain lipoprotein and HDL proteome watch (50, 51). We compared proteins enriched in astrocyte and microglia ApoE lipoprotein to proteins identified in both datasets (Supplemental Fig. S2A). In addition to the brain lipoprotein and HDL proteome watch databases, which were enriched for lipoproteins, we also compared our results to proteins that were enriched in AD patient CSF (52). These results showed that 50 proteins in the astrocyte ApoE lipoprotein proteome overlapped with the brain lipoprotein proteome watch, 112 with the HDL proteome watch, and 6 with the AD-associated CSF proteome out of 429 total proteins (Supplemental Fig. S2B). Similarly, 43 proteins in the microglia ApoE lipoprotein proteome overlapped with the brain lipoprotein proteome watch, 101 with the HDL proteome watch, and 12 with the AD-associated CSF proteome out of 495 total proteins (Supplemental Fig. S2B). The astrocyte and microglia ApoE lipoprotein proteome were most similar, followed by the HDL proteome watch, then the brain lipoprotein proteome watch, and lastly the AD-associated CSF proteome (Supplemental Fig. S2C). The similarity of the proteome from astrocyte and microglia-immunopurified ApoE lipoproteins with lipoproteins from human plasma and CSF suggests that despite differences due to cell culture conditions and species differences, these results are translatable to humans. The isolation of ApoE lipoproteins from astrocytes and microglia allows for the detection of low-abundance proteins associated with ApoE that would not be detected even with enrichment for lipoproteins in CSF and plasma. After the application of strict filtering criteria, 15 and 19 proteins were identified with high confidence as being present in astrocyte and microglial-secreted ApoE lipoproteins, respectively (Supplemental Table S5). Astrocyte-secreted ApoE lipoproteins were enriched in several proteins important in lipid metabolism, including Clu, apolipoprotein C1 (ApoC1), and phospholipid transfer protein (Pltp) (Fig. 3A). PLTP, which is enriched in the astrocyte ApoE lipoprotein, is also found in the brain lipoprotein proteome watch, HDL proteome watch, and AD-associated CSF proteome (Supplemental Table S12). Clu has been previously implicated as a risk factor for AD (53, 54). Clu, which is enriched in the astrocyte ApoE lipoprotein, is also found in the brain lipoprotein proteome watch and HDL proteome watch (Supplemental Table S12). Clu also cooperatively suppresses Aβ plaque deposition in mouse models of Aβ amyloidosis (55). ApoC1 is known to be expressed by astrocytes and microglia and to modulate the immune response (56). ApoC1, which is enriched in the astrocyte ApoE lipoprotein, is also found in the microglia ApoE lipoprotein proteome and HDL proteome watch (Supplemental Table S12). Pltp is known to be present in ApoE lipoprotein, and Pltp activity is reduced in AD patients compared to controls (57). Analysis of proteins enriched in astrocyte-secreted ApoE lipoprotein shows minimal differences based on ApoE isoform (Fig. 3B). Muscular lamin A/C interacting protein (Mlip) appears to be consistently enriched in astrocyte-secreted ApoE2 lipoprotein; however, the function of Mlip is unclear (Fig. 3C, D, Supplemental Table S6). Mutations in Mlip have been associated with myopathy, with reports suggesting that it could play a role in myoblast differentiation (58). The proteome of ApoE4 and ApoE3 astrocyte lipoproteins showed no differences in protein abundance based on significance and threshold criteria (Fig. 3E). The proteome of ApoE4 astrocyte lipoproteins is most distinct compared to ApoE2 and ApoE3 astrocyte lipoproteins (Fig. 3F); however, the PCA suggests substantial overlap in the proteome of all astrocyte ApoE lipoproteins (Fig. 3B).Fig. 3A: Histogram of the most abundant proteins measured in astrocyte ApoE lipoproteins. Abundance was determined by excluding common contaminants, showing differential expression compared to proteins measured in ApoE^−/−^ astrocyte-conditioned media, and the protein abundance being greater than the total mean + 0.5 standard deviation threshold. Error bars represent mean ± SEM. For proteomics analysis, the following groups were analyzed: ApoE2 astrocytes (n = 5), ApoE3 astrocytes (n = 5), and ApoE4 astrocytes (n = 5). B: PCA of the proteome of astrocyte ApoE lipoproteins. Circles are colored based on the ApoE isoform and represent the 95% confidence interval. Astrocyte samples largely overlap in their proteome, suggesting that the ApoE isoform is not a major determinant of the proteome of astrocyte-secreted ApoE lipoproteins. C: Volcano plot showing differential protein abundance of proteins bound to astrocyte ApoE2 immunopurified lipoproteins versus astrocyte ApoE3 immunopurified lipoproteins. Lysozyme 2 (Lyz2) was increased in ApoE3 astrocyte lipoproteins, and Mlip was increased in ApoE2 astrocyte lipoproteins. D: Volcano plot showing differential protein abundance of proteins bound to astrocyte ApoE2 immunopurified lipoprotein versus astrocyte ApoE4 immunopurified lipoprotein. Transmembrane glycoprotein NMB (Gpnmb) was found to be increased in ApoE4 astrocyte lipoproteins, and Mlip was increased in ApoE2 astrocyte lipoproteins. E: Volcano plot showing differential protein abundance of proteins bound to astrocyte ApoE3 immunopurified lipoprotein versus astrocyte ApoE4 immunopurified lipoprotein. No proteins were differentially abundant between ApoE3 and ApoE4 astrocyte lipoproteins. F: Heatmap of the most abundant proteins bound to ApoE lipoproteins isolated from astrocyte-conditioned media.

Proteomics of microglia-secreted ApoE lipoproteins show clear differences based on the ApoE isoform

In contrast to astrocytes, the protein composition of microglia-secreted ApoE lipoproteins showed differences based on the ApoE isoform. The most abundant proteins in microglia-secreted ApoE lipoproteins include proteins involved in lipid metabolism, such as ApoC1 and Lpl, and proteins involved in the complement pathway, namely the subunits of C1q, C1qa, and C1qb (Fig. 4A). C1qb, which is enriched in the microglia ApoE lipoprotein, is also found in the brain lipoprotein proteome watch and HDL proteome watch (Supplemental Table S12). Lpl, which is enriched in the microglia ApoE lipoprotein, is also found in the HDL proteome watch and AD-associated CSF proteome (Supplemental Table S12). Milk fat globule-EGF factor 8 protein (Mfge8) is one of the most abundant proteins in microglial ApoE lipoprotein (Fig .4A). This protein has been previously reported to be secreted by microglia and to promote the phagocytosis of apoptotic neuronal cells by the microglia-like BV2 cell line (59). PCA of microglia-secreted lipoproteins shows clear separation of the samples based on ApoE isoform (Fig. 4B). ApoE lipoproteins secreted from ApoE2 microglia consistently have lower levels of Lpl compared to ApoE lipoproteins secreted from ApoE3 and ApoE4 microglia (Fig. 4C, D, Supplemental Table S7). C1qa and C1qb are enriched in ApoE4 microglia-secreted lipoprotein, suggesting the possibility that ApoE secreted from ApoE4 microglia may be more proinflammatory than ApoE2 and ApoE3 microglia (Fig. 4D, E, Supplemental Table S7). ApoE4 lipoprotein is also enriched for Lpl and AarF domain-containing kinase 2 (Fig. 4D, E, Supplemental Table S7). ApoC1 is also lower in ApoE4 lipoprotein compared to ApoE2 and ApoE3 lipoproteins (Fig. 4D, E, Supplemental Table S7). The differential abundance of these proteins is shown in the heatmap with clear differences in the proteome of microglia ApoE lipoproteins depending on the ApoE isoform (Fig. 4F).Fig. 4A: Histogram of the most abundant proteins measured in microglia ApoE lipoproteins. Abundance was determined by excluding common contaminants, showing differential expression compared to proteins measured in ApoE^−/−^ microglia-conditioned media, and the protein abundance being greater than the total mean + 0.5 standard deviation threshold. Error bars represent mean ± SEM. For proteomics analysis, the following groups were analyzed: ApoE2 microglia (n = 3), ApoE3 microglia (n = 3), and ApoE4 microglia (n = 4). B: PCA of the proteome of microglia ApoE lipoproteins. Circles are colored based on the ApoE isoform and represent the 95% confidence interval. Microglia samples show clear differences in the proteome of microglia-secreted ApoE lipoproteins based on the ApoE isoform. This finding is in contrast to the proteome of astrocyte ApoE lipoproteins, suggesting that the ApoE isoform plays a greater role in the proteome of microglia ApoE lipoproteins. C: Volcano plot showing differential protein abundance of proteins bound to microglial ApoE2 versus ApoE3 immunopurified lipoprotein. Complement C1q B chain (C1qb) and Lpl are increased in ApoE3 versus ApoE2 microglial lipoprotein. D: Volcano plot showing differential protein abundance of proteins bound to microglial ApoE2 versus ApoE4 immunopurified lipoprotein. Multiple proteins are differentially abundant between ApoE2 microglia and ApoE4 microglia lipoproteins, showing a strong contribution of the ApoE isoform to the proteome of microglia-secreted ApoE lipoprotein. ApoC1, aldolase fructose-bisphosphate B (Aldob), and Ankk1 are increased in ApoE2 microglial lipoproteins. AarF domain-containing kinase 2 (Adck2), Lpl, fibronectin 1 (Fn1), C1qb, C1qa, pyruvate kinase M1/2 (Pkm), and lysozyme 2 (Lyz2) are increased in ApoE4 microglial lipoproteins. E: Volcano plot showing differential protein abundance of proteins bound to microglial ApoE3 versus ApoE4 immunopurified lipoprotein. Similar to ApoE2 microglia lipoproteins, Ankk1 and ApoC1 are increased in ApoE3 versus ApoE4 microglia lipoproteins. Adck2, Pkm, C1qa, and Lpl are increased in ApoE4 versus ApoE3 microglia lipoproteins, showing a similar pattern compared to ApoE2 microglia lipoproteins. F: Heatmap of the most abundant proteins bound to ApoE lipoproteins isolated from microglia-conditioned media. Differences in protein abundance are seen based on the ApoE isoform.

Proteomics reveals differential protein abundance between ApoE lipoproteins immunopurified from astrocytes and microglia

Comparing astrocyte and microglia immunopurified ApoE lipoproteins, several proteins are found to be differentially abundant depending on the cellular source of ApoE (Fig. 5B, D). Mfge8 is highly enriched in microglia ApoE lipoproteins compared to astrocyte ApoE lipoproteins (Fig. 5B, Supplemental Table S8). The complement proteins, C1qa, C1qb, and complement factor I, are also enriched in microglia (Fig. 5B, Supplemental Table S8). Apolipoprotein D is increased in astrocyte-secreted ApoE lipoproteins (Fig. 5B, Supplemental Table S8). Apolipoprotein D is an apolipoprotein primarily produced in the brain and is associated with mediating the interaction between HDL and LDL in the plasma (60, 61). Pla2g7 was increased in astrocyte ApoE lipoproteins (Fig. 5B, Supplemental Table S8) and is associated with AD (62). Pltp was also enriched in astrocyte ApoE lipoproteins (Fig. 5B, Supplemental Table S8) and has been previously reported to induce secretion of ApoE from astrocytes (57). Assessing the proteins enriched in astrocyte ApoE lipoproteins (Fig. 5B), GO term analysis of the biological processes showed an enrichment of proteins involved in glucose metabolic processes and acute phase responses (Supplemental Table S9). GO term analysis of the cellular components of differentially abundant astrocyte proteins showed an enrichment in HDL particles and extracellular space (Supplemental Table S9). GO term analysis of the proteins enriched in microglia ApoE lipoproteins showed an enrichment of proteins involved in synapse pruning, complement activation, proteolysis, and the innate immune response (Supplemental Table S9). This is driven by the presence of C1qa, C1qb, and complement factor I in microglial ApoE lipoproteins in contrast to astrocyte-secreted ApoE lipoproteins. GO term analysis of the cellular compartment revealed proteins associated with the extracellular space, complement complex, and LDL (ApoC1 and Lpl) (Supplemental Table S9). PCA of the proteome identified in microglia and astrocyte ApoE lipoproteins showed clear separation (Fig. 5C). These results are in line with the lipidomics data and suggest that the cellular source of ApoE lipoprotein is the primary determinant in lipoprotein composition compared to ApoE isoform. A heatmap showing the abundance of proteins identified in microglia and astrocyte ApoE lipoprotein also shows a clear difference in the protein composition of ApoE lipoprotein secreted from microglia and astrocytes (Fig. 5D). A comparison of astrocyte and microglia proteins based on the ApoE isoform shows the same trends as comparing all astrocyte and microglia samples together irrespective of the ApoE isoform (Supplemental Fig. S5).Fig. 5A: Volcano plot showing differential protein abundance of proteins bound to astrocyte versus microglia immunopurified lipoprotein. Proteins involved in glucose metabolism (Pkm, HECT domain E3 ubiquitin protein ligase 4 [Hectd4], and ApoD) and acute phase response (fibronectin 1 [Fn1] and Serpina3n) are differentially abundant in astrocyte ApoE lipoproteins. Proteins involved in complement activation (C1qa, C1qb, and complement factor I [CfI]), synapse pruning (C1qa and C1qb), proteolysis (matrix metallopeptidase 12 [Mmp12], serine carboxypeptidase 2 [Cpb2], CfI, and cathepsin B [Ctsb]), and the innate immune response (C1qa, C1qb, and CfI) are differentially abundant in microglia ApoE lipoprotein. For proteomics analysis, the following groups were analyzed: ApoE2 astrocytes (n = 5), ApoE3 astrocytes (n = 5), ApoE4 astrocytes (n = 5), ApoE2 microglia (n = 3), ApoE3 microglia (n = 3), and ApoE4 microglia (n = 4). B: PCA of the proteome of astrocyte and microglia ApoE lipoproteins. Circles are colored based on cell type and represent the 95% confidence interval. There is a clear separation of samples by cellular source. The proteome of ApoE4 microglial lipoproteins is also clearly separated from other samples. The cellular source of the particles is the primary contributor to the differential proteome of the ApoE lipoproteins. C: Heatmap of the most abundant proteins bound to ApoE lipoproteins isolated from astrocyte and microglia-conditioned media. Differences in protein abundance are observed comparing astrocyte-secreted ApoE lipoproteins and microglia-secreted ApoE lipoproteins.

Treatment of ApoE4 astrocytes with an LXR agonist leads to a significant reduction in CEs in secreted ApoE lipoproteins

Treatment of astrocytes with LXR agonists has been previously reported to upregulate the expression of ABCA1 and ABCG1, promoting the expression and secretion of ApoE from astrocytes (63, 64). Previous research has shown that the overexpression of ABCA1 ameliorated Aβ pathology in a mouse model of Aβ amyloidosis and that treatment with an LXR agonist was protective in a mouse model of tau-mediated neurodegeneration (65, 66). In order to better understand how treatment of cells with an LXR agonist could alter ApoE lipoprotein composition, we treated ApoE4 astrocyte cultures with T0901317. ApoE lipoprotein was isolated as described previously from cell cultures treated with T0901317 or vehicle control. Isolated lipoprotein was analyzed by proteomics and lipidomics as described above. Since ApoE^−/−^ astrocyte cells were not treated in this experiment, we did not filter proteins of treated samples based on there being a significant enrichment compared to ApoE^−/−^ control media. Lipidomics of immunopurified ApoE lipoproteins revealed a significant decrease of several CE species in ApoE lipoprotein isolated from T0901317-treated astrocytes (Fig. 6A, G). PCA of the lipidome showed partial separation of the two classes (Fig. 6B). Proteomics analysis of ApoE lipoproteins revealed no significantly different proteins associated with ApoE lipoprotein, with the exception of ATPase H+ transporting accessory protein 2 being significantly enriched in ApoE lipoprotein from vehicle-treated astrocytes (Fig. 6C, H). Negative-stain electron microscopic analysis of ApoE lipoproteins isolated from T0901317- and vehicle-treated astrocytes reveals that ApoE particles isolated from T0901317-treated astrocytes were significantly larger than those from vehicle-treated astrocytes (Fig. 6E, F).Fig. 6A: Volcano plot showing decreased CE species in ApoE lipoproteins immunopurified from T0901317-treated astrocyte-conditioned media compared to ApoE lipoproteins immunopurified from vehicle-treated astrocyte-conditioned media. For lipidomics analysis, the following groups were analyzed: ApoE lipoprotein immunopurified from T0901317-treated ApoE4 astrocyte-conditioned media (n = 3) and ApoE lipoprotein immunopurified from vehicle-treated ApoE4 astrocyte-conditioned media (n = 3). B: PCA of the lipidome of ApoE lipoproteins immunopurified from T0901317-treated astrocyte-conditioned media compared to ApoE lipoproteins immunopurified from vehicle-treated astrocyte-conditioned media shows partial separation of samples based on the treatment. Circles represent the 95% confidence interval based on the cell type. C: Volcano plot showing differential protein abundance of proteins bound to ApoE lipoprotein immunopurified from T0901317-treated astrocyte-conditioned media compared to ApoE lipoproteins immunopurified from vehicle-treated astrocyte-conditioned media. D: PCA of the proteome of ApoE lipoproteins immunopurified from T0901317-treated astrocyte-conditioned media compared to ApoE lipoproteins immunopurified from vehicle-treated astrocyte-conditioned media. Circles are colored based on cell type and represent the 95% confidence interval. E: Negative stain of ApoE lipoproteins immunopurified from T0901317-treated astrocyte-conditioned media compared to ApoE lipoproteins immunopurified from vehicle-treated astrocyte-conditioned media. F: Histogram of the diameter of ApoE lipoproteins immunopurified from T0901317-treated astrocyte-conditioned media compared to ApoE lipoproteins immunopurified from vehicle-treated astrocyte-conditioned media. Three hundred ApoE lipoproteins were measured for each group. Statistical comparison between groups was performed by Chi-square test. G: Heatmap of measured lipid species in ApoE lipoproteins immunopurified from T0901317-treated astrocyte-conditioned media compared to ApoE lipoproteins immunopurified from vehicle-treated astrocyte-conditioned media. H: Heatmap of the most abundant proteins bound to ApoE lipoproteins immunopurified from T0901317-treated astrocyte-conditioned media compared to ApoE lipoproteins immunopurified from vehicle-treated astrocyte-conditioned media.

Proteins identified in secreted ApoE lipoproteins are present in astrocytes and microglia in cell culture and in vivo

In order to validate our findings, we assessed whether the proteins we identified as being highly enriched in astrocytes or microglia were enriched in the respective cell types. For microglia, we assessed for the presence of C1q, which was highly enriched in ApoE lipoprotein immunopurified from microglia-conditioned media. For astrocytes, we assessed for the presence of Clu (ApoJ), which was found to be enriched in ApoE lipoproteins from astrocyte-conditioned media and has been previously reported to be secreted by astrocytes (1). First, we assessed astrocytes and microglia in cell culture. C1q was found to only be present in microglia in cell culture and not in astrocytes (Fig. 7A). In contrast, Clu (ApoJ) was found to only be present in astrocytes in cell culture and not in microglia (Fig. 7B). Staining for GFAP, a marker for astrocytes, and Iba1, a marker for microglia, also demonstrates the purity of the cell cultures (Fig. 7A, B). To determine whether this relationship held in a physiological setting, we stained sections of the frontal cortex from mice with knock-in of the human APOE3 or APOE4 allele (28). Lpl, which was highly enriched in ApoE lipoproteins from microglia, was also found to colocalize with microglia in both ApoE3 and ApoE4 mouse frontal cortex (Fig. 8A). This was also found to be the case in the context of a mouse model of Aβ amyloidosis, where Lpl colocalized in microglia in both APPPS1-APOE3 and APPPS1-APOE4 knock-in mice (Supplemental Fig. S6A). Clu (ApoJ) was found to be clearly colocalized with astrocytes in both ApoE3 and ApoE4 mouse frontal cortex (Fig. 8B). Clu (ApoJ) also colocalized with astrocytes in APPPS1-APOE3 and APPPS1-APOE4 knock-in mice (Supplemental Fig. S6B). Together, these results reinforce the validity of our previous results and the cell specificity of the proteins secreted on ApoE lipoproteins from these various cell types.Fig. 7A: Purity of astrocyte cultures is demonstrated by positive staining of GFAP, an astrocyte marker, and negative staining for Iba1, a microglial marker. Purity of microglial cultures is demonstrated by negative staining of GFAP, an astrocyte marker, in astrocyte culture and positive staining for Iba1, a microglial marker. Representative images of C1q staining in astrocyte and microglia cultures. C1q is present in microglia and not in astrocytes. Arrows demonstrate colocalization of C1q with microglia as demonstrated by colocalization of C1q with Iba1 and not GFAP. B: Purity of astrocyte cultures is demonstrated by positive staining of GFAP, an astrocyte marker, and negative staining for Iba1, a microglial marker. Purity of microglial cultures is demonstrated by negative staining of GFAP, an astrocyte marker, in astrocyte culture and positive staining for Iba1, a microglial marker. Representative images of Clu (ApoJ) staining in astrocyte and microglia cultures. Clu (ApoJ) is present in astrocytes and not in microglia. Arrows demonstrate colocalization of Clu with astrocytes as demonstrated by colocalization of Clu with GFAP and not Iba1.Fig. 8A: Representative images of Lpl staining in ApoE3 and ApoE4 knock-in mice frontal cortical sections. Astrocytes are stained using the astrocyte marker GFAP. Microglia are stained using the microglial marker, Iba1. Lpl is present in microglia of both ApoE3 and ApoE4 knock-in mice. Arrows point to the colocalization of Lpl with Iba1 in both ApoE3 and ApoE4 knock-in mice frontal cortical sections. B: Representative images of Clu (ApoJ) staining in ApoE3 and ApoE4 knock-in mice frontal cortical sections. Astrocytes are stained using the astrocyte marker GFAP. Microglia are stained using the microglial marker Iba1. Clu (ApoJ) is present in astrocytes of both ApoE3 and ApoE4 mice. Arrows point to colocalization of Clu with GFAP in both ApoE3 and ApoE4 knock-in mice frontal cortical sections.

Discussion

Although ApoE is a critical driver in the disease pathogenesis of AD, limited studies have described the lipid and protein composition of ApoE-containing lipoproteins. In the CNS, astrocytes and microglia are the two primary cell types that secrete ApoE. Astrocytes account for the bulk of ApoE secretion under homeostatic conditions (14). In the context of neurodegeneration, microglia strongly upregulate ApoE expression (4, 11). In this study, we found that the lipidome and proteome of ApoE lipoprotein are altered based on the cellular source and ApoE isoform, with the cellular source of ApoE being the primary determinant. Previous research has shown that astrocytes under inflammatory conditions secrete ApoE lipoproteins containing long-chain fatty acids that are neurotoxic in culture (20). ApoE isoform has also been reported to alter cell metabolism in human iPSC-derived astrocytes (16). The proteome of ApoE lipoproteins secreted from microglia showed differences based on ApoE isoform. ApoE lipoproteins from APOE4 microglia were enriched in C1q compared to ApoE lipoproteins from APOE2 and APOE3 microglia, suggesting that ApoE lipoproteins from APOE4 microglia may be more inflammatory. However, the mechanism by which the ApoE isoform affects the association of certain proteins with the lipoproteins is unclear. Previous research has hypothesized that differences in the secretome of APOE4 microglia may be due to greater flux of ApoE4 through the endolysosomal pathway and accumulation of lipid droplets in APOE4/4 microglia (67). The ApoE4 isoform has also been reported to drive disruption of the tricarboxylic acid cycle and lipid metabolism in microglia (68).

An important finding of this study is the identification of additional proteins on nascent ApoE lipoproteins secreted from primary astrocytes and microglia. In the CNS, ApoE is initially secreted as a discoidal particle that is similar in size to HDL in plasma (1, 2, 49). In plasma, ApoA1 is the main protein component of HDL and is known to associate with a wide range of proteins that alter its function and metabolism (51, 69). A particularly interesting example is trypanosome lytic factor (70, 71, 72), which is a complex consisting of ApoA1, haptoglobin-related protein (Hrp), and apolipoprotein L1 (ApoL1) (72). ApoA1 acts as a scaffold protein, organizing this lipoprotein complex to carry Hrp and ApoL1. Hrp binds to receptors on trypanosomes, causing the internalization of the complex, whereafter ApoL1 can form a pore in the plasma membrane of the trypanosome, causing cell lysis. Based on the current study, it is intriguing to hypothesize that ApoE could act as a scaffold protein to organize protein complexes on individual particles for carrying out distinct functions throughout the body. Although ApoE is the primary constituent of ApoE lipoproteins secreted by astrocytes and microglia, the proteins found with ApoE lipoproteins may regulate lipoprotein trafficking and function.

Compared to astrocytes, microglia lipoproteins contained higher levels of ApoC1, which inhibits binding of ApoE to LRP1 and LDLR (73). Additionally, Ltf is highly enriched in ApoE lipoproteins secreted by microglia. Ltf is known to have an antimicrobial function, suggesting that ApoE secreted by microglia could be induced by microbial infection (74). Mfge8, which is enriched in microglial lipoproteins, could also be involved in promoting antimicrobial activity and clearance of cell debris due to the role of Mfge8 in promoting phagocytosis (75). Mfge8 is one of the most enriched proteins in microglia ApoE lipoproteins compared to astrocyte lipoproteins (Fig. 5B) and is thought to function as a cell adhesion protein connecting smooth muscle to elastic fibers (75, 76). Mfge8 has also been found to coaggregate with vascular Aβ plaques, cerebral amyloid angiopathy, in AD (77). Mfge8 can bind phosphatidylserine and has been reported to be secreted as an “eat-me” signaling molecule (78). It has also been identified as a biomarker for stress cardiomyopathy along with ApoE (79). Ankyrin repeat and kinase domain-containing 1 (Ankk1) is a Ser/Thr protein kinase most closely associated with dopamine receptor D2 (80). Mutations in Ankk1 have been related to reductions in autoantibodies against Aβ in AD, suggesting a possible link between Ankk1 and ApoE in the pathogenesis of AD (81). ApoE lipoproteins secreted by microglia may be more proinflammatory than astrocyte ApoE, as the former are enriched in C1q and matrix metallopeptidase 12. ApoE has been previously reported to form a complex with C1qa and C1qb (82, 83), and both proteins were enriched in microglial ApoE lipoproteins. Previous reports show a correlation between CSF levels of C1q and AD (84). Furthermore, C1q and ApoE have been shown to compete for binding to Trem2, potentially suggesting that C1q bound to ApoE may prevent interaction with Trem2 (85). The enrichment of Lpl with microglia ApoE lipoprotein could suggest differences in lipolysis of lipids contained in microglia-secreted ApoE compared to lipids in astrocyte-secreted ApoE. Lpl may also alter the binding of ApoE lipoproteins to their receptors, as Lpl has been previously reported to mediate the binding of lipoproteins to LRP1 (86, 87, 88).

One of the most enriched proteins in astrocyte ApoE is PLTP, which has been reported to have reduced activity in the context of AD and induces the secretion of ApoE from astrocytes (57, 89). Reduced PLTP activity has also been reported in the CSF of multiple sclerosis patients (57). This suggests that under homeostatic conditions, the association of PLTP and ApoE in astrocyte ApoE lipoprotein plays an important role in brain lipid metabolism. Pla2g7 (Lp-PLA_2_) is a phospholipase A2 enzyme that degrades oxidized phospholipids and is thought to promote the antiatherogenic effects of HDL. Mutations in Lp-PLA_2_ have been associated with increased risk for AD (62) and cardiovascular disease (90, 91). Serpina3n has been reported to inhibit granzyme B activity, suggesting that secretion of Serpina3n on astrocyte ApoE lipoprotein may prevent extracellular matrix degradation (92). This contrasts with microglia-secreted ApoE lipoprotein, which is enriched in matrix metallopeptidase 12. LCAT was detected in astrocyte ApoE lipoprotein samples but was not significantly enriched in astrocyte ApoE lipoprotein compared to ApoE^−/−^ astrocyte-conditioned media. The secretion of LCAT into the media by astrocytes is consistent with previous reports reporting that LCAT is produced by astrocytes (93). This result may reflect the lack of substrates for LCAT due to the cells being cultured in serum-free conditions. Alternatively, the elution methods used to dissociate ApoE from the immunoaffinity column may have also caused dissociation of LCAT from the ApoE lipoprotein. ApoE has also been reported to have significantly less affinity for LCAT than ApoA1, suggesting that it may be easily dissociated from ApoE lipoprotein (94). Despite the presence of LCAT in astrocyte media, astrocyte-secreted ApoE lipoproteins contained less CEs than microglia-secreted ApoE lipoprotein. The biological mechanism underlying the increased ratio of CEs in microglia-secreted ApoE lipoprotein is unknown and is an important area for future work. Future work investigating ApoE isolated from animals only expressing ApoE from specific cell types would yield more physiological insights into the protein composition of ApoE lipoprotein in an in vivo setting.

The lipid composition of ApoE lipoproteins was also found to be altered in a manner dependent on the cellular source. The enrichment of FC and SM in astrocyte ApoE lipoproteins suggests reduced membrane fluidity, which could impact the conformation of ApoE and other proteins on the ApoE lipoprotein. For example, the lipolysis of lipids in VLDL-containing ApoE4 results in increased particle fluidity and an expansion of ApoE4 conformation (95). The lipid composition of ApoE lipoproteins may also be important in driving disease pathogenesis. ApoE2, due to its inability to bind LDLR, has been reported to be neuroprotective as it leads to less uptake of CEs, which can be neurotoxic in excess, by neurons (96). The addition of polyunsaturated fatty acids to mouse brains resulted in lipofuscinosis, showing that the lipid content of ApoE lipoprotein can be neurotoxic (96). The enrichment of CE species that we reported in microglial ApoE lipoprotein suggests that microglial ApoE lipoprotein may have the potential to be more neurotoxic than astrocyte-secreted ApoE lipoproteins, even if the effects of CEs are not directly neurotoxic but are associated indirectly with toxicity. Our finding that treatment of astrocytes with T0901317 leads to a reduction in the CE content of ApoE lipoproteins may also suggest a mechanism for the protective effect of LXR agonists or ABCA1 overexpression in mouse models of Aβ amyloidosis and tau-mediated neurodegeneration. Microglia may also play a direct role in the fibrillization of Aβ through the uptake of Aβ bound to ApoE, where both molecules aggregate in the endolysosomal pathway (12). Future work will focus on better understanding the implications of these findings, namely, understanding how the proteins and lipids cosecreted with ApoE lipoprotein alter the function of ApoE. Although this discussion has focused on the implications for AD, many of the described proteins are also relevant for cardiovascular disease, and our work provides the foundation for mechanistic inquiries into the specific functions of ApoE lipoproteins secreted by different cell types.

There are some limitations to the current work that will need to be addressed by future studies. The elution of ApoE lipoprotein using 3 M sodium thiocyanate was chosen as it has been shown not to alter particle size and preserves ApoE in its native, lipidated form (97). However, this may result in disruption of proteins loosely bound to ApoE lipoprotein. Additionally, to purify a sufficient quantity of secreted ApoE lipoprotein, astrocytes and microglia cells were isolated and cultured in vitro. In order to promote cell culture expansion and health, serum was added to the culture media in the initial stages prior to the collection of ApoE lipoproteins from the cell-conditioned media. Although we took steps to mitigate the influence of serum on ApoE lipoprotein composition by switching cultures to serum-free media and excluding bovine-specific peptides, culturing cells in serum conditions likely altered the composition of ApoE lipoprotein secreted from microglia and astrocytes. Considering the differences of mice and men, species differences also are an important consideration in interpreting these results. CSF lipoproteins isolated from humans and mice differ partly due to the presence of cholesterol ester transfer protein in humans, which is not present in mice. This results in spherical ApoE lipoprotein isolated from human CSF, whereas mouse ApoE lipoproteins are discoidal (2, 98). How this alters the lipidome and proteome of ApoE lipoproteins is a subject for future research. Future work purifying ApoE from animals expressing ApoE only in particular cell types coupled with ultra-high sensitivity proteomics and lipidomics would provide more physiological analysis of the ApoE lipoprotein proteome and lipidome. An additional area of future study is the study of the biological mechanisms underlying the association of these proteins with ApoE lipoproteins. Our current study characterizes proteins and lipids associated with microglia and astrocyte-secreted ApoE lipoprotein but does not distinguish whether these proteins are incorporated into the particle during the secretion pathway or become associated following secretion by the cell. How the incorporation of proteins into ApoE lipoproteins alters their function is currently unknown. LPS-binding protein, which can transfer LPS to CD14 to initiate inflammatory Toll-like receptor 4/myeloid differentiation factor 2 signaling, becomes anti-inflammatory when associated with ApoA1-containing HDL (69, 99, 100). Thus, understanding the biological mechanisms of the protein complexes identified in this study will be an important area of future research.

Data Availability

Data are included in the supplemental data.

Supplemental Data

This article contains supplemental data.

Author Contributions

M. R. S. and D. M. H. conceptualization; M. R. S., Z. W., and Y. R. formal analysis; M. R. S., Z. W., L. R. G., H. W., P. L., and G. T. T. investigation; M. R. S. resources; M. R. S. writing–original draft; J. D. U. and D. M. H. writing–review & editing; M.R.S. visualization; N. Z., X. H., J. P., and D. M. H. supervision; D. M. H. project administration; D. M. H. funding acquisition.

Ethical Approval

All animal procedures and experiments were performed under guidelines approved by the Institutional Animal Care and Use Committee at Washington University School of Medicine.

Conflict of Interest

D.M.H. has equity and is on the scientific advisory board of C2N Diagnostics and is on the scientific advisory boards of Denali, Genentech, Acta, Cajal Neuroscience, and Switch Therapeutics, and consults for Pfizer and Roche. All other authors declare that they have no conflicts of interest with the contents of this article.