Lipidomics reveals TREM2-associated dysregulation of plasma lipid metabolism in Alzheimer's disease

Abstract

Background

Triggering receptor expressed on myeloid cells 2 (TREM2) is a genetic risk factor for Alzheimer's disease (AD). While TREM2 facilitates central nervous system lipid clearance, its influence on peripheral lipid metabolism remains unclear.

Objective

To investigate the association between plasma sTREM2 and peripheral lipid profiles in AD and to explore the mechanistic role of TREM2 in peripheral lipid regulation.

Methods

We conducted a cross-sectional study of 59 AD patients and 54 healthy controls and measured plasma biomarkers including sTREM2 as well as performed targeted lipidomics profiling. Mechanistic exploration was performed via plasma and hippocampal lipidomics in Trem2 knockout and APP/PS1 mice.

Results

Plasma sTREM2 levels were elevated in AD and were negatively correlated with the plasma p-tau217/Aβ₄₂ ratio and p-tau217. Multivariate analysis revealed a distinct lipidomics signature in AD, in which 30 lipid species were significantly altered. We prioritized significantly altered biomarkers to inform a composite biomarker panel combining sTREM2 with a set of sphingomyelins, phosphatidylinositols, diacylglycerols, fatty acids, and cholesteryl esters, which showed strong discrimination between AD and controls (AUC = 0.93). In a mouse model of APP/PS1, we found that Trem2 knockout partially normalized plasma sphingomyelins and hexosylceramide levels. Finally, cross-tissue comparisons further suggested that TREM2 exerted distinct effects on peripheral sphingolipid metabolism that were less evident in hippocampal tissue.

Conclusions

Our findings associate TREM2 with lipid dysregulation in AD and support development of a plasma sTREM2–lipid panel for patient classification.

Keywords

Alzheimer's disease lipidomics plasma sphingolipids TREM2

Introduction

Alzheimer's disease (AD) accounts for 60–70% of dementia cases and represents a growing global health crisis, with over 55 million affected individuals worldwide.¹ Clinically, AD patients present with progressive memory loss, cognitive impairment, and behavioral changes.² Pathologically, AD is characterized by three pillars: (1) extracellular deposition of amyloid-β (Aβ), which forms neuroinflammatory plaques; (2) intracellular hyperphosphorylation of tau protein, which leads to neurofibrillary tangles accompanied by synaptic loss and neurodegeneration³; and (3) dysregulation of lipid metabolism, which interacts closely with amyloid and tau pathologies.⁴

The contribution of lipid dysregulation to AD pathogenesis has become increasingly appreciated in recent years. Genome-wide association studies have identified lipid metabolism genes such as APOE and ABCA7 as prominent AD risk factors.⁵ However, the upstream genetic regulators of lipid perturbations and their mechanistic connections to AD pathologies, such as amyloid precursor protein processing, remain poorly understood. Furthermore, validation of these findings in large, independent cohorts is essential to direct clinical translation. In addition, multiple studies implicate lipid metabolic reprogramming as both a driver and a consequence of AD pathogenesis.⁶ For example, membrane phospholipid catabolism generates bioactive metabolites that modulate Aβ aggregation kinetics and promote tau hyperphosphorylation, while sphingolipid imbalance amplifies neuroinflammation via ceramide-mediated apoptotic signaling. Dysregulated lipid metabolism, particularly that of phospholipids and sphingolipids, has been documented in both central nervous system (CNS) tissues and peripheral biofluids during early stages of AD,^7–9 positioning lipidomics profiling as a useful approach for dissecting disease mechanisms and identifying stage-specific biomarkers.¹⁰

TREM2 is a lipid-sensing immunoreceptor expressed on microglia in the CNS and on macrophages in the periphery and altered TREM2 is a recognized genetic risk factor for AD.¹¹ As a lipid-binding receptor in the CNS, TREM2 recognizes phospholipids and sulfatides—including those exposed on apoptotic cells—thereby coupling lipid cues to microglial responses.¹² In its membrane-bound form, TREM2 signals through the DAP12–SYK axis to enhance phagocytic clearance of lipid-rich debris and myelin by microglia. Its soluble ectodomain (sTREM2), generated by ADAM10/17-mediated shedding, supports microglial survival, proliferation, migration, and phagocytosis.^13,14 TREM2 also modulates cellular lipid handling, coordinating programs linked to cholesterol transport and lipid droplet metabolism during chronic myelin phagocytosis.¹⁵

By contrast, TREM2's contribution to systemic lipid metabolism in the periphery remains poorly defined. It is also unclear how peripheral sTREM2 may relate to brain pathology. Current evidence mainly derives from animal models and in vitro studies, leaving potential correlations between plasma lipid levels and TREM2 across different AD stages and APOE genotypes largely unexplored.^16,17 Therefore, an examination of the relationship between plasma lipid metabolism and TREM2 in AD patients could unravel pathogenic pathways and accelerate the development of mechanism-based diagnostics and targeted interventions.

The present study aimed to (1) assess the potential association between plasma sTREM2 levels and primary AD pathological features in a cross-sectional cohort; (2) characterize alterations in plasma lipidomics in AD using targeted LC-MS/MS; (3) evaluate the diagnostic potential of sTREM2 and plasma lipid biomarkers; and (4) experimentally validate the role of TREM2 in lipid metabolism in a Trem2-knockout (KO) APP/PS1 mouse model of AD. We found that plasma sTREM2 was correlated with AD pathology and, when combined with lipid classes, achieved high discriminatory power. In plasma, lipid alterations arising from TREM2 deficiency were primarily related to sphingolipid and glycerophospholipid metabolism. These findings demonstrate associations between TREM2 and lipid dysregulation in AD and underscore the value of peripheral lipid biomarkers.

Methods

Study participants

This study enrolled a total of 113 participants between January 2021 and June 2022, comprising 59 AD patients and 54 healthy controls (HC). AD participants were recruited from the Xuanwu Hospital, Capital Medical University, and met the diagnostic criteria established by the National Institute on Aging–Alzheimer's Association (NIA–AA), including biological classification based on the ATN framework (amyloid (A), tau (T), and neurodegeneration (N)).¹⁸ All AD diagnoses were confirmed through positive amyloid PET imaging and/or reduced cerebrospinal fluid (CSF) Aβ₄₂ levels, ensuring amyloid positivity in accordance with research diagnostic standards. HCs were community-dwelling individuals recruited from nearby communities in the Baizhifang District. To be included, HCs were determined to have normal global cognition, defined as a Clinical Dementia Rating (CDR) score of 0 and a Montreal Cognitive Assessment (MoCA) score ≥26 or Mini-Mental State Examination (MMSE) score between 24 and 30, with no subjective memory complaints or history of cognitive decline. Exclusion criteria for both groups included: (1) any current major psychiatric disorder or a major psychiatric disorder diagnosed within the past two years, (2) history of traumatic brain injury, (3) structural brain lesions, (4) active systemic inflammatory or autoimmune conditions, (5) chronic corticosteroid or immunosuppressive medication use, and (6) genetic mutations associated with early-onset familial dementia. All procedures followed the principles of the Declaration of Helsinki and received approval from the Xuanwu Hospital (Capital Medical University) ethics committee (Ethics Approval Number: (2020) 097). Written informed consent was obtained from each participant prior to inclusion.

Animals

Trem2-KO (C57BL/6JSmoc-Trem2^em1Smoc) mice were generated by Shanghai Model Organisms Center using the CRISPR-Cas9 system. APP/PS1 mice (B6/J-Tg (APPswe, PSEN1dE9)) were purchased from Nanjing Junke Bioengineering Co. APP/PS1-Trem2-KO mice were generated by crossing APP/PS1 heterozygous mice (APP/PS1^+/–) with Trem2-KO mice (Trem2^+/–), followed by interbreeding of APP/PS1^+/– Trem2^+/– offspring (Supplemental Figure 1). All experimental animals were on a C57BL/6 genetic background. Unless otherwise noted, we used male mice 10 months of age (20–30 g). Mice were housed in SPF facilities at the Xuanwu Hospital Animal Experiment Center with unrestricted access to chow and water, a 12-h light/12-h dark schedule, and ambient temperature of 22°C. Experiments followed the NIH Guide for the Care and Use of Laboratory Animals and were approved by the Xuanwu Hospital Institutional Animal Care and Use Committee (Ethics Approval Number: 1123031700237).

Sample preparation

Venous blood samples were collected once from human participants via peripheral venipuncture. All blood samples were collected using sodium heparin-coated vacutainer tubes. Immediately after collection, tubes were gently inverted 5–10 times to ensure proper anticoagulant mixing. Samples were processed within 30 min or temporarily stored at 4 °C (not exceeding 4 h) prior to centrifugation. Plasma was separated by centrifugation at 3000 rpm for 10–15 min at 4 °C. The supernatant was carefully aliquoted into pre-labeled microcentrifuge tubes to avoid multiple freeze–thaw cycles and stored at −80°C until further analysis. CSF was collected using a drip method, with CSF directly dripping from the puncture needle into low-adsorption polypropylene (PP) collection tubes to minimize protein loss. After collection, fluid samples were centrifuged at 3000 rpm for 15 min at 4°C, and resulting aliquots were immediately frozen at −80°C until further analysis.

For animal experiments, 10-month-old mice were anesthetized with an intraperitoneal injection of Avertin (22,2-tribromoethanol; 250 mg/kg body weight). Adequate depth of anesthesia was confirmed by the absence of pedal withdrawal and corneal reflexes before blood collection. While under deep anesthesia, blood was collected from the retro-orbital venous plexus using a capillary tube and transferred into sodium heparin-coated vacutainer tubes. Following blood collection, animals were euthanized by cervical dislocation while still under deep anesthesia, and brains were rapidly removed. Hippocampi from both hemispheres were dissected immediately on an ice-cold surface in HBSS, blotted dry, snap-frozen in liquid nitrogen, and stored at −80°C until analysis.

Measurement of plasma biomarkers

Plasma sTREM2 levels were measured using commercial ELISA kits from Abcam (ab224881) by trained personnel in accordance with the manufacturer's instructions. All samples were analyzed in duplicate under blinded conditions. The intra-assay coefficient of variation for sTREM2 was 1.5%. Plasma levels of Aβ₄₂, Aβ₄₀, p-tau217, total tau (T-tau), and APOE4 protein were measured using the fully automated Lumipulse G600II platform (Fujirebio Europe NV, Ghent, Belgium), following the manufacturer's protocols as previously described.¹⁹

Targeted lipidomics of plasma samples

Plasma for LC–MS/MS-based targeted lipidomics was snap-frozen in liquid nitrogen immediately after isolation and shipped on dry ice to APTBIO Co., Ltd (Shanghai, China) for analysis. Lipids were extracted using a methyl tert-butyl ether (MTBE) protocol. Briefly, 200 µL methanol and 10 µL of an internal standard mixture were added to each sample, followed by 800 µL MTBE. Tubes were vortexed vigorously, ultrasonicated for 20 min at 4°C, then incubated 30 min at room temperature. Phase separation was induced by adding 200 µL LC–MS–grade water; samples were vortexed and centrifuged (14,000 rpm, 15 min, 4°C). The upper organic layer was collected, dried under a nitrogen stream, and reconstituted in 200 µL isopropanol/acetonitrile (9:1, v/v). After a clarifying spin (14,000 rpm, 15 min, 4°C), the supernatant was used for LC–MS analysis. Chromatography was performed on a Shimadzu LC-30AD UHPLC, and detection on a SCIEX 6500+ QTRAP mass spectrometer.

Chromatography conditions

For chromatography, a C18 column was maintained at 45°C. Solvent A was acetonitrile/water (70:30, v/v) with 5 mM ammonium acetate; solvent B was isopropanol. At a flow of 0.35 mL min^–1, the gradient started at 20% B (0 min), reached 60% B (5 min), progressed to 100% B (13 min), and then was returned to 20% B from 13.1–17 min for re-equilibration. Samples were kept at 10°C during analysis.

For Amino separation, the column was held at 40°C. Solvent A: 2 mM ammonium acetate in methanol/acetonitrile (1:1). Solvent B: 2 mM ammonium acetate in acetonitrile/water (1:1). The program was 3% B for 0–3 min, a linear ramp to 100% B over 3–13 min, a 100% B hold from 13–17 min, and re-equilibration to 3% B from 17.1–22 min, at 0.40 mL/min.

Mass spectrometry

Analyses were acquired on a QTRAP 6500 + mass spectrometer (AB SCIEX) with rapid polarity switching. For ESI in positive mode, source settings were: 400°C temperature, GS1 = 50, GS2 = 55, CUR = 35, and IS = + 3000 V. For ESI in negative mode, the source was held at 400°C with GS1 = 50, GS2 = 55, CUR = 35, and IS = −2500 V. Quantification used multiple reaction monitoring (MRM). Pooled QC samples were injected throughout the run to track instrument stability, and data were processed in Sciex OS. QC and biological specimens were handled in parallel; metabolites showing CV < 30% in QC were considered reproducibly measured.

Data analysis

Categorical variables were presented as frequencies (percentages), and group comparisons were performed using the chi-square test. Continuous variables with normal distribution were expressed as mean ± standard deviation (SD), and comparisons between groups were conducted using Student's t-test. For non-normally distributed continuous variables, data were described as median and interquartile range (M (P25, P75)), and group comparisons were carried out using the Mann–Whitney U test. Differences in lipid abundances were tested using analysis of covariance (ANCOVA), adjusted for age, sex, education level, and body mass index (BMI). To control for multiple comparisons, p-values were adjusted using the Benjamini–Hochberg false discovery rate (FDR) method, and lipids with FDR-adjusted q < 0.05 were considered statistically significant. Pearson correlation was employed for normally distributed data, while Spearman rank correlation was used for non-normally distributed data to calculate correlations between blood markers and sTREM2. Correlations among biomarker concentrations, lipids, and neuropsychological scores were assessed via partial correlation analysis, adjusted for age, sex, education level, and BMI.

Receiver-operating characteristic (ROC) curves were generated to determine the diagnostic accuracy of blood markers. Area under the ROC curve (AUC) values were computed by binary logistic regression, and cut-off values that maximized Youden indices were determined to assess sensitivity and specificity. All analyses were performed using SPSS Statistics software (IBM, version 25). Lipidomics data were processed with MetaboAnalyst 6.0. Features with >50% missing values across all samples were removed. For the remaining features, missing values were imputed by k-nearest neighbors. The data were analyzed using SIMCA-P 14.1 for multivariate analyses, including principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and orthogonal partial least squares discriminant analysis (OPLS-DA). Pathway and Enrichment analysis functions were performed to integrate pathway enrichment analysis and pathway topology analysis, starting from significant lipids identified by statistical analyses.

Results

Demographic and clinical characteristics

Demographic and clinical characteristics of the participants are presented in Table 1. The cohort included 59 AD patients and 54 HCs. There were no significant differences between groups in terms of age, sex distribution, or years of education (p > 0.05). The BMI value was significantly lower in the AD group compared with the HC group (23 [21–25] versus 24.9 [22.8–26.9], p = 0.021). No significant differences were detected between groups for the prevalence of hypertension, diabetes, coronary heart disease, stroke, or aspirin use (p > 0.05), nor were there significant differences for plasma lipid parameters, including high-density lipoprotein (HDL), low-density lipoprotein (LDL), total cholesterol (TC), or triglycerides (TG). Both the MMSE and MoCA scores were significantly reduced in the AD group compared to HCs (MMSE: 18 [14–21] versus 28 [28–29]; MoCA: 13 [8–17] versus 26 [24–27]; p < 0.001), indicating impaired global cognition and consistent with the clinical diagnosis of AD.

Table 1.

Demographics.

Variable	HC (n = 54)	AD (n = 59)	t/z/χ²	p
Age (years)	66.24 ± 6.86	63.90 ± 8.68	1.599	0.113
Sex (male/female)	20/34	24/35	0.157	0.692
Education (years)	12(9,15)	12(7,14)	−1.884	0.06
BMI (kg/m²)	24.9(22.8,26.9)	23(21,25)	−2.306	0.021
Hypertension, n (%)	23(42.59%)	19(32.20%)	1.303	0.254
Diabetes, n (%)	7(12.96%)	5(8.85%)	0.598	0.439
CHD, n (%)	3(5.56%)	4(6.78%)	0.073	1
Stroke, n (%)	4(7.41%)	2(3.39%)	0.905	0.423
Aspirin intake, n (%)	10(18.52%)	6(10.17%)	1.526	0.217
LDL (mmol/L)	2.75(1.95,3.74)	2.80(2.01,3.39)	−0.543	0.587
HDL (mmol/L)	1.41(1.25,1.69)	1.25(1.17,1.58)	−1.663	0.096
TC (mmol/L)	4.98 ± 0.94	4.64 ± 1.10	1.745	0.086
TG (mmol/L)	1.54(1.07,2.17)	1.21(0.85,1.88)	−1.881	0.06
MMSE (score)	28(28,29)	18(14,21)	−9.099	＜0.001
MoCA (score)	26(24,27)	13(8,17)	−8.973	＜0.001

Categorical variables are presented as frequency (percentage), and comparisons between groups were performed using the chi-square test. Continuous variables with a normal distribution are expressed as mean ± standard deviation (SD), with comparisons between groups conducted using Student's t test. For continuous variables not following a normal distribution, data are described as median and interquartile range (M (P25, P75)), and comparisons between groups were carried out using the Mann–Whitney U test.

AD: Alzheimer's disease; BMI: body mass index; CHD: coronary heart disease; HDL: high-density lipoprotein; LDL: low-density lipoprotein; MMSE: Mini-Mental State Examination; MoCA: Montreal Cognitive Assessment; HC: healthy controls; TC: total cholesterol; TG: triglyceride.

Plasma sTREM2 is elevated in AD patients and associated with AD biomarkers and circulating lipids

We assessed the potential relationship between plasma sTREM2 and AD pathological features in the cohort. Plasma sTREM2 levels were significantly higher in the AD group than in the HC group (p = 0.0256; Figure 1A). Within the AD cohort, plasma sTREM2 levels were positively correlated with MMSE scores (r = 0.4421, 95% CI 0.04–0.72, p = 0.0346; Figure 1B). We detected negative correlations for plasma sTREM2 levels with AD pathological markers, including plasma p-tau217 (r = −0.4330, 95% CI −0.72– −0.02, p = 0.0345; Figure 1C) and the p-tau217/Aβ₄₂ ratio (r = −0.5313, p = 0.0075, 95% CI −0.78– −0.158; Figure 1D), both quantified using the Lumipulse^® platform in AD patients. sTREM2 levels positively correlated with several lipid-related markers, including TC, TG, LDL, and apolipoprotein B (APOB) (TC, r = 0.450, p = 0.028; TG, r = 0.450, p = 0.027; LDL, r = 0.423, p = 0.039; ApoB, r = 0.437, p = 0.033; Figure 1E). There was no apparently informative trend between plasma sTREM2 levels and HDL or apolipoprotein A (APOA).

Figure 1.

Correlation between plasma sTREM2, AD biomarkers, and circulating lipids. (A) Plasma sTREM2 concentrations were higher in patients with AD compared with HCs (p = 0.0256). (B) Correlation between plasma sTREM2 and MMSE scores within the AD cohort (r = 0.4421, 95% CI 0.04–0.72, p = 0.0346). (C, D) Spearman correlation showed that plasma sTREM2 was negatively correlated with plasma p-tau217 levels and p-tau217/Aβ₄₂ ratio, both measured by Lumipulse (r = −0.4330, 95%CI −0.72− −0.02, p = 0.0345; r = −0.5313, 95% CI −0.78− −0.158, p = 0.0075). (E) Heatmap showing correlation between plasma sTREM2 and lipid profiles, including TC, TG, HDL, LDL, APOA, and APOB. All correlations were adjusted for age, sex, BMI, and education. p-tau217: phosphorylated tau at threonine 217; Aβ₄₂: amyloid beta 42; TC: total cholesterol; TG: triglycerides; HDL: high-density lipoprotein; LDL: low-density lipoprotein; APOA: apolipoprotein A; APOB: apolipoprotein B. *p < 0.05, **p < 0.01, ****p < 0.0001. (Color figure available online).

AD patients exhibit a distinct plasma lipidomics signature

We employed targeted lipidomics of plasma samples from AD patients and HCs as an initial evaluation of lipid contents. Briefly, an OPLS-DA indicated clear separation between the two groups (Figure 2A). Permutation testing (n = 200) supported the robustness of the OPLS-DA model (Figure 2B; Supplemental Table 1). After adjusting for age, sex, education level, and BMI, 30 lipid species were identified as significantly different between the AD and HC groups based on the following criteria: VIP > 1, FC > 2 or < 0.5, and FDR-adjusted q < 0.05 (Benjamini–Hochberg correction), all of which were decreased (Table 2).

Figure 2.

Distinct plasma lipidomics signature identified in AD samples. (A) OPLS-DA score plot of AD patients versus HCs. Dots of the same color represent various biological replicates within the group. The distribution of dots reflects the degree of difference between and within groups. (B) Permutation test of the OPLS-DA model for AD patients versus HCs. (C) Volcano plots of the FC (x-axis) and −log10(adjusted p-value) (y-axis) for each detected lipid in the comparison of AD patients versus HCs. Red dots (right side) represent significantly upregulated (FC > 1.5) molecules in AD patients; blue dots (left side) represent downregulated (FC < 0.67) molecules in AD patients. (D) Bar plot showing VIP-ranked discriminatory lipid species (VIP > 1, p < 0.05, FC > 2 or FC < 0.5). VIP, Variable Importance in Projection—a unitless measure of each variable's contribution to the OPLS-DA model (larger values indicate greater influence). (E) Correlation between lipid species and AD biomarkers. Warm colors (red) indicate positive correlations, and cool colors (blue) indicate negative correlations. Data were adjusted for age, sex, education level, and BMI. Statistical significance was determined using FDR-adjusted q < 0.05. (Color figure available online).

Table 2.

Significantly different lipids between the AD and HC groups.

Lipid	Class	RT (min)	Fold change	VIP	q-value
Cer(d18:1/20:0)	Cer	5.55	0.470975	1.565148	0.042911
ChE(16:0)	ChE	9.57	0.465427	1.583547	0.011247
ChE(20:3)	ChE	9.11	0.429463	1.71359	0.037299
DHCer(d18:0/20:0)	DHCer	5.94	0.243458	1.650086	0.023379
DHCer(d18:0/22:0)	DHCer	6.44	0.359323	1.545597	0.008976
DHCer(d18:0/24:0)	DHCer	6.94	0.317073	1.684816	0.003710
Hex2Cer(d18:1/22:0)	Hex2Cer	5.52	0.306518	1.665563	0.003697
Hex2Cer(d18:1/24:0)	Hex2Cer	6.02	0.398053	1.641365	0.001524
LPC(22:4)	LPC	6.77	0.488634	1.599927	0.001346
LPE(18:0)	LPE	8.09	0.433971	1.967004	0.000064
LPE(20:0)	LPE	8.03	0.202577	1.950793	0.000009
LPE(20:1)	LPE	8.03	0.330098	1.861578	0.000063
LPE(20:2)	LPE	8.03	0.470614	1.243554	0.001282
LPE(22:4)	LPE	8.06	0.244555	2.040135	0.000053
LPE(22:5)	LPE	8.06	0.489569	1.785856	0.002155
PA(16:0/14:0)	PA	5.9	0.406719	1.807698	0.004981
PA(16:0/16:0)	PA	5.9	0.451568	1.737017	0.000196
PE(18:0/22:4)	PE	7.03	0.397589	1.859682	0.000880
PE(18:1/22:4)	PE	7.03	0.396094	1.720978	0.002029
PE(18:2/22:4)	PE	7.03	0.321291	1.691437	0.000343
PE(O-18:0/22:4)	PE	7.15	0.406323	1.612873	0.000725
PE(P-16:0/22:4)	PE	7.15	0.374204	1.685031	0.000449
PE(P-18:0/22:4)	PE	7.15	0.254513	1.85455	0.000046
PI(18:1/22:4)	PI	10.8	0.498913	1.569991	0.000550
PI(18:2/20:3)	PI	10.8	0.411135	1.297794	0.000875
PI(18:2/20:4)	PI	10.8	0.412986	1.391959	0.006141
PS(18:0/18:1)	PS	11.99	0.065294	1.914841	0.000005
PS(18:0/18:2)	PS	11.99	0.201137	1.671522	0.000019
PS(18:1/18:1)	PS	11.99	0.093152	1.820289	0.000018
SM(d18:1/26:0)	SM	6.7	0.484359	1.555723	0.016497

Adjusted for age, sex, education level, and BMI; FDR-corrected q < 0.05.

To visualize informative changes in lipids between groups, volcano plots were generated based on FC and p-values (FC > 1.5 or FC < 0.67) (Figure 2C). An OPLS-DA identified 16 lipid species with high discriminatory power (i.e., VIP score) (Figure 2D), which notably included phospholipids and sphingolipids. These findings collectively define a plasma lipidomics signature in AD patients distinct from HCs.

We subsequently explored correlations between biomarkers or cognitive scores and various lipid types within the AD group. Briefly, LPE (22:4), PE (18:0/22:4), PI (18:1/22:4), and ChE (20:3) lipids were each positively correlated with CSF levels of Aβ₄₀ (r = 0.482, 95% CI 0.22–0.68, p < 0.01; r = 0.348, 95% CI 0.06–0.59, p < 0.05; r = 0.368, 95% CI 0.08–0.60, p < 0.05; r = 0.385, 95% CI 0.10–0.61, p < 0.01). Hex2Cer(d18:1/24:0) was positively correlated with CSF t-tau (r = 0.309, 95% CI 0.006–0.56, p < 0.05), while PS (18:1/18:1) was positively correlated with plasma p-tau217 and with the p-tau217/Aβ₄₂ (r = 0.571, 95% CI 0.10–0.83, p < 0.05). PE (18:1/22:4) showed a positive correlation with plasma APOE4 protein (r = 0.397, 95% CI 0.002–0.68, p < 0.05), and LPE (20:1) exhibited a similar association with plasma APOE4 protein (r = 0.421, 95% CI 0.03–0.70, p < 0.05). Conversely, PI (18:2/20:3) was negatively correlated with MMSE (r = −0.360, 95% CI −0.59– −0.05, p < 0.05) and MoCA (r = −0.44, 95% CI −0.66– −0.16, p < 0.01) scores (Figure 2E). Collectively, these lipid profile data reveal informative alterations in lipid metabolism in the AD group—particularly in phospholipids—that are associated with AD biomarkers and cognitive scores, indicating a potential role for peripheral lipid dysregulation in AD pathogenesis.

sTREM2 and plasma lipids can discriminate AD from HC

Given the aforementioned sTREM2 data and our plasma lipidomics signature findings for AD patients, we speculated that it may be possible to reliably distinguish AD patients from HCs on the basis of a panel of analytes in plasma samples. Pursuing this goal, we first examined correlations between plasma sTREM2 levels in AD patients and lipids. This showed that plasma sTREM2 was positively correlated with the levels for a set of lipid types, including sphingomyelins (SM), phosphatidylinositol (PI), diacylglycerols (DG), fatty acids (FA), and cholesteryl esters (ChE) (SM, r = 0.458, p = 0.025; PI, r = 0.536, p = 0.007; ChE, r = 0.410, p = 0.047; DG, r = 0.475, p = 0.019; FA, r = 0.427, p = 0.037) (Figure 3A).

Figure 3.

Correlation between plasma sTREM2 and lipids and diagnostic performance of a combined panel. (A) Heatmap showing the correlation between plasma sTREM2 and lipid classes, including SM, PI, ChE, DG, and FA. Warm colors (red) indicate positive correlations, and cool colors (blue) represent negative correlations. (B-C) ROC analysis for the diagnostic performance of sTREM2 and lipid classes (SM, PI, sTREM2, DG, FA, ChE) in distinguishing AD patients from HCs (SM, AUC = 0.899; PI, AUC = 0.855; sTREM2, AUC = 0.804; ChE, AUC = 0.784; DG, AUC=0.701; FA, AUC = 0.635; combined panel, AUC = 0.93). ROC: receiver-operating characteristic; AUC: area under the ROC curve. *p < 0.05, **p < 0.01, ****p < 0.0001. (Color figure available online).

Given these correlations, we next evaluated the individual performance of each for discriminating AD patients from HCs (Figure 3B) using ROC analysis, which generated the following AUCs: SM 0.8987 (95% CI 0.83–0.97; p < 0.001), PI 0.8547 (95% CI 0.79–0.92; p < 0.001), sTREM2 0.8043 (95% CI 0.69–0.91; p < 0.001), ChE 0.7837 (95% CI 0.70–0.87; p < 0.001), DG 0.7006 (95% CI 0.60–0.80; p < 0.001), and FA 0.6350 (95% CI 0.53–0.74; p = 0.0134). We found that a panel comprising the sTREM2 data with data for these 5 lipid types demonstrated excellent performance for distinguishing AD patients from HC (AUC = 0.93, 95% CI = 0.85–1, p < 0.001), exceeding all individual biomarkers for disease prediction (Figure 3B, C), suggesting that plasma-based measurement of these analytes may provide a practical, minimally invasive approach to facilitate diagnostic workflows.

Trem2 KO increases plasma SM and Hex1Cer in APP/PS1 mice

Previous studies reported that microglial TREM2 promotes cholesterol-ester efflux, and we observed elevated plasma sTREM2 in AD patients, which correlated with specific lipid classes, including SM, PI, DG, FA, and ChE. Therefore, we next examined whether TREM2 contributes to the peripheral lipid profiles by performing plasma lipidomics in 10-month wild type (WT) mice and in the APP/PS1 AD mouse model. Consistent with our findings in human samples, compared with WT, APP/PS1 mice had significantly reduced plasma levels of multiple sphingolipids and phospholipids, including SM, Hex1Cer, Hex2Cer, phosphatidic acid (PA), and ChE (Figure 4A).

Figure 4.

Plasma lipid profiles of WT, APP/PS1, and APP/PS1-Trem2-KO mice. (A) Heat map and trend analyses showing lipid profiles of WT, APP/PS1, and APP/PS1-Trem2-KO mice. Warm colors indicate higher relative abundance, cool colors lower. (B) Volcano plot of differentially expressed plasma lipids in APP/PS1-Trem2-KO versus APP/PS1 (FC > 1.2 or FC < 0.83). Red indicates the upregulation of lipid molecules in the APP/PS1-Trem2-KO group, blue indicates downregulation, and gray represents no significant difference. (C–E) Quantification of individual Cer, SM, and Hex1Cer species significantly altered in plasma (FC > 2, p < 0.05). Statistical analysis was performed using an unpaired t-test (n = 4−5 per group). Data are presented as mean ± SEM. *p < 0.05, **p < 0.01, ****p < 0.0001. (Color figure available online).

We next generated Trem2-KO mice in the APP/PS1 background by crossing Trem2-KO mice with APP/PS1 and repeated the experiment (Supplemental Figure 1). Compared with APP/PS1 mice, APP/PS1-Trem2-KO mice had increased SM and Hex1Cer levels and reduced phosphatidylserine (PS) levels (Supplemental Table 2, Figure 4A). To exclude the possibility that the observed alterations in SM, Hex1Cer, and PS were caused by Trem2 deficiency itself, we also performed plasma lipidomics in Trem2-KO mice on a wild-type background. No significant differences in SM, Hex1Cer, or PS levels were detected between Trem2-KO and WT mice (Supplemental Figure 2), indicating that Trem2 primarily affects these lipid classes in the context of amyloid pathology rather than under basal physiological conditions.

To identify which lipid species accounted for these lipid class changes, we generated volcano plots, which identified 74 lipid species that differed between APP/PS1 and APP/PS1-Trem2-KO animals, including 46 species that decreased and 28 species that were increased in the Trem2-KO group (FC > 1.2 or FC < 0.83; Supplemental Table 3). Notably, a substantial proportion of the upregulated lipids belonged to the sphingolipid class, including seven species—SM (d18:1/20:1), Cer (d18:1/20:0), Cer (d18:1/22:1), Hex1Cer (d18:1/20:0), Hex1Cer (d18:1/22:0), Hex1Cer (d18:1/22:1), and Hex1Cer (d18:0/20:0)—each increased by more than twofold (p < 0.05; Figure 4B–E). These observations suggest that TREM2 deficiency preferentially affects plasma sphingolipid metabolism.

TREM2 differentially regulates lipid metabolism in the CNS and periphery

In our study, we observed that lipid classes significantly altered in the plasma of APP/PS1 mice were similarly dysregulated in AD patient plasma. Specifically, lipids such as Hex1Cer, SM, ChE, and PA, which were reduced in APP/PS1 mouse plasma (Figure 4A), also exhibited reductions in AD patients, suggesting shared patterns of lipid dysregulation between APP/PS1 mice and AD patients (Figure 5A). Notably, Trem2 KO in APP/PS1 mice significantly perturbed lipid classes including Hex1Cer, SM, and PS, which are also significantly altered in AD patients (Figure 5A). We then compared the lipid species significantly altered in human AD plasma and APP/PS1-Trem2-KO mouse plasma, identifying four overlapping lipids: Cer (d18:1/20:0), LPE (20:2), PE (P-16:0/22:4), and PI (18:2/20:3) (Figure 5B, C). KEGG pathway enrichment analysis revealed that glycerophospholipid metabolism and sphingolipid metabolism pathways were significantly enriched in human AD plasma (Figure 5D). A similar pattern was observed in APP/PS1-Trem2-KO mouse plasma, where these pathways were also notably enriched (Figure 5E; Supplemental Table 4). However, the degree of enrichment in the APP/PS1-Trem2-KO mouse plasma was less pronounced compared to human AD plasma.

Figure 5.

Cross-species comparison reveals TREM2-associated lipid dysregulation in plasma and brain. (A) Comparison of lipid class content in human plasma from HC and AD patients. (B) Venn diagram illustrating the overlap of lipid species significantly dysregulated in human AD plasma (30 lipids) and mouse plasma from APP/PS1-Trem2-KO mice. (C) Four lipid species (Cer (d18:1/20:0), LPE (20:2), PE (P-16:0/22:4), and PI (18:2/20:3)) were significantly dysregulated in both human AD plasma and APP/PS1-Trem2-KO mouse plasma. (D-E) Bubble plots from KEGG analysis showing significantly enriched pathways in human AD plasma and APP/PS1-Trem2-KO mouse plasma. (F) Volcano plot of hippocampal lipid species (FC > 1.2 or FC < 0.83). Red indicates the upregulation of lipid species in the APP/PS1-Trem2-KO group, blue indicates downregulation, and gray represents no significant difference. (G) Venn diagram showing 74 plasma and 85 brain lipids significantly dysregulated in APP/PS1-Trem2-KO mice. (H) Seven lipids—including PE 20:2/20:3, PI 15:0/18:1, PI 18:1/20:2, Hex1Cer d18:1/22:0, Hex1Cer d18:0/24:0, Cer d18:1/22:1, PE(P-16:0/22:5)—were altered in both plasma and brain (FC > 1.2 or FC < 0.83). (I) KEGG pathway analysis of hippocampus in APP/PS1-Trem2-KO mice. Each bubble represents one KEGG pathway; only the top 20 pathways by p-value are shown. The x-axis is the rich factor, the ratio of significantly changed lipids mapped to a pathway to the total number of lipids annotated in that pathway. Bubble color encodes enrichment significance as −log10(p); warmer colors indicate smaller p values (more significant enrichment). Bubble size is proportional to the number of significantly changed lipids mapped to that pathway. (Color figure available online).

To test whether TREM2-mediated lipid regulation in the CNS is similar to that observed in our plasma profiling experiments, we performed targeted lipidomics on hippocampal tissue collected from 10-month-old APP/PS1 and APP/PS1-Trem2-KO mice. In total, 85 hippocampal lipids differed between the two groups, including 32 species that were increased and 53 species that were decreased (FC > 1.2 or FC < 0.83) in the Trem2-KO group (Figure 5F; Supplemental Table 5). Notably, 57 of these lipid species belonged to the glycerophospholipid class, indicating that phospholipids predominated in the hippocampal lipidomic profile.

Compared with APP/PS1 mice, APP/PS1-Trem2-KO mice have seven species altered in both plasma and hippocampus: PE (20:2/20:3), PI (15:0/18:1), PI (18:1/20:2), Hex1Cer (d18:1/22:0), Hex1Cer (d18:0/24:0), Cer (d18:1/22:1), and PE (P-16:0/22:5) indicating partial overlap between central and peripheral lipid remodeling (Figure 5G, H). KEGG pathway analysis indicated that lipids involved in glycerophospholipid metabolism were enriched in the hippocampus of Trem2-KO mice (Figure 5I; Supplemental Table 6). Together, these findings suggest that TREM2 deficiency disrupts membrane phospholipid remodeling in both the CNS and periphery, while additionally perturbing sphingolipid metabolism in the periphery, reflecting a broader impact of TREM2 on systemic lipid homeostasis.

Discussion

Altered lipid metabolism is a hallmark of AD, but the precise mechanisms through which lipid dysregulation contributes to AD pathogenesis are still being defined. Moreover, little is known about whether peripheral lipid profiles may be associated with the development or progression of AD. In this study, we examined plasma lipid profiles as well as concentrations of sTREM2 in a clinical cohort of AD patients and HCs. To provide mechanistic context, we further performed lipidomic analyses in 10-month-old APP/PS1 and APP/PS1-Trem2-KO mice. At this age, APP/PS1 mice typically exhibit substantial Aβ plaque burden in the cortex and hippocampus, pronounced glial activation, and reported deficits in learning and memory, together with synaptic dysfunction and neuronal loss.⁹ This pathological context broadly corresponds to the symptomatic stage of our amyloid-positive AD cohort and therefore strengthens the rationale for cross-species comparison of peripheral lipidomic signatures. We found that sTREM2 levels were higher in AD patients and correlated with biomarkers indicative of lower disease burden. In AD patients, sTREM2 correlated positively with several lipid classes, including sphingolipid- and glycerophospholipid-related classes. Finally, we demonstrated that a panel combining plasma sTREM2 with five lipid classes distinguished AD from controls. Our findings describe a novel role of TREM2 in regulating specific lipid classes in the periphery and indicate that TREM2 and plasma lipid profiles warrant further evaluation as possible blood-based biomarkers for AD.

sTREM2 and Alzheimer's disease

Mutations in TREM2 are an established risk factor for AD, and its soluble ectodomain (sTREM2) is detectable in CSF and blood.²⁰ Although CSF sTREM2 has been widely studied as a proxy for microglial activation and neuroinflammation in AD,¹⁴ data on plasma sTREM2 remain scarce. Our findings demonstrate that plasma sTREM2 levels are significantly elevated in AD patients. Within the AD cohort, higher plasma sTREM2 levels were positively correlated with MMSE scores and negatively correlated with plasma p-tau217 levels and the p-tau217/Aβ₄₂ ratio. Together, these findings suggest that although plasma sTREM2 is increased at the disease stage, relatively higher sTREM2 levels within symptomatic AD may be associated with a lower pathological burden and better-preserved cognitive function. These observations are in line with previous reports showing that plasma sTREM2 levels inversely correlate with CSF tau concentrations in amyloid-positive individuals.²¹

Studies using tau-PET imaging revealed that higher sTREM2 levels are associated with reduced cerebral tau burden: elevated plasma sTREM2 attenuated the positive association between Aβ-PET positivity and increased tau deposition,²² implying that peripheral sTREM2 may modulate or suppress tau aggregation and exert a neuroprotective effect. Moreover, loss of membrane TREM2 on monocytes—due to its cleavage into sTREM2—impairs Aβ₄₂ phagocytosis in AD patients.²³ Notably, longitudinal analyses also suggest that both CSF and plasma sTREM2 levels can predict cognitive decline and the transition from MCI to AD,^24,25 reinforcing its potential as a dynamic biomarker across disease stages. Taken together, our results support the hypothesis that elevated plasma sTREM2 reflects a microglial-driven compensatory response to accumulating tau pathology and cognitive deterioration in AD, further underscoring the diagnostic and prognostic utility of this soluble immune receptor in clinical settings.

Glycerophospholipids and Alzheimer's disease

Lipid dysregulation is increasingly implicated in AD pathophysiology, alongside amyloid and tau. In our cohort, targeted LC–MS/MS showed altered phospholipids in plasma—most notably changes in PE and LPE species—and these lipids associated with core AD biomarkers. These results indicate that phospholipid remodeling tracks with biomarker burden in vivo. Phospholipid changes in AD patients are primarily driven by oxidative stress, altered enzyme activity, and interactions with Aβ peptides.^26,27 These alterations affect both the structure and abundance of phospholipids in brain and blood, with potential consequences for membrane stability and synaptic integrity.²⁸ High LPE/PE ratios and increased LPE are associated with early AD pathology and may reflect increased phospholipase A₂ activity and inflammation.²⁹ Prior work provides mechanistic context: oxidative stress and phospholipase A₂ activity promote peroxidation and cleavage of membrane glycerophospholipids, selective depletion of plasmalogens, and accumulation of free polyunsaturated fatty acids; these processes can influence Aβ aggregation kinetics and tau phosphorylation, and decreased PE could reduce membrane fluidity, affecting the function of membrane proteins—including TREM2—whose lipid microenvironment is important for efficient Aβ clearance.^30–33 PI and its phosphorylated derivatives (phosphoinositides) are critical secondary messengers in neurons and glia. In AD, abnormal protein deposits such as Aβ and tau disrupt PI signaling, particularly in microglia, leading to impaired actin cytoskeleton remodeling and defective phagocytosis of pathological proteins. This disruption hampers microglial clearance of Aβ and contributes to neuroinflammation and disease progression.³⁴ In addition, we compared the targeted lipidomics results between human AD plasma and APP/PS1-Trem2-KO mouse plasma, identifying four overlapping lipid species, three of which are glycerophospholipids. Notably, PI (18:2/20:3) was significantly altered in both species and was found to be negatively correlated with MMSE and MoCA scores. Multiple studies have demonstrated that TREM2 activates PI3K–AKT and PLCγ2 signaling pathways, both of which depend on PI availability and turnover. Disruption of this axis may impair cytoskeletal remodeling, phagocytosis, and lipid droplet processing in microglia, thereby linking peripheral PI dysregulation to central immune dysfunction in AD.^35,36 This highlights the potential role of TREM2 in modulating PI signaling and suggests that disruptions in PI metabolism may contribute to AD pathology, particularly in relation to neuroinflammation and impaired microglial function.

Sphingolipids and Alzheimer's disease

In addition to glycerophospholipid alterations, we identified changes in sphingolipids between AD patients and controls, including a decrease in dihydroxy ceramide (DHCer) species and hexosyl ceramide (Hex2Cer) with fatty-acid chains longer than 20 carbons in AD patients. Notably, Hex2Cer(d18:1/24:0) showed a positive correlation with CSF t-tau, whereas most monohydroxy and trihydroxy ceramides remained unchanged, which might suggest opposing pathobiological effects of different Cer subgroups. Independent studies echo these observations. Han et al. reported that 8 of 33 SM species—particularly those with C22–C24 acyl chains—were diminished in AD plasma.³⁷ In the Mayo Clinic Study of Aging (n = 1255), network analysis identified 51 lipid correlations that shifted with cognitive decline and amyloid deposition, with DHCer/C16:0 and PC/C40:5 most closely linked to amyloid status.³⁸ Also, the APOE4 allele is associated with higher circulating Cer d18:1/20:0 and Cer d18:1/22:0,³⁹ and pharmacological modulation of the sphingosine-1-phosphate receptor improved memory in both APP/PS1 and E4FAD mice.^39,40 Collectively, clinical, genetic and pre-clinical evidence converge on the ceramide/sphingomyelin axis as a promising therapeutic target for AD.

Cholesterol esters and Alzheimer's disease

Our study demonstrated a significant correlation between ChE (20:3) and CSF Aβ₄₀ levels. ChE are primarily synthesized in plasma through lecithin-cholesterol acyltransferase (LCAT)-mediated transfer of fatty acids from phosphatidylcholine to cholesterol,⁴¹ thereby facilitating lipoprotein assembly and reverse-cholesterol transport. The altered ChE profile observed in AD may therefore reflect dysregulation of the LCAT axis. Consistent with this interpretation, reduced LCAT activity has been reported in AD plasma, a defect that limits cholesterol loading onto HDL and depletes phosphatidylcholine substrates.⁴² Other work indicates that ChE modulate Aβ biogenesis via the cholesterol-binding domain of amyloid precursor protein: lowering ChE stores through CYP46A1 activation decreases Aβ and tau pathology in pre-clinical models.⁴³ Taken together, these observations link peripheral ChE metabolism to central Aβ production and highlight the LCAT–ChE pathway as a potential metabolic target in AD.⁴⁴

TREM2 and lipid metabolism

Growing evidence shows that sTREM2 is tightly entwined with peripheral lipid homeostasis. In coronary artery disease, high plasma sTREM2 predicts cardiovascular mortality by signaling plaque instability and impending rupture.⁴⁵ It rises early in metabolic-dysfunction-associated steatohepatitis, mirroring hepatic TREM2 expression, macrophage infiltration and histological disease activity,⁴⁶ and even drives hepatocyte lipid uptake via CD36 in non-alcoholic steatohepatitis.⁴⁷ Together, these data position sTREM2 as both biomarker and effector along the cardiometabolic axis. In this study, we show that in plasma of AD sTREM2 associates not only with global lipid indices but also with entire lipid classes including SM, PI, ChE, DG and FA. When sTREM2 is integrated with these lipid classes, the resulting panel yields outstanding diagnostic accuracy for distinguishing AD patients from HC (AUC = 0.93), highlighting a multiplex blood-based signature that captures the intersection of neuroinflammation and lipid metabolism.

To test whether the lipid signatures linked to plasma sTREM2 in patients reflect TREM2 signaling rather than late-stage AD pathology, we profiled lipidomes in APP/PS1 mice lacking Trem2, thereby eliminating both membrane-bound TREM2 and sTREM2. In plasma, Trem2 KO restored the APP/PS1-associated depletion of SM and Hex1Cer. Meanwhile, our cross-species analysis revealed that human AD plasma exhibits significant alterations in SM and Hex1Cer, two key sphingolipids. KEGG pathway enrichment analysis further showed that both human AD plasma and APP/PS1-Trem2-KO mouse plasma were significantly enriched in sphingolipid and glycerophospholipid metabolism pathways, indicating preliminary pathway-level similarities in lipid metabolic disturbances across species. Notably, TREM2 has been shown to directly bind anionic phospholipids and sphingolipids, including ceramides and sphingomyelins, which act as damage-associated lipid ligands during myelin injury and neurodegeneration.^48,49 Consistent with this framework, dysregulation of sphingolipid classes—particularly SM and Hex1Cer—in both human AD plasma and APP/PS1-Trem2-KO mouse plasma may reflect disrupted TREM2-dependent control of sphingolipid synthesis and turnover. Indeed, previous metabolomic and pharmacologic studies in Trem2-deficient mice have demonstrated that loss of TREM2 drives pathological accumulation of long-chain ceramides and sphingomyelins, highlighting a central role for sphingolipid metabolism in TREM2's metabolic regulatory function.⁵⁰ Cross-species comparisons of plasma lipidomics in AD patients and Trem2-deficient mouse models remain scarce. By integrating human plasma lipid signatures with targeted lipidomics in APP/PS1-Trem2-KO mice, our study provides a hypothesis-generating translational link to contextualize TREM2-related lipid signatures alongside late-stage AD pathology. However, this should not be interpreted as a direct, head-to-head human–mouse plasma lipidomic comparison. A true head-to-head comparison would require harmonized lipid features, matched sample collection and preprocessing, aligned quantification, and effect-size concordance testing across shared lipid species. Because our human and mouse plasma lipidomes were generated in independent cohorts and under different experimental contexts, our cross-species analyses are limited to overlap and pathway-level patterns and are therefore hypothesis-generating, with causal inference remaining constrained.

In APP/PS1-Trem2-KO mice, seven lipid species were altered in both plasma and hippocampus (four glycerophospholipids and three sphingolipids), and KEGG analyses indicated that hippocampal changes were enriched in glycerophospholipid metabolism, whereas plasma changes involved both sphingolipid and glycerophospholipid pathways. Taken together, Trem2 KO produced compartment-specific lipid remodeling rather than a uniform shift, with a relative emphasis on sphingolipid metabolism in the plasma. The divergence between plasma and hippocampal lipid alterations likely reflects tissue-specific metabolic demands and cellular composition: microglial lipid metabolism in the CNS is dominated by myelin and membrane turnover, whereas circulating lipids are shaped by hepatic synthesis and macrophage lipid handling. Thus, TREM2-dependent lipid remodeling may operate in parallel yet directionally distinct pathways across compartments.

In the periphery, TREM2 marks lipid-associated macrophages (LAMs). Germline or hematopoietic Trem2 KO collapses the LAM transcriptional program and provokes adipocyte hypertrophy, hypercholesterolemia, and insulin resistance in high-fat–fed mice.^51,52 In atherosclerotic plaques, TREM2 enhances CD36-mediated cholesterol uptake and foam-cell formation, and its absence lowers lesional lipid burden.⁵³ Consistent with a disease-modifying role in peripheral tissues, in bleomycin-induced pulmonary fibrosis TREM2 is selectively upregulated in monocyte-derived macrophages; genetic deletion or pharmacologic blockade of TREM2 mitigates inflammation and fibrosis, coincident with suppression of sphingolipid signaling and reduced sphingomyelin synthesis and abundance in bronchoalveolar lavage fluid.⁵⁴ These observations portray TREM2 as a dynamic lipid-immune rheostat, modulated by tissue milieu and metabolic state. Integrating our human biomarker panel with this preclinical evidence positions TREM2 as a lipid–immune checkpoint in AD and suggests that therapeutic strategies that restore TREM2 function, or bolster downstream lipid-efflux pathways, may normalize lipid metabolism and modify disease trajectory.

Limitations of the study

Our study has several limitations. First, its cross-sectional design prevents us from establishing temporal or causal relationships between metabolite alterations, sTREM2 dynamics, and AD progression; longitudinal cohorts will be essential to determine whether these changes are drivers or consequences of disease. Second, we focused exclusively on blood biomarkers and did not assess biofluid or tissue crosstalk, such as liver or gut–brain interactions, which may shape systemic lipid profile. Third, while the TREM2 R47H variant represents a common AD risk allele in humans, our human cohort was not enriched for nor stratified by specific TREM2 risk variants, and our experimental validation was conducted in a Trem2 KO mouse model. Therefore, genotype-matched human–mouse comparisons were not performed, and it remains to be determined whether the observed lipid alterations fully reflect variant-specific TREM2 signaling in human AD. Future studies leveraging TREM2 variant-carrier cohorts and corresponding mouse models will be essential to enable genotype-matched analyses and more precise causal inference. Finally, APOE genotyping was not available in our clinical cohort, which prevented us from stratifying participants by APOE ε4 carrier status. Given the strong influence of APOE ε4 on lipid metabolism and AD risk, future studies in larger cohorts that include APOE genotyping will be needed to better account for APOE-related variability in plasma lipid profiles.

Conclusions

This study combines clinical profiling with preclinical modeling to uncover a TREM2-associated lipid signature in AD. Targeted plasma lipidomics identified 30 differentially expressed molecules; when integrated with sTREM2, this panel distinguished AD patients from cognitively normal controls with high diagnostic accuracy (AUC = 0.93). In APP/PS1 mice, Trem2 KO restored the APP/PS1-associated reduction in plasma SM and Hex1Cer and revealed seven shared lipid changes in both hippocampus and plasma, predominantly within glycerophospholipid and sphingolipid classes, indicating compartment-specific effects. Taken together, these findings support the hypothesis that TREM2 contributes to lipid dysregulation in AD and indicate that assessment of plasma sTREM2 combined with lipid classes warrants further evaluation as a blood-based tool for disease classification and monitoring. Future studies incorporating TREM2 risk variants, longitudinal human cohorts, and cell-type–resolved lipidomics will be required to clarify the causal role of TREM2-dependent lipid remodeling in AD progression.

Supplemental Material

sj-docx-1-alz-10.1177_13872877261450934 - Supplemental material for Lipidomics reveals TREM2-associated dysregulation of plasma lipid metabolism in Alzheimer's disease

Supplemental material, sj-docx-1-alz-10.1177_13872877261450934 for Lipidomics reveals TREM2-associated dysregulation of plasma lipid metabolism in Alzheimer's disease by Tongtong Zhang, Miao He, Yan Wang, Peiyang Gao, Xinyi Xia, Jiangting Li, Yunsi Yin, Guodong Zhao, Ouyang Chen, Miao Qu, Yi Tang and Qi Qin in Journal of Alzheimer's Disease

Footnotes

Acknowledgements

The authors thank all the participants for their participation.

ORCID iDs

Yan Wang

Yi Tang

Qi Qin

Ethical considerations

All procedures followed the principles of the Declaration of Helsinki and received approval from the Xuanwu Hospital (Capital Medical University) ethics committee (Ethics Approval Number: (2020) 097). Written informed consent was obtained from each participant prior to inclusion. Experiments followed the NIH Guide for the Care and Use of Laboratory Animals and were approved by the Xuanwu Hospital Institutional Animal Care and Use Committee (Ethics Approval Number: 1123031700237).

Consent to participate

Written informed consent was obtained from all participants.

Consent for publication

Not applicable.

Author contribution(s)

Tongtong Zhang: Data curation; Formal analysis; Investigation; Visualization; Writing – original draft.

Miao He: Data curation; Investigation; Visualization.

Yan Wang: Methodology.

Peiyang Gao: Conceptualization.

Xinyi Xia: Visualization.

Jiangting Li: Formal analysis.

Yunsi Yin: Formal analysis.

Guodong Zhao: Visualization.

Ouyang Chen: Supervision.

Miao Qu: Resources; Writing – review & editing.

Yi Tang: Resources; Writing – review & editing.

Qi Qin: Conceptualization; Data curation; Formal analysis; Funding acquisition; Resources; Supervision; Visualization; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by Beijing Natural Science Foundation (grant number L251023), National Key Research and Development Program of China (grant number 2025YFE0206600), Capital's Funds for Health Improvement and Research (grant number 2024-2-1032), Beijing Nova Program (grant number 20240484566), Xuanwu Hospital Talent Convergence Program (grant number HZ2025PYYX005), Beijing High Level Innovation Talent Program—Young Elite Talents Initiative (grant number 2-1-008-0257).

Declaration of conflicting interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Yi Tang and Qi Qin are Editorial Board Members of this journal but were not involved in the peer-review process of this article nor had access to any information regarding its peer-review. The remaining authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Supplemental material

Supplemental material for this article is available online.

References

Scheltens

De Strooper

Kivipelto

, et al. Alzheimer's disease. Lancet 2021; 397: 1577–1590.

Knopman

Amieva

Petersen

, et al. Alzheimer disease. Nat Rev Dis Primers 2021; 7: 33.

LaFerla

Oddo

. Alzheimer's disease: abeta, tau and synaptic dysfunction. Trends Mol Med 2005; 11: 170–176.

Di Paolo

Kim

. Linking lipids to Alzheimer's disease: cholesterol and beyond. Nat Rev Neurosci 2011; 12: 284–296.

Jansen

Savage

Watanabe

, et al. Genome-wide meta-analysis identifies new loci and functional pathways influencing Alzheimer's disease risk. Nat Genet 2019; 51: 404–413.

Rustam

Masters

, et al. Characterization of brain-derived extracellular vesicle lipids in Alzheimer's disease. J Extracell Vesicles 2021; 10: e12089.

Chua

Chai

Chew

, et al. Immunomodulatory sphingosine-1-phosphates as plasma biomarkers of Alzheimer's disease and vascular cognitive impairment. Alzheimers Res Ther 2020; 12: 122.

Chua

Torta

Chong

, et al. Lipidomics profiling reveals distinct patterns of plasma sphingolipid alterations in Alzheimer's disease and vascular dementia. Alzheimers Res Ther 2023; 15: 214.

Ferré-González

Lloret

Cháfer-Pericás

. Systematic review of brain and blood lipidomics in Alzheimer's disease mouse models. Prog Lipid Res 2023; 90: 101223.

10.

Wong

Braidy

Poljak

, et al. The application of lipidomics to biomarker research and pathomechanisms in Alzheimer's disease. Curr Opin Psychiatry 2017; 30: 136–144.

11.

Schlepckow

Morenas-Rodríguez

Hong

, et al. Stimulation of TREM2 with agonistic antibodies-an emerging therapeutic option for Alzheimer's disease. Lancet Neurol 2023; 22: 1048–1060.

12.

Wang

Cella

Mallinson

, et al. TREM2 Lipid sensing sustains the microglial response in an Alzheimer's disease model. Cell 2015; 160: 1061–1071.

13.

Lin

Kong

Chen

, et al. Decoding sTREM2: its impact on Alzheimer's disease - a comprehensive review of mechanisms and implications. Front Aging Neurosci 2024; 16: 1420731.

14.

Zhang

Xiang

, et al. TREM2 And sTREM2 in Alzheimer's disease: from mechanisms to therapies. Mol Neurodegener 2025; 20: 43.

15.

Nugent

Lin

van Lengerich

, et al. TREM2 Regulates microglial cholesterol metabolism upon chronic phagocytic challenge. Neuron 2020; 105: 837–854.e839.

16.

Gouna

Klose

. TREM2-dependent Lipid droplet biogenesis in phagocytes is required for remyelination. J Exp Med 2021; 218: e20210227.

17.

Filipello

You

Mirfakhar

, et al. Defects in lysosomal function and lipid metabolism in human microglia harboring a TREM2 loss of function mutation. Acta Neuropathol 2023; 145: 749–772.

18.

Jack Jr

Bennett

Blennow

, et al. NIA-AA research framework: toward a biological definition of Alzheimer's disease. Alzheimers Dement 2018; 14: 535–562.

19.

Martínez-Dubarbie

Guerra-Ruiz

López-García

, et al. Diagnostic performance of plasma p-tau217 in a memory clinic cohort using the Lumipulse automated platform. Alzheimers Res Ther 2025; 17: 68.

20.

Zhong

Zhuo

, et al. Soluble TREM2 ameliorates pathological phenotypes by modulating microglial functions in an Alzheimer's disease model. Nat Commun 2019; 10: 1365.

21.

Park

Lee

Kim

, et al. The relationship of soluble TREM2 to other biomarkers of sporadic Alzheimer's disease. Sci Rep 2021; 11: 13050.

22.

Lan

. Higher plasma soluble TREM2 correlates with reduced cerebral tau accumulation in Alzheimer's disease. Mol Psychiatry 2025; 30: 3988–3997.

23.

La Rosa

Agostini

Piancone

, et al. TREM2 Expression and amyloid-beta phagocytosis in Alzheimer's disease. Int J Mol Sci 2023; 24: 8626.

24.

Chen

Lin

Huang

, et al. Microglial activation, tau pathology, and neurodegeneration biomarkers predict longitudinal cognitive decline in Alzheimer's disease continuum. Front Aging Neurosci 2022; 14: 848180.

25.

Zhao

Jiao

, et al. Soluble TREM2 levels associate with conversion from mild cognitive impairment to Alzheimer's disease. J Clin Invest 2022; 132: e158708.

26.

Frisardi

Panza

Seripa

, et al. Glycerophospholipids and glycerophospholipid-derived lipid mediators: a complex meshwork in Alzheimer's disease pathology. Prog Lipid Res 2011; 50: 313–330.

27.

Singh

Kukreti

Saso

. Oxidative stress: a key modulator in neurodegenerative diseases. Molecules 2019; 24: 1583.

28.

Senanayake

Goodenowe

. Plasmalogen deficiency and neuropathology in Alzheimer's disease: causation or coincidence? Alzheimers Dement (N Y) 2019; 5: 524–532.

29.

Kurano

Saito

Uranbileg

, et al. Modulations of bioactive lipids and their receptors in postmortem Alzheimer's disease brains. Front Aging Neurosci 2022; 14: 1066578.

30.

Igarashi

Gao

, et al. Disturbed choline plasmalogen and phospholipid fatty acid concentrations in Alzheimer's disease prefrontal cortex. J Alzheimers Dis 2011; 24: 507–517.

31.

Yamashita

Kiko

Fujiwara

, et al. Alterations in the levels of amyloid-β, phospholipid hydroperoxide, and plasmalogen in the blood of patients with Alzheimer's disease: possible interactions between amyloid-β and these lipids. J Alzheimers Dis 2016; 50: 527–537.

32.

Varma

Oommen

Varma

, et al. Brain and blood metabolite signatures of pathology and progression in Alzheimer disease: a targeted metabolomics study. PLoS Med 2018; 15: e1002482.

33.

Huang

Wang

Shen

. Lipoprotein-associated phospholipase A2: the story continues. Med Res Rev 2020; 40: 79–134.

34.

Desale

Chinnathambi

. Phosphoinositides signaling modulates microglial actin remodeling and phagocytosis in Alzheimer's disease. Cell Commun Signal 2021; 19: 28.

35.

Qin

Yang

, et al. TREM2 In the pathogenesis of AD: a lipid metabolism regulator and potential metabolic therapeutic target. Mol Neurodegener 2022; 17: 40.

36.

Liu

Chen

Liu

, et al. PI3K/AKT Pathway as a key link modulates the multidrug resistance of cancers. Cell Death Dis 2020; 11: 797.

37.

Han

Rozen

Boyle

, et al. Metabolomics in early Alzheimer's disease: identification of altered plasma sphingolipidome using shotgun lipidomics. PLoS One 2011; 6: e21643.

38.

Wang

Bui

Vemuri

, et al. Lipidomic network of mild cognitive impairment from the Mayo Clinic Study of Aging. J Alzheimers Dis 2021; 81: 533–543.

39.

Crivelli

Luo

Kruining

, et al. FTY720 Decreases ceramides levels in the brain and prevents memory impairments in a mouse model of familial Alzheimer's disease expressing APOE4. Biomed Pharmacother 2022; 152: 113240.

40.

Baloni

Arnold

. Multi-omic analyses characterize the ceramide/sphingomyelin pathway as a therapeutic target in Alzheimer's disease. Commun Biol 2022; 5: 1074.

41.

Ouimet

Barrett

Fisher

. HDL And reverse cholesterol transport. Circ Res 2019; 124: 1505–1518.

42.

Turri

Conti

Pavanello

, et al. Plasma and cerebrospinal fluid cholesterol esterification is hampered in Alzheimer's disease. Alzheimers Res Ther 2023; 15: 95.

43.

van der Kant

Langness

Herrera

, et al. Cholesterol metabolism is a druggable axis that independently regulates tau and amyloid-β in iPSC-derived Alzheimer's disease neurons. Cell Stem Cell 2019; 24: 363–375.e369.

44.

Rudajev

Novotny

. Cholesterol-dependent amyloid β production: space for multifarious interactions between amyloid precursor protein, secretases, and cholesterol. Cell Biosci 2023; 13: 171.

45.

Cuciuc

Tshori

Grib

, et al. Circulating soluble TREM2 and cardiovascular outcome in cohort study of coronary atherosclerosis patients. Int J Mol Sci 2022; 23: 13121.

46.

Mahmoudian

Clavero-Mestres

Rusu

, et al. sTREM2 evaluation in women with metabolic dysfunction-associated steatotic liver disease: advancing diagnostic approaches for early MASH. Eur J Intern Med 2025; 141: 106418.

47.

Kothari

Savard

. sTREM2 is a plasma biomarker for human NASH and promotes hepatocyte lipid accumulation. Hepatol Commun 2023; 7: e0265.

48.

Wang

Kyauk

Shen

, et al. TREM2-dependent Microglial function is essential for remyelination and subsequent neuroprotection. Glia 2023; 71: 1247–1258.

49.

Scott-Hewitt

Perrucci

. Local externalization of phosphatidylserine mediates developmental synaptic pruning by microglia. EMBO J 2020; 39: e105380.

50.

Sharif

Brunner

Korosec

, et al. Beneficial metabolic effects of TREM2 in obesity are uncoupled from its expression on macrophages. Diabetes 2021; 70: 2042–2057.

51.

Jaitin

Adlung

Thaiss

, et al. Lipid-associated macrophages control metabolic homeostasis in a Trem2-dependent manner. Cell 2019; 178: 686–698.e614.

52.

Hendrikx

Porsch

Kiss

, et al. Soluble TREM2 levels reflect the recruitment and expansion of TREM2(+) macrophages that localize to fibrotic areas and limit NASH. J Hepatol 2022; 77: 1373–1385.

53.

Guo

Wen

, et al. TREM2 Promotes cholesterol uptake and foam cell formation in atherosclerosis. Cell Mol Life Sci 2023; 80: 137.

54.

Kang

Cao

, et al. Blockade of TREM2 ameliorates pulmonary inflammation and fibrosis by modulating sphingolipid metabolism. Transl Res 2025; 275: 1–17.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

2.14 MB

0.00 MB