Male subjects were randomly selected from the ongoing PRAEDIAS prevention study in the Department of Internal Medicine III at Carl Gustav Carus Medical School, Technical University Dresden as previously described [27]. Each participant provided written informed consent and the study was approved by the Dresden Ethics Committee (EK139092001) in accordance with the Declaration of Helsinki principles. All test subjects underwent a standardised clinical examination including a comprehensive metabolic characterization. Only individuals who did not manifest acute inflammatory processes (C-reactive protein <5 mg/L), diabetes mellitus or severe renal or hepatic diseases, participated in the study. Subjects should not have received antihypertensive medication or therapy that might have affected lipid metabolism. The study population consisted of 70 men including 19 subjects with hypertension. Basal anthropometric and clinical data are presented in Supplemental Table S1 (Supplementary Materials).

Blood samples for lipid profiling were taken after overnight fasting and EDTA-plasma was prepared by 10 min centrifugation at 4°C and 3000 g. All samples were immediately shock-frozen in liquid nitrogen and stored at −80°C until analysed.

Homeostasis model assessment of insulin resistance (HOMA-IR) index was calculated as (fasting insulin [µU/mL]×fasting glucose [mM])/22.5 [28].

Blood pressure was measured in accordance with WHO guidelines. The diagnosis of hypertension was based on systolic blood pressure ≥140 mmHg and/or diastolic blood pressure 90≥mmHg.

Plasma triglycerides, total cholesterol, HDL and LDL cholesterol were determined by standard enzymatic methods on a MODULAR analyser (Roche, Indianapolis, IN), free fatty acid on a COBAS MIRA analyser (Global Medical Instrumentation Inc, Ramsey, MN), and plasma glucose on a DX80 analyser (Beckman-Coulter, Fullerton, CA). HbA1C was measured by HPLC (Bio-Rad Laboratories, Richmond, CA). Plasma insulin levels were determined by an enzyme-linked immunosorbent assay (Asbach Medical Products, Obrigheim, Germany).

Synthetic lipid standards were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL). Water (LC-MS grade) was purchased from Fisher Scientific (Loughborough, United Kingdom); chloroform, methanol and ammonium acetate were of Liquid Chromatography grade and purchased from Fluka (Buchs SG, Switzerland). Methyl-tert-butylether (MTBE) and 2-propanol were purchased from Sigma-Aldrich Chemie GmbH (Munich, Germany).

Plasma samples were thawed and extracted with MTBE as described in [20]. Briefly, 20 µL of EDTA plasma was placed in a 2 mL vial (Eppendorf, Hamburg, Germany). 320 µL of the internal standards mixture consisting of 1,2-di-O-octadecyl-sn-glycero-3-phosphocholine, 1,2-di-O-phytanyl-sn-glycero-3-phosphoethanolamine and N-heptadecanoyl-D-erythro-sphingosylphosphorylcholine with a concentration of 3.2 nM for each lipid in 94% methanol was added and the solution was thoroughly mixed. Afterwards, 1 mL of MTBE was added and the mixture was vortexed at 20°C for one hour. Then 250 µL of water were added and tube thoroughly vortexed. After centrifuging for 1 minute at 4000 rpm on a Minispin centrifuge (Eppendorf, Hamburg, Germany,) 800 µL of the upper organic phase was transferred into a new vial and stored at −20°C until analysis. For mass spectrometric analysis, 8 µL of the extract were diluted with 80 µL CHCl₃/MeOH/2-propanol 1/2/4 (v/v/v) containing 7.5 mM ammonium acetate in a well of a 96 well plate (Eppendorf, Hamburg, Germany) and then sealed with aluminum foil.

Mass spectrometric analysis was performed on a hybrid LTQ Orbitrap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) equipped with a robotic nanoflow ion source TriVersa (Advion BioSciences Ltd, Ithaca NY) using chips with 4.1 µm nozzle diameter. The ion source was controlled by Chipsoft 6.4. software (Advion BioSciences) and operated at the ionization voltage of 0.95 kV and gas pressure 1.25 psi. Plates with lipid extracts were chilled down to 12°C.

MS survey scans were acquired in positive ion mode using the Orbitrap analyzer operated under the target mass resolution of 100,000 (Full Width at Half Maximum, FWHM), defined at m/z 400 under automatic gain control set to 1.0×10⁶ as the target value [24]. Where specified, targeted MSⁿ experiments were performed using collision-induced dissociation (CID) mode using the linear ion trap analyzer of the LTQ Orbitrap machine [24], [29].

Raw data files acquired from analyzed samples were converted into *.mzXML format by readw.exe utility (a tool of Trans-Proteomic Pipeline software collection, downloaded from http://tools.proteomecenter.org/wiki/index.php?title=Software:ReAdW). Mass spectra were further processed by LipidX software developed in-house. Altogether, the data set consisted of 151 high-resolution survey mass spectra including 12 blank controls. Spectra acquired within 28 s to 120 s from the start of sample infusion (timing required to stabilize the analyte flow and electrospray, as was judged by total ion current (TIC) trace) were averaged and recalibrated using m/z of synthetic standards SM 35∶1 and PC –O 20∶0/-O 20∶0 as references. Recalibrated spectra were further aligned such that related peaks were matched within the full dataset. Only peaks detected at the signal-to-noise ratio above the factor of 5 and recognized in more than 20% of all spectra were further considered. Identification of lipid species relied on accurately determined masses [24] considering a mass accuracy of better than 4 ppm and a retrieval rate of 90% for all plasma samples. Peaks also recognized in blank controls were excluded. Lipidomics analysis covered 10 major lipid classes: cholesterol (Chol), cholesterylesters (Chol-FA), phosphatidylcholines (PC, PC-O), lysophosphatidylcholines (LPC), phosphatidylethanolamines (PE, PE-O), sphingomyelins (SM), diacylglycerols (DAG) and triacylglycerols (TAG).

Species of PC, PC-O and LPC were quantified by the intensity ratios of their peaks to the internal standard PC-O 18∶0/-O 18∶0; PE and PE-O species – to the internal standard PE-O 20∶0/-O 20∶0; SM species – to the internal standard SM 35∶1. The abundance of individual TAG, DAG and Chol-FA species was determined by dividing their absolute intensities with the average of the absolute intensities of the three internal standards.

The abundance of free cholesterol was determined from the intensity of the positively charged ammonium adduct at m/z 404.3892. As an integrated abundance measure of cholesteryl esters, we considered the intensity of the peak with m/z 369.3521 (Chol- moieties). This fragment is produced in the ion source from ammonium adducts of cholesteryl esters via neutral loss of acyl chains [18].

Two-tailed bivariate Pearson correlation tests were applied to evaluate the correlation between plasma levels of total cholesterol and triglycerides as measured by routine clinical chemistry methods and sums of plasma levels of cholesterol and triglyceride species obtained by mass spectrometry.

To obtain a multivariate preliminary data survey a “Principal Component Analysis” of all 95 lipid species and Chol-moieties integral index was performed in order to assemble highly correlated lipid species into common factors. Using “Eigenvalues” over 3 and “Varimax” rotation with Kaiser Normalization five factors were extracted. Factor composition is shown in Supplemental Table S3. The association between lipidomic profile and hypertension was subsequently examined using a model of multivariate analysis of variance (MANOVA) with these factors as dependent variables and hypertension status as an independent factor, thereby regarding BMI and HOMA as covariates. The specific effect of each factor on hypertension status was then tested by univariate analysis of variance (ANCOVA) with p-values corrected for multiple testing (Bonferroni procedure).

In order to assess the impact of BMI and HOMA-IR on the plasma lipid profile of the study population, cut-offs for maximal discrimination of two groups were calculated by repeated discriminant function analysis using triglyceride species as discriminating variables. The resulting cutoffs were 27.5 kg/m² for BMI (60^th percentile; 87.1% correct classification), and 3.5 for HOMA-IR (67^th percentile; 80.0% correct classification).

For comparisons among groups (insulin-sensitive vs. insulin-resistant and normotensive vs. hypertensive) a general linear model was applied with lipid species as dependent variables and HOMA or the hypertension status as fixed factor. Since most of the lipid species were correlated (by bivariate correlation analysis) to BMI, this parameter was used as covariate in both models evaluating the effects of insulin resistance and hypertension on lipid species. Additionally, HOMA was used as a covariate in the model calculating the effect of hypertension on lipid profile. Adjusted means of lipid species were taken from both models for further calculations.

Data are given as mean percentage changes or as mean with a 95% interval of confidence unless otherwise stated. A value of p<0.05 was considered statistically significant. All statistical analyses were performed with the SPSS statistical package (v.16.0 for Windows; SPSS, Chicago, IL, USA).

Basal clinical data show that men with hypertension had significantly higher BMI, WHR, fasting plasma glucose and HOMA (Table 1). The prevalence of hypertension was 16.7% (7/42) in lean and 42.9% (12/28) in obese men. Conversely, the prevalence of obesity (BMI>27.5 kg/m²) was 63.2% (12/19) in the hypertensive and 31.4% (16/51) in the control group (Figure 1). Furthermore, 47.4% (9/19) of the hypertensive group had a HOMA-IR index >3.5, whereby 8 were obese (Figure 1). In comparison, the prevalence of insulin resistance in the control group was 27.5% (14/51), including 11 obese individuals (Figure 1). Cumulatively, these data confirm an increased prevalence of obesity and insulin resistance in patients with hypertension, which were associated with significantly higher plasma triglyceride levels (Table 1). These differences disappeared after controlling the triglyceride values for BMI and HOMA.

A top-down shotgun lipidomics workflow [24] was optimized for high-throughput clinical screens (Figure 2). Internal standards were spiked into plasma samples prior to one-step lipid extraction by MTBE. Total lipid extracts were directly infused into a LTQ Orbitrap mass spectrometer and survey mass spectra acquired within less than 3 min time at the target mass resolution of 100,000 (FWHM). High mass resolution, better than 4 ppm mass accuracy and practical compositional constraints identified lipid species directly, solely relying on their accurately determined m/z [19], [24]. Each plasma sample was analyzed with 2 technical (independent analysis of the same extract) replicas. The dataset, organized in a form of a flat-file database by LipidX software, comprised 151 survey mass spectra acquired from 70 individual plasma samples. Within the imposed compositional constraints, LipidX quantified 95 lipids from 10 major classes: 5 lipids belonging to Chol-FA, 4 to DAGs, 5 to LPCs, 13 to PCs and 13 to PC-Os, 8 to PEs and 5 to PE-Os, 10 to SMs, 31 to TAGs, and free cholesterol. The total quantities of TAGs and cholesterol determined by mass spectrometry were strongly correlated with clinical indices (Figure 3).

Multiple correlations of individual lipid species were analyzed by the method of PCA. Highly correlated lipid species were assembled into five common factors with Eigenvalues over 3 (Supplemental Table S3). Factor 1 included PCs, LPCs, and TAGs with short saturated fatty acid moieties; Factor 2 included all PC-O and PE-O species except PE-O [40∶5]; Factor 3 mostly included TAGs with polyunsaturated fatty acid moieties; Factor 4 mainly included SMs and cholesterol esters; Factor 5 mainly included TAGs and DAGs with unsaturated fatty acid moieties (Supplemental Table S3). The individual regression coefficients of these factors were saved and used as dependent variables in a model of multivariate analysis of variance (MANOVA).

Based on clinical chemistry data obese individuals had significantly higher triglycerides and lower HDL-cholesterol, whereas LDL- and total cholesterol remained unchanged (Figure 4A). Corroborating these findings, lipidomics analysis revealed that the abundance of nearly all lipids classes was generally increased (Figure 4B) in obese subjects. The content of saturated TAG and DAG species (Figure 4B) having, on average, less than one double bond per fatty acid moiety, increased 2- to 3- fold.

According to clinical indices, individuals with insulin resistance (HOMA-IR >3.5) had a significant decrease in HDL-cholesterol, while changes in triglycerides and total cholesterol were insignificant (Figure 5A). The lipidomic screen (Figure 5B), controlled for the effects of BMI, revealed a pattern of predominantly decreased levels of LPC, PC, PC-O, PE, and SM, reaching significance only for PE 38∶2 species. There was also a clear trend toward increasing the content of PE-O species, while the content of TAGs varied considerably (Figure 5B). We therefore concluded that moderate insulin resistance is accompanied by specific, but minor changes of the lipidome, which might be partially concealed by the effects if obesity.

Altogether, shotgun lipidomics revealed that in the investigated cohort obesity and insulin resistance affect the lipidome in different ways. Since obesity-related changes are massive, special statistical considerations should be taken to reveal the effects of concomitant metabolic disorders.

Contrary to the effect of BMI both clinical indices and shotgun profiling did not reveal major differences in plasma lipidome of normotensive and hypertensive subjects (Figure 6A and B), while changes of moderate magnitude that were observed among almost all lipid classes. The most notable exception, however, was a statistically confident and coherent decrease in the content of free cholesterol and PC-O species (Figures 6 and 7). Interestingly, the PE-O content was also decreased (significantly for the three out of five detected species, Figures 6 and 7), while PE and PC remained almost unchanged – which is, again, in sharp contrast to lipidomics changes instigated by BMI increase. Therefore, we concluded that hypertension is specifically associated with the deficiency of free cholesterol, PC-O, and PE-O species in male plasma lipidomes and it is seemingly unrelated to obesity-dependent alterations.

A multivariate model (MANOVA) comprising PCA factors as dependent variables, hypertension status as independent factor, and BMI and HOMA as covariates indicated significant and independent effects of BMI (p< 0.001) and hypertension status (p = 0.032). The subsequent univariate analysis with p-values corrected for multiple tests (Bonferroni procedure) appointed to the significant difference between normotensive and hypertensive individuals for Factor 2 (p = 0.018), which emphasized the role of ether lipids in the pathogenesis of hypertension.

Further, structural analysis of these PC-Os using with MS³ in the negative ion mode [30] revealed that all affected highly unsaturated species, such as PC-O 36∶4, PC-O 38∶5, PC-O 36∶5, PC-O 38∶4 and PC-O 38∶6 comprised the arachidonic acid moiety (Supplemental Table S2). We further determined that PC-O species were present in the concentration of approx. 32.2 µM (individual concentrations ranging from 9.6 to 64.8 µM) in the whole studied population. Individuals with hypertension had significantly lower levels of these PC-O species than normotensive controls (26.2±2.8 vs. 34.4±1.7 µM, p = 0.014). Structural analysis of downregulated PE-O species by MS² in positive ion mode [31] revealed that these species were plasmalogens with high content of arachidonic acid moieties (Supplemental Table S2).