One-year effects of high-intensity statin on bioactive lipids: Findings from the JUPITER trial

Abstract

Background:

Statin effects extend beyond LDL-C reduction, potentially modulating the metabolism of bioactive lipids (BALs), crucial for biological signaling and inflammation. These bioactive metabolites may serve as metabolic footprints, helping uncover underlying processes linked to statins pleiotropic effects and yielding a better understanding of their cardioprotective properties.

Aim:

To investigate the impact of high-intensity statin therapy vs. placebo on plasma BALs in the JUPITER trial (NCT00239681), a randomized primary prevention trial involving individuals with LDL-C <130 mg/dL and high-sensitivity C-reactive protein ≥ 2 mg/L.

Methods:

Using a non-targeted mass spectrometry approach, over 11,000 lipid features were assayed from baseline (Y0) and 1-year (Y1) plasma samples from cardiovascular disease (CVD) non-cases from 2 non-overlapping nested sub-studies: JUPITER_discovery (N=589) and JUPITER_validation (N=409). The effect of randomized allocation of rosuvastatin 20mg vs. placebo on BALs was examined by fitting a linear regression with delta values (Δ=Y1 – BL) adjusted for age and baseline levels of each feature. Significant associations in discovery were analyzed in the validation cohort. Multiple comparisons were adjusted using two-stage overall False Discovery Rate (FDR).

Results:

We identified 610 lipid features associated with statin randomization with significant replication (overall FDR<.05), including 26 with annotations. Statin therapy significantly increased levels of 276 features, including BALs with anti-inflammatory activity and arterial vasodilation properties. On the other hand, 334 features were significantly lowered by statin therapy, including arachidonic acid and pro-inflammatory and pro-platelet aggregation BALs. By contrast, statin therapy reduced an EPA-derived HEPE metabolite which may be related to impaired glucose metabolism. Additionally, we observed sex-related differences in 6 lipid metabolites and 6 unknown features.

Conclusion:

Statin allocation was significantly associated with upregulation of BALs with anti-inflammatory, anti-platelet aggregation, and antioxidant properties and downregulation of BALs with pro-inflammatory and pro-platelet aggregation activity, supporting the pleiotropic effects of statins beyond LDL cholesterol reduction.

Graphical Abstract

INTRODUCTION

The pleiotropic effects of statins (HMG-CoA reductase inhibitors) go beyond lipid-lowering.^1,2 Studies have suggested that statins have several atheroprotective properties, including anti-inflammatory,^3,4 antioxidant,⁵ antiplatelet,⁶ improve endothelial function,⁷ atherosclerotic plaques stabilization,⁸ and immunomodulatory effects.⁹ Although these extended properties of statins have been described, the mechanisms involved are not fully uncovered. Since statins act early in the intracellular cholesterol biosynthesis, they can modify multiple pathways of the metabolic profile, influencing the lipid metabolism at a biomolecular level, including the modulation of bioactive lipids (BALs)^10–12.

Bioactive lipids are a diverse group of lipid signaling molecules that act through interactions with specific receptors or directly affect cellular processes, with versatile and bioactive roles in various biological functions.^13–16 These biomolecules are involved in pro- and anti-inflammatory responses and inflammation resolution,^15,16 hemostasis and platelet aggregation activity,¹⁷ endothelial integrity and vascular permeability,¹⁸ control of vascular tone,¹⁹ and diabetogenic pathways.²⁰

We hypothesized that randomized statin treatment favorably alters levels of BALs and other lipid features with cardioprotective properties compared to placebo, helping to elucidate the spectrum of pleiotropic effects of statins. In this study, we leveraged data and blood samples from a randomized trial of apparently healthy participants, the Justification for the Use of statins in Prevention: an Intervention Trial Evaluating Rosuvastatin (JUPITER; NCT 00239681),²¹ to investigate the impact of high-intensity statin therapy versus placebo on plasma BALs over a one-year period

METHODS

Data Availability

Because the data collected for this study was nested within a randomized clinical trial, requests to access the dataset from qualified researchers trained in human subject confidentiality protocols should be sent to the Steering Committee of the parent trial.

Study samples and design

We assayed BALs and other lipid features within baseline (BL; data collected before randomization) and one-year (Y1) blood plasma samples from two non-overlapping sets of participants from the JUPITER trial. JUPITER was a landmark randomized, double-blinded, and placebo-controlled trial that investigated the effects of 20 mg/day rosuvastatin vs. placebo on the rate of cardiovascular outcomes in 17,802 participants (median follow-up 1.9 years) with average to low levels of low-density lipoprotein cholesterol (LDL-C < 130 mg/dl) and elevated high-sensitivity C-reactive protein (hs-CRP ≥ 2 mg/L). Patients with diabetes, uncontrolled hypertension, or using post-menopausal hormone-replacement therapy were excluded from the trial. The JUPITER trial inclusion and exclusion criteria are fully described elsewhere.²¹

The discovery and validation cohorts for the current analysis comprised different sets of JUPITER participants who had provided blood at baseline and year 1 follow-up. The discovery cohort (JUPITER_discovery; N=589) included individuals who remained free of incident CVD during the study duration and had provided blood samples as part of a prior genome wide association study.²² Incident CVD was defined as incident myocardial infarction, stroke, coronary revascularization, unstable angina requiring hospitalization, or death. The validation cohort (JUPITER_validation; N=409) comprised the age- and sex-matched control group from a nested CVD case-control sub-study within JUPITER, who had undergone other biomarker assays.²³ Only participants with BAL and other lipid features measurements in the two time-points were included in the current study. Associations of statins with delta values of each lipid feature (Δ, difference between Y1 and BL levels) were primarily investigated in the discovery cohort (JUPITER_discovery). Then, features significantly associated with statin randomization were analyzed in the validation cohort (JUPITER_validation).

All participants provided written informed consent at the time of enrollment. Institutional review board approvals for both studies were obtained from Partners HealthCare (Boston, MA). The first and senior authors had full access to all the data in the study and take responsibility for their integrity and data analysis.

Clinical and biomarker risk factors

Baseline questionnaires were used to collect sex, age, race/ethnicity, use of non-randomized supplements or medications, smoking, and other relevant aspects of health history. Body weight and height were measured during physical examination by the study personnel, and standard fasting lipid panels and fasting glucose were obtained in a central laboratory as part of the clinical trial. LDL-C concentrations were calculated by the Friedewald equation when triglycerides were <400 mg/dL and measured by ultracentrifugation when ≥400 mg/dL.²⁴ Lp(a) concentrations were measured in a blinded manner at Quest Diagnostics Nichols Institute (San Juan Capistrano, CA) with a commercially available assay (Randox Laboratories; Crumlin, Co. Antrim, United Kingdom) that is not affected by kringle IV type 2 repeats.²⁵ hs-CRP was measured using a high-sensitivity assay (Behring Nephelometer).²⁴ A glycoprotein acetylation biomarker (GlycA) on acute phase reactants was measured using nuclear magnetic resonance spectroscopy (NMR). GlycA signals were quantified at LipoScience Inc (Raleigh, NC) from plasma NMR spectra obtained from the automated NMR Profiler system.²⁶ We also measured secretory phospholipase A2 (sPLA2) at Quest Diagnostics Nichols Institute (San Juan Capistrano, CA) with a commercially available enzyme immunoassay (Cayman assay; Cayman Chemical Co. Ann Arbor MI) based on a double-antibody sandwich technique that is specific for sPLA2-IIA.²⁷ Concentrations of lipoprotein-associated phospholipase A2 (LpPLA2) mass were determined by a latex particle–enhanced turbidimetric immunoassay for LpPLA2 run on the Roche P-modular analyzer (PLACTM test, diaDexus). LpPLA2 activity was measured in a research-use automated enzyme assay system, run on the Roche P-modular analyzer (CAM assay, diaDexus) with a colorimetric substrate that is converted upon hydrolysis by the phospholipase enzyme.²⁸

BAL profiling

Blood samples were collected in EDTA tubes at BL and Y1 and stored at the Brigham and Women’s Hospital in vapor-phase liquid nitrogen (−170°C). Samples were thawed, separated into aliquots, refrozen, and shipped on dry ice to and sent to the University of California, San Diego for metabolomics preprocessing and profiling. Approximately 11,000 features were extracted using a directed non-targeted high-throughput liquid chromatography-mass spectrometry (LC-MS) using high mass accuracy for measurement of bioactive lipid species and fully described elsewhere.^29,30 Several steps and procedures were employed for compounds identification, including a chemical networking of MS/MS spectral fragments, system analysis of chemical patterning, and three databases of commercial standards.³⁰ Samples from BL and Y1 of the same participant were randomly placed in tandem wells on a plate to avoid batch effects and were blindly assayed. Samples were sealed under argon after preparation to reduce oxidation in the plate while waiting to be injected into the instrument. To avoid experimental biases, blinded quality control samples were additionally embedded throughout the experiment within each batch. Moreover, the performing laboratory had a multi-tiered quality control approach that allows for close monitoring of sample-to-sample variation and preparation, as well as system performance and any potential system drift over the sample run.

We detected and removed adducts and carbon-13 (¹³C) isotopes, i.e., features with correlation coefficient > 0.95, up to ± 0.01 minutes retention time from each other, and difference in mass to charge ratio equal to one or more ¹³C atom mass. Then, we excluded features that were detected in less than 80% of samples. To further reduce redundant fragments (due to contaminants, fragments, or other MS artifacts), we examined remaining highly correlated features (r>0.95), and those with the lower median relative intensity within each pairwise comparison were removed.¹³ Missing values were imputed to a random number between zero and 0.25 of the lowest observed value, thus preserving variability. The outlier threshold was 4 standard deviations (SDs) from the median of each feature. Then, outlier observations for the upper tail of the distribution were shrunk into a proportional value between median+4SD and median+5SD, thus preserving the rank order of these observations. Modifying the outlier threshold to 3SD from the median and, therefore, shrinking high values into a proportional value between median+3SD and median+4SD did not change the results.

By design, batch effects and instrument drifts had minimal effect on one-year changes in assayed lipid features since BL and Y1 plasma samples were placed on neighboring wells. Therefore, to avoid overcorrecting, we did not apply batch correction methods for the analysis involving time-points changes. Baseline readings were then shifted to zero-mean and unit-SD and Y1 readings were scaled to baseline SDs. Subsequent Δ calculation was defined as Y1-BL. As for analyses focused on BL levels, relative concentrations of all LC-MS features were corrected for plate effects³¹ using COMBAT (R-package sva³²).

Statistical analysis

Continuous variables distributions are summarized as medians and interquartile ranges and the Mann–Whitney U test was used to examine contrasts between placebo and statin allocated participants. Categorical variables were cross-tabulated as numbers and percentages between-group comparisons conducted using the X² test.

The mean difference in metabolite concentration changes (outcome) between the statin group and the placebo group was assessed by linear regression models adjusted for age and BL values of each feature. No adjustments for sex were made since this study was nested within a clinical trial, and, unless covariates have a strong association with the primary outcome, adjustments potentially decrease precision or reduce the statistical power in randomized trials.^33,34 All regressions were controlled for multiple testing using a two-stage procedure proposed by Benjamini & Yekutieli.³⁵ Briefly, this approach guarantees an overall False Discovery Rate (FDR) correction < 0.05 by the multiplication of the FDR levels from each stage (discovery and validation). For the discovery stage, FDR level was set at 0.20 and changes associated with statin therapy below this threshold were analyzed independently in a second stage in JUPITER_validation. In this second stage, FDR level was set at 0.25 and lipid features were selected if associations were significant, and β-coefficients had same direction of effects as in discovery. Linear regression estimates obtained from each cohort separately were meta-analyzed using the inverse variance weighted method.³⁶ We assessed the goodness of fit of selected features by regressing their predicted changes, based on estimates obtained in the discovery stage, by their observed changes. Next, linear regression assumptions were checked for selected features in both cohorts (Supplemental Methods). Findings were finally considered statistically validated if their association with statin therapy met the FDR threshold in the validation cohort after using the appropriate procedure, which was decided based on tests of linear regression assumptions (Supplemental Methods).

We also explored interactions of randomization arm with sex, age (below and above the median of the sample), and non-randomized aspirin use in the relationship with BALs and other lipid features by including a multiplicative term in the regression model. Models for interactions were adjusted for BL readings (and age in sex and aspirin models) and multiple comparisons were adjusted at FDR level of 0.1.

We performed partial Spearman correlation analysis (partial_Spearman; R package PResiduals³⁷) of baseline BALs with baseline levels of LDL-C, high-density lipoprotein cholesterol (HDL-C), total cholesterol (TC), triglycerides, Lp(a), glucose, body mass index (BMI), systolic blood pressure, hs-CRP, GlycA, sPLA2, LpPLA2 mass, and LpPLA2 activity in a pooled dataset comprising both discovery and validation cohorts and adjusting for sex, age and randomization arm. Additionally, we performed partial Spearman correlation analysis of ΔBALs and other lipid features with Δbiomarkers.

All statistical analyses were performed using the R software (version 3.6.0).³⁸ Please see the Major Resources Table in the Supplemental Material.

RESULTS

Baseline Characteristics

Table 1 displays baseline demographic characteristics and clinical biomarkers in discovery and validation JUPITER cohorts. Placebo and active participants were overall well-balanced with respect to clinical risk factors within each sub-study, except for age in JUPITER_discovery and smoking in JUPITER_validation. Compared to the validation cohort, the discovery cohort was overall younger (p = 5.58e⁻⁰⁵), comprised more women (p=5.18e⁻¹²) and less hypertensive subjects (p = 0.039), and had higher median levels of TC (p = 0.030) and triglycerides (p = 0.046). Regarding the representation of study samples to the entire JUPITER trial population, both discovery and validation cohorts were slightly older (but with overlapping IQR) and less diverse in terms of race (more whites); the frequency of women in discovery was higher in discovery and lower in validation; median HDL-C levels were slightly higher in discovery. A comparison of the baseline characteristics of the total JUPITER trial population with the current sub-samples are presented in Supplemental Table S1.

Table 1.

Baseline characteristics of the participants, according to randomized treatment.

	JUPITERdiscovery				JUPITERvalidation
	Overall (N = 589)	Placebo (N = 310)	Active (N = 279)	p-value	Overall (N = 409)	Placebo (N = 220)	Active (N = 189)	p-value
Age (years)*	67 [62, 72]	68 [64, 73]	66 [61, 72]	0.013	70 [64, 75]	69 [64, 75]	70 [64, 74]	0.974
Women (%)	294 (50)	159 (51)	135 (48)	0.535	114 (28)	64 (29)	50 (27)	0.630
White (%)	589 (100)	310 (100)	279 (100)	1.000	376 (92)	197 (90)	179 (95)	0.084
Smoker (%)	66 (11)	36 (12)	30 (11)	0.842	47 (12)	18 (8)	29 (15)	0.035
Hypertension (%)	301 (51)	166 (54)	135 (48)	0.243	237 (58)	126 (57)	111 (59)	0.844
BMI (kg/m2)*	29 [25, 33]	29 [25, 33]	28 [25, 32]	0.792	28 [26, 31]	28 [26, 32]	29 [26, 31]	0.764
HDL-C (mg/dl)*	52 [43, 62]	52 [43, 63]	51 [42, 61]	0.458	49 [40, 61]	50 [40, 62]	48 [40, 60]	0.543
LDL-C (mg/dl)*	110 [96, 120]	111 [97, 120]	108 [94, 120]	0.217	110 [94, 120]	108.50 [94, 120]	111 [94, 119]	0.982
TC (mg/dl)*	189 [176, 202]	191 [178, 202]	188 [173, 202]	0.247	187 [168, 201]	187 [165, 202]	187 [169, 200]	0.934
Triglycerides (mg/dl)*	120 [89, 172]	119 [87, 165]	123 [92, 176]	0.248	116 [81, 162]	115 [84, 159]	117 [80, 166]	0.871
hs-CRP (mg/L)*	4.0 [2.7, 6.5]	4.0 [2.9, 6.2]	4.0 [2.7, 6.9]	0.877	4.2 [2.9, 6.9]	4.2 [2.8, 6.9]	4.2 [2.9, 6.9]	0.837

^*median [IQR]; BMI – body mass index; HDL-C – high-density lipoprotein cholesterol; LDL-C – low-density lipoprotein cholesterol; TC – total cholesterol; hs-CRP – high-sensitivity C-reactive protein.

Statin treatment effects on BALs

After ¹³C isotopes, adducts, and redundant fragments removal, the final untargeted dataset comprised 8060 features detected in more than 80% of samples. In the discovery stage, we detected changes in 3306 features significantly associated with statin randomization adjusted for age and BL levels, of which 690 were also statistically significant after the second stage in the validation cohort (overall FDR<0.05). The goodness of fit for predicted changes contrasted with observed changes was significant for all 690 selected features (FDR < 0.05), and Pearson correlation coefficients for annotated ones are provided in Supplemental Table S2.

Then, we performed supplemental analysis to ensure statistical rigor using the appropriate procedure when a linear regression assumption was violated. A total of 610 features were considered validated, including 26 annotated (Figs 1 and 2, Supplemental Figures S1, S2, and Supplemental Table S3). Significant increase with statin therapy was observed for 276 features, including hepoxilins A3 and B3 (HXA3 and HXB3), 18-hydroxyeicosapentaenoic acid (HEPE), and 19-hydroxyeicosatetraenoic acid (HETE). On the other hand, statistically significant reduction was observed in levels of 334 features, including arachidonic acid (AA), 8-HEPE, 15-hydroxyeicosatrienoic acid (HETrE), and 11-dehydro-thromboxane B2 (d-TXB2) (Figures 1 and 2). Results also indicated activation of the 12- and 15-HETE pathways from AA as their peroxide markers were noted, and inhibition of linoleic and alpha-linolenic metabolic pathways as their oxidation products were reduced (Supplemental Figure S2).

Figure 1. After one year of randomized treatment, statin therapy associated with changes in 610 lipid features:

Volcano plot with fold changes from baseline to year one in JUPITER_discovery for significantly validated lipid features in response to randomized statin (adjusted for age and BAL baseline value; overall FDR<.05). For better visualization, the plot is zoomed in for values between −1.0 and 1.0 log₂ fold change and −log₁₀ (p-value) < 20. A plot containing all significant lipid features can be found in Supplemental Figure S1. Fold changes were calculated as ${\mathit{\text{median}}}_{\mathit{\text{statin}}}\frac{Y1}{BL}/{\mathit{\text{median}}}_{\mathit{\text{placebo}}}\frac{Y1}{BL}$.

Abbreviations (in alphabetic order): DiHDPA-dihydroxydocosapentaenoic acid; diHETE - dihydroxyeicosatetraenoic acid; d-TX - dehydro-thromboxane; FAHFA - fatty acyl esters of hydroxy fatty acid; HEPE - hydroxyeicosapentaenoic acid; HETE - hydroxyeicosatetraenoic acid; HETrE - hydroxyeicosatrienoic acid; Hx - hepoxilin; LPE - lysophosphatidylethanolamine; PUFA - polyunsaturated free fatty acids.

Figure 2. Linear regression results in discovery and validation cohorts for randomized statin treatment effects on bioactive lipids (BALs) and pathway markers:

β-coefficients and 95%CIs per 20% baseline change in the main 12 annotated features, including main eicosanoids and pathway markers for the statin group vs. placebo, adjusted for age and baseline levels of each feature (overall FDR<.05). Same abbreviations as in figure 1.

Linear regression point estimates for discovery and validation cohorts were reported, showing consistency in terms of directionality and significance. Figure 2 summarizes these results per 20% change for the 12 main eicosanoids and pathway markers identified in our analysis, adjusted for age and BL levels of each feature. Supplemental Figure S2 displays linear regression results for all 26 annotated features for the statin group vs. placebo in both cohorts. The 95% confidence intervals (95%CIs) for annotated features overlapped in discovery and validation, except for Lysophosphatidylethanolamine (LPE 20:0), and linear association magnitudes across cohorts were meta-analyzed using the inverse variance weighted method, revealing a multiple R²=0.90 (p-value = 2.24e⁻¹³) with a slope of the fit of 0.58 (95%CI = 0.49 – 0.66).

Supplemental Table S3 displays point estimates and 95%CIs for annotated features after rerunning an appropriate procedure based on linear regression assumptions. Overall, point estimates after procedures were similar to linear regression estimates in both cohorts, and 95% CIs overlapped.

Considering that extreme values may represent relevant information³⁹ about effects on metabolites, we dealt with potential outliers (delta values > 4 SDs from the median of each feature) by shrinking them to a proportional value between median+4SD and median+5SD, preserving the rank order of these observations. Comparatively, the number of statistically significant features in a dataset without outliers was 530, corresponding to 87% of the total 610 detected in the shrunk dataset.

Spearman correlations of BAL levels with clinical biomarkers

To gain physiological insights about these lipid features, Spearman correlation analysis of baseline levels with cardiometabolic traits adjusted for age, sex, and randomization arm was performed in a pooled dataset comprising both discovery and validation cohorts. Figure 3 shows partial Spearman correlation coefficients for 12 annotated features, including main eicosanoids and pathway markers. In figure 3A we can observe that baseline levels of features downregulated with statin therapy (red labels) were consistently positively correlated with higher levels of baseline triglycerides and TC. The strongest positive correlations considering baseline levels were for 11d-TxB2 vs. BMI (R = 0.20, p-value = 2.53e⁻¹¹) and AA vs. triglycerides (R = 0.19, p-value = 7.70e⁻¹⁰). The strongest negative correlations were for 11d-TxB2 vs. HDL-C (R = −0.14, p-value = 4.19e⁻⁶) and HxB3 vs. triglycerides (R = −0.15, p-value = 1.32e⁻⁶). Figure 3B consistently shows that one-year changes in metabolites and pathway markers downregulated with statin (red labels) have the same directionality of one-year changes in LDL-C, TC, and inflammatory biomarkers GlycA, and LpPLA2 activity. Concurrently, changes in tetranor 12-HETE inversely correlated with Δhs-CRP, ΔsPLA2, and ΔLpPLA2 mass, also inflammatory markers. Delta 18-HEPE correlated negatively with ΔLDL-C. Supplemental Figure S4 displays the correlation of baseline levels and delta values of all annotated features with clinical biomarkers.

Figure 3. Partial Spearman correlation (R) for the main 12 annotated BALs and pathway markers with clinical biomarkers.

Spearman rank correlation coefficients were adjusted for age, sex, and statin randomization. A) Correlations between BAL baseline levels vs. clinical biomarker baseline levels. B) Correlations between BAL delta values vs. clinical biomarker delta values. Abbreviations (alphabetic order): DiHDPA-dihydroxydocosapentaenoic acid; BMI - body mass index; diHETE - dihydroxyeicosatetraenoic acid; d-TX - dehydro-thromboxane; GlycA - glycoprotein acetylation; HDL - high density lipoprotein cholesterol; HEPE - hydroxyeicosapentaenoic acid; HETE - hydroxyeicosatetraenoic acid; HETrE - hydroxyeicosatrienoic acid; hs-CRP - high-sensitivity C-reactive protein; Hx - hepoxilin; LDL - low density lipoprotein cholesterol; Lp(a) - lipoprotein (a); LpPLA2 - lipoprotein-associated phospholipase A2; SBP - Systolic blood pressure; sPLA2 - Secretory Phospholipase A2; TC - total cholesterol; Blue labels: Increased with statin randomization; Red labels: decreased with statin randomization.

Interaction analyses

For the prespecified exploratory interaction analysis with sex, we separated the features with significant association with statin randomization in two groups: annotated and unannotated. Considering the 26 annotated lipid metabolites, sex was an effect modifier (FDR for interaction < 0.1) in 2 associations in JUPITER_discovery, including 18-HEPE that was increased only in women, and 3 associations in JUPITER_validation. Although not significant, the directionality of the interaction effects in one cohort showed replication in the other cohort. Supplemental Figure S2 indicates significant interactions with sex in discovery and validation cohorts and Supplemental Figure S3 displays sex-stratified analysis for these annotated BALs. As for the 584 unannotated significant associations in both cohorts, sex was a modifier of statin effects for 5 features in JUPITER_discovery and 2 in JUPITER_validation (FDR for interaction < 0.1; same directionality of the interaction effect in both cohorts).

None of the 610 significantly validated associations were modified by interactions with age (below and above the median) or non-randomized aspirin use after adjustment for multiple comparisons (FDR < 0.1).

DISCUSSION

The present study showed that participants randomized to statin in the JUPITER trial exhibited modifications in 610 non-targeted lipid features over a one-year period compared to placebo. Of these, 26 features had annotations, of which 16 were identified as polyunsaturated fatty acids (PUFAs), BALs (eicosanoids and docosanoids), or pathway markers. PUFAs are long fatty acid chains from omega-3 and omega-6 from which lipid metabolites such as eicosanoids and docosanoids, molecules containing a 20 and 22 carbon backbone, respectively, are derived.⁴⁰ Based on features with annotations, current results suggest alterations in BALs that potentially contribute to the pleiotropic effects conferred on statins, mostly favoring beneficial aspects associated with cardioprotective properties. We observed downregulation of pro-inflammatory mediators, oxidative stress markers, pro-platelet aggregation factors, and vasoconstrictor agents, and an increase in anti-inflammatory and pro-resolution lipid mediators.⁴¹

The anti-inflammatory effects of statins have been widely described and are independent of LDL-C.⁴ Therefore, combined lipid- and inflammation-lowering effects are additive to the overall vascular benefit of this class of medication.³ Consistently, current findings indicate reduction in AA, which has primarily pro-inflammatory activity and is a precursor of pro-inflammatory mediators,⁴¹ and increase in molecules with pro-resolution and/or anti-inflammatory-related bioactivity, including 18-HEPE,⁴² 5,6 DiHETE,¹⁸ and 19-HETE⁴³ in the statin-randomized group compared to placebo.

Downregulated AA is likely linked to statin-induced inhibition of platelet phospholipase A2⁴⁴ or to a higher conversion rate of AA into its end products. Several previous randomized investigations have reported increased AA levels with statin, but differences should be noted. For example, the current study samples consisted of apparently healthy non-hyperlipidemic participants with elevated hs-CRP, while other investigations included patients with lipid disorders,⁴⁵ atherosclerosis,^46,47 or chronic inflammatory conditions.⁴⁸ Further, we examined the effects of high-dose rosuvastatin, classified as a high-potency statin, whereas previous findings mostly focused on randomized low to moderate doses of low-potency statins such as simvastatin and pravastatin.^45,46 Therefore, statin effects on the formation and conversion of PUFAs, particularly AA, may be statin-specific and dose-dependent, as also speculated by a previous study.⁴⁹ Additionally, underlying risk factors and clinical conditions can also limit or interfere with statin activity on AA formation and metabolism.

Specialized pro-resolving mediators (SPMs) constitute a wide array of molecules that restrain inflammation and resolve the infection, limiting acute inflammatory responses and facilitating the clearance of tissue pathogens and dying cells from sites of inflammation.^15,50 Our results suggest that statin therapy increased the conversion of 18-HEPE, an SPM derivative from EPA and a precursor of E-series resolvins.⁴² This BAL upregulation with statin therapy may have cardioprotective effects by preventing pro-inflammatory activation of cardiac fibroblasts that produce IL-6 and monocyte chemoattractant protein-1 macrophages, and subsequent fibrosis, as these effects were observed after isolated 18-HEPE administration in mice.⁴² Interestingly, current sex-stratified models indicated that the 18-HEPE was upregulated by statin therapy only in women.

We also observed upregulation of 5,6-DiHETE, likely from EPA metabolism via the 5,6-epoxyeicosatetraenoic acid intermediate. Vasculature inflammatory responses induced by histamine application have been inhibited both in-vivo and in-vitro with 5,6-DiHETE pre-treatment.¹⁸ Through cell-specific Ca²⁺ channels modulation, 5,6-DiHETE attenuated smooth muscle cells dilation and endothelial barrier dysfunction, reducing vascular hyperpermeability.¹⁸ Baseline levels of this BAL correlated negatively with hs-CRP and glycA, a pro-inflammatory glycoprotein,²⁶ and its delta values correlated positively with delta HDL-C values. Moreover, 5,6-DiHETE may have a crucial role in the promotion of healing by inhibiting excessive or sustained inflammation.¹⁸

Despite decreased levels of AA, we detected upregulation of its metabolite 19-HETE in the statin group. It is possible that the minimal use of rosuvastatin as substrate of CYP enzymes^51,52 lowers the competition for CYP, potentially increasing the availability of this enzyme that may utilize AA preferentially in comparison to other PUFAs. Physiological actions 19-HETE, a product from AA metabolism, include a cardioprotective effect against angiotensin II-induced cardiac hypertrophy and renal effects. Importantly, 19-HETE is an endogenous antagonist of many of the physiological actions of its the structurally related 20-HETE.⁴³ 19-HETE reverses the pro-inflammatory activity that involves the secretion of inflammatory cytokines (IL-6, IL-8, TNFα) and renal vasoconstriction caused by 20-HETE.^19,53

Our findings also corroborate previous regarding antiplatelet effects of statins.⁶ The biosynthesis of the marker of platelet activation thromboxane A2 (TxA2) is reflected by its downstream inactivated product 11-d TxB2,⁵⁴ significantly decreased in statin-allocated patients of the current study as well as in a previous report with randomized hypercholesterolemic patients.⁵⁴ Furthermore, the anti-inflammatory effects of statins may also suppress 11-d TxB2 generation through transcellular synthesis in leukocytes and endothelial cells that occurs in the presence of inflammation.⁵⁵

Antioxidant effects of statins⁵ may be associated with an increase in hepoxilins. Hepoxilins have been suggested as oxidative stress-induced protective lipid mediators in cells with high glutathione peroxidase capacity.⁵⁶ This hypothesis seems reasonable since this population has high hs-CRP levels, which is associated with increased oxidative stress.⁵⁷ Moreover, statin treatment significantly increases circulating concentrations of glutathione peroxidase enzyme.⁵⁸

Statin immunomodulatory activity⁹ potentially contributes to the polarization of macrophages to alternatively activated M2 states, which have pro-resolving phenotype.⁵⁹ Previous work showed that M2-polarized macrophages have the ability to transform 12-HETE, a potent proinflammatory lipid mediator that enhances angiotensin II-induced vascular constriction, into products that no longer have this activity. One of these products is the tetranor 12-HETE, which was increased in the statin group in this study. Tetranor 12-HETE is a degraded product of 12-HETE that does not activate vasoconstriction and contributes to the anti-inflammatory functions of alternatively activated M2 macrophages.⁵⁹ Both tetranor 12-HETE and hepoxilins are end products derived from the AA 12-lipoxygenase pathway.⁶⁰ This may indicate that high-intensity statin therapy potentially activates alternative routes and processes that suppress the formation or accumulation of 12-HETE.

Although the current study focused on one-year statin effects on molecules with bioactive properties, the statin-induced decrease in LPE (18:2) and increase in LPE (20:2) were noteworthy. These glycerophospholipids showed the most pronounced associations with statin therapy among the metabolites with annotation (Supplemental Figure S2 and Supplemental Table S3). LPEs have been described as a reservoir for arachidonic acid as a preferable site for enzymatic binding.⁶¹ Specifically, LPE (18:2) has been significantly associated with an increased odds of coronary artery disease.⁶² Changes in LPE (18:2) correlated positively with ΔLDL-C and ΔTC (Supplemental Figure S4). No reports particularly regarding LPE (20:2) have been found, but our results showed that one-year changes in LPE (20:2) inversely correlated with ΔLDL-C and ΔLpPLA2 mass and activity. Particularly, there was a substantial difference in LPE (20:2) point estimates between discovery and validation. This discrepancy may indicate a higher sensitivity of this marker to the sample’s characteristics since the validation cohort was statistically older, comprised more men and more hypertensive subjects, and had lower levels of TC and triglycerides.

Contrasting with the cardioprotective and beneficial findings regarding statin effects on the bioactive lipidome, current results showed downregulated 12-HEPE, product from EPA metabolism. 12-HEPE is a SPM and promotes glucose uptake in adipocytes and muscle.⁶³ The reduction in this lipid mediator involved in circulating glucose may provide insights into the modestly increased risk of diabetes onset or progression associated with statin use,^64,65 but further extensive investigations are necessary.

Our results had several strengths, including the novel BAL assay, large sample sizes, validation of findings in a different cohort, a study design that minimized laboratory drifts, and longitudinal data. Further, there is no similar investigation with such an extensive panel of BALs and other lipid features in a randomized statin study. The study samples were determined by the rigor and eligibility criteria of a landmark clinical trial, which, by design, ensures no systematic differences between treatment groups that could influence the outcome of interest⁶⁶.

As for potential limitations, the current discovery cohort is not entirely representative of the total JUPITER trial population since it is restricted to a small portion (3.3%) randomly selected for BALs and other lipid features profiling. On the other, the validation cohort comprises 79% of all control individuals from a nested CVD case-control study within the JUPITER trial,²³ with comparable distributions of age, LDL-C, HDL-C, and sex. Additionally, the generalizability of the findings is limited, and future studies should examine more diverse populations.

One may argue that the effects in the bioactive lipidome predominantly result from statin-induced modifications in the cholesterol profile rather than a direct effect of the drug, or that these changes reflect statin effects on other factors beyond lipid metabolism, such as mononuclear leukocytes (i.e., lymphocytes and macrophages)^67,68 or coagulation proteins.⁶⁹ However, it is reasonable that a complex interaction within concurrent metabolic changes interplays with effects on multiple structures and systems, contributing to the clinical pleiotropic impact of statin treatment. Ultimately, the incidence of major cardiovascular events is likely reduced by a multifactorial process.⁷⁰ Another caveat to be considered is the impact of freezing and thawing, as well as ex-vivo lipids oxidation, on the identification and concentration of plasma bioactive lipids. These conditions have been tested by previous work using the same LC-MS method.²⁹ The authors showed that artifactual, non-enzymatic modification during sample collection, storage, processing, and analysis had minimal impact on the presence or levels of these metabolites.

In summary, current results suggest that high-dose rosuvastatin therapy upregulates BALs with anti-inflammatory and pro-resolution bioactivity. Concurrently, there was a decrease in levels of factors with vasoconstrictor, pro-inflammatory, and pro-platelet aggregation bioactivity. The untargeted LC-MS approach used in this study indicates evidence of statin effects on several unknown biomolecular structures, demonstrating that there is so much to be explored in the bioactive lipidome and highlighting the potential of these markers to understand drug effects better. In conclusion, the present study showed that after one year, participants randomized to statin exhibited modifications in 610 plasma lipid features compared to placebo, including 324 positive and 366 negative associations. We comprehensively linked identified molecular features to their bioactivity in order to provide information that potentially explains the protective role of statins against atherosclerosis and cardiovascular disease as the first step to pave a better understanding of underlying physiological-related mechanisms.

Highlights.

• Bioactive lipids can serve as metabolic footprints helping to uncover biological processes underlying the pleiotropic effects of statin.

• High-dose rosuvastatin therapy upregulated BALs with anti-inflammatory and pro-resolution bioactivity.

• High-dose rosuvastatin therapy decreased the levels of vasoconstrictor, pro-inflammatory, and pro-platelet aggregation factors.

• The analysis of bioactive lipids provided information that potentially explains in-depth the protective role of statins against atherosclerosis and cardiovascular disease.

Non-standard Abbreviations and Acronyms

BAL

Bioactive Lipid

BL

Baseline

BMI

Body mass index

CVD

Cardiovascular disease

d-TX

Dehydro-thromboxane

diHETE

Dihydroxyeicosatetraenoic acid

FDR

False Discovery Rate

GlycA

Glycoprotein acetylation

HDL-C

High-density lipoprotein cholesterol

HEPE

Hydroxyeicosapentaenoic acid

HETE

Hydroxyeicosatetraenoic acid

HETrE

Hydroxyeicosatrienoic acid

HMG-CoA

3-hydroxy-3-methylglutaryl coenzyme A

hs-CRP

High-sensitivity C-reactive protein

Hx

Hepoxilin

JUPITER

Justification for the Use of statins in Prevention: an Intervention Trial Evaluating Rosuvastatin

LC-MS

Liquid chromatography-mass spectrometry

LDL-C

Low-density lipoprotein cholesterol

Lp(a)

Lipoprotein (a)

LPE

Lysophosphatidylethanolamine

LpPLA2

Lipoprotein-associated phospholipase A2

NMR

Nuclear magnetic resonance

PUFA

polyunsaturated free fatty acids

sPLA2

Secretory phospholipase A2

SPM

Specialized pro-resolving mediators

TC

Total cholesterol

Y1

Year-one