Arachidonic Acid Mobilization and Peroxidation Promote Microglial Dysfunction in Aβ Pathology

Abstract

Aberrant increase of arachidonic acid (ARA) has long been implicated in the pathology of Alzheimer's disease (AD), while the underlying causal mechanism remains unclear. In this study, we revealed a link between ARA mobilization and microglial dysfunction in Aβ pathology. Lipidomic analysis of primary microglia from App^NL-GF mice showed a marked increase in free ARA and lysophospholipids (LPLs) along with a decrease in ARA-containing phospholipids, suggesting increased ARA release from phospholipids (PLs). To manipulate ARA-containing PLs in microglia, we genetically deleted lysophosphatidylcholine acyltransferase 3 (Lpcat3), the main enzyme catalyzing the incorporation of ARA into PLs. Loss of microglial Lpcat3 reduced the levels of ARA-containing PLs, free ARA and LPLs, leading to a compensatory increase in monounsaturated fatty acid (MUFA)-containing PLs in both male and female App^NL-GF mice. Notably, the reduction of ARA in microglia significantly ameliorated oxidative stress and inflammatory responses while enhancing the phagocytosis of Aβ plaques and promoting the compaction of Aβ deposits. Mechanistically, scRNA seq suggested that LPCAT3 deficiency facilitates phagocytosis by facilitating de novo lipid synthesis while protecting microglia from oxidative damage. Collectively, our study reveals a novel mechanistic link between ARA mobilization and microglial dysfunction in AD. Lowering brain ARA levels through pharmacological or dietary interventions may be a potential therapeutic strategy to slow down AD progression.

Significance Statement

This study revealed a novel mechanistic link between the increase of arachidonic acid and microglial dysfunction in Alzheimer's disease. We discovered that microglia in an AD mouse model show heightened free ARA, pointing to increased ARA release from phospholipids. By targeting lysophosphatidylcholine acyltransferase in microglia, we effectively reduced ARA levels, leading to decreased oxidative stress and inflammation, and enhanced clearance of Aβ plaques. This study suggests that lowering brain ARA levels could be a viable approach to slow AD progression.

Introduction

Arachidonic acid (ARA), an n6 polyunsaturated fatty acid (PUFA), is abundantly present in the brain and exerts critical functions in various aspects of brain physiology (Rapoport, 2008; Tallima and El Ridi, 2018). Dysregulation of ARA metabolism and signaling has been implicated in the pathogenesis of neurodegenerative diseases, particularly Alzheimer's disease (AD; Thomas and Olivier, 2016). Increased levels of ARA and its derivatives have been observed in brain tissues of AD patients, suggesting increased ARA mobilization (Rao et al., 2011). Moreover, the inflammatory mediators derived from ARA, such as prostaglandins and leukotrienes, contribute to neuroinflammation in AD pathology. Genome-wide association studies have identified genes associated with ARA metabolism and signaling, such as PTGS2 and PLA2G4E, that are linked to AD risk (Ma et al., 2008; Chen et al., 2016; Hu et al., 2020; Pérez-González et al., 2020). Despite accumulating evidence supporting the involvement of ARA in AD pathology, the precise mechanisms and contributions of ARA dysregulation in AD pathogenesis remain elusive.

ARA-derived peroxidation products may contribute to AD pathogenesis through increasing oxidative stress. ARAs are highly susceptible to lipid oxidation mediated by reactive oxygen species (ROS). ROS can target both esterified ARA in PLs and free ARA in the cytoplasm. The peroxidation of ARA leads to the formation of diverse carbonyl-containing lipid peroxides, such as acrolein, 4-hydroxyhexenal (4-HHE), and 4-hydroxydodecadienal (4-HDDE; Sultana et al., 2013; Sousa et al., 2017). These reactive aldehydes can form adducts with proteins, PLs, and nucleic acids, impairing cellular function and integrity, acting as potent mediators of oxidative stress, exacerbating neuronal damage, and promoting neuroinflammation. Lipid peroxidation also renders neuron and glial cells more vulnerable to ferroptosis, a specific form of regulated cell death (Chen et al., 2021). The accumulation of lipid peroxidation products, alongside other factors like amyloid-beta (Aβ) plaques and tau pathology, has been consistently observed in the brains of AD patients (Tönnies and Trushina, 2017; Butterfield and Halliwell, 2019). However, due to the limited availability of tools to manipulate ARA levels in the brain, the precise contribution of ARA-derived lipid peroxides to AD pathology remains uncertain.

In the brain, the majority of ARA is esterified into membrane PLs at sn2 position and actively remodeled through the “deacylation-reacylation cycle,” also known as the “Lands cycle.” In this cycle, ARA released from membrane PLs through phospholipase A2 (PLA2) and subsequently re-esterified into the membrane PLs through the lysophospholipid acyltransferases (LPLATs; Wang and Tontonoz, 2019). Previous studies have shown that ARA is mainly incorporated into PLs during the Land's cycle (Hishikawa et al., 2014). Therefore, the reacylation step catalyzed by LPLATs likely determines the levels of free ARA and ARA-containing phospholipids (PLs). Among LPLATs, LPCAT3 shows strong substrate preference toward ARA and likely mediates the incorporation of ARA into PLs. Indeed, genetic deletion of Lpcat3 causes a selective reduction of ARA-containing PLs in in peripheral tissues such as liver and intestine (Rong et al., 2013, 2015; Hashidate-Yoshida et al., 2015; Wang et al., 2016). Manipulation of Lpcat3 expression in the brain cells offers a new approach to interrogate the contributions of ARA mobilization in AD.

In this study, our lipidomic analysis revealed an increase of free ARA and lysophospholipids (LPLs) in microglia isolated from App^NL-GF mice. These alterations are likely a consequence of increased ARA release from PLs. By leveraging the substrate preference of LPCAT3 for ARA, we selectively modulate the abundance of free ARA and ARA-containing PLs in microglia. Our results revealed a positive correlation between ARA levels and lipid peroxides within microglia. Furthermore, we demonstrated that ARA influences microglial phagocytosis and inflammatory response in response to Aβ pathology. Together, our study uncovered a mechanistic connection between ARA mobilization, lipid peroxidation, and microglial dysfunction in AD.

Materials and Methods

Ethics approval and consent to participate

All experiments were performed in compliance with the institutional guidelines (The Ohio State University). All animal work was performed following NIH guidelines and protocols approved by The Ohio State University Institutional Animal Care and Use Committee (IACUC).

Animals

The generation of Lpcat3^f/f was described previously (Rong et al., 2015). Vav1-Cre was purchased from The Jackson Laboratory (Strain #:035670). APP^NL-GF mice were a gift from Dr. Saido's lab in RIKEN (Saito et al., 2014). Lpcat3^f/f mice were crossed to Vav1-Cre mice to obtain Lpcat3^f/f;Vav1-Cre mice which were further crossed with APP^NL-GF mice until reaching homozygous for both hAPP and Lpcat3 flox. APP^NL-GF; Lpcat3^f/f; Vav1-Cre and its littermate control APP^NL-GF; Lpcat3^f/f were used in the experiments. Mice were housed in a specific pathogen-free facility with ad libitum access to food and water and 72°F and 12 h light on/off cycle. Mice of both sexes were used for all experiments.

Antibodies and reagents

Primary antibodies used in this study are as follows: Anti-Human Amyloid β (N; 82E1) from Immuno-Biological Laboratories, Anti-Human Amyloid β (H31L21) from Thermo Fisher Scientific, Anti 4-HNE (HNEJ-2) from Abcam, Anti-Acrolein (10A10) from Abcam, Anti-CD45 (30-F11) from BioLegend, Anti-LAMP1 antibody (ab25245) from Abcam. All secondary antibodies were purchased from Thermo Fisher Scientific or Jackson ImmunoResearch. For Thioflavin S staining, sections were stained with 0.05% Thioflavin S in 50% ethanol for 8 min and differentiated in two changes of 80% ethanol for 10 s each. Sections were then washed in large volumes of distilled water three times and incubated in a high concentration of 3× PBS buffer for 30 min. Sections were then briefly rinsed with distilled water and coverslipped for imaging.

Histological analysis

Brains were sectioned on a cryostat at 40 mm thickness. For immunofluorescence staining, free-floating sections were blocked with PBS/T containing 10% normal goat serum (NGS) at room temperature for 60 min, incubated with primary antibody in blocking solution (PBS/T with 1% NGS) at 4°C for 24–48 h, and then incubated with secondary antibody at room temperature for 2 h. Sections were mounted on slides with ProLong Diamond (Life Technologies). Images were captured on a ZEISS Axio Observer and quantified using ImageJ software. Auto Threshold methods “Otsu” or “Triangle” were used to define the region of interests (ROI). Statistical analyses were conducted using a two-tailed unpaired t test or one-way ANOVA.

Aβ40 and Aβ42 ELISA

Proteins were sequentially extracted from brain tissues with RIPA, and 5 M guanidine buffer in the presence of protease inhibitors as described previously (Gao et al., 2020). For quantifying Aβ40 and Aβ42, Human Amyloid Beta (1–42) ELISA Kit, 448707 and Human Amyloid Beta 1–40 ELISA Kit, 449007 from BioLegend were used. Statistical analyses were conducted using a two-tailed unpaired t test or one-way ANOVA.

Lipid extraction

Lipids were extracted from isolated microglia by homogenizing with 180 µl of isopropanol using a vortex mixer. The mixture was spiked with 20 µl of isotope-labeled Equisplash Lipidomix (Avanti Polar Lipids), containing a mix of 13 isotope-labeled lipid species, including 15:0−18:1(d7) PC, 18:1(d7) LysoPC, 15:0− 18:1(d7) PE, 18:1(d7) LysoPE, 15:0−18:1(d7) PG (Na salt), 15:0−18:1(d7) PI (NH4 salt), 15:0−18:1(d7) PS (Na salt), 15:0−18:1(d7)−15:0 TAG, 15:0−18:1(d7) DAG, 18:1(d7) MAG, 18:1(d7) CE, d18:1−18:1(d9) SM, and C15 ceramide-d7, dissolved in methanol, used for internal standard purposes. The mixture was subsequently sonicated in an ice water bath for 20 min followed by incubation at 4°C for 30 min. Finally, it was centrifuged at 13,000 rpm for 10 min at 4°C, and 150 µl of the supernatant containing extracted lipids was collected for LC-MS injection. To ensure instrument stability throughout the analysis, a pooled quality control sample was prepared by mixing ∼10 µl aliquots of every biological sample analyzed in an LC-MS vial and homogenizing the QC sample by vortexing.

LC-MS-based untargeted lipidomic analysis

LC-MS/MS analyses were performed on a Vanquish ultrahigh-performance liquid chromatography (UHPLC) system coupled to a Q Exactive Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific). The electrospray ionization source (ESI) on the Q Exactive was operated in both positive and negative ion modes. The ion-spray voltage was set at 4 kV with a capillary tube temperature of 320°C. The sheath gas rate was set to 10 arbitrary units. Full MS scans of 1 ms were performed at a resolution of 70,000. The automatic gain control (AGC) target was set to 3 × 10⁶, and the maximum IT was set to 200 ms (Wang et al., 2016). Data-dependent MS/MS (dd-MS²) and selected ion monitoring (dd-SIM) scans were conducted to improve compound identification at a resolution of 17,500, an AGC target of 1 × 10⁵, and maximum IT of 50 ms (Wang et al., 2016). Normalized collision energies (NCEs) were set to 20, 50, and 80.

A sample volume of 5 µl was injected onto an Acquity UPLC CSH C18 1.7 µm 2.1 mm × 100 mm column (Waters). The mobile phases were (A) 100% water, containing 5 mM ammonium acetate and 0.1% formic acid, and (B) acetonitrile and H2O at a proportion of 95/5 (v/v), containing 5 mM ammonium acetate and 0.1% formic acid. Mobile phases were delivered at a flow rate of 0.35 ml/min for a 22 min run with the following stepwise gradient for Solvent B: initially 30%, 0–5 min at 30%, 5–15 min at 62%, 15–16.5 min at 82%, 16.5–18.5 min at 99%, and 18.5–22 min at 30% (Chen et al., 2022). Pooled QC samples were injected among every 10 injections of biological samples to monitor instrument performance. LC-MS spectra were aligned and annotated by MS/MS matching to online databases using MSDial software (http://prime.psc.riken.jp/compms/msdial/main.html). Successfully annotated compounds were exported along with each compound's chemical formula and structure, retention time, m/z, adduct type, and MS/MS spectrum. A coefficient of variance (CV) was calculated for QC samples, and only annotated lipids with a CV < 20% were included in further analysis. To ensure further reproducibility, CVs were also calculated for isotope-labeled internal standards. All annotated internal standards displayed a CV < 20%, suggesting high analytical reproducibility. Peak areas for each annotated lipid were normalized to cell count for each replicate, and statistical analyses were conducted in MetaboAnalyst 5.0. Data were further analyzed for lipid class differentials using LipidSig Web-based software for statistical analysis (http://www.chenglab.cmu.edu.tw/lipidsig/Profiling/).

Primary microglia culture and treatments

Primary microglia were prepared as described previously (Gao et al., 2017). Briefly, mouse cortices from 1- to 3-d-old newborn pups were isolated in DPBS. After removing the meninges, brain tissues were cut into small pieces and transferred to 50 ml of conical tube with 5 ml of DMEM/F12 complete medium. Tissues were then titrates 20 times with 10 ml of pipette, then 20 times with 1 ml of pipette before filtering through 70 µM filter, and centrifuged at 300 g for 5 min. After resuspension in 10% FBS/DMEM, cells were plated onto poly-d-lysine (PDL)–coated T-75 flasks to generate mixed glial cultures in the medium of 10% FBS/DMEM containing 10 µg/ml GM-CSF. When confluent on Day 12, microglia were separated from the glia layer by shaking the flask at 200 rpm for 3 h. Floating microglia were seeded at 75,000 cells/cm² in PDL-coated plates. Once attached to the plates, microglia were cultured in serum-free TIC medium containing TGF-b2: 2 ng/ml, IL-34: 100 ng/ml, and cholesterol: 1.5 mg/ml as described previously (Bohlen et al., 2017). Primary microglia were treated with BSA-arachidonate polyunsaturated fatty acid complex (34931, Cayman Chemical), BSA-oleate monounsaturated fatty acid complex (29557, Cayman Chemical), or BSA control for BSA-fatty acid complexes (29556, Cayman Chemical) at various concentrations for 24 h before adding Aβ fibrils (1 µM). For pHrodo-myelin phagocytosis assay, primary microglia were treated with 4-HNE at concentration indicated for 2 h before adding the pHrodo-labeled myelin for and incubate for another 2 h before evaluating phagocytosis.

Adult microglia isolation

Adult microglia were isolated using magnetic-activated cell sorting (MACS) as described previously (Gao et al., 2020). Briefly, mice were anesthetized and perfused with cold PBS to remove circulating blood cells. Dissected brains were chilled on ice and dissociated using brain dissociation kit (#130-107-677, Miltenyi Biotec) and gentleMACS Tissue Dissociator. After dissociation, cell suspensions were filtered through SmartStrainers 70 µm and centrifuged at 300 g for 10 min to pellet the cells. Myelin in the cell suspension was depleted by myelin removal beads (130-096-733, Miltenyi Biotec) and subjected to flow cytometry or further purified by CD11b MicroBeads (130-049-601, Miltenyi Biotec) and magnetic LD columns (130-042-901, Miltenyi Biotec) to obtain microglia for ScRNA seq.

Flow cytometry to evaluate phagocytic microglia

To prepare methoxy-X04 (Tocris, 4920) solution, methoxy-X04 was first dissolved in DMSO. Methoxy-X04 DMSO solution was then added to the mix of propylene glycol and PBS (1:1), and the solution was stirred at 4°C until a yellowish green emulsion was obtained. Methoxy-X04 solution was prepared freshly on the day of injection and was injected into mice at the dose of 10 mg/kg. For measuring microglial cells with Aβ content, animals at 6–7 months were intraperitoneally injected with methoxy-X04 (10 mg/kg) 24 h before tissue collection. Animals were processed as described above, and dissociated cells were incubated in Fixable Viability Dye 780 (eBioscience, 1:1,000) and anti-CD16/32 (clone 2.4G2, hybridoma, ATCC, 1;200) to label dead cells and prevent binding of fluorescent antibodies to Fc receptors. Cells were then stained with fluorescently labeled monoclonal antibodies for CD11b (Invitrogen, M1/70, 1:500), CD45 (Invitrogen, 30-F11, 1:100), and CX3CR1 (R&D Systems, 1:50). Live CD45⁺ CD11b⁺ CX3CR1⁺ microglia cells were gated as X04+ or X04− using the DAPI excitation channel to determine the percentage of microglia with ingested amyloid content. Wild-type animals injected with methoxy-X04 solution were used as negative controls to construct an appropriate gating strategy for analysis and sorting.

Bulk RNA seq and data analysis

The hippocampus was dissected from brain tissues, submerged in RNAlater solution (Thermo Fisher Scientific), and stored in −80°C freezer until further processing. Total RNA was extracted using Quick-RNA Miniprep (R1055, Zymo Research). RNA quality was evaluated by TapeStation using high Sensitivity RNA ScreenTape (5067-5579, Agilent). RNA samples with RNA integrity numbers >8 were used for cDNA library construction. RNA seq libraries were prepared using SMART Seq mRNA LP Kit (Takara Bio) following the manufacturer's instructions. The qualities of the cDNA library were assessed using TapeStation using High Sensitivity D5000 ScreenTape (5067-5592, Agilent). cDNA library samples were then pooled and sequenced with the HiSeq 4000 System (Illumina).

Demultiplexed FASTQ files of bulk RNA sequencing data were aligned to the mouse genome (Mus_musculus.GRCm39) using STAR (version 2.7.10a). Adapters were trimmed using flexbar (version 3.5.0.). Reads mapped to genomic features were counted using featureCounts (version 2.0.3). The count matrix was imported in R (version 4.2.0) for analysis. Gene expression signatures of different experimental conditions were compared using DeSeq2 (version 1.36.0), using the Wald test for hypothesis testing and the “apeglm” for LFC shrinkage. Genes with adjusted p values (using Bonferroni’s correction) <0.05 and log2 fold changes >|0.25| were considered significantly differentially expressed.

Single-cell seq and data processing

Library construction

Single-cell RNA-seq libraries were generated using the 10× Genomics Chromium NEXT GEM Single Cell 3′ Reagent Kit. Briefly, primary microglia isolated from adult mice were loaded onto chromium chips with a capture target of 10,000 cells per sample. Libraries were prepared following the provided protocol and sequenced on an Illumina NovaSeq with a targeted sequencing depth of 50,000–100,000 reads per cell. FASTQ files from sequencing were then used as inputs to the 10× Genomics Cell Ranger pipeline.

Read processing, quality control, and filtering

Gene expression matrices were generated with the Cell Ranger Pipeline v7.0.0 (10× Genomics) and aligned to the Mouse (mm10) reference transcriptome. The resulting digital gene expression matrix was filtered, normalized, and clustered using R version 4.2.0 and Seurat version 4.1.1. Genes that are expressed in <10 cells, and cells with >5% of reads mapped to mitochondrial genes or with <500 features and 1,000 UMIs were removed. DoubletFinder (version 2.0.3) was used to identify false-negative Demuxlet classifications caused by doublets formed from cells with identical SNP profiles, and an average of 7.1% of cells per sample were confidently predicted as doublets and removed. In total, 37,487 cells were retained for downstream analysis.

Normalization, integration, and clustering

The gene expression matrix was normalized and scaled using the Seurat function SCTransform which also identifies the most variable genes, of which the top 3,000 were used for dimensionality reduction. During normalization, we also removed confounding sources of variation, including the effects of mitochondrial expression, ribosomal expression, and the difference between the G2M and S phase scores. Six samples were integrated to correct for any potential library batch effect by using the Seurat functions FindIntegrationAnchors and IntegrateData based on reciprocal PCA with n = 5 neighbors (k.anchor). Integrated matrix was used for downstream analysis. Cells were clustered using the Louvain algorithm based on the first 40 principal components with a resolution of 0.5. The Uniform Manifold Approximation and Projection (UMAP) was used for nonlinear reduction and two-dimensional data visualization.

Cluster annotation, subsetting, and reclustering

Cell-type annotations were assigned to each cluster based on two levels of evidence. First, the Seurat function FindAllMarkers was used to identify cluster marker genes based on one-versus-all Wilcoxon rank sum differential expression tests for each cluster. Second, the cell-type identity of all cells was predicted based on a panel of curated marker genes in a published ScSeq dataset of the App^NL-G-F mouse brain (Sala Frigerio et al., 2019). This led to the identification of 18 clusters, 14 of which represented unique microglial/myeloid cell identities/states. For the final analysis of high-quality microglia, we kept putative microglia and removed other cell types. After in silico purification, 34,265 microglia were retained for downstream analysis. To recluster the dataset of putative microglia, we normalized the dataset using SCTransform, performed PCA, and selected 40 dimensions for dimensionality reduction by UMAP, as described above. To identify clusters, we first used the Seurat function FindNeighbors based on the first 40 PCs and then performed unbiased clustering by using the Seurat function FindClusters, with the resolution of granularity set to 0.2. This led to the identification of eight clusters, each representing a cell state defined by unique or transitory profiles (Fig. 7A).

Figure 7.

Microglial LPCAT3 deficiency enhances de novo lipid synthesis and MUFA incorporation in DAM. A, UMAP plots of different microglial clusters. Each cell was color-coded based on its cluster affiliation. TRM, Transiting response microglia; HM, homeostatic microglia; RM, ribosomal microglia; DAM, disease-associated microglia; CRM, cytokine response microglia; SM, stressed microglia; IRM, interferon response microglia; CPM, cycling/proliferating microglia. See Extended Data Figure 7-1 for more details. B, Violin plot of signature genes for the major clusters. HM (Tmem119, P2ry12, Cx3cr1); DAM (ApoE, Cst7, Itgax, Cd74, H2-Aa, H2-Ab1); IRM (Ifit2, Ifit3, Ifitm3), CPM (Top2a, Mcm2, Mki67). C, Heat map of top 10 differentially expressed genes from each cluster. D, Average percentage of microglia in each cluster for each genotype (n = 3 for each group). E, Heat map and clustering of differentially expressed (DE) genes from pseudobulk analysis of DAM. See Extended Data Figure 7-2 for more details. F, Volcano plot of DE genes related to PL and fatty acid synthesis, cholesterol synthesis, and oxidative phosphorylation in DAM. See Extended Data Figure 7-3 for more details in TRM.

Differential gene expression analysis

Differential gene expression analysis was performed in two aspects. First, differentially expressed genes of specific cell states were found by applying the Seurat function FindAllMarkers for overall DE and FindMarkers for side-by-side comparisons. Second, gene expression signatures of different genotypes within each cell state were compared using the Pseudobulk differential expression analyses with DeSeq2 using the Wald test for hypothesis testing and the “apeglm” for LFC shrinkage, as described above. Genes with adjusted p values (using Bonferroni’s correction) <0.05, and log2 fold changes >|0.25| were considered significantly differentially expressed. g:Profiler was used to test for gene sets enriched in significant DE genes in the following databases: Wikipathways Mouse (2022), KEGG Mouse (2022), and Reactome Mouse (2022).

Gene coexpression module analysis

Using the Seurat function AddModuleScore, gene module scores were computed for previously published microglia response signatures of App^NL-G-F mice (Sala Frigerio et al., 2019; Extended Data Fig. 7-1) including homeostatic microglia (HM) markers (Tmem119, P2ry12, Cx3cr1), activated response microglia markers (Apoe, Cst7, Itgax, Lpl, Spp1, Gpnmb, Dkk2, Cd74, H2-ARA, H2-Ab1), transiting response microglia markers (Apoe, Cst7, Itgax, Cd74, H2-ARA, H2-Ab1), interferon response microglia markers (Ifit2, Ifit3, Ifitm3, Oasl2, and Irf7), cycling and proliferating microglia (Top2a, Mcm2, Tubb5, Mki67, Cdk1), and an MHC class II signature (Chen et al., 2021; Cd74, H2-Ab1, H2-ARA, H2-D1, H2-K1, H2-Eb1, H2-DMa).

Results

ARA mobilization is increased in the microglia of App^NL-GF mice

To investigate whether Aβ pathology influences ARA metabolism in microglia, we conducted a lipidomic analysis on microglia acutely isolated from App^NL-GF mice at 9 months old, a stage characterized by extensive Aβ plaque deposition. Our analysis revealed a marked increase in unsaturated fatty acids (UFAs), particularly polyunsaturated fatty acids (PUFAs) such as ARA (FA 20:4), docosatetraenoic acid (FA 22:4), and docosahexaenoic acid (DHA, FA 22:6), in microglia from App^NL-GF mice compared with age-matched wild-type controls (Fig. 1A). Concurrently, we observed elevated levels of LPLs, including lysophosphatidylcholine (LPC, 16:0, 18:0, 18:1) and lysophosphatidylethanolamine (LPE, 18:0 and 18:1), in the microglia of App^NL-GF mice (Fig. 1B). This increase in both UFAs and LPLs, likely generated from the hydrolysis of PLs at the sn-2 position, indicates a shift in PL metabolism. Indeed, our comparison of acyl chain composition of the most abundant PLs in mammalian cells, phosphatidylcholine (PC) and phosphatidylethanolamine (PE), revealed a decline in PUFA-containing PLs (e.g., PC 36:4, PC 38:6, PE 38:4) and an increase in monounsaturated (MUFA) and saturated fatty acid (SFA)-containing PLs (e.g., PC 34:0, PC 38:1, PE 36:1) in microglia from App^NL-GF mice (Fig. 1C,D). Collectively, our lipidomic data suggest an enhanced hydrolysis of PUFA-containing PLs and subsequent release of ARA in microglia in response to Aβ pathology.

Figure 1.

ARA mobilization is increased in the microglia of App^NL-GF mice. A, Lipidomic profiling of free fatty acids (A) and LPLs (B) in acutely isolated microglia from 9-month-old App^NL-GF mice and age-matched wild-type mice. Acyl chain composition of PC (C) and PE (D) in acutely isolated microglia from 9-month-old App^NL-GF mice and age-matched wild-type mice. *p < 0.05; **p < 0.01; ***p < 0.001. All the lipid species listed in C and D have FDR < 0.05.

LPCAT3 catalyzes the incorporation of ARA into PLs in microglia

Previous studies indicate that ARA is incorporated into PLs mainly during the remodeling process through LPCAT3 (Wang and Tontonoz, 2019). Building on this, we hypothesized that chronic knock-out of Lpcat3 in microglia would impede the incorporation of ARA into PLs, thus allowing us to manipulate and interrogate ARA metabolism and signaling in the microglia of App^NL-GF mice. To test this, we crossed Lpcat3^flox/flox mice with a Cre mouse line driven by Vav1 promoter, enabling Lpcat3 deletion in hematopoietic cells, including brain microglia. Primary microglia from Lpcat3^f/f, vav1-Cre mice showed a >95% reduction in Lpcat3 mRNA levels (Fig. 2A), confirming the efficiency of Vav1-Cre–driven Lpcat3 knock-out. In primary microglia cultured from neonates, LPCAT3 deficiency notably decreased levels of ARA-containing 36:4 PC and 38:4 PC, while not affecting linoleate-containing 34:2 PC. This resulted in a compensatory increase in 32:1 PC levels (Fig. 2B). Similar reductions were also observed in ARA-containing 36:4 PE and 38:4 PE (Fig. 2C).

Figure 2.

LPCAT3 catalyzes the incorporation of ARA into PLs in microglia. A, Vav1-Cre mediated knock-out of Lpcat3 in primary microglia cultured from newborn pups. Acyl chain composition of PC (B) and PE (C) in primary microglia cultured from newborn pups of Lpcat3^f/f and Lpcat3^f/f; Vav1-Cre mice. Acyl chain composition of PC (D), PE (E), phosphatidylserine, and phosphatidylinositol (F) in primary microglia acutely isolated from 2-month-old ALp^f/f-Cre and littermate control ALp^f/f mice. See Extended Data Figure 2-1 for more details. * All the lipid species listed in D, E, and F have FDR < 0.05.

Figure 2-1

(A) Partial least squares discriminant analysis (PLS-DA) of lipidome, and (B) the abundance of major lipid species from lipidome analysis of primary microglia acutely isolated from 2-month-old ALp^f/f-Cre and littermate control ALp ^f/f mice. Download Figure 2-1, TIF file ^{(13.5MB, tif)} .

To extend our findings from ex vivo to in vivo, we bred Lpcat3^f/f, vav1-Cre mice into App^NL-GF background (abbreviated as ALp^f/f-Cre thereafter). At 2 months, prior to Aβ plaque deposition, we isolated primary microglia from ALp^f/f-Cre mice and its littermate controls ALp^f/f mice for lipidomic analysis using ultrahigh-performance liquid chromatography-high-resolution mass spectrometry (UPLC-HRMS). The partial least squares discriminant analysis (PL-SDA) distinctively segregated Lpcat3-null microglia from wild-type controls, with only minor differences in total signals of major lipid classes, except for a mild increase in PC and a decrease in ether-linked PC (PC O-; Extended Data Fig. 2-1). LPCAT3 deficiency resulted in a significant reduction of ARA in all major PL classes including PC, PE, PS, and PI (Fig. 2D–F), highlighted by the decrease in PC 36:4, PC 38:4, PC O-38:5, PE 36:4, PE O-40:5, and PS 38:4. A compensatory rise in MUFA-containing PLs, such as PC 32:1, PC 34:2, PE 34:1, and PE 36:2, were observed, likely occurred to maintain membrane fluidity. Moreover, an increase in 22:4-containing PLs like PC 16:0_22:4 and PE 16:0_22:4 was observed, potentially resulting from the carbon chain elongation of ARA (Fig. 2D,E). Our data collectively suggest that LPCAT3 is the pivotal acyltransferase-mediating ARA incorporation into key PL species in microglia.

LPCAT3 deficiency reduces free ARA and LPLs in the microglia of App^NL-GF mice

To investigate how Lpcat3 deficiency impacts the dynamics of microglial lipidome, especially ARA and ARA-containing lipid species, in response to Aβ pathology, we performed lipidomic analysis on primary microglia acutely isolated from ALp^f/f-Cre mice and its littermate controls (ALp^f/f) at 9 months of age, when the Aβ plaque deposition reaches a plateau. PL-SDA on all lipid species easily separated between genotypes, and the abundance of major lipid classes was comparable between genotypes, except for a mild decrease of ether-linked PC (PC O-) in the Lpcat3 null microglia (Extended Data Fig. 3-1).

Notably, the absence of Lpcat3 selectively reduced ARA-containing PLs such as PC O-16:0_20:4, PC O-16:1_20:4, PE 16:0_20:4, and PS 18:0_20:4, while elevating levels of MUFA-containing PLs, including PC 32:1, PC 34:2, PE 34:1, and PE 36:2 (Fig. 3A–C). Importantly, Lpcat3 deletion reduced the levels of LPLs, including LPC 16:0 and LPC 18:0, and free ARA without affecting the levels of other PUFAs such as FA 22:4 and FA 22:6 (Fig. 3D,E). These results imply that the elevated ARA and LPLs in the microglia of App^NL-GF mice predominantly stem from their release from PLs, and chronic knock-out of microglia Lpcat3 blocks ARA mobilization in microglia. Furthermore, blocking the incorporation of ARA into microglial PLs does not affect overall brain ARA levels. Lipidomic analysis of cortical tissue from WT, ALp^f/f, and ALp^f/f-Cre mice at 9 months of age revealed comparable levels of ARA and other fatty acids across all genotypes (Extended Data Fig. 3-2).

Figure 3.

Loss of Lpcat3 reduced ARA mobilization in the microglia from 9-month-old App^NL-GF mice. Lipidome profiling of PC (A) and PE (B), phosphatidylserine (C), LPLs (D), and free fatty acids (E) in primary microglia acutely isolated from 9-month-old ALp^f/f-Cre and littermate control ALp^f/f mice. See Extended Data Figures 3-1 and 3-2 for more details. ** All the lipid species listed in A, B, and C have FDR < 0.05.

Figure 3-1

(A) PLS-DA of lipidome, and (B) the abundance of major lipid species from lipidome analysis of primary microglia acutely isolated from 9-month-old ALp^f/f-Cre and littermate control ALp ^f/f mice. Download Figure 3-1, TIF file ^{(15.7MB, tif)} .

Figure 3-2

The relative abundance of major FFAs in the cortex of 9-month-old WT, ALp^f/f-Cre, ALp ^f/f mice. Download Figure 3-2, TIF file ^{(5.3MB, tif)} .

Loss of microglial LPCAT3 enhances the microglial phagocytosis of fibrillar Aβ and ameliorates Aβ pathology

As microglia are the primary cell type responsible for clearing Aβ plaques in the brain, we evaluated whether inhibiting ARA mobilization in microglia affects the progression of Aβ pathology. By 9 months of age, ALp^f/f-cre mice showed a significant reduction in total Aβ plaque areas (82E1) compared with their littermate control ALp^f/f mice (Fig. 4A,B). Notably, these mice exhibited increased levels of dense-core plaques, as indicated by Thioflavin S staining (Fig. 4A,C), suggesting that LPCAT3 deficiency promotes microglia-mediated compaction of Aβ plaques. Consistently, ELISA showed reductions in both soluble (RIPA fraction) and insoluble (GuHCl fraction) forms of Aβ42 in the brain of ALp^f/f-cre mice compared with its ALp^f/f littermates (Fig. 4E,F). Furthermore, the area of dystrophic neurites, marked with LAMP1 antibody around Aβ plaques, was reduced in ALp^f/f-cre mice (Fig. 4A,D). These findings indicate that loss of LPCAT3 in microglia decreases Aβ deposition, enhances plaque compactness, and alleviates Aβ-associated neuropathology.

Figure 4.

Loss of microglia Lpcat3 decreases Aβ deposition and ameliorates plaque-associated dystrophic neurites in App^NL-GF mice. A, Representative images of brain sections stained with Aβ (82E1, red), Thioflavin S (green), LAMP1 (magenta) from 9-month-old ALp^f/f-Cre and littermate control ALp^f/f mice. B, Quantification of area positive for Aβ (82E1; B), Thio-S (C), and LAMP1 (dystrophic neurites; D) between genotypes. Aβ42 ELISA of soluble (RIPA) fraction (E) and insoluble (GuHCL) fraction (F) of total cortical lysate. n = 8 for each group. Scale bar, 200 µM. G, Representative images of brain sections stained with Iba1 (green), CD68 (red), and 82E1 (magenta) from 9-month-old ALp^f/f-Cre and littermate control ALp^f/f mice. H, Quantification of area positive for Iba1 and CD68. *p < 0.05; **p < 0.01; ***p < 0.001. Scale bar, 50 µM.

We hypothesize that loss of LPCAT3 enhances microglial clearance of Aβ plaques. Immunostaining of brain sections revealed no significant difference in microglia recruitment and clustering around plaques between genotypes, as indicated by similar percentages of Iba1+ areas (Fig. 4G,H). Interestingly, despite lower overall Aβ plaque areas, ALp^f/f-Cre mice showed comparable percentages of CD68+ areas to ALp^f/f littermate (Fig. 4G,H). Since CD68 is a lysosomal marker of phagocytic activity, this suggests that Lpcat3 deficiency may boost microglial phagocytosis of Aβ plaques. To directly assess microglial phagocytosis of fibrillar Aβ in vivo, we injected mice intraperitoneally with the fluorescent dye methoxy-X04 to label Aβ plaques and isolated microglia 16 h later for flow cytometry analysis. We observed an increased percentage of phagocytic microglia (X04+) in ALp^f/f-Cre mice compared with its ALp^f/f littermates at 6–7 months old (Fig. 5A,B). Interestingly, in Lpcat3-deficient microglia, the expression of CD45, indicative of inflamed microglia, was lower in the X04+ microglia but not in the X04- population (Fig. 5C,D). This suggests that LPCAT3 loss ameliorates inflammatory responses in microglia while enhancing their capacity to phagocytose Aβ.

Figure 5.

Loss of LPCAT3 enhanced the microglial phagocytosis of fibrillar Aβ. A, Representative images of flow cytometry quantifying methoxy-x04 positive and negative microglia in 6-month-old wild type, ALp^f/f-Cre and littermate control ALp^f/f mice. B, Quantification and comparison of methoxy-x04 positive microglia between genotypes. Representative images and quantification of mean fluorescence intensity of CD45 in methoxy-x04 positive (C) and negative microglia (D) from ALp^f/f-Cre and littermate control ALp^f/f mice. n = 5 for each group, *p < 0.05.

Blocking microglial ARA mobilization reduces the inflammatory response and oxidative stress associated with Aβ plaques

Aligning with the reduction in CD45 expression in X04+ microglia identified in flow cytometry analysis, we observed a notable decrease in CD45 positive area around Aβ plaques in the brains of ALp^f/f-Cre mice compared with its littermates ALp^f/f mice (Fig. 6A,B). Pursuing the underlying mechanisms, we hypothesized that microglia might undergo a respiratory burst while engulfing fibrillar Aβ, leading to the production of ROS. These ROS are capable of oxidizing ARA, resulting in lipid peroxides that could exacerbate inflammatory responses.

Figure 6.

Reduction of microglial ARA levels ameliorates oxidative stress and inflammatory response associated with Aβ plaque. A, Quantification of areas positive for CD45, 4-HNE, and acrolein in the cortical regions between 9-month-old ALp^f/f-Cre and littermate control ALp^f/f mice. Representative images of brain sections stained with amyloid plaques (white) and CD45 (red; B), or 4-HNE (green; C), or acrolein (green; D) in 9-month-old ALp^f/f-Cre and littermate control ALp^f/f mice. n = 8 for each group. *p < 0.05; ***p < 0.001; ****p < 0.0001. Scale bar, 50 uM. Volcano plot of genes related to the “oxidative stress and redox pathway” (E), as well as selected genes related to microglial activation and neuroprotection (F). n = 6 for each group.

To test our hypothesis that elevated ARA levels in phagocytic microglia correlate with increased oxidative stress and inflammation, we measured the levels of 4-hydroxynonenal (4-HNE), a major aldehyde produced during ARA peroxidation. In 9-month-old ALp^f/f-Cre, we observed accumulation of 4-HNE around Aβ plaques (Fig. 6B), indicating that Aβ plaques trigger oxidative stress. Notably, LPCAT3 deficiency in ALp^f/f-Cre mice led to a significant reduction in 4-HNE levels around Aβ plaques (Fig. 6A,B), suggesting that lower ARA levels in microglia mitigate Aβ-induced oxidative stress. Similarly, acrolein, another toxic aldehyde generated from ARA lipid peroxidation, showed increased presence around Aβ plaques in ALp^f/f mice but significantly decreased in ALp^f/f-Cre mice (Fig. 6A,C).

Transcriptome analysis of the hippocampus revealed notable differences between genotypes. In ALp^f/f-Cre mice, there was a significant downregulation in genes associated with the “oxidative stress and redox” pathway (P_adj= 0.0035). This category includes genes encoding antioxidant enzymes such as glutathione S-transferases (Gstm1, Gstm5), peroxiredoxin (Prdx1, Prdx2, Prdx3), ferroxidase (Fth1, Ftl1), and glutathione peroxidase (GPX4; Fig. 6E). The induction of these enzymes typically occurs in response to elevated oxidative stress, so their decreased expression in ALp^f/f-Cre mice implies lower oxidative stress levels. In addition, genes linked to microglial activation, such as ApoE, Cst3, Cst7, and Ccl6, were downregulated, while genes related to neuroprotective effects, such as Gpr68, Inhba, and Nptx2, were upregulated in ALp^f/f-Cre mice (Fig. 6F). Together, our results suggest a direct link between ARA-induced oxidative stress and inflammatory responses in microglia.

Microglial LPCAT3 deficiency increases de novo lipid synthesis and MUFA incorporation in disease-associated microglia

To determine the impact of Lpcat3 deficiency on microglial heterogeneity in response to Aβ pathology, we isolated microglia (CD11b⁺) from 9-month-old Alp^f/f-cre mice and their littermate controls (Alp^f/f) using MACS. We then prepared RNA sequencing libraries using a 10× Genomics platform. After stringent quality control, including the removal of peripheral immune cells and predicted doublets, our analysis included 37,487 microglia (21,069 microglia from Alp^f/f mice and 16,418 microglia from Alp^f/f-cre mice).

Unsupervised clustering revealed eight distinct microglial clusters (Fig. 7A–C). HM exhibited high levels of canonical markers like Tmem119, P2ry12, and Cx3cr1. Disease-associated microglia (DAM) showed reduced expression of homeostatic genes but increased expression in immune response (Itgax, Cst7) and MHC class II presentation (Cd74, H2-ARA) genes. Transitioning response microglia (TRM) shared transcriptomic profiles with DAM but expressed lower levels of DAM markers (Spp1, Cd74, Gpnmb). Cytokine response microglia (CRM) had similar profiles to TRM but with elevated stress- and cytokine-related genes (ATF3, Nfkb, Tnf). Interferon response microglia (IRM) predominantly expressed interferon response type I pathway genes (Ifit2, Ifit3, Ifitm3). Ribosomal microglia (RM) were characterized by higher ribosomal and lower stress-response gene expression. A small cluster of cycling/proliferating microglia (CPM) was identified, characterized by elevated DNA replication and cell cycle genes (Top2a, Mcm2, Mki67). These clusters align well with previous classifications of microglial states (Extended Data Fig. 7-1; Sala Frigerio et al., 2019). No significant differences in the proportion of each microglial cluster were observed between genotypes (n = 3 for each genotype; Fig. 7D), suggesting Lpcat3 loss does not alter microglia state transition in response to Aβ pathology.

Figure 7-1

Gene co-expression module analysis based on previously published microglial signatures of App^NL-G-F mice (Sala Frigerio et al., 2019). HM (Tmem119, P2ry12, Cx3cr1), ARM (Apoe, Cst7, Itgax, Lpl, Spp1, Gpnmb, Dkk2, Cd74, H2-ARA, H2-Ab1), TRM (Apoe, Cst7, Itgax, Cd74, H2-ARA, H2-Ab1), IRM (Ifit2, Ifit3, Ifitm3, Oasl2, and Irf7), CPM (Top2a, Mcm2, Tubb5, Mki67, Cdk1), and an MHC class II signature (Cd74, H2-Ab1, H2-ARA, H2-D1, H2-K1, H2-Eb1, H2-DMa). Download Figure 7-1, TIF file ^{(19MB, tif)} .

Figure 7-2

Functional enrichment analysis of differentially expressed genes from Pseudobulk analysis of DAM (ALp^f/f-Cre vs ALp^f/f). Download Figure 7-2, TIF file ^{(8.5MB, tif)} .

Figure 7-3

(A) Volcano plot of DE genes related to pathways “‘fatty acid synthesis’, ‘cholesterol synthesis’(B), and oxidative phosphorylation (C) from Pseudobulk analysis of TRM (ALp^f/f-Cre vs ALp^f/f, 9-month-old). Download Figure 7-3, TIF file ^{(18.6MB, tif)} .

Pseudobulk differential expression analysis within each cluster revealed that Lpcat3 loss in DAM led to significant changes in 56 genes, predominantly upregulated (Fig. 7E). Majority of upregulated genes were associated with de novo synthesis of fatty acid (Scd2, Fasn, Insig1, etc.) and cholesterol synthesis (Sqle, Msmo1, Dhcr24, etc.; Fig. 7F, Extended Data Fig. 7-2), potentially driven by Srebf1 and Srebf2, the master regulators of fatty acid and cholesterol synthesis. This suggests that Lpcat3 deficiency enhances lipid synthesis to support membrane expansion during phagocytosis. Additionally, genes linked to fatty acyl chain desaturation (Fads1) and MUFA incorporation into membrane PLs (Acsl3) were upregulated, aligning with increased MUFA-containing PLs in Lpcat3 null microglia (Figs. 2, 3) and potentially protecting DAM against oxidative stress from Aβ phagocytosis. Furthermore, mitochondrial genes involved in ATP production (Atp8, Nd5, Nd4l) were also induced in Lpcat3-deficient DAM (Fig. 7F). Similar trends in lipid synthesis and oxidative phosphorylation gene induction were noted in TRM (Extended Data Fig. 7-3). Overall, our single-cell RNA sequencing analysis suggests that Lpcat3 deficiency induces expression in genes related to lipid synthesis, MUFA incorporation, and ATP production in DAM, potentially facilitating phagocytosis by providing energy and lipids for membrane expansion while protecting microglia from oxidative damage during Aβ phagocytosis.

Aβ-induced ARA mobilization from PLs exacerbates oxidative stress and inhibits phagocytosis in primary microglia

To directly assess whether increased levels of ARA-containing PLs exacerbate Aβ-induced oxidative stress and inflammation in microglia, we pretreated primary microglia cultures with various concentrations of ARA for 24 h before exposing them to Aβ42 fibrils. We found that treatment of ARA dose-dependently increased levels of 4-HNE and CD45 upon Aβ challenge (Fig. 8A,B). In contrast, oleate, a MUFA less susceptible to peroxidation, did not significantly affect levels 4-HNE and CD45 (Fig. 8A,B). This supports our in vivo findings that ARA mobilization from ARA-containing PLs intensifies oxidative stress and inflammation in response to Aβ phagocytosis, whereas MUFA-containing PLs protect microglia against oxidative stress.

Figure 8.

Aβ-induced ARA mobilization from PLs increases 4-HNE production, exacerbates inflammatory response, and inhibits phagocytosis. A, Representative images of 4-HNE and CD45 staining in WT primary microglia pretreated with various concentrations of ARA, oleate, or BSA (control) for 24 h, followed by adding Aβ fibrils (2 µM) for another 24 h. B, Quantification of the relative intensity of 4-HNE and CD45 at different conditions. C, Representative images of 4-HNE staining in WT and Lpcat3 KO microglia pre-treated with ARA or oleate for 24 h, followed by Aβ fibrils (2 µM) treatment for another 24 h. D, Quantification of the relative intensity of 4-HNE at across different genotypes and conditions. E, Representative images of 4-HNE staining and pHrodo-myelin in primary microglia pretreated with various concentrations of 4-HNE for 2 h, followed by adding pHrodo-labeled myelin for 2 h to evaluate phagocytosis. F, Quantification of 4-HNE intensity and percentage of microglia containing pHrodo-myelin at different conditions indicated in the figure.

Further, to validate whether the LPCAT3-mediated incorporation of ARA is necessary for Aβ-induced 4-HNE production, we pretreated wild-type or Lpcat3 knock-out primary microglia with ARA or oleate before challenging with fibril Aβ42. Our experiments showed that ARA significantly increased 4-HNE levels in wild-type microglia, but not in Lpcat3 knock-out cells. Conversely, oleate treatment effectively mitigated the induction of 4-HNE levels by fibril Aβ42 (Fig. 8C,D). These observations imply that the incorporation of ARA into PLs is essential for the subsequent mobilization of ARA and the formation of its peroxidation derivative, 4-HNE.

Considering the role of ARA-derived 4-HNE and its potential impact on microglial functionality, we also explored whether 4-HNE directly inhibits microglial phagocytosis. Primary microglia were pretreated with varying concentrations of 4-HNE for 2 h before being challenged with pHrodo-conjugated myelin, a phagocytic substrate. As the levels of cellular 4-HNE increased, we observed a notable decline in myelin phagocytosis accompanied by an increase in CD45 expression (Fig. 8E,F). These findings indicate that 4-HNE, a principal aldehyde product from ARA peroxidation, could be a critical mediator inhibiting microglial phagocytosis.

Discussion

In our study, we discovered that microglia in App^NL-GF mice, a model of amyloidosis, exhibit raised levels of free ARA and LPLs. These lipid species, likely produced via PL hydrolysis, indicate increased ARA mobilization during microglia activation. To target ARA-containing PLs specifically in microglia, we genetically deleted Lpcat3, capitalizing on its substrate preference for ARA. Our findings show that Lpcat3 deletion effectively reversed the increased free ARA and LPLs in microglia, confirming PLs as the primary ARA source in microglia responding to Aβ pathology. Notably, Lpcat3 deletion in microglia resulted in reduced oxidative stress and inflammatory responses, enhanced Aβ plaque phagocytosis, and mitigated Aβ-associated neuropathology. Mechanistically, our scRNA seq revealed that LPCAT3 deficiency facilitates phagocytosis by supplying energy and lipids for membrane expansion while protecting microglia from oxidative damage. These findings reveal a novel mechanism linking ARA mobilization to microglial dysfunction in AD, enhancing our understanding of the complex interplay between lipid metabolism, neuroinflammation, and AD pathology.

Previous studies have consistently reported dysregulated brain ARA metabolism in AD. AD patients exhibit elevated ARA levels in gray matter, particularly in regions with high senile plaque and activated microglia densities (Esposito et al., 2008; Rao et al., 2011; Bazinet and Laye, 2014; Snowden et al., 2017). Furthermore, a higher ARA to DHA ratio in cerebrospinal fluid (CSF) and plasma, especially in APOE4 carriers with mild AD, predicts a higher risk of transitioning to mild cognitive impairment (Abdullah et al., 2017; Tomaszewski et al., 2020). The increased cPLA2 activity in the brain and CSF of AD patients has been implicated in this process (Stephenson et al., 1996; Fonteh et al., 2013), with higher levels noted in ApoE4 human brains alongside elevated LTB4, ROS, and neuroinflammation (Wang et al., 2022). Indeed, genetic ablation of cPLA2 in hAPP mice protected against Aβ-dependent deficits, suggesting its pathogenic role in AD (Sanchez-Mejia et al., 2008). Building on these insights, our research further demonstrates that LPCAT3 is another pivotal enzyme in mediating ARA incorporation into microglial PLs, establishing a causal link between ARA mobilization and microglial dysfunction in AD.

Our results indicated that elevated ARA in microglia leads to increased lipid peroxide production, including acrolein and 4-HNE, in plaque-associated microglia. In AD brains, activated microglia upregulate enzymes like NADPH oxidase, producing ROS that react with PUFAs to generate lipid peroxides (Montine et al., 2002). Indeed, the levels of 4-HNE and acrolein, two major lipid peroxidation products from ARA, increased in the brains of mild cognitive impairment and early AD patients (Williams et al., 2006) and further propagated and amplified in the late AD stage (Bradley et al., 2012). Consistent with this notion, we observed increased levels of ARA-derived 4-HNE and oxidized PLs in plaque-associated microglia. A limitation of this study is that we mainly focus on how ARA mobilization and metabolism impact microglial responses. Future investigation is needed to explore whether ARA metabolism is also altered in other brain cells, such as astrocytes and neurons, in AD, and how aberrant ARA-induced oxidative stress contributes to cognitive impairment during the disease progression.

Our results suggested that 4-HNE is likely responsible for ARA-mediated inflammatory responses and impairment of Aβ phagocytosis in microglia, though the precise mechanisms remain to be explored. 4-HNE forms adducts with proteins, PLs, and nucleic acids, altering membrane properties and impacting the functions of membrane-bound enzymes and transporters (Yehuda et al., 2002; Bacot et al., 2003; Sultana et al., 2013; Castro et al., 2017; Sousa et al., 2017). Redox proteomics has identified numerous oxidatively modified proteins in AD brains involved in key cellular processes, a critical aspect to decipher the role of oxidative stress in AD pathogenesis (Castegna et al., 2002; Lu et al., 2004; Poon et al., 2004; Reed et al., 2008; Perluigi et al., 2009; Butterfield et al., 2012). Besides ARA-derived lipid peroxides, free ARA may contribute to AD pathogenesis by synthesizing eicosanoids (Heneka et al., 2015), potentiating neurotransmitter release (Williams et al., 1989; Latham et al., 2007), and possibly inducing tau polymerization (King et al., 2000; Ingham et al., 2022), although its effects on tau in vivo are yet to be established.

Interestingly, scRNA seq revealed that LPCAT3 deficiency led to increased de novo synthesis of fatty acids and cholesterol in DAM. Increased lipid synthesis is required for pathogen-stimulated phagocytosis in macrophages (Lee et al., 2018). The induction of fatty acid and cholesterol biosynthesis is likely driven by Srebf1 and Srebf2, two master regulators of lipid synthesis (Horton et al., 2002). It remains unclear how ARA and lipid peroxidation impact Srebf1/2 activities. Oxidative stress can lead to the oxidation of specific cysteine residues on critical proteins involved in the SREBP activation pathway. One such protein is Insig (insulin-induced gene), which plays a crucial role in regulating SREBP processing and activation. Oxidation of specific cysteine residues on Insig may disrupt its interaction with SREBP and prevent the proteolytic cleavage and activation of SREBPs (Zhou et al., 2020).

LPCAT3 has recently emerged as a critical regulator of ferroptosis, a unique form of regulated cell death characterized by iron-dependent lipid peroxidation (Li and Li, 2020; Reed et al., 2022). Ferroptosis is initiated by the accumulation of lipid hydroperoxides, resulting from the uncontrolled peroxidation of PUFAs within cellular membranes (Dixon et al., 2012; Xie et al., 2016). LPCAT3 plays a pivotal role in this process by controlling the PUFA composition in membranes, thereby influencing cell susceptibility to lipid peroxidation and subsequent ferroptotic death. Research has shown that cells deficient in LPCAT3 exhibit reduced levels of PUFA-containing PLs, making them less prone to ferroptosis (Ichu et al., 2020; Reed et al., 2022). Conversely, elevated LPCAT3 expression increases cell susceptibility to ferroptosis through the buildup of PUFA-rich PLs (Kagan et al., 2017). Consistent with previous research, our results showed that LPCAT3 impacts lipid peroxidation and oxidative stress in microglia. Specifically in the context of AD, where disrupted lipid metabolism and heightened oxidative stress are prominent features, altered ARA metabolism might increase the vulnerability of microglia and neuronal cells to ferroptosis. Therefore, further investigations are needed to elucidate whether and how LPCAT3 modulates ferroptosis in AD pathogenesis. This could lead to new therapeutic strategies targeting LPCAT3-mediated lipid metabolism to alleviate or prevent the progression of AD.