Serum choline plasmalogens, particularly those with oleic acid in sn-2, are associated with proatherogenic state.

Serum plasmalogens (Pls) (1-O-alk-1'-enyl-2-acyl glycerophospholipids) are of particular interest for studies on metabolic disorders associated with oxidative stress and chronic inflammation. Serum levels of Pls are known to correlate positively with HDL-cholesterol (HDL-C); however, few studies have examined serum Pls molecular species in association with pathophysiological conditions and their clinical significance. To clarify these, we determined serum levels of individual ether glycerophospholipids in Japanese asymptomatic cohorts (n = 428; 362 male and 66 female subjects) by LC/MS/MS, and examined their correlations with clinical parameters. We found that the proportion of choline Pls (PlsCho) among total serum phospholipids was significantly lower in the male group over 40 years old and was associated with multiple risk parameters more strongly than HDL-C. The abundance of serum PlsCho with oleic acid (18:1) in sn-2 exhibited the strongest positive correlation with serum concentrations of adiponectin and HDL-C, while being inversely associated with waist circumference and the serum levels of TG and small dense LDL-cholesterol. The characterization of serum ether glycerophospholipids verified the specificity of PlsCho, particularly the ones with 18:1 in sn-2, as a sensitive biomarker for the atherogenic state.

was obtained from all participants, and the studies were approved by the ethics committees of Hokkaido University, Teikyo University School of Medicine, and ADEKA Corporation, Ltd. Sera were separated from fasting blood in a blood collection tube containing coagulation accelerating agent by centrifugation at 1,500 g for 10 min, and then an antioxidant was added (dibutyl hydroxytoluene, fi nal concentration of 10 M) and metal chelator (EDTA, fi nal concentration of 1 mM), followed by immediate freezing at Ϫ 80°C until analysis. Measurements of body weight, stature, waist circumference, and blood pressure were recorded. Participants were asked to complete health questionnaires consisting of 24 questions with regard to their medical history, as well as their dietary, smoking, drinking, and exercise habits.

Measurement of ether glycerophospholipids
Serum lipids were extracted by chloroform/methanol (1:2, v/v) after the serum was freeze-dried. Briefl y, 1 ml of methanol, including internal standards [1,2-dimyristoyl-phosphatidylcholine (DMPC) and 1,2-dimyristoyl-phosphatidylethanolamine (DMPE)] and dibutyl hydroxytoluene, was added to the freezedried serum (100 l) and shaken for 30 s. Then, 0.5 ml of chloroform was added to each tube and shaken for 30 s. After letting the tube stand at room temperature for 30 min, the tubes were shaken, centrifuged, and the chloroform/methanol layer was harvested. This procedure was repeated three times, and the collected chloroform/methanol layer was evaporated until dry. The dried samples, including extracted lipids, were dissolved with methanol. Ether glycerophospholipids, including Pls, in ex- used not only to generate a standard curve for quantifying individual molecular species of ether glycerophospholipids but also to assess the precision and accuracy of the detection, as described previously ( 17 ). The mean recovery of 18:0-18:1 and 18:0-20:4 for PlsCho and PlsEtn, respectively, as well as DMPC, was 91.8%. Intra-assay (n = 10) coeffi cient of variation percentage of these phospholipids ranged from 5.1 to 8.5%.
Synthetic 16:0-18:1 1-O -alkyl-2-acyl-sn -glycero-3-phosphoethanolamine (PakEtn) synthesized enzymatically from the corresponding PakCho by exchanging choline to ethanolamine was used to generate a standard curve for quantifying PakEtn molecular species. The concentrations of individual molecular species of ether glycerophospholipids were determined using corresponding internal standards (DMPC for PlsCho and PakCho, and DMPE for PlsEtn and PakEtn; Avanti Polar Lipids, Inc.). The numbers of molecular species recorded were 24 and 30 for each class of alkylphospholipids (Pak) and Pls, respectively, all of which were major molecules determined by LC/MS/MS.

Condition of LC/MS/MS
LC/MS/MS was conducted following a slightly modifi ed method described previously ( 16 ). LC separation was performed using an Accela UPLC system (Thermo Fisher Scientifi c, Inc., Waltham, MA) with a BEH C8 column (1.7 m, 100 mm × 2.1 mm; Waters Corp., Milford, MA) at 60°C and a flow rate of 450 l/min. Mobile phase A consisted of water containing 5 mM ammonium formate, while mobile phase B consisted of acetonitrile. Mobile phase A was set at 80% at 0 min, decreased linearly to 20% at 1.5 min, to 10% at 18 min, and then to 5% at 20 min, and maintained at 5% for 1 min before being returned to starting conditions. The MS analysis was performed using a TSQ Quantum Access Max (Thermo Fisher Scientifi c, Inc.) equipped with an HESI probe in positive ion mode. The MS operating conditions were containing DHA (22:6) could be a useful biomarker for the early detection and long-term follow-up of AD.
Recently, particular attention has been paid to the involvement of Pls in metabolic diseases associated with oxidative stress and chronic infl ammation ( 6 ). Pls may be involved in the etiology of both T1D and T2D through oxidative stress. Patients with T1D exhibit absolute insulin defi ciency due to autoimmune destruction of insulin-producing pancreatic ␤ cells. In a longitudinal study, children who developed T1D were later found to have reduced serum levels of ether type choline glycerophospholipids (ChoGpl), including PlsCho at birth and consistently throughout medical follow-ups ( 7 ). Pls defi ciency may cause enhanced susceptibility of pancreatic ␤ cells to oxidative damage and trigger progression to diabetes. Oxidative stress may also be associated with T2D and related cardiovascular diseases. Notably, decreased PlsEtn levels and increased lipid peroxidation were found in LDL particles from obese patients with metabolic syndrome (MetS) and patients with T2D ( 8 ), which may be a risk factor for atherosclerosis. Plasma Pls may be a good marker for inferring the level of oxidative stress and predicting cardiovascular mortality in patients with end-stage renal disease ( 9 ). Plasma concentration of Pls also decreases with hypertension ( 10 ), as well as with age and chronic infl ammatory-related diseases such as atherosclerosis and AD ( 11,12 ). Pls function as reservoirs for precursor fatty acids, such as arachidonic acid and DHA, which generate bioactive lipid mediators. Eicosanoids derived from arachidonic acid play an important role in the regulation of vascular tone and possibly link Pls with hypertension via blood pressure regulation. Studies have postulated that Pls serve as endogenous antioxidants and protect membrane lipids and lipoprotein particles from excessive oxidation by scavenging reactive oxygen species via the vinyl-ether moiety (13)(14)(15).
To date, however, available data regarding the relationship between serum Pls and clinical disease entities are still limited, and the association of Pls molecular species with preclinical pathophysiological conditions is yet to be clarifi ed. Previously, we performed an advanced comprehensive analysis of serum ether glycerophospholipids, including Pls, by LC/MS/MS, which has allowed the accurate quantifi cation of each molecular species ( 16,17 ). The present study provides the reference data from 428 Japanese asymptomatic subjects on their serum levels of individual ether glycerophospholipid species. By analyzing the correlation of each molecular species with clinical parameters or biochemical measurements, we found that PlsCho with 18:1 in the sn -2 position strongly correlated with a wide range of risk factors for atherosclerosis and MetS. Our fi ndings suggest that serum PlsCho, like HDL-cholesterol (HDL-C), may serve as a sensitive antiatherogenic biomarker.

Study subjects and design
Asymptomatic subjects (n = 428; 362 male and 66 female subjects; age, 22-66 years old) who were referred for a routine health examination were enrolled in the study. Written informed consent optimized as follows: a spray voltage of 3,000 V, and capillary and vaporizer temperatures of 250 and 500°C, respectively. Nitrogen was used as both the sheath and auxiliary gas. Argon was used as the collision gas at a pressure of approximately 1.0 mTorr. The collision energy was 32 eV for PakCho and PlsCho, 20 eV for PlsEtn, and 18 eV for PakEtn. Data acquisition was performed with Xcalibur 1.3 software (Thermo Fisher Scientifi c, Inc.). Mass fragments used for identifying each class of Pls and Pak molecular species are summarized in supplementary Tables I and II, respectively. PlsEtn was quantifi ed according to the procedure described by Zemski Berry and Murphy ( 18 ). In brief, fragment ions at m/z 364, 391, and 392 were used for the identifi cation of the sn -1 position of PlsEtn containing hexadecanol (16:0), octadecanol (18:0), and octadecenol (18:1), respectively. PakCho and PlsCho, identifi ed by three characteristic fragments (precursor 184 derived from phosphocholine, as well as sn -1 origin and sn -2 origin fragments), were quantifi ed using a fragment ion at m/z 184, following the separation of each molecular species by UPLC. PakEtn, identifi ed by three characteristic fragments (neutral loss 141 derived from phosphoethanolamine, sn -1 origin fragments, and fatty acids derived from the sn -2 position), was quantifi ed using a fragment ion at m/z 141, following the separation of each molecular species by UPLC ( 19 ). The presence of Pls was further confi rmed by the disappearance of the peak upon treatment with acid.

Statistical analysis
Statistical analysis was performed using Excel (2010; Microsoft Corporation, Redmond, WA) with the add-in software Statcel 3 (OMS, Tokyo, Japan). Data are presented as mean ± SD or median and interquartile range. Results were analyzed by Tukey-Kramer post hoc tests for parametric data (supplementary Table III), and Steel-Dwass ( Table 1 ; supplementary Tables III-V) as well as Mann-Whitney U tests ( Table 1 ; supplementary Tables III-VI) for nonparametric data. Correlations among measurements were assessed with Spearman's rank tests ( Figs. 3, 4, 6 ; Tables 2, 3 ).

Background of asymptomatic subjects
Clinical parameters and serum biochemical measurements of the subjects were classifi ed by age (under 39 or over 40 years old) and gender (supplementary Table III).   The data indicate that male subjects over 40 years old ( у 40) exhibited various abnormalities in clinical and biochemical parameters, including waist circumference, systolic and diastolic blood pressures, blood glucose, the serum levels of UA, AST, ALT, ␥ -GTP, and TGs, as well as the atherogenic index of plasma (AIP) [calculated as log (TG/HDL-C)]. To confi rm that these у 40-year-old subjects belonged to the population at risk, serum concentrations of sdLDL, hsCRP, Hcy, and adiponectin were determined. The data of certain male subjects у 40 years old indicated that they were near the upper or lower limit of a normal serum range, suggesting that male subjects over 40 years old are associated with proatherogenic status (supplementary Table III).

Characteristics of serum ether glycerophospholipids
The serum concentrations of each class of Pls (i.e., PlsCho and PlsEtn) and Pak (i.e., PakCho and PakEtn), along with their class ratios (PlsCho/PlsEtn and PakCho/PakEtn) and their proportions (mole percent) in total phospholipids (PlsCho/PL, PlsEtn/PL, total Pls/PL, PakCho/PL, PakEtn/ PL, and total Pak/PL), were summarized according to gender and age ( Table 1 ). The results show that while the serum levels of PlsCho and PlsEtn were almost equal, the serum concentrations of PakEtn were much lower than those of PakCho. In addition, the PlsCho/PL ratio for the у 40-year-old male group was signifi cantly lower than those for other groups, whereas the PlsCho/PlsEtn ratio and total Pls/PL (mole percent) were higher among the р 39-and у 40-year-old whereas among the ethanolamine class, three alcohol species were distributed almost equally. Arachidonic acid (20:4), the major fatty acid in the sn -2 position, was common in ether glycerophospholipids. In addition, DHA was rich among the ethanolamine class, and linoleic acid (18:2) was specifi cally rich in PlsCho. The interrelationship among serum ether glycerophospholipids is summarized in Fig. 3 . A strong positive correlation was observed between the serum concentrations of PakCho and PakEtn ( r = 0.714, P < 0.01), presumably because they share a common biosynthetic precursor, 1-alkyl-2-acyl-glycerol. The synthetic pathway of PlsEtn from PakEtn appears to be supported by both the strong positive correlation between their serum levels ( r = 0.630, P < 0.01) and the marked similarity between their molecular profi les ( Fig. 1 ). However, the biosynthetic route of PlsCho is considered to be derived from PlsEtn, but not from PakCho ( 20,21 ). As such, the correlation between the serum levels of PlsCho and PlsEtn ( r = 0.611, P < 0.01) was higher than that between PlsCho and PakCho ( r = 0.521, P < 0.01). However, the molecular profi les of PlsCho and PlsEtn were remarkably different from each other ( Fig. 1 ), suggesting that a part of serum PlsCho species may have different origins than other serum ether glycerophospholipids .

Relationship between ether glycerophospholipids and risk factors for the atherogenic state
We found that the PlsCho/PL (mole percent) was signifi cantly associated with multiple risk factors for the atherogenic status and that these correlations were at similar, or even higher, levels than those of HDL-C ( Table 2 ). In addition, we found that among PlsCho species, PlsCho with 18:1 and 18:2 in sn -2 correlated positively with the serum adiponectin concentration ( r = 0.465 and 0.340, respectively), but inversely with waist circumference ( r = Ϫ 0.420 and Ϫ 0.371, respectively), the serum concentrations of TG ( r = Ϫ 0.528 and Ϫ 0.479, respectively), sdLDL-C ( r = Ϫ 0.311 and Ϫ 0.261, respectively), and AIP ( r = Ϫ 0.656 and Ϫ 0.574, respectively) ( Fig. 4 ). In contrast, no species of Pak or PlsEtn showed signifi cant associations with these parameters (data not shown ). The correlations with these parameters were compared among versatile classes or female groups, respectively ( Table 1 ). The statistical comparisons of age groups were adjusted by confounding factors such as diet, smoking, drinking, and exercise. Although the signifi cant differences in some of these confounding factors were observed among age groups, these confounding factors only had signifi cant effects on PL, but not other Pls-related parameters. Heavy drinkers (more than three times per week) showed signifi cantly higher serum concentrations of PL, but there were no signifi cant differences related to the frequency of drinking among groups (data not shown). The proportion of female subjects in the study population (15.4%) was exceedingly low, so the gender comparisons in this study might be limited.

DISCUSSION
Here, we report for the fi rst time that the PlsCho/PL (mole percent) not only was signifi cantly lower in the у 40year-old male group ( Table 1 ) but also showed stronger correlations with multiple antiatherogenic parameters than subclasses of glycerophospholipids containing 18:1 in sn -2 ( Table 3 ). The results indicate that among the corresponding glycerophospholipids, only PlsCho showed signifi cant correlations with all of these parameters. In contrast, diacyl choline (PtdCho) and diacyl ethanolamine (PtdEtn) glycerophospholipids with 18:1 in sn -2, as well as total 18:1 lipids in sera, correlated positively with TG, sdLDL-C, and AIP ( Table 3 ).
Of note, the serum levels of both classes of Pls molecular species were found to highly correlate with the serum HDL-C concentration, as described previously ( 22 ). The comparison of correlation coeffi cients among various molecular species in the у 40-year-old male subjects (n = 186) demonstrates that the amount of serum PlsCho with 18:1 in the sn -2 position correlated most strongly with the HDL-C concentration ( r = 0.741, P < 0.001) ( Fig. 6 ). Notably, the levels of both classes of Pls with docosapentaenoic acid (22:5) or 22:6 in sn -2 were also found to be associated with the HDL-C levels.   ( 26 ). Interestingly, a polymorphism of the human PEMT gene is associated with diminished activity and may confer susceptibility to nonalcoholic fatty liver disease (NAFLD) ( 27,28 ). NAFLD shares common clinical manifestations with MetS, such as abdominal obesity, T2D, dyslipidemia, and insulin resistance ( 29 ). Indeed, it has been demonstrated that NAFLD is associated with peroxisomal dysfunction and decreased plasma Pls, as observed in a comprehensive analysis of plasma lipids in NAFLD patients ( 30 ).
Moreover, malnutrition of methyl donors, such as methionine or choline, may cause the suppression of PEMT methylation, resulting in decreased PlsCho and elevated homocysteine levels, which is a risk factor for cardiovascular disease. These considerations suggest that an insufficient PEMT methylation process is, at least in part, associated with MetS and atherosclerosis. Our previous intervention study on hyperlipidemic subjects has shown that increase in plasma PlsCho by ingestion of myoinositol, which is a known nutrient to improve fatty liver disease, is accompanied by reduction in atherogenic sdLDL ( 31 ). This fi nding may provide direct evidence for the validity of plasma PlsCho as a biomarker for MetS and atherosclerosis.
Alternatively, the enhancement of enzymatic degradation of PlsCho may be considered the cause of decreased plasma PlsCho in a proatherogenic state. Exposure of hypoxia as well as thrombin stimulation to endothelial cells induce the activation of membrane-associated calciumindependent phospholipase A 2 , which selectively hydrolyzes membrane Pls and preferentially PlsCho with arachidonic acid in sn -2 ( 32 ).
PlsCho with 18:1 in sn -2 occupies only one-tenth of the whole PlsCho population. The comparison of the correlations of clinical and biochemical parameters with the class HDL-C ( Table 2 ). In addition, serum PlsCho with 18:1 or 18:2 in the sn -2 position also showed strong inverse correlations with a wide range of risk factors, such as age ( Figs. 4,5 ). Given the high susceptibility of their vinyl-ether moiety to oxidation ( 23 ), serum Pls may represent systemic markers for oxidative stress. However, only specifi c molecular species, i.e., PlsCho with 18:1 or 18:2 in sn -2, can be antiatherogenic indicators, despite the shared possession of vinyl-ether linkage in all Pls molecules. It is likely that besides serving as an oxidative index, PlsCho with 18:1 or 18:2 in sn -2 plays additional roles; it may be associated with different origins of other serum ether glycerophospholipids, which is evidenced by their specifi c molecular profi les ( Figs. 1-3 ).
Plasma Pls are primarily synthesized in the liver and intestine in addition to dietary sources of Pls, and are secreted in the circulating blood as lipoprotein components. Because the liver contains only low amounts of Pls, a large part of the Pls synthesized by the liver seem to be preferentially transported to the blood stream ( 24 ). In particular, PlsCho are predominantly localized in lipoproteins in blood milieu, suggesting their involvement in lipoprotein metabolism and/or function. PlsCho represent 3-5% of serum ChoGpl. ChoGpl are synthesized in the liver via two major pathways, i.e., the CDP-choline and phosphatidylethanolamine methylation pathways, and the latter is catalyzed by the hepatocyte-specifi c enzyme phosphatidylethanolamine N -methyltransferase (PEMT) through three steps during reaction with S -adenosylmethionine ( 25 ).
The CDP-choline pathway is the dominant route for ChoGpl synthesis in all mammalian tissues, while the PEMT methylation pathway is responsible for ‫ف‬ 30% of ChoGpl generated in the liver. Given that PlsCho are principally synthesized via PlsEtn, a part of plasma PlsCho may be generated through the hepatic PEMT methylation pathway Dashed lines indicate the r value of 0.18, which is taken to be statistically signifi cant with P < 0.01. species-dependent manner. Pls are relatively abundant in HDL fractions compared with other lipoprotein fractions ( 37 ). However, this cannot fully explain the observation that the correlation between the serum levels of PlsCho and HDL-C ( r = 0.699, P < 0.01) was considerably higher than that between PlsEtn and HDL-C ( r = 0.411, P < 0.01; Fig. 6 ), because the Pls class distribution is not significantly different among various lipoprotein fractions ( 37 ). The reason for the highly positive correlation between the serum levels of Pls and HDL-C remains elusive. A recent lipidomic analysis demonstrated that individuals with low HDL-C displayed changes in the quality of HDL particles, along with decreased Pls contents ( 38 ). In addition, phospholipids have also been suggested to play important roles in HDL-mediated RCT (39)(40)(41)(42). Interestingly, Pls-defi cient cells have been shown to exhibit reduced cholesterol effl ux by HDL ( 43 ), which may be associated with impaired cellular processing of cholesterol ( 44,45 ). These notions suggest the relevance of serum Pls to the physiologically important functions of HDL, including RCT. Further studies or subclass of glycerophospholipids containing 18:1 in sn -2 reveals the specifi city of PlsCho with 18:1 ( Table 3 ). These results verify the prominence of PlsCho with 18:1 as a biomarker for the atherogenic state. Interestingly, 1-palmitoyl-2-oleoyl-sn -glycerol-3-phosphocholine (16:0-18:1 PtdCho) was previously identifi ed as an endogenous ligand for the PPAR ␣ in the liver ( 33 ). We speculate that 16:0-18:1 PlsCho, the main species among PlsCho with 18:1 in sn -2, may have a similar function. Activation of PPAR ␣ exhibits pleiotropic benefi cial effects on plasma lipid profi le and vascular wall infl ammation ( 34 ). For example, its activation promotes cholesterol efflux from macrophages and enhances reverse cholesterol transport (RCT) by strongly inducing expression of genes that encode the cholesterol transporter ABCA1 ( 35 ). Moreover, PPAR ␣ activation facilitates the biosynthesis of Pls ( 36 ) as well as apolipoprotein A-I, as seen by the resultant increased plasma HDL and Pls levels.
Our study reveals that a wide range of Pls molecular species correlate positively with HDL-C in a class-or molecular Values represent Spearman's correlation coeffi cient by rank test, and the values over 0.18 are taken to be statistically signifi cant with P < 0.01. Identifi cation and quantifi cation of individual ether glycerophospholipid molecular species were performed with the method as described in the Methods. Diacylglycerophospholipid molecular species were identifi ed using the same mass fragments as used for identifi cation of Pak molecular species. Synthetic 16:0-18:1 of each PtdCho and PtdEtn were used for generating a standard curve to quantify individual diacylphospholipid molecular species.