Association of eicosanoids and eicosanoid-related metabolites with obesity and cardiometabolic traits

Abstract

Aims:

The molecular pathways by which obesity contributes to systemic inflammation are unclear. Eicosanoids are bioactive lipids that govern the upstream initiation of pro- and anti-inflammatory activity. We sought to investigate the association of eicosanoids with obesity, adiposity and cardiometabolic traits.

Methods:

We conducted a cross-sectional analysis of the Multi-Ethnic Study of Atherosclerosis (MESA) study with external validation in two community-based cohorts: Framingham Heart Study and Atherosclerosis Risk in Communities Study. We measured 811 eicosanoids using a directed, non-targeted mass spectrometry-based platform and examined their associations with BMI and eight related adiposity and cardiometabolic traits using multivariable-adjusted linear regression.

Results:

Among 5,101 MESA participants (mean age 63 years, 53% women), we found 255 eicosanoids and related metabolites associated with BMI (FDR q<0.01 for all). Among these, 82 showed consistent associations across all three cohorts, including 18 metabolites with known identities: 9 were associated with lower BMI including docosahexaenoic acid derivative 19,20-DiHDPA, arachidonic acid derivatives (20-carboxy-arachidonic acid, 5,6-diHETrE, 14,15-diHETrE, 15-HpETE), and eicosapentaenoic acid derivative 5-HpEPE. By contrast, 9 metabolites associated with higher BMI including hydroxyoctadecenoic acid (HOME), 11-hydroxy-9-octadecenoic acid (11-HOME), and adrenic acid. Many BMI-associated metabolites overlapped with other adipose traits including 240 (94%) with waist circumference and 93 (36%) with visceral adipose tissue.

Conclusion:

We identified 82 BMI-associated eicosanoids, including docosanoids and arachidonic acid derivatives associated with lower BMI, while octadecanoids and adrenic acid derivatives associated with higher BMI. Our findings highlight the role of bioactive lipids and specific pro- and anti-inflammatory pathways that may underlie adverse cardiometabolic consequences of obesity.

INTRODUCTION

Obesity affects more than two in five adults in the United States and is projected to impact nearly one in two adults by 2030.^1,2 It plays a central role in the pathogenesis of cardiometabolic disease including development of type 2 diabetes, hypertension, dyslipidemia, and metabolic dysfunction–associated steatotic liver disease (MASLD).³ Obesity is also an independent risk factor for cardiovascular disease^4,5, which underscores the urgent need to better understand underlying biological mechanisms. While visceral adiposity is thought to incite systemic inflammation as one of the major contributors to adverse obesity-related health consequences, exact molecular pathways remain unclear.

Eicosanoids and eicosanoid-related metabolites, including prostaglandins, leukotrienes, lipoxins, and thromboxanes, are bioactive lipids derived from polyunsaturated fatty acids.⁶ They regulate the upstream activation of pro- and anti-inflammatory pathways and have been linked to physiological functions such as vascular tone, endothelial function, and inhibition of platelet aggregation.⁷ Notably, eicosanoids and related metabolites have been implicated in the pathophysiology of obesity-related traits, such as impaired glucose metabolism, insulin resistance, and adipose tissue dysfunction.⁸ However, most existing evidence is derived from animal models or studies focused on individual eicosanoids, and comprehensive profiling of eicosanoid metabolites across large human samples are scarce.

To address this gap, we conducted large-scale profiling of over 800 eicosanoids and related metabolites using a liquid chromatography / mass spectrometry-based platform (LC-MS) to examine their associations with obesity and adipose traits. We hypothesized that delineation of pro- and anti-inflammatory bioactive lipids associated with obesity might shed light on the role of these lipid metabolites in obesity-related cardiometabolic risk. Further, to gain broader insight into overlap with other obesity-related traits, we examined associations with additional measures of cardiometabolic health, including dyslipidemia and glycemic measures.

METHODS

Study Population

We utilized individual-level data from three prospective, community-based cohort studies with available plasma samples for metabolite profiling: The Multi-Ethnic Study of Atherosclerosis (MESA) Exam 2 (2002–2004; n = 6,228), Framingham Heart Study (FHS) Offspring Cohort Exam 8 (2005–2008; n = 3,021), and the Atherosclerosis Risk in Communities (ARIC) Study Visit 2 (1990–1992; n = 14,291). The MESA cohort served as the derivation sample, with external validation performed in the FHS and ARIC cohorts. Individuals with a history of heart failure, myocardial infarction, end-stage renal disease, or missing key covariates were excluded. The final sample included 5,101 participants from MESA, 2,447 from FHS, and 8,882 from ARIC. Details of exclusions for each cohort are presented in Supplemental Figure S1.

Further cohort-specific details are provided in the Supplemental Methods. All participants provided written informed consent, and all protocols were approved by the institutional review boards of the respective institutions.

Clinical Assessment

All participants underwent a detailed medical history, physical examination, and laboratory testing as outlined in the study protocols.^9–11 Sex was self-reported across all cohorts. Active smoking was defined as self-reported daily cigarette use. BMI was calculated as weight in kilograms divided by height in meters squared and expressed as kg/m². Waist-to-hip ratio was calculated using waist and hip circumference measurements (in centimetres) as provided by each cohort. Hypertension was defined as a systolic blood pressure ≥ 130 mmHg, a diastolic blood pressure ≥ 80 mmHg, or the use of antihypertensive medications. Diabetes was defined as self-reported or physician-diagnosed diabetes, a fasting glucose ≥126 mg/dl, hemoglobin A1c ≥ 6.5%, or use of glucose-lowering medications. Estimated glomerular filtration rate (eGFR) was calculated using the creatinine-based race-free Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) 2021 equation. Fasting lipid panels included assessment of total and high-density lipoprotein (HDL) cholesterol concentrations.

Visceral and subcutaneous fat areas were measured in a subset of participants in the MESA study (n = 1,947). These measurements were derived from abdominal computed tomography (CT) scans obtained during the MESA Exams 2 and 3 using standardized image analysis protocols.¹² After applying prespecified exclusions, 1,301 participants were included in our analysis. Additional details are provided in the Supplemental Methods.

Profiling of Eicosanoids and Eicosanoid-related Metabolites

Plasma eicosanoids and eicosanoid-related metabolite profiling for all cohorts was conducted in the same laboratory at the University of California, San Diego, USA, using a directed, non-targeted LC-MS platform in combination with computational spectral fragmentation analysis for metabolite identification. Detailed descriptions of the profiling methodology have been published previously.^13,14 In short, fasting blood samples were collected following at least 8 hours of fasting, centrifuged at 4 °C, and stored at −80 °C. Plasma extracts (20 μL) were injected onto a Phenomenex Kinetex C18 column (2.1 × 100 mm, 1.7 μm particle size) maintained at 50 °C. Chromatographic separation was performed using a Thermo Vanquish UPLC. Detection was performed on a Q Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific) equipped with a heated electrospray ionization (HESI) source operating in negative ion mode, with further details outlined in Supplemental Methods. Metabolite identification was based on mass-to-charge ratio (m/z), retention time, and fragmentation pattern, using both authentic standards and spectral networking. Annotations and lipid classifications were referenced using the LIPID MAPS database.¹⁵ Data files were converted from RAW to mzXML format using MSConvert (version 3.0.9393), and feature extraction and alignment were performed using in-house R scripts and mzMine 2.21.¹⁶ Lipids with known molecular identities were prioritized in the main tables, whereas detected lipid species that could not be assigned definitive molecular names using available standards, spectral libraries, and the present annotation framework were classified as putative or unknown lipids.

After excluding metabolites with low detection rates, defined as metabolites with more than 90% undetectable values (3.0% in MESA, 1.0% in FHS, 2.3% in ARIC), there were 811 metabolites identified in MESA, 661 in FHS, and 644 in ARIC. We then aligned metabolites from validation cohorts (FHS and ARIC) with the derivation cohort (MESA) based on LC-MS spectral features. Additional quality-control procedures for the directed, non-targeted LC-MS platform have been described previously.¹⁴ In prior validation of this platform, replicate plasma analyses demonstrated a median coefficient of variation of 7.9%. A total of 652 metabolites from FHS and 644 from ARIC were successfully aligned with corresponding metabolites in MESA and were the focus of validation analyses. Putative metabolite identities were further verified by manual inspection of spectral fragmentation pattern networking and manual verification.

Statistical Analysis

Baseline clinical characteristics are summarized separately for each study cohort. Continuous variables are presented as mean ± standard deviation (SD), and categorical variables as counts and percentages. To prepare metabolite data, we imputed values below the detection threshold as 25% of the minimum observed concentration for that metabolite for each study cohort. Metabolite data were natural logarithm (log)-transformed due to right-skewed distributions and then standardized to mean = 0 and SD = 1.

In the primary analyses, we assessed cross-sectional associations of each eicosanoid and related metabolite with BMI using linear regression models in MESA. All models were adjusted for age, sex, race/ethnicity, study site, plate number to account for potential batch effects, systolic blood pressure, antihypertensive medication use, diabetes, active smoking, statin use, aspirin use, estimated glomerular filtration rate (eGFR), log-transformed triglycerides, total cholesterol, and HDL cholesterol. For replication, metabolites significantly associated with BMI in MESA were examined in FHS and ARIC using similarly adjusted linear regression models. As the majority of participants in the FHS cohort were White (99%), race/ethnicity was not included as a covariate in the FHS analyses. Multiple testing was addressed using the false discovery rate (FDR) correction, with an FDR q-value < 0.01 considered statistically significant. Baseline descriptive statistics and regression analyses were performed using SAS software (version 9.4, SAS Institute Inc.) and R (version 4.3.2).

To explore the broader relevance of these metabolites across cardiometabolic traits, we conducted additional analyses in MESA. Using the same analytical approach, we examined associations of all metabolites with eight additional cardiometabolic traits: waist circumference (WC), waist-to-hip ratio (WHR), visceral adipose tissue (VAT), subcutaneous adipose tissue (SAT), fasting glucose (FG), hemoglobin A1c (HbA1c), triglycerides (TG), and HDL cholesterol (Figure 1). Results are presented for metabolites that were significantly associated with both BMI and the respective traits.

Figure 1.

MESA: Multi-Ethnic Study of Atherosclerosis; FHS: Framingham Heart Study; ARIC: Atherosclerosis Risk in Communities Study; BMI: body mass index; WC: waist circumference; VAT: visceral adipose tissue. Created in BioRender. Ho, J. (2026) https://BioRender.com/srgxo5i.

To investigate shared patterns in the association of metabolites with cardiometabolic traits, we conducted principal component analysis (PCA) on the matrix of effect estimates for eicosanoids significantly associated with BMI across all nine traits. PCA was performed using the PCA module from the scikit-learn library in Python (version 3.12.4, packaged by Anaconda, Inc.; scikit-learn version 1.4.2).¹⁷ The first three components, which explained the largest proportion of total variance, were retained for interpretation and visualization. A loadings plot was used to illustrate the contribution of each trait to the principal components. To assess how BMI-associated metabolites aligned with these axes, we generated a PCA score plot in which each point represented a metabolite (FDR < 0.01). All plots were generated using the Matplotlib library (version 3.8.4) in Python.¹⁸ Additionally, we created a heatmap to display the direction and magnitude of effect estimates for all metabolites across the nine traits. The heatmap was generated using the Seaborn data visualization library (v0.13.2) and was sorted by decreasing BMI effect size, with row labels (metabolite names) suppressed for clarity.¹⁹

RESULTS

For analyses including anthropometrics, our primary cohort included n = 5,101 MESA participants with mean age 63 ± 10 years, 53% women, and BMI 28.3 ± 5.4 kg/m². Validation samples included n = 2,447 FHS (mean age 66 ± 9 years, 56% women, BMI 28.1 ± 5.4 kg/m²) and n = 8,882 ARIC participants (mean age 57 ± 6 years, 57% women, BMI 27.7 ± 5.2 kg/m²) with other cardiometabolic traits presented in Table 1.

Table 1.

Clinical Characteristics by Cohorts

	MESA	FHS	ARIC
	(n= 5,101)	(n = 2,447)	(n= 8,882)
Age, years	63 (10)	66 (9)	57 (6)
Women	2,681 (53)	1,376 (56)	5,033 (57)
Race/Ethnicity *
White, Caucasian	2,051 (40)	2,403 (99)	6,928 (78)
Asian	606 (12)	6 (<1%)	0 (0)
Black, African-American	1,346 (26)	4 (<1%)	1,954 (22)
Other	0	20 (1%)	0
Hispanic	1,098 (22)	0	0
BMI, kg/m2	28.3 ± 5.4	28.1 ± 5.4	27.7 ± 5.2
WC, cm	97.7 ± 14.3	101.18 ± 14.6	97.1 ± 14.0
WHR, cm	0.93 ± 0.08	0.96 ± 0.09	0.92 ± 0.08
SBP, mmHg	124 ± 20	128 ± 17	121 ± 18
DBP, mmHg	70 ± 10	74 ± 10	72 ± 10
Hypertension	2,497 (49)	1,636 (67)	4,211 (47)
Diabetes mellitus	698 (14)	199 (9)	1,050 (12)
Current smoking	556 (11)	215 (9)	1,890 (21)
Anti-hypertensive medication	2,084 (41)	1,132 (46)	2,434 (27)
Lipid-lowering medication	1,134 (22)	972 (40)	483 (5)
Statin use	1,033 (20)	906 (37)	176 (2)
Aspirin use	1,576 (31)	966 (40)	4,378 (49)
Fasting glucose, mg/dl	99 ± 28	106 ± 22	110 ± 34
Hb A1c, %	5.7 ± 0.9	5.7 ± 0.6	5.7 ± 1.0
Triglycerides, mg/dl	130 ± 80	116 ± 66	131 ± 83
Total Cholesterol, mg/dl	191 ± 35	188 ± 36	211 ± 39
HDL Cholesterol, mg/dl	52 ± 15	58 ± 18	50 ± 17
eGFR, ml/min/1.73 m2	80 ± 16	83 ± 15	66 ± 11

Variables are presented as n (%) or mean ± SD.

^*Race/ethnicity was missing for 14 participants in FHS.

Abbreviations: BMI = body mass index; DBP = diastolic blood pressure; eGFR = estimated glomerular filtration rate; HDL = high density lipoprotein; SBP = systolic blood pressure; WC = waist circumference; WHR = waist to hip ratio.

Association of eicosanoids and eicosanoid-related metabolites with BMI

Among 811 eicosanoids and eicosanoid-related metabolites measured, 255 (31%) were associated with BMI in MESA (FDR q-value < 0.01). Among these, 109 (43%) metabolites were associated with lower BMI, and by contrast, 146 (57%) were associated with higher BMI (Figure 2). Of 255 metabolites associated with BMI in MESA, 206 (80%) were aligned with corresponding metabolites in FHS and 209 (82%) in ARIC and were eligible for external validation. Across all three cohorts, 171 metabolites (67%) were aligned. Among these, 143 (70%) were significantly associated with BMI in FHS, and 119 (57%) in ARIC (Supplemental Table S1). Further, 82 metabolites were associated with BMI across all three cohorts (FDR q<0.01 for all), with 18 metabolites with known identities displayed in Table 2, some with known associations with obesity and cardiometabolic traits across both experimental models and human samples (prior evidence summarized in Supplemental Table S2, full results in Supplemental Table S3).

Figure 2.

Volcano plot showing associations of eicosanoid and eicosanoid-related metabolites with BMI. A total of 255 out of 811 metabolites were significantly associated as shown in blue (FDR q<0.01). X-axis displays beta coefficients, which indicate SD-unit change in natural log-transformed BMI per 1-SD change in log-transformed metabolite. See Table 2 for full metabolite nomenclature.

Table 2.

Association of 18 eicosanoids and eicosanoid-related metabolites with BMI and overlap with VAT

									BMI-associated metabolites that share overlap with VAT
			MESA		FHS		ARIC		MESA		FHS
m/z	Retention time, min	Putative metabolite ID	β (SE)	p-value	β (SE)	p-value	β (SE)	p-value	β (SE)	p-value	β (SE)	p-value
Associated with higher BMI
379.2469	4.958	HOME	0.24 (0.02)	6.2×10−44	0.16 (0.02)	7.0×10−16	0.08 (0.01)	5.1×10−13			0.10 (0.03)	0.000156
379.2488	5.1985	11-HOME	0.22 (0.02)	5.2×10−40	0.17 (0.02)	1.9×10−17	0.03 (0.01)	0.001			0.13 (0.03)	1.2×10−6
353.2313	4.621917	Hydroxyhexadecanoic acid	0.24 (0.02)	3.2×10−37	0.18 (0.02)	2.3×10−20	0.06 (0.01)	5.3×10−8			0.16 (0.03)	8.4×10−9
381.2622	6.0125	Hydroxyoctadecanoic Acid	0.23 (0.02)	9.6×10−32	0.15 (0.02)	1.9×10−12	0.06 (0.01)	5.9×10−8			0.14 (0.03)	7.9×10−7
379.247	5.284833	Putative 11-Hydroxy-9-octadecenoic Acid	0.17 (0.02)	2.8×10−22	0.16 (0.02)	1.2×10−16	0.06 (0.01)	3.5×10−10			0.13 (0.03)	5.0×10−7
395.2432	5.247833	Putative FAHFA 3:0/18:2	0.11 (0.01)	2.8×10−15	0.13 (0.02)	1.4×10−13	0.05 (0.01)	1.0×10−7	0.09 (0.03)	0.000649	0.15 (0.03)	2.0×10−9
331.2613	6.53975	Adrenic Acid	0.08 (0.01)	1.6×10−8	0.08 (0.02)	6.6×10−5	0.03 (0.01)	0.001	0.08 (0.03)	0.002846	0.09 (0.03)	0.000703
361.2385	6.320833	Eicosapentaenoic Acid	0.05 (0.01)	3.0×10−5	0.12 (0.02)	1.2×10−10	0.06 (0.01)	5.9×10−11	0.08 (0.02)	0.000366	0.08 (0.03)	0.001484
335.2195	4.970333	Putative 12-HpETE	0.04 (0.01)	0.001	0.05 (0.02)	<0.01	-0.03 (0.01)	<0.001
Associated with lower BMI
361.2385	3.656833	19,20-DiHDPA	−0.21 (0.01)	5.5×10−53	−0.20 (0.02)	6.4×10−28	−0.21 (0.01)	1.2×10−105	−0.18 (0.02)	1.0×10−12	−0.11 (0.02)	4.2×10−6
333.2079	4.033	20-carboxy-Arachidonic Acid	−0.12 (0.01)	2.1×10−21	−0.09 (0.02)	2.0×10−7	−0.13 (0.01)	1.4×10−42	−0.09 (0.02)	0.000148
311.2232	4.612667	Octadec-9-enedioic acid	−0.11 (0.01)	1.9×10−16	−0.10 (0.02)	2.2×10−7	−0.10 (0.01)	6.6×10−26	−0.12 (0.02)	2.1×10−7
295.2281	5.112167	Analog of HODE	−0.11 (0.01)	1.3×10−15	−0.10 (0.02)	5.1×10−7	−0.17 (0.01)	1.1×10−63	−0.13 (0.02)	7.4×10−8
337.2383	4.3845	5,6-diHETrE	−0.12 (0.02)	5.2×10−15	−0.14 (0.02)	1.0×10−13	−0.06 (0.01)	2.3×10−10
337.2386	3.7185	14,15-diHETrE	−0.08 (0.01)	5.1×10−10	−0.07 (0.02)	7.4×10−5	−0.10 (0.01)	8.3×10−23
333.205	4.618834	5-HpEPE	−0.10 (0.02)	7.5×10−10	−0.07 (0.02)	0.0002	−0.07 (0.01)	1.0×10−11	−0.12 (0.03)	2.8×10−5
335.2233	4.557167	15-HpETE	−0.08 (0.01)	3.4×10−9	−0.07 (0.02)	3.7×10−5	−0.10 (0.01)	1.6×10−23	−0.08 (0.03)	0.001746
311.2231	4.729833	13-HpODE	−0.05 (0.01)	5.1×10−5	−0.08 (0.02)	6.9×10−6	−0.14 (0.01)	1.1×10−44	−0.09 (0.02)	0.000149

The table includes eicosanoid metabolites with confirmed molecular identity that were significantly associated with BMI in all three cohorts (MESA, FHS, and ARIC). All metabolites met FDR q-value < 0.01. For each metabolite, β coefficients (standard error) from linear regression models are shown for MESA, FHS, and ARIC. β values represent the SD-unit change in natural log-transformed BMI per 1-SD increase in natural log-transformed eicosanoid levels. Models were adjusted for age, sex, race/ethnicity (except in FHS), study site, plate number, diabetes status, smoking, systolic blood pressure, antihypertensive medication, aspirin use, statin use, total cholesterol, HDL-cholesterol, log-transformed triglycerides, and estimated glomerular filtration rate (eGFR).

Abbreviations: m/z: mass-to-charge ratio; HOME: hydroxyoctadecamonoenoic acid; FAHFA: fatty acid ester of hydroxy fatty acid; HpETE: hydroperoxyeicosatetraenoic acid; DiHDPA: dihydroxydocosapentaenoic acid; HODE: hydroperoxyoctadecadienoic acid; diHETrE: dihydroxyeicosatrienoic acid; HpEPE: hydroperoxyeicosapentaenoic acid; HpODE: hydroperoxyoctadecadienoic acid

Of 18 metabolites with confirmed identities, 9 were associated with lower BMI across all three cohorts and included docosahexaenoic acid (DHA) derivative 19,20-dihydroxydocosa-pentaenoic acid (DiHDPA), arachidonic acid derivatives (20-carboxy-arachidonic acid, 5,6-dihydroxyeicosatrienoic acid [diHETrE], 14,15-diHETrE, 15-hydroperoxy-eicosatetraenoic acid [HpETE], and eicosapentaenoic acid (EPA) derivatives (5-hydroperoxy-eicosapentaenoic acid [HpEPE]). The largest magnitude of effect was observed for 19,20-DiHDPA, with a 1-SD higher level associated with a 0.20 SD ± 0.01 lower BMI in MESA, with similar effect sizes across other cohorts.

By contrast, 9 metabolites associated with higher BMI across all three cohorts including fatty acids such as hydroxyoctadecenoic acid (HOME), 11-hydroxy-9-octadecenoic acid (11-HOME), hydroxyhexadecanoid acids, and adrenic acid. Of note, the majority of metabolites associated with BMI displayed consistent directionality across derivation and validation cohorts (99.3% concordance among MESA and FHS, 91% concordance among MESA and ARIC, Figure 3). The largest magnitude of effect was observed for HOME, with a 1-SD higher level associated with a range of 0.08 to 0.23 SD higher BMI in MESA, FHS and ARIC, respectively.

Figure 3.

Scatter plots comparing beta coefficients for the association of eicosanoids and eicosanoid-related metabolites with body mass index (BMI) in the MESA cohort versus the FHS (left panel) and ARIC (right panel) cohorts. Each point represents a metabolite with available data in both cohorts (of n=255 BMI-associated metabolites in MESA, n = 143 were significantly associated in FHS and n = 119 in ARIC). The x-axis displays effect estimates from MESA, and the y-axis shows the corresponding estimates from FHS and ARIC, respectively.

In exploratory analyses, additional adjustment for obstructive sleep apnea and physical activity did not substantively change results. We did not detect effect modification by age or lipid lowering medication use. There were 4 of 18 metabolites that showed effect modification by sex, all of which showed directionally consistent effects in sex stratified analyses. Similarly, 4 metabolites showed effect modification by race, and 2 by diabetes status (Supplemental Table S4). Lastly, associations with obesity were similar across individuals with and without insulin resistance (Supplemental Table S5, Supplemental Figure S2).

Overlap with BMI-associated metabolites and visceral adiposity

We next examined the association of n=255 BMI-related metabolites across other adiposity-related traits ascertained among MESA participants including waist circumference (WC), waist-to-hip-ratio (WHR), visceral and subcutaneous adipose tissue depots (VAT, SAT). We found that the majority of BMI-related metabolites overlapped with other adipose traits including 240 (94%) with WC, 146 (57%) with WHR, 93 (36%) with VAT, and 61 (24%) with SAT. Of note, n = 49 metabolites were uniquely associated with VAT (Supplemental Table S6) and not SAT, while n=44 overlapped between VAT and SAT, with overlap for VAT and SAT shown in Supplemental Figure S3.

Because VAT is metabolically the most active adipose depot, we examined associations with VAT in the subset of individuals with available VAT measures in the external sample of FHS participants (n=1,150). Among the 18 BMI-associated metabolites with known molecular identities highlighted in Table 2, 10 overlapped with VAT in MESA, and 9 validated in additional analyses in FHS.

Relations of BMI-associated eicosanoid metabolites across other cardiometabolic traits

We next examined the association across other cardiometabolic traits, including fasting glucose, HbA1c, HDL cholesterol, and triglyceride (TG) concentration, as shown in Figure 4 and Supplemental Table S7. Overlap with glycemic traits included 113 (44%) associated with fasting glucose and 60 (24%) with HbA1c. With respect to lipid measures, 149 (58%) were associated with HDL cholesterol, and 152 (60%) with TG levels. Of note, 7 metabolites were significantly associated across all traits. Directionality across traits was generally consistent (see heatmap, Figure 3). For example, 19,20-DiHDPA was associated with lower BMI, in addition to adverse adipose, lipid and glycemic traits.

Figure 4.

Heatmap of n=255 BMI-associated eicosanoid and eicosanoid-related metabolites, and associations across other cardiometabolic traits. Color coding represents the standardized β coefficient in the primary model (SD-unit change in obesity or cardiometabolic trait per 1 SD-unit change in eicosanoid metabolite). Color scale indicates positive associations in red and negative associations in blue of each metabolite with the respective trait. A cluster dendrogram above the heatmap depicts hierarchical clustering. BMI: body mass index, WC: waist circumference; HbA1c: hemoglobin A1c; SAT: subcutaneous adipose tissue; VAT: visceral adipose tissue; HDL: high-density lipoprotein; WHR: waist-to-hip ratio; TG: triglycerides; FG: fasting glucose.

Using hierarchical clustering we observed that the following cardiometabolic traits clustered based on eicosanoid profiles: (1) BMI, WC, VAT, and SAT, (2) WHR, TG, fasting glucose, HbA1c, and (3) HDL cholesterol. Using PCA, we examined vector plots to elucidate correlation between cardiometabolic traits (Figure 5), with the first three PCs explaining up to 96.8% of the total variance (PC1:81.0%, PC2:12.5%, PC3:3.3%). This similarly demonstrated that eicosanoid profiles of BMI, WC, and WHR were more closely correlated, whereas lipid and glycemic traits (HDL cholesterol, A1c) were less correlated with BMI.

Figure 5.

Vector plot displaying correlation of cardiometabolic traits based on principal component analysis (PCA) of eicosanoid and eicosanoid-related metabolites. Arrows represent cardiometabolic traits, with arrow length reflecting the strength of correlation with the first two principal components (PC1 and PC2) and arrow orientation indicating similarity in eicosanoid patterns. Percent variance explained by each component is indicated on the axes. BMI: body mass index, WC: waist circumference; HbA1c: hemoglobin A1c; SAT: subcutaneous adipose tissue; VAT: visceral adipose tissue; HDL: high-density lipoprotein; WHR: waist-to-hip ratio; TG: triglycerides; FG: fasting glucose.

Of 18 metabolites with known identities shown in Table 2, we found that 12 had genome-wide significant metabolite quantitative trait loci (mQTLs) from a prior eicosanoid metabolite-focused genome-wide association study (GWAS) shown in Supplemental Table S8.²⁰ Among these mQTLs, we found two were associated with BMI in the GIANT consortium (rs738408 (PNPLA3, P=9.5 × 10⁻⁵), and rs9288513 (PECR, P=0.01)), and 13 were associated with other cardiometabolic traits in prior GWAS, including waist circumference (rs2900478), bulk lipid traits, among others.

DISCUSSION

We examined the association of obesity, adiposity, and related cardiometabolic traits with specific eicosanoids and eicosanoid-related metabolites, a class of bioactive lipids involved in the upstream regulation of systemic inflammation. Our main findings are as follows: first, we identified 255 metabolites, including docosanoids, octadecanoids and oxylipins, that were associated with BMI in the MESA cohort, with 32% that replicated across both FHS and ARIC cohorts. We delineated overlapping pro-inflammatory eicosanoid metabolites with adiposity-related traits including VAT, thought to underlie adverse cardiometabolic consequences of obesity. We specifically find linoleic acid-derived oxylipins, adrenic acid, and 12-HETE that mark an active inflammatory phenotype associated with abdominal adiposity. Equally important, we show that EPA-related (19,20-DiHDPA) and arachidonic acid-derived metabolites (5,6-diHETrE, 14,15-diHETrE) with known anti-inflammatory properties were inversely associated with adiposity. In turn, VAT-related eicosanoid profiles were closely correlated with glycemic and metabolic traits (HbA1c, FG, triglycerides), and less correlated with BMI and HDL cholesterol. These findings point toward linoleic acid derivatives and others that might represent future potential therapeutic targets and identify at-risk individuals in the face of the growing burden of adiposity-related cardiovascular disease.

While systemic inflammation is implicated as a central mechanism of adverse cardiovascular effects of obesity, exact biological pathways are not fully elucidated. Our study demonstrates a potential role of specific eicosanoids and related bioactive lipids derived from arachidonic acid and other polyunsaturated fatty acids (PUFAs), which are known to govern the upstream initiation of both pro- and anti-inflammatory activity. Indeed, prior studies have examined specific eicosanoids, including arachidonic acid-derived hydroxyeicosatetraenoic acids (HETEs) and prostaglandin E2 (PGE2), which are thought to contribute to visceral adipose tissue inflammation and are strongly associated with measures of obesity and adiposity such as BMI and WC.^21–23 More recent lipidomic analyses highlight that specific oxylipin ratios, such as increased 12,13-EpOME to DiHOME in white adipose tissue, may serve as biomarkers of worsening metabolic syndrome in individuals with obesity.²⁴ Lastly, our group has previously demonstrated that specific eicosanoids including prostaglandins, oxylipins, epoxides, docosanoids, and ω-3 PUFA may mediate the association of BMI with heart failure with preserved ejection fraction (HFpEF), one of the end-stage CV manifestations of obesity and dysfunctional adiposity.²⁵ We now build on these findings to provide a detailed look at the association of obesity and cardiometabolic traits with bioactive lipids using a novel platform spanning hundreds of eicosanoids and related metabolites, ascertained across three large community-based cohorts.

We identified 82 unique eicosanoid and related metabolites associated with BMI across all three cohorts, including docosanoids, octadecanoids, hydroxy fatty acids and other classical and non-classical eicosanoids. Notably, our findings suggest that higher BMI is associated with greater elaboration of pro-inflammatory metabolites, including linoleic acid-derived oxylipins, adrenic acid, and 12-HETE. In total, 57% were associated with higher BMI, including the top hits HOME and 11-HOME, both oxylipins related to linoleic acid. The biological effects of 11-HOME in obesity remain incompletely characterized. However, the related hydroxy-monounsaturated fatty acid, (9Z)-11-HOME, has been shown in vitro to activate PPAR α, δ, and γ, suggesting a potential link between HOME metabolites and PPAR-mediated lipid metabolic pathways.²⁶ Separately, targeted activation of PPAR α/γ in adipose tissue macrophages has been shown to improve lipid metabolism and obesity-induced metabolic dysfunction in diet-induced obese mice. ²⁷ Importantly, prior short-term dietary intervention studies targeting linoleic acid intake have demonstrated suppression of inflammation and suggest this pathway may be modifiable³⁰. We also found that adrenic acid, an arachidonic acid derivative, was associated with higher BMI. Interestingly, higher adrenic acid levels have been observed in overweight and obese children and are implicated in inflammation, lipid metabolism, and oxidative stress. In animal studies, the enzyme iPLA2β, which releases adrenic acid, has been shown to promote obesity and diabetes.^28–30 Similarly, 12-HETE was associated with higher BMI in our study, and has been observed in adipose tissue of obese individuals and in a high-fat diet model of obesity-related glomerulopathy, further upregulating TNF-α, MCP-1, and IL-6.^31,32

By contrast, 43% of BMI-associated metabolites were associated with lower BMI, including EPA-related and arachidonic acid-derived metabolites with known anti-inflammatory properties. For example, the top metabolite docosanoid 19,20-DiHDPA known to be an eicosapentaenoic acid (EPA) derivative which has previously been inversely associated with adiposity, metabolic syndrome prevalence, fatty liver index, insulin resistance, and dyslipidemia, in line with other EPA-derived metabolites.³³ Further, treatment with EPA and the broader class of ω‑3 PUFAs have been studied in clinical trials demonstrating reductions in hospitalizations and mortality in patients with chronic heart failure compared to those receiving a placebo.^34–37 We also identified 5-HpEPE as a key eicosanoid metabolite associated with lower BMI. 5-HEPE is a downstream product of 5-HpEPE, which in turn improved metabolic outcomes in high-fat diet mouse models by reducing adiposity, enhancing insulin sensitivity, increasing energy expenditure, and promoting browning of white adipose tissue.³⁸ Lastly, we found that docosanoid derivatives 14,15-diHETrE and 5,6-diHETrE were associated with lower BMI, with prior studies demonstrating similar associations with lower BMI, white adipose tissue, glucose and metabolic syndrome.^24,39

We also investigated the overlap of metabolite profiles with other adiposity measures and cardiometabolic traits, including WC, WHR, specific adipose depots including VAT and SAT, glucose, HbA1c, HDL cholesterol, and TG. Overlapping metabolites included the EPA derivatives 19,20-DiHDPA and 14,15-DiHETE, which in addition to being associated with lower BMI, were also associated with lower adipose traits and higher HDL, consistent with their previously reported inverse associations with metabolic syndrome and adverse lipid profiles.³³ Similarly, HODE, a BMI-related omega-6 PUFA, was significantly associated with lower adiposity traits and higher cholesterol, but not with glycemic traits. We also present cluster-based analyses, which highlight that eicosanoid profiles of adiposity-related traits (e.g. WC, SAT, VAT, TG) were more closely correlated, whereas glycemic traits (HbA1c, glucose) and HDL cholesterol displayed distinct eicosanoid patterns relative to BMI.

This study has several limitations that should be considered. First, as a cross-sectional and observational analysis, causal inferences cannot be drawn, and we cannot determine whether the observed eicosanoids represent upstream contributors to obesity, downstream consequences of obesity, or correlated biomarkers. Future longitudinal and experimental studies will be needed to clarify underlying biological mechanisms, or potential clinical utility as diagnostic or prognostic biomarkers, and relation to other inflammatory biomarkers. Second, the cohorts examined were focused on community-dwelling US adults, and generalizability to other settings may be limited. Third, we leveraged a novel directed non-targeted platform and pursued focused verification of molecular identities among metabolites of greatest interest including tandem mass spectrometry (MS2) analyses. However, comprehensive verification of all other metabolites beyond the salient molecules highlighted herein would require further targeted validation and structural confirmation and are beyond the scope of our current work. Therefore, our understanding of eicosanoids as both substrates and products within interconnected enzymatic pathways remains incomplete, and our analyses may not fully capture the complexity of these biological networks - future pathway-based analyses will be helpful when a greater number of metabolites have been verified. Fourth, although 82 BMI-associated metabolites were replicated across all three cohorts, only 18 had confirmed molecular identities. The remaining replicated metabolites remain unannotated, which limits biological interpretation and pathway-level inference. Future studies using expanded spectral libraries and targeted approaches for metabolite characterization will be needed to clarify their identities and potential mechanistic roles in obesity-related inflammation and cardiometabolic risk. Fifth, eicosanoids often act locally through autocrine and paracrine signaling, and circulating plasma levels that were assessed in our study may not fully reflect tissue-specific concentrations.

In sum, we identified 82 eicosanoids and eicosanoid-related metabolites that were associated with BMI across three large community-based studies. Specifically, we found that octadecanoids, adrenic and eicosapentaenoic acid derivatives were associated with higher BMI, while docosanoids and arachidonic acid derivatives were associated with lower BMI. Our findings highlight the role of bioactive lipids and specific pro- and anti-inflammatory metabolites that may underlie adverse cardiovascular consequences of obesity and adiposity. Importantly, these findings highlight potentially modifiable pathways which have previously been targeted via dietary interventions and other strategies, to modulate downstream health consequences of obesity for future study.

Data Availability Statement:

The data underlying this article are available in appropriate controlled-access repositories (dbGaP) for all three cohorts (MESA, FHS, and ARIC), at https://www.ncbi.nlm.nih.gov/gap/.