Application of comparative lipidomics to elucidate postprandial metabolic excursions following dairy milk ingestion in individuals with prediabetes

Li Chen; Shiqi Zhang; Xiaowei Sun; Joshua D McDonald; Richard S Bruno; Jiangjiang Zhu

doi:10.1021/acs.jproteome.0c01009

. Author manuscript; available in PMC: 2022 May 7.

Published in final edited form as: J Proteome Res. 2021 Mar 15;20(5):2583–2595. doi: 10.1021/acs.jproteome.0c01009

Application of comparative lipidomics to elucidate postprandial metabolic excursions following dairy milk ingestion in individuals with prediabetes

Li Chen ^1,², Shiqi Zhang ¹, Xiaowei Sun ¹, Joshua D McDonald ¹, Richard S Bruno ¹, Jiangjiang Zhu ^1,^2,^*

PMCID: PMC8106868 NIHMSID: NIHMS1693298 PMID: 33719448

Abstract

Nutrient-dense dairy foods are an important component of a healthy diet. Recommendations, however, advise non- and low-fat dairy foods despite controversy concerning whether full-fat dairy foods adversely impact cardiometabolic health. Therefore, in this study, our objective is to examine the differential plasma lipidomic responses to non-fat or full-fat milk ingestion during postprandial hyperglycemia. Seven adults with prediabetes completed a randomized cross-over study in which glucose was consumed alone or with non-fat or full-fat dairy milk. Plasma collected at 30 min intervals for 3 h was used to perform untargeted lipidomics analysis. A total of 332 lipids from 20 classes and five lipid categories were detected at different time points during the postprandial period. Dairy milk, especially non-fat milk, protected against lipid changes otherwise induced by glucose ingestion. Co-ingestion of dairy milk with glucose, regardless of fat content, significantly altered lipid profiles although full-fat milk more substantially modulated lipid profiles. For the identified lipid biomarkers, 68.0% and 66.7% of the lipids significantly increased at 90 min and 180 min, respectively, while phosphatidylcholines (GPs) contributed most for the significant increase. Comparative lipidomics analysis indicated that both dairy milk induced significant changes in several lipid pathways, including glycerophospholipid metabolism and alpha-linolenic acid metabolism, to protect against postprandial hyperglycemia. In summary, our comparative lipidomics results suggested that dairy milk-mediated lipid modulation may be an effective dietary approach to reduce the risk of metabolic diseases among those with prediabetes.

Keywords: lipidomics, prediabetes, dairy milk, cardiovascular disease, biomarkers

1. Introduction

Prediabetes, which is diagnosed based on marginally elevated fasting glucose, increases the risk for diabetes. Importantly, 5–10% of persons per year with prediabetes will progress to diabetes and related cardiovascular diseases (1). Cardiovascular disease (CVD) is the leading cause of death in the US and worldwide (2, 3). Greater dairy milk consumption has been reported to associate with a lower risk of CVDs, and chronic intervention studies have shown enhanced effects of cardiovascular protection by dairy milk (4). Although low-fat and fat-free dairy foods are often suggested as components of healthy eating patterns, the adverse effects of full-fat dairy on cardiometabolic health have not been clearly demonstrated (5).

Significant progress has been made in the past two decades in the identification and quantification of lipid species of milk (6). Bovine milk generally contains 3–5% (w/v) of fat and several thousand of lipid species, with the dominant portion (about 98%) is in the form of triacylglycerols (TG) (6). TGs can be degraded by most cells, therefore, the cholesterol content of TG-rich lipoproteins (remnant cholesterol) is more likely to be the cause of CVD rather than triglyceridemia (7). Furthermore, a number of large-scale epidemiological studies have shown that elevated TG levels are independently associated with increased incidence of cardiovascular events (8). Whereas, glycerophospholipids (GP), sphingolipids (SP) and glycolipids are minor components (0.5–1%) of milk fat but are the major constituents of milk fat globule membrane. GP and SP are quantitatively the most important phospholipids (PLs) in milk (9). There is considerable evidence that PLs have important functions for human health, such as anti-inflammatory activity and reducing the risk of CVDs (10, 11). A controlled study revealed that co-ingesting non-fat or full-fat dairy milk with a high-carbohydrate breakfast similarly attenuated postprandial hyperglycemia that was otherwise induced by the breakfast alone (12). However, it is still unclear how dairy milk intake can alter the host plasma lipid profiles, and whether these profile changes influence cardiometabolic health outcomes.

Lipid biomolecules are critical for many biological processes, including the major constituents of biological membranes, the energy storage, the signaling in extra- and intracellular, and digestion (13–17). Hence, the identification and quantification of lipids are one of the fastest-growing research fields (13). Lipidomics, originally classified under the broader umbrella of metabolomics, is the study of the lipid changes that are induced by various factors (18, 19). Lipids consist of several structurally and functionally diverse molecular species that cover a broad range of polarity, from nonpolar to polar, and are present in biological samples at significantly varying levels, from the femtomole level to the micromole level (20). HRMS-based lipidomics studies of plasma samples have been powerful approaches to understand pathophysiological processes and develop diagnostic tests for metabolic diseases such as CVDs (21, 22). Therefore, this study aimed to perform a non-targeted lipidomics analysis to elucidate the fat content-dependent postprandial metabolic modifications of dairy milk in persons with prediabetes. To achieve this goal, a collection of archived samples from our previous study (23) was used, and a non-targeted, HRMS-based lipidomics workflow was applied. As illustrated in Figure 1, plasma samples were collected from adults with prediabetes who completed a three-arm, randomized, crossover trial in which they ingested a glucose challenge alone or in combination with an equal volume of non-fat or full-fat milk. Selected samples from three time points, namely the beginning (0 min), the middle point (90 min), and the end (180 min) of each 3-h postprandial period were analyzed for this study. Temporal changes of lipid profiles of dairy milk interventions were assessed, and the impact of low-fat or full-fat dairy milk on hyperglycemia-induced changes in the plasma lipidome were compared.

Figure 1. — The lipidomics workflow used in this study.

2. Experimental Section

2.1. Chemicals and reagents

High-performance liquid chromatography (HPLC)-grade solvents (chloroform, methanol, and acetonitrile) and chemicals (ammonium acetate and formic acid) were purchased from Fisher Scientific (Waltham, MA, USA). 15:0/18:1(d7) PE, 18:1(d7) LPC, and 18:1(d7) LPE were purchased from Avanti polar lipids, Inc. (Alabaster, AL, USA). 15:0/18:1(d7) PE, 18:1(d7) LPC, and 18:1(d7) LPE were dissolved in methanol and prepared as internal standards for quality control purposes.

2.2. Study subjects and design

The study protocol was approved by the Institutional Review Board at The Ohio State University (2015H0088) and registered at ClinicalTrials.gov (NCT02482610). Written informed consent was obtained from all subjects before their enrollment in the study. Complete details of the study, along with its primary and secondary outcome variables, have been reported previously (23). Briefly, adults with confirmed prediabetes completed a three-arm, randomized, crossover clinical trial following an overnight fast (10–12 h). Participants were provided all foods and beverages for 3 days prior to each study visit. Energy requirements were estimated using the Harris–Benedict equation and adjusted for physical activity (24). Archived samples from seven study participants (5 males and 2 females, Table S1) in the previous clinical trial were used for this study, and a total of sixty-three samples were used for lipidomics analysis. At each visit, they were randomized to ingest 75 g glucose dissolved in 473 ml water (GLU group), 75 g glucose dissolved in 473 ml of non-fat milk (NFM group), or 75 g glucose dissolved in 473 ml of full-fat milk (FFM group). Blood was collected prior to (t=0 min, T_{0 min}) and post-ingestion of each beverage at 90 min (T_{90 min}) and 180 min (T_{180 min}). Plasma was obtained by centrifugation (4°C, 15 min, 1500×g), aliquot into cryovials, and snap-frozen in liquid nitrogen. All samples were maintained at −80°C until analyses were performed.

2.3. Sample preparation

Plasma (20 μL) sample was mixed with 165 μL of chloroform/methanol (2/1, v/v) and 20 μL of an internal standard solution (a mix of 15:0/18:1(d7) PE, 18:1(d7) LPC, and 18:1(d7) LPE prepared in methanol). The mixture was vortexed for 2 min, incubated at room temperature for 30 min, and centrifuged at 22051×g for 10 min. Then, the lower phase was collected for analysis using an UPLC-Q-Exactive Orbitrap/MS system. In addition, 20 μL of each prepared sample was pooled to establish quality control (QC) samples to ensure the measurement quality of the lipid profiles. During the sequenced sample analyses, the QC sample was injected at the beginning and the end of the sequence, as well as after every ten injections, to monitor and evaluate the stability of the system and the reliability of the experiment data.

2.4. Instrumentation

A Thermo Vanquish UPLC system coupled with Q-Exactive Orbitrap mass spectrometer equipped with a heated electrospray ionization (HESI) probe (Thermo Fisher, CA, USA) was used in this study. An Acquity UPLC CSH C18 1.7μm 2.1×100mm column (Waters Corp, Milford, MA, USA) was applied for lipid separation in both negative and positive ionization modes with separate injections. Mobile phase A was 100% water, containing 5 mM ammonium acetate and 0.1% formic acid, and mobile phase B was acetonitrile and H₂O at a proportion of 95/5 (v/v), containing 5 mM ammonium acetate and 0.1% formic acid. A linear gradient elution program was set as 35% B from 0–1 min, increasing to 95% B over 20 min; after another 10 min at 95% B, the mobile phase composition was then returned to 35% B. The total run time was 35 min. The flow rate was 0.4 mL/min and the column temperature was 40°C. The MS/MS data were acquired at the ranges of m/z 100–1500. The resolution for data collection in full scan and fragment spectra were 140000 and 17500, respectively. AGC target was 3e⁶ and the maximum IT was 50ms for full scan mode. While the MS/MS mode parameters were set as automatic gain control (AGC) target at 1e⁵, maximum IT of 50ms, loop count at 10, isolation window of 1.2 m/z, and normalized collision energy (NCE) of 20, 50, and 80 eV.

2.5. Data processing and statistical analysis

The raw data obtained from plasma samples were identified using the MS-DIAL software (http://prime.psc.riken.jp/compms/msdial/main.html). The exported database contained lipid groups with key information including the metabolite name, formula and chemical structure, retention time, m/z, adduct type, and MS/MS spectrum. The repeatability of the developed extraction method was assessed by QC samples and the coefficient of variation (CV) less than 30% was used for multivariate and univariate statistical analysis. A multivariate statistical analysis was conducted using MetaboAnalyst 4.0 (https://www.metaboanalyst.ca). Principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were used to estimate the significant differences and the predictive ability of lipidomics profiles between different groups or time-points. The parameters R² (goodness-of-fit) and Q² (predictability) (25) were used to assess the quality of PLS-DA models and to avoid the risk of over-fitting (26). The values of R² > 0.8, Q² > 0.4, and both close to 1.0 were considered the signal with the outstanding predictive ability and fitness (27, 28). The univariate statistical analysis of repeated measure ANOVA (p<0.05) with post-hoc test was performed to evaluate whether the significant differences in lipid molecules exist among the three treatment groups. Potential lipid molecular biomarkers were selected based on the cutoff of a variable importance for the projection (VIP) value > 1 and a p < 0.05 (28, 29). Hierarchical clustering and correlation analyses of each sample were performed using Heml 1.0 (30). Lipid biosynthesis pathway analysis was performed by MetaboAnalyst 4.0 (https://www.metaboanalyst.ca) and KEGG based on the screened lipids with significant differences (p <0.05 and VIP>1). All mass spectra data have been deposited to MassIVE with access ID MSV000086825.

3. Results

3.1. Bioanalytical characteristics of the plasma lipidome of persons with prediabetes

Herein, our analyses combined the results from both positive and negative experiments to ensure full coverage of the most dominant plasma lipid species (Figure 2). After filtering out duplicate lipids that were detected in both positive and negative ionization modes, a total of 332 lipids from twenty classes and five lipid categories were detected in all treatment arms (Figure 2C and Table S2). Due to the limitation from available databases, certain peaks with excellent intensity and clear MS/MS fragmentation patterns were not recognized by MS-DIAL. These features were labeled as ‘unknowns. However, in recognition of their strong signals, they were retained in the analysis for potential future discovery and validation. As shown in Figure 2D, among the five lipid categories, namely 44.58% GP, 17.77% SP, 12.35% glycerolipids (GL), 11.75% fatty acyls (FA) and 2.11% sterol lipids (ST), and 11.45% unknown lipids, there were twelve classes of GPs (phosphatidylcholines, PC; lysophosphatidylcholines, LPC; phosphatidylcholines ether, EtherPC; phosphatidylethanolamines, PE; phosphatidylinositols, PI; lysophosphatidylethanolamines, LPE; lysophosphatidylcholines ether, EtherLPC; lysophosphatidylethanolamines ether, EtherLPE; phosphatidylethanolamines ether, EtherPE; phosphatidylinositols, PS; lysophosphatidylinositols, LPS; and phosphatidylglycerols, PG), one class of SP (sphingomyelins, SM), two classes of GL (diacylglycerols, DG and triacylglycerols, TG), three classes of FA (fatty acids, FFA; N-acyl ethanolamines, NAE, and acylcarnitines, CAR), two classes of ST (bile acids, BA and Sulfated ST) and 38 unknown lipids were detected in the pooled QC samples. These findings demonstrated that our approach exhibited broad coverage of plasma metabolites in our samples. This method was also used to analyze non-fat and full-fat milk beverages that were provided to the study participants. The results are shown in Table S3, the primary lipids in non-fat milk and full-fat milk were GPs that occupied 46.67% and 44.86% of total lipids, respectively. Despite that the number of GLs have a significant difference between non-fat and full-fat milk, within which 22 TGs can be detected from non-fat milk while 31 TGs can be detected in full-fat milk, the total numbers of lipids detected from both milk were quite similar (315 from non-fat milk vs. 321 from full-fat milk).

Figure 2. — Representative UPLC-MS results and the classification of lipids identified in this study. The representative TIC of pooled QC samples in positive (A) and negative mode (B), and the lipidome coverage (C) obtained by UPLC/Q-exactive orbitrap-MS. (D) The percentage distribution of 332 identified lipids in 20 classes (outer ring) and 5 categories (inner ring).

3.2. Dairy milk protects against lipid changes induced by glucose ingestion

To investigate the alterations in the levels of lipid species by dairy milk ingestion, lipidomics analysis of plasma samples was conducted and the results were compared by ANOVA, followed by PCA and PLS-DA analyses. PCA is an unsupervised modeling approach that provides an overview of any grouping patterns, while PLS-DA, as a supervised statistical model for multivariable analysis, helps to find the most influential differentiating features for separating the known sample groups based on PCA analysis (26). It can be seen that the GLU (Figure 3A and 3B), NFM (Figure 3E and 3F), and FFM (Figure 3I and 3J) at the different time points (T_{0 min} vs T_{90 min} vs and T_{180 min}) were partially or fully separated (only with PLS-DA). As shown in Figure 3A, 3E and 3I, PCA analysis showed a relative better separation in GLU group than that in NFM and FFM groups when comparing all three time points tested. However, PLS-DA analysis showed a better separation of the lipidomes between T_{0 min} and T_{180 min} in all GLU, NFM, and FFM groups (Figure 3B, 3F, and 3J).

Figure 3. — Comparative lipidomics detecting temporal changes of GLU, NFM and FFM groups during the dairy milk intervention of prediabetes. (A, E and I) Scores scatter plot for PCA of plasma metabolic profiling of the group GLU, NFM and FFM at the different time points (0, 90 min and 180 min), respectively. (B, F and J) Scores scatter plot for PLS-DA of plasma metabolic profiling of the group GLU, NFM and FFM at different time points, respectively. (C, G and K) Cross validation of the group GLU, NFM and FFM at different time points, respectively. (D, H and L) One-way ANOVA of different lipids in the group GLU, NFM and FFM at different time points, respectively. The red dot represents p<0.05。

In our model, Q² and R² have displayed a pairwise score plot for the top five components. The goodness-of-fit (R²>0.8) and good predictability (Q²>0.4) were shown in all five components of GLU groups (Figure 3C). The parameters for the classification of R²<0.8 and Q²<0.4 were observed in NFM (Figure 3G) and FFM (Figure 3K) indicating the moderate modeling quality of PLS-DA to these two groups. Furthermore, the univariate statistical analysis reported that thirty-four, two, and twenty-two lipids displayed significant (p<0.05) changes in GLU, NFM, and FFM groups among the three time points, respectively (Figure 3D,3H, and 3L).

Subsequently, the significantly changed lipids among the different groups were visualized in heatmaps (Figure 4A, 4B, and 4C; Table S3). The 28 significantly changed lipids from the GLU group including eight SM, eight PC, seven CAR, and five unknowns (Figure 4A). Meanwhile, a total of 17 lipids including seven SM, three PC, four CAR, and three unknowns were significantly changed (p<0.05) in the FFM group (Figure 4B). However, in the NFM group, only FFA 18:2 and an unidentified lipid (unknown 1) were significantly changed (p<0.05; Figure 4C). Figure 4D summarizes the number of changed lipids from seven classes, four lipid categories, as well as several lipids with unknown identifications from all three groups, as each dot represent one altered lipid from the labeled classes.

3.3. Lipid biomarkers identification among GLU, NFM and FFM groups

Before investigating the lipidomic differences among three groups after the beverage ingestion, we first performed the analysis of baseline data from T_{0 min} timepoint to verify whether any between-treatment differences in the plasma lipidome occurred at fasting. Comparing the lipid profiles of individuals with prediabetes in GLU, NFM and FFM groups at 0 min, PCA and PLS-DA showed that at the initial timepoint, the plasma lipid profiles of participants from three interventions were largely overlapped (Figure S3), which indicated that the lipid profile difference discovered at later timepoints were indeed induced by the different beverages, rather than an excursion in profiles due to simple time effect. Our T_{0 min} data also reflected the stability of the fasting lipidome, which may also be used as good references for future studies.

3.3.1. Lipid biomarkers identification among GLU, NFM and FFM at 90 min

As we were monitoring the accurate response of postprandial lipid profiles to dairy milk, 90 min and 180 min post-intervention were selected to check the mid-point and end-point changes in the lipidome. First, we compared the post-ingestion profiles at 90 min in the GLU, NFM, and FFM groups. PCA analysis of lipid metabolites showed a relatively clear separation between GLU-T_{90 min} and FFM- T_{90 min} groups (Figure 5A), but NFM- T_{90 min} partially overlapped with both of these groups. In contrast, PLS-DA analysis showed a clear separation among the GLU- T_{90 min}, NFM- T_{90 min}, and FFM- T_{90 min} groups (Figure 5B). These separation patterns suggested that while the ingestion of both milk products can drive the lipid profiles away from the GLU group, lipid profiles from FFM ingestion was more pronounced than those of the NFM group compared to the GLU group. High Q² and R² values (Figure S1) also indicated the good predictability of PLS-DA, which indicated the marked differences in regulating the lipid metabolism of individuals with prediabetes via GLU, NFM or FFM interventions at 90 min. After ANOVA analysis, 29 lipids from six classes (four CAR, five NAE, one PC, one PI, seven LPC, four LPE), two lipid categories (FA and GP), and seven unknowns were to identified to be significantly changed between treatments (Figure S1). Considering these findings, the heatmap was constructed to illustrate the relative levels of these lipids from three different groups at 90 min (Figure 5C). Based on the biomarker selection cutoffs of p< 0.05 (Talbe S5) and) and VIP>1(Table S6) and) (28, 29), 26 lipid biomarkers were identified, consisting of nine FA (four CAR and five NAE), ten GP (five LPC, four LPE and one PC) and seven unknowns (Table 1). Among these, the abundance of twelve lipids were increased, and six lipids were decreased in NFM compared with those in GLU, while seventeen lipids were increased, and nine lipids were decreased in FFM compared with that in GLU (Figure 5D). Furthermore, ten lipids were increased, and five lipids were decreased in FFM compared with those in NFM.

Figure 5. — Comparing the lipid profiles of prediabetes in GLU, NFM and FFM groups at 90 min post-intervention. (A) Scores scatter plot for PCA of plasma lipid profiles of the Group GLU, NFM and FFM at 90 min. (B) PLS-DA of plasma lipid profiles of patients collected from Group GLU, NFM and FFM at 90 min. (C) Heatmap of the significant changes of lipidomes in Group GLU, NFM and FFM at 90 min. (D) One-way ANOVA analysis of different lipids in three groups at 90 min post-intervention. Different labels of letters indicate statistical significance (p<0.05).

Table 1.

Identification of significantly different lipids among GLU, NFM and FFM at 90min and 180min

lipid name	RT (min)	m/z	Formula	Adduct type	Lipid Categories	90min		180min
lipid name	RT (min)	m/z	Formula	Adduct type	Lipid Categories	p.value ^a	VIP(Comp.1)^b	p.value^c	VIP(Comp.1)^d

CAR 10:0	3.281	316.24878	C17H34NO4	[M]⁺	FA	2.95E-03	1.5715
CAR 11:0	3.292	330.26456	C18H36NO4	[M]⁺	FA	5.53E-05	1.8822	2.46E-04	1.4649
CAR 18:1	13.399	426.35745	C25H48NO4	[M]⁺	FA	7.22E-03	1.5282	1.22E-03	1.2968
CAR 18:2	11.187	424.34232	C25H46NO4	[M]⁺	FA	7.01E-09	2.2211	3.47E-09	1.8533
CAR 9:0	1.814	302.23331	C16H32NO4	[M]⁺	FA			6.10E-03	1.2362
NAE 18:0	2.968	328.32178	C20H41NO2	[M+H]⁺	FA	4.25E-05	1.6677	1.11E-03	1.4302
NAE 18:1	14.771	326.30545	C20H39NO2	[M+H]⁺	FA	8.62E-05	1.2253	1.57E-02	1.163
NAE 19:1	22.047	340.32178	C21H41NO2	[M+H]⁺	FA	7.46E-04	1.8566	4.43E-09	1.8539
NAE 20:0	26.081	356.35254	C22H45NO2	[M+H]⁺	FA			3.02E-04	1.4766
NAE 20:1	10.92	354.33664	C22H43NO2	[M+H]⁺	FA	8.42E-05	1.8354	3.71E-05	1.5953
NAE 22:0	15.027	384.38306	C24H49NO2	[M+H]⁺	FA	3.96E-08	1.8667	4.06E-06	1.5269
DG 27:0	19.361	521.42065	C30H58O5	[M+Na]⁺	GL			8.68E-06	1.6368
LPC 16:1/0:0	10.81	496.34128	C24H50NO7P	[M+H]⁺	GP			1.17E-03	1.4299
LPC 18:1	10.562	522.35565	C26H52NO7P	[M+H]⁺	GP	1.21E-04	1.2605	3.53E-05	1.597
LPC 18:2/0:0	10.353	520.33893	C26H50NO7P	[M+H]⁺	GP			1.70E-02	1.1757
LPC 20:0/0:0	15.9	552.40131	C28H58NO7P		GP	2.30E-05	1.1715	3.05E-05	1.5047
LPC 20:1/0:0	13.874	550.38727	C28H56NO7P	[M+H]⁺	GP	1.11E-05	2.0135	7.00E-05	1.5883
LPC 20:2/0:0	12.218	548.36951	C28H54NO7P	[M+H]⁺	GP	2.39E-03	1.8868	2.18E-02	1.1591
LPC O-18:0	13.978	510.39142	C26H56NO6P	[M+H]⁺	GP			4.13E-06	1.7012
LPC O-18:1	12.031	508.37552	C26H54NO6P	[M+H]⁺	GP	1.33E-03	1.8863
LPE 18:0	13.643	482.32507	C23H48NO7P	[M+H]⁺	GP	3.78E-05	1.9959
LPE O-16:1	11.798	438.29779	C21H44NO6P	[M+H]⁺	GP	2.87E-04	1.6586
LPE O-18:1	14.364	466.32947	C23H48NO6P	[M+H]⁺	GP	1.78E-04	1.1302
LPE O-3:0	0.588	258.11038	C8H20NO6P	[M+H]⁺	GP	9.32E-04	1.9927
PC 18:2/20:4	24.008	806.57001	C46H80NO8P	[M+H]⁺	GP			2.09E-07	1.6058
PC 34:4	23.146	754.53748	C42H76NO8P	[M+H]⁺	GP			6.40E-04	1.2048
PC 35:3	26.258	770.57104	C43H80NO8P	[M+H]⁺	GP			7.80E-03	1.2328
PC 35:4	24.531	768.55548	C43H78NO8P	[M+H]⁺	GP			1.21E-06	1.5718
PC 36:4	25.626	782.56525	C44H80NO8P	[M+H]⁺	GP			4.94E-09	1.7695
PC 36:5	23.608	780.55322	C44H78NO8P	[M+H]⁺	GP			2.41E-04	1.5151
PC 37:4	27.927	796.58307	C45H82NO8P	[M+H]⁺	GP			7.51E-05	1.1144
PC 38:6	25.361	828.54822	C46H80NO8P	[M+Na]⁺	GP	8.15E-03	1.7377
PC O-34:1	29.289	746.59583	C42H84NO7P	[M+H]⁺	GP			7.00E-09	1.4643
PC O-36:3	29.835	770.60236	C44H84NO7P	[M+H]⁺	GP			3.97E-03	1.0479
PC O-40:8	28.659	816.58752	C48H82NO7P		GP			1.33E-04	1.10220
PI 16:0/18:2	23.113	833.52051	C43H79O13P	[M-H]^-	GP			2.50E-05	1.47740
PI 16:0/20:4	22.79	857.51971	C45H79O13P	[M-H]^-	GP			9.45E-04	1.44130
SM 18:2_2O/18:0	24.003	729.59094	C41H81N2O6P	[M+H]⁺	SP			1.53E-06	1.14780
SM 25:2_2O/11:0	23.819	729.59143	C41H81N2O6P	[M+H]⁺	SP			2.67E-09	1.64270
SM 28:3_2O/14:1	28.781	809.65729	C47H89N2O6P	[M+H]⁺	SP			3.96E-08	1.67070
SM 34:0_2O	24.382	705.58154	C39H81N2O6P	[M+H]⁺	SP			2.30E-02	1.72020
SM 34:1_3O	22.606	719.57007	C39H79N2O7P	[M+H]⁺	SP			1.92E-09	1.37700
SM 38:3_3O	29.692	771.60675	C43H83N2O7P	[M+H]⁺	SP			2.61E-06	1.65430
SM 40:3_2O	28.138	783.63812	C45H87N2O6P	[M+H]⁺	SP			6.14E-07	1.71550
Unknown 1	19.271	381.17398	Unknown	[M+Cl]^-	Unknown			1.09E-02	1.20590
Unknown 2	21.411	383.18958	Unknown	[M-H]^-	Unknown	2.19E-03	1.1872
Unknown 5	11.777	506.32498	Unknown	[M-H]^-	Unknown	2.47E-04	1.7508
Unknown 16	10.22	311.16953	Unknown	[M-H]^-	Unknown	1.91E-08	2.0327
Unknown 18	26.475	820.52661	Unknown	[M+H]⁺	Unknown			2.10E-02	1.06860
Unknown 20	17.188	326.37845	Unknown	[M+H]⁺	Unknown	4.36E-06	2.0273	1.38E-05	1.66310
Unknown 21	25.585	340.358	Unknown	[M+H]⁺	Unknown			1.48E-04	1.19460
Unknown 23	23.308	1077.34009	Unknown	[M+2H]²⁺	Unknown			3.73E-10	1.74300
Unknown 25	7.489	415.21237	Unknown	[M+H]⁺	Unknown	1.54E-05	1.0811	5.56E-10	1.45000
Unknown 27	7.606	460.26874	Unknown		Unknown	1.06E-04	1.5869	1.61E-07	1.71040
Unknown 28	21.479	489.31244	Unknown	[M+H]⁺	Unknown	6.70E-08	1.8258	4.34E-07	1.68260

Open in a new tab

^a,b,c, and ^d represent that more detail information such as p value, f value, FDR value and VIP value for component 1, 2, 3, 4, and 5 supplied in supporting information.

3.3.2. Lipid biomarkers identification among GLU, NFM and FFM at 180 min

The 180 min post-prandial timepoint was selected to check the end-point changes in the plasma lipidome of the studied prediabetic persons. PCA analysis (Figure 6A) and PLS-DA analysis (Figure 6B) showed a clear separation in all groups of GLU- T_{180 min}, NFM- T_{180 min}, and FFM- T_{180 min} groups. Both analyses also showed an improved group separations among the three study groups based on their lipidomics profiles. High Q² and R² values (Figure S2) from the PLS-DA analysis also indicated good predictability of PLS-DA model in differentiating the three groups at 180 min based on their lipid profiles, which suggested that both dairy milk still exert some level of lipid modulating effects after 180 min of ingestion, potentially due to the slower pharmacokinetic absorption of lipids. Furthermore, similar to the separation patterns at 90 mins, significant differences regarding the lipid modulating effects between non-fat milk and full-fat milk can be observed and FFM group distanced away from the GLU group even further compared to NFM. Based on one-way ANOVA analysis, fifty-nine lipids from eight classes and four lipid categories and eight unknowns were detected at significantly different levels among the three groups at 180 min (Figure S2). Based on these findings, a heatmap was constructed to illustrate the relative levels of these lipids (Figure 6C). When applying a biomarker selection cutoffs with p< 0.05 (Table S7) and VIP >1(Table S8), forty-four lipid molecular biomarkers were selected, which including ten FA (four CAR and six NAE), nineteen GP (seven LPC, ten PC and two PI), one GL (one DG), seven SP (seven SM) and seven unknowns (Table 1). Among these, twenty-seven lipids were increased, and nine lipids were decreased in NFM compared with that in GLU, while twenty-eight lipids were increased, and fourteen lipids were decreased in FFM compared with that in GLU (Figure 6D). However, there were less number of significant different lipids detected between NFM and FFM, with only five lipids were increased, and six lipids were decreased in FFM compared with those in NFM. This result suggested that the difference of lipid profiles between the two was smaller than their difference to the GLU group.

Figure 6. — Comparing the lipid profiles of prediabetes in GLU, NFM and FFM groups at 180 min post-intervention. (A) Scores scatter plot for PCA of plasma lipid profiles of the Group GLU, NFM and FFM at 180 min. (B) PLS-DA of plasma lipid profiles of patients collected from Group GLU, NFM and FFM at 180 min. (C) Heatmap of the significant changes of lipidomes in Group GLU, NFM and FFM at 180 min. (D) One-way ANOVA analysis of different lipids in three groups at 180 min post-intervention. Different labels of letters indicate statistical significance (p<0.05).

3.4. Dairy milk regulate the biosynthesis of PC and SM among other lipid metabolic pathways

After comparing the lipid profiles and analyzing individual lipid biomarkers, we performed lipid-based metabolic pathway analysis of these individuals with prediabetes post interventions. The metabolic pathways with impact value (IV)>0.1 were considered as relevant pathways involved in the conditions studied (31). Six disturbed metabolic pathways of the detected lipids from the pairwise comparison of GLU, NFM, and FFM groups were identified (Figure S4), including glycerophospholipid metabolism, ether lipid metabolism, arachidonic acid metabolism, glycosylphosphatidylinositol-anchor biosynthesis, sphingolipid metabolism, and alpha-linolenic acid metabolism. According to the KEGG database, there are two (glycerophospholipid metabolism and alpha-linolenic acid metabolism) and three (glycerophospholipid metabolism, sphingolipid metabolism, and alpha-linolenic acid metabolism) disturbed metabolic pathways of the potential biomarkers (VIP>1 and p<0.05) in both NFM and FFM groups at 90 min and 180 min, respectively, compared to GLU group at the same time points (Figure 7 and Figure S5). Glycerophospholipid metabolism (IV>0.1) was the pathway with the most significantly different lipids involvement, followed by alpha-linolenic acid metabolism (IV<0.1) and sphingolipid metabolites (IV<0.1).

Figure 7. — The overview of lipid pathway analysis based on the results of this study. Possible disturbed metabolic pathways of the potential biomarkers for GLU, NFM and FFM. Red and yellow color boxes indicated increased and decreased lipid classes during the three hours of dairy milk intervention, respectively, in both NFM and FFM groups. The green color box indicates the decreased level of lipids at 90 min and the increased level of lipids at 180 min in both NFM and FFM groups. Lipids without boxes are intermediates that were not detected but were used as background for enrichment analysis based on the KEGG database.

4. Discussion

4.1. Bioanalytical characteristics of the plasma lipidome of persons with prediabetes

MS-based lipidomics is well-established as a sensitive and effective approach for detecting biological responses via the comprehensive investigation of many lipid species at once (32). A nontargeted, high-throughput, HRMS-based lipidomics analysis of human plasma samples can detect and identify a suite of lipids species, and therefore is the choice of method used in this study. To evaluate the effects of dairy milk intervention on the plasma lipidome of individuals with prediabetes who completed a glucose challenge, a total of 63 samples and eight QC samples were analyzed in both positive and negative ion modes. While some lipids easily carried a positive charge as demonstrated in our representative chromatogram negatively charged lipid species were also covered in this study. Different from our results, lipidomics profiles of plasma of healthy humans after overnight fasting were reported to be 34.69% SP, 27.21% GP, 18.20% FA, 12.41% GL, 6.12% ST and 1.36% PR (30). The overwhelming majority of GP in healthy human plasma are PC and PE (30), while the top 2 classes of GP were PC and LPC in our results. Furthermore, the majority of FA in healthy human plasma were eicosanoids in one previous study (30), while it was CAR in our results. However, similar to our results, SM has been accounted for the largest fraction of SP in plasma and TG were the most abundant members of GL in another report (30). However, serum lipidomics analysis in type 1 diabetes revealed that PC, LPC, TG, and SM were among the lipids that were primarily identified (31). Although cholesterol was the most abundant ST in the plasma of healthy humans (30), only BA and some derivatives were detected in most of our specimens. Meanwhile, no PR was detected in our samples. This may due to the difference of plasma samples used in different studies (such as healthy individuals vs. individuals with prediabetes, overnight fasting vs. dairy milk ingestion). Meanwhile, additional optimization of our workflow could potentially increase the sensitivity and the coverage to specific groups of lipids as well. Overall, the results indicated apparent different lipidomics profiles between healthy individuals and individuals with prediabetes from different studies. It is also indicative that many bioanalytical factors, from the different sample preparation process to the different mass spectrometer instrument utilized, may greatly influence the detected lipidome among different studies.

4.2. Dairy milk protects against lipid changes induced by glucose ingestion

To investigate the alterations in the levels of lipid species by dairy milk ingestion, lipidomics analysis of plasma samples was conducted and the results were compared in each group. Taken together, these data suggested a time-dependent trend of lipidomic changes after the beverage ingestions. And glucose ingestion alone could already induce substantial changes of lipid profile over time. However, non-fat or full-fat milk could partially abrogate these significant changes at both 90 min and 180 min post-ingestion, consistent with observations from PCA and PLS-DA figures showing closer distance or some level of overlaps between T_{0 min} to T_{90 min} and T_{180 min} for both NFM and FFM groups. Also, non-fat milk showed a stronger effect of minimizing the lipid changes compared to full-fat milk, as demonstrated with less-drifted overall profiles and the lower number of significantly changed lipids. Interestingly, there were other studies with similar observations, and one has assessed the blood lipid effects of the consumption of whole milk or skim milk in diets for adult male subjects, and suggested that whole milk increases total cholesterol and low-density lipoproteins (LDL)-cholesterol more than skim milk (33). While our previous results (23) demonstrated that NFM and FFM similarly protected against downstream lipid peroxidation otherwise induced by glucose ingestion, and consequently protect against postprandial hyperglycemia (PPH) - mediated impairments in vascular endothelial function at similar levels.

4.3. Lipid biomarkers identification among GLU, NFM and FFM groups

As we reported previously in these same persons with prediabetes (23), the postprandial insulin responses, relative to 0 min, were higher at 30 −120 min in GLU group and at 30 −150 min in NFM and FFM groups. However, there was no significant difference in plasma insulin among the GLU, NFM, and FFM groups at either 90 min or 180 min (23). CARs, especially those with long chains, were suggested as early biomarkers of type 2 diabetes even before detected insulin resistance in animal studies, and were significantly associated with increased risk of type 2 diabetes (34). In this analysis, four of the twenty-one detected CARs were also significantly altered in NFM and FFM groups compared to GLU group. Similar to our results, significantly different acylcarnitine profiles were observed among subjects with normal glucose tolerance, prediabetic states, or type 2 diabetes (35). As initial metabolites of β-oxidation, the accumulation of long-chain CARs might reflect relatively severe conditions of fatty acid oxidation dysregulation and mitochondrial overload with an impaired tricarboxylic acid cycle (36). In our previous report, significantly lower plasma malondialdehyde (MDA) concentrations were detected in the NFM and FFM groups compared to the GLU group (23). Our findings here could indicate that a significant decrease of CARs (such as CAR 18:2) is related to the attenuated lipid peroxidation effects after milk ingestion. As the docosahexaenoic acid and eicosapentaenoic acid converted to N-acyl ethanolamines with anti-inflammatory properties (37), the increased NAEs (such as NAE 18:1 and NAE 19:1) could limit postprandial inflammatory responses after milk ingestion (38). On the other hand, we identified that LPC and LPE were significantly increased in both NFM and FFM groups in comparison to the GLU group (Figure 5D). Phospholipids have been reported to be an important component of biological membranes, and involved in many physiological functions such as anti-inflammatory activity, reducing the risk of CVD, and reducing cholesterol absorption (39). Park et al. (40) reported that LPCs and LPEs containing unsaturated fatty acids and free fatty acids contributed to an increased risk of coronary artery disease, and LPCs containing saturated fatty acids were correlated with a reduced risk. Furthermore, LPCs act as an inflammatory medium that modulates proliferation and apoptosis of endothelial cells, thereby affecting the development of atherosclerosis (41). LPC and LPE contain only one fatty acid moiety in their structures, while both of them in human serum have more saturated fatty acids than unsaturated fatty acids (42). Besides, there were 147 (44.67%) and 144 (44.86%) of phospholipids analyzed in non-fat milk and full-fat milk, respectively (Table S3). Therefore, the significantly increased LPE 18:0 and LPC 20:0/0:0 and decreased LPC 18:1 in NFM and FFM groups may play an important role in reducing the risk of CVD in persons with prediabetes.

When considering the number of lipid biomarkers idneitified at 90 min and 180 min (twenty-six vs. forty-five), the effect of milk ingestion became pronounced following more significant changes in lipid classes and categories at 180 min than that at 90 min. The significantly decreased CAR 18:2, NAE 18:1, and NAE 19:1 were also observed when comparing NFM and FFM to GLU at 180 min, suggesting the antioxidative and anti-inflammatory properties of dairy milk could last till 180 min. PC, LPC, and SM were the lipids species with the most significant changes. When taking a close look at our data, we noticed that DG 27:0 was significantly increased in both NFM and FFM compared to GLU, and the accumulation of DGs has been linked to the development of lipid-induced insulin resistance (43), which is in contrast to the benifitial effects as expected. However, not all DG isomers may activate protein kinase C to induce insulin resistance (44). PCs, which are mainly synthesized in the endoplasmic reticulum (ER), are one of the most abundant lipids in eukaryotic cell membranes and key players in membrane-mediated cell signaling (45, 46). Dysregulated lipid homeostasis in cardiac cell membranes of diabetic patients may be contributing to disrupted receptor tyrosine kinases signaling (such as insulin receptor) (47). Therefore, the significantly increased PC and LPC in NFM and FFM groups could play a role in reducing the risk of CVD for individuals with prediabetes. It was reported that PC 35:4, LPC 20:0 and LPC 18:0 inversly associated with future cardiovascular events (nonfatal myocardial infarction, nonfatal stroke, and cardiovascular death) and improve prediction of future CVD events and death (48). Therefore, the increased PC 35:4, LPC 20:0/0:0 and LPC 18:0/0:0 in NFM and FFM comparing to GLU at 180 min in our study may predict a positive effect of milk in protecting against CVD. Sphingolipids mainly include ceramides, sphingomyelin (SM), and glycosphingolipids (49) and exhibit a wide spectrum of biological functions depending on the cellular context (50). Accumulation of sphingolipid levels of plasma was shown to elevate the risk of cardiovascular complications in diabetics (47). After 180 min milk intervention, in NFM and FFM groups, four significantly down-regulated SMs (SM 18:2_2O/18:0, SM 28:3_20/14:1, SM 34:0_2O and SM 40:3_2O) were detected, which might indicate dairy milk could protect patients with prediabetes by reducing the SM leves and consequentially lowering the CVD risk. Based on our previously report (23), dairy milk, regardless of fat content, protects against postprandial hyperglycemia-mediated impairments in vascular endothelial function in adults with prediabetes by limiting oxidative stress responses that reduce nitric oxide bioavailability. Therefore, we speculated that the significantly changed lipids we reported here collectively could play an active role in reducing the risk of CVD for individuals with prediabetes through a variety of biochemical/biological mechanisms.

4.4. Dairy milk regulate the biosynthesis of PC and SM among other lipid metabolic pathways

In the study, the dairy milk regulated lipid metabolic pathways were largely in agreement with other previous studies that revealed how the functional lipids extracted from dairy milk exerting therapeutic properties by significantly regulating the glycerophospholipid or sphingolipid metabolisms involved in modulating signaling conduction and building permeability barriers, as well as alleviating oxidative stress and inflammation in inflammatory bowel diseases models (51, 52). Comparing to GLU, the biosynthesis of PC increased in both NFM and FFM groups at 90 min and 180 min, while the biosynthesis of SM decreased in both NFM and FFM groups only at 180min.

Based on KEGG-based analysis, we also summarized the lipid metabolic pathways and showed that PC could convert to LPC; SM and ceramide can transform into each other during the dairy milk intervention. Although we identified specific plasma lipids/lipid pathways relative to dairy milk-mediated attenuation of CVD risk factors, the present study was limited primarily to correlation analyses that can not establish causality. We also acknowledge that the size of our study cohort is relatively small, and the duration of intervention monitoring is only at 180 min, which is not directly correlated to clinical endpoint of CVD. However, these findings are expected to guide future hypothesis generation in animal and clinical studies to define the mechanism by which dairy milk or other dietary approaches alters postprandial metabolic responses. Additionally, our plasma lipidome analysis does not preclude the possible and profound lipidome changes in other important biological specimens such as liver and colon. Such analyses of dietary interventions along with the consideration for the health benefits of lifestyle changes or other risk reduction measures will collectively add values to the efforts of prediabetes and CVD interventions. Overall, our results are expected to support investigations that intend to utilize novel bioanalytical tools, such as MS-based lipidomics, to unravel the complex lipid modulation processes that are involved in the management of cardiovascular health of persons with prediabetes.

5. Conclusion

To our best knowledge, this is the first study to apply lipidomics to the study of dairy fat ingestion induced lipidome responses in the postprandial period. Our investigation of the plasma lipidome from a group of prediabetic individuals suggested that lipidomics analysis is a sensitive bioanaltyical tool to help elucidate the postprandial lipidomic responses to dairy milk. Both shared and unique lipid biomarkers from NFM and FFM groups can be observed over three-hours of interventions. Furthermore, our results may suggest that the increased PC and decreased SM indicated the dairy milk could reduce the risk of CVD for prediabetic individuals, while the difference in the fat content of milk could have differential effects on lipid regulations.

Supplementary Material

supporting info

Figure S1. Statistical analysis for the data obtained at 90min. (A) significantly changed lipids after one-way ANOVA analysis (p<0.05). (B) Q2 and R2 values after PLS-DA analysis.

Figure S2. Statistical analysis for the data obtained at 180min. (A) significantly changed lipids after one-way ANOVA analysis (p<0.05). (B) Q2 and R2 values after PLS-DA analysis.

Figure S3. PCA score plot comparing the lipid profiles of prediabetes in GLU, NFM and FFM groups at 0 min, which indicated indifference in baseline measurement at the beginning of each intervention.

Figure S4. Possible disturbed metabolic pathways by total lipids (RSD<30%) for GLU, NFM and FFM. 1. Glycerophospholipid metabolism. 2. Ether lipid metabolism. 3. Arachidonic acid metabolism. 4. Glycosylphosphatidylinositol-anchor biosynthesis. 5. Sphingolipid metabolism. 6. Alpha-linolenic acid metabolism. T2 represents 90 min, T3 represents 180 min.

Figure S5. Possible disturbed metabolic pathways of the potential biomarkers fo GLU, NFM and FFM. 1. Glycerophospholipid metabolism (pink box). 2. Alpha-linolenic acid metabolism. 3. Sphingolipid metabolism. T2 represents 90 min, T3 represents 180 min.

Table S1. Participant characteristics (n=7, 5males and 2 females)

Table S2. A list of detected lipids in this study (pos and neg)

Table S3. The comparison of lipid categories in non-fat and full-fat milk

Table S4. Detailed differential lipid biomarkers that corresponding to the temporal changes within GLU, NFM and FFM groups.

Table S5. Detailed differential lipid biomarkers that corresponding to the group difference among GLU, NFM and FFM groups at 90 min.

Table S6. A list of differential lipids among GLU, NFM and FFM groups at 90 min (with the cutoff of VIP>1).

Table S7. Detailed differential lipid biomarkers that corresponding to the group difference among GLU, NFM and FFM groups at 180 min.

Table S8. A list of differential lipids among GLU, NFM and FFM groups at 180 min (with the cutoff of VIP>1).

NIHMS1693298-supplement-supporting_info.docx^{(5.3MB, docx)}

Acknowledgments

The present study was partially supported by the National Institute of General Medical Sciences of the National Institutes of Health under Award Number R35GM133510 (to JZ), and by the National Dairy Council (to RSB).

Footnotes

Conflict of interest

The authors claim no conflict of interest.

Conflict of Interest Statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

ReferencesUncategorized References

1.Tabák AG; Herder C; Rathmann W, et al. , Prediabetes: a high-risk state for diabetes development. The Lancet 2012, 379, (9833), 2279–2290. [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Benjamin EJ; Virani SS; Callaway CW, et al. , Heart Disease and Stroke Statistics—2018 Update: A Report From the American Heart Association. Circulation 2018, 137, (12), e67–e492. [DOI] [PubMed] [Google Scholar]
3.Yang L; Fish AF; Zhu Y, et al. , Sex differences in 10-year ischemic cardiovascular disease risk prediction in Chinese patients with prediabetes and type 2 diabetes. BMC Cardiovascular Disorders 2019, 19, (1), 1–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Soedamah-Muthu SS; Ding EL; Al-Delaimy WK, et al. , Milk and dairy consumption and incidence of cardiovascular diseases and all-cause mortality: dose-response meta-analysis of prospective cohort studies. The American journal of clinical nutrition 2011, 93, (1), 158–171. [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Hirahatake KM; Bruno RS; Bolling BW, et al. , Dairy foods and dairy fats: new perspectives on pathways implicated in cardiometabolic health. 2020, 11, (2), 266–279. [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Liu Z; Rochfort S; Cocks B, Milk lipidomics: What we know and what we don’t. Progress in lipid research 2018, 71, 70–85. [DOI] [PubMed] [Google Scholar]
7.Nordestgaard BG; Varbo AJTL, Triglycerides and cardiovascular disease. 2014, 384, (9943), 626–635. [DOI] [PubMed] [Google Scholar]
8.Toth P. P. J. V. h.; management, r., Triglyceride-rich lipoproteins as a causal factor for cardiovascular disease. 2016, 12, 171. [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Contarini G; Povolo M. J. I. J. o. M. S., Phospholipids in milk fat: composition, biological and technological significance, and analytical strategies. 2013, 14, (2), 2808–2831. [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Merrill AH Jr; Schmelz E-M; Wang E, et al. , Role of dietary sphingolipids and inhibitors of sphingolipid metabolism in cancer and other diseases. The Journal of nutrition 1995, 125, (suppl_6), 1677S–1682S. [DOI] [PubMed] [Google Scholar]
11.Cilla A; Diego Quintaes K; Barberá R, et al. , Phospholipids in human milk and infant formulas: benefits and needs for correct infant nutrition. Critical reviews in food science and nutrition 2016, 56, (11), 1880–1892. [DOI] [PubMed] [Google Scholar]
12.Li J; Janle E; Campbell WW, Postprandial glycemic and insulinemic responses to common breakfast beverages consumed with a standard meal in adults who are overweight and obese. Nutrients 2017, 9, (1), 32. [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Züllig T; Trötzmüller M; Köfeler HC, Lipidomics from sample preparation to data analysis: a primer. Analytical and Bioanalytical Chemistry 2020, 412, (10), 2191–2209. [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Züllig T; Köfeler HC, HIGH RESOLUTION MASS SPECTROMETRY IN LIPIDOMICS. Mass Spectrom Rev 2020. [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Sarkar A; Zhang S; Holmes M, et al. , Colloidal aspects of digestion of Pickering emulsions: Experiments and theoretical models of lipid digestion kinetics. Advances in colloid and interface science 2019, 263, 195–211. [DOI] [PubMed] [Google Scholar]
16.Eghlimi R; Shi X; Hrovat J, et al. , Triple Negative Breast Cancer Detection Using LC-MS/MS Lipidomic Profiling. J Proteome Res 2020, 19, (6), 2367–2378. [DOI] [PubMed] [Google Scholar]
17.Buas MF; Gu H; Djukovic D, et al. , Identification of novel candidate plasma metabolite biomarkers for distinguishing serous ovarian carcinoma and benign serous ovarian tumors. Gynecol Oncol 2016, 140, (1), 138–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Wenk MR, Lipidomics: New Tools and Applications. Cell 2010, 143, (6), 888–895. [DOI] [PubMed] [Google Scholar]
19.Wang R; Li B; Lam SM, et al. , Integration of lipidomics and metabolomics for in-depth understanding of cellular mechanism and disease progression. Journal of Genetics and Genomics 2020, 47, (2), 69–83. [DOI] [PubMed] [Google Scholar]
20.Bakhytkyzy I; Hewelt-Belka W; Kot-Wasik A, The dispersive micro-solid phase extraction method for MS-based lipidomics of human breast milk. Microchemical Journal 2020, 152, 104269. [Google Scholar]
21.Laaksonen R; Ekroos K; Sysi-Aho M, et al. , Plasma ceramides predict cardiovascular death in patients with stable coronary artery disease and acute coronary syndromes beyond LDL-cholesterol. European heart journal 2016, 37, (25), 1967–1976. [DOI] [PMC free article] [PubMed] [Google Scholar]
22.McFadyen JD; Meikle PJ; Peter K, Platelet lipidomics: a window of opportunity to assess cardiovascular risk? European Heart Journal 2017, 38, (25), 2006–2008. [DOI] [PubMed] [Google Scholar]
23.McDonald JD; Mah E; Dey P, et al. , Dairy milk, regardless of fat content, protects against postprandial hyperglycemia-mediated impairments in vascular endothelial function in adults with prediabetes by limiting oxidative stress responses that reduce nitric oxide bioavailability. The Journal of Nutritional Biochemistry 2019, 63, 129–139. [DOI] [PubMed] [Google Scholar]
24.Harris JA; Benedict F. G. J. P. o. t. N. A. o. S. o. t. U. S. o. A., A biometric study of human basal metabolism. 1918, 4, (12), 370. [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Li H; Song Y; Zhang H, et al. , Comparative lipid profile of four edible shellfishes by UPLC-Triple TOF-MS/MS. Food Chemistry 2020, 310, 125947. [DOI] [PubMed] [Google Scholar]
26.Szymańska E; Saccenti E; Smilde AK, et al. , Double-check: validation of diagnostic statistics for PLS-DA models in metabolomics studies. Metabolomics 2012, 8, (1), 3–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Wan J-B; Bai X; Cai X-J, et al. , Chemical differentiation of Da-Cheng-Qi-Tang, a Chinese medicine formula, prepared by traditional and modern decoction methods using UPLC/Q-TOFMS-based metabolomics approach. Journal of pharmaceutical and biomedical analysis 2013, 83, 34–42. [DOI] [PubMed] [Google Scholar]
28.Zhang W; Huo T; Li A, et al. , Identification of neurotoxicity markers induced by realgar exposure in the mouse cerebral cortex using lipidomics. Journal of Hazardous Materials 2020, 389, 121567. [DOI] [PubMed] [Google Scholar]
29.Sun L-M; Zhu B-J; Cao H-T, et al. , Explore the effects of Huang-Lian-Jie-Du-Tang on Alzheimer’s disease by UPLC-QTOF/MS-based plasma metabolomics study. Journal of pharmaceutical and biomedical analysis 2018, 151, 75–83. [DOI] [PubMed] [Google Scholar]
30.Deng W; Wang Y; Liu Z, et al. , HemI: A Toolkit for Illustrating Heatmaps. PLOS ONE 2014, 9, (11), e111988. [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Gong L.-l.; Yang S; Zhang W, et al. , Discovery of metabolite profiles of metabolic syndrome using untargeted and targeted LC–MS based lipidomics approach. Journal of pharmaceutical and biomedical analysis 2020, 177, 112848. [DOI] [PubMed] [Google Scholar]
32.Feng S; Dai Z; Liu AB, et al. , Intake of stigmasterol and β-sitosterol alters lipid metabolism and alleviates NAFLD in mice fed a high-fat western-style diet. Biochimica et Biophysica Acta (BBA)-Molecular and Cell Biology of Lipids 2018, 1863, (10), 1274–1284. [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Steinmetz KA; Childs MT; Stimson C, et al. , Effect of consumption of whole milk and skim milk on blood lipid profiles in healthy men. The American journal of clinical nutrition 1994, 59, (3), 612–618. [DOI] [PubMed] [Google Scholar]
34.Sun L; Liang L; Gao X, et al. , Early prediction of developing type 2 diabetes by plasma acylcarnitines: a population-based study. 2016, 39, (9), 1563–1570. [DOI] [PubMed] [Google Scholar]
35.Mai M; Tönjes A; Kovacs P, et al. , Serum levels of acylcarnitines are altered in prediabetic conditions. 2013, 8, (12), e82459. [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Koves TR; Ussher JR; Noland RC, et al. , Mitochondrial overload and incomplete fatty acid oxidation contribute to skeletal muscle insulin resistance. 2008, 7, (1), 45–56. [DOI] [PubMed] [Google Scholar]
37.Balvers MG; Verhoeckx KC; Plastina P, et al. , Docosahexaenoic acid and eicosapentaenoic acid are converted by 3T3-L1 adipocytes to N-acyl ethanolamines with anti-inflammatory properties. 2010, 1801, (10), 1107–1114. [DOI] [PubMed] [Google Scholar]
38.Rogers TS; Demmer E; Rivera N, et al. , The role of a dairy fraction rich in milk fat globule membrane in the suppression of postprandial inflammatory markers and bone turnover in obese and overweight adults: an exploratory study. 2017, 14, (1), 36. [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Küllenberg D; Taylor LA; Schneider M, et al. , Health effects of dietary phospholipids. Lipids in health and disease 2012, 11, (1), 3. [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Park JY; Lee SH; Shin MJ, et al. , Alteration in metabolic signature and lipid metabolism in patients with angina pectoris and myocardial infarction. PLoS One 2015, 10, (8), e0135228. [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Akerele O; Cheema S, Fatty acyl composition of lysophosphatidylcholine is important in atherosclerosis. Medical hypotheses 2015, 85, (6), 754–760. [DOI] [PubMed] [Google Scholar]
42.Williams J; Kuchmak M; Witter R, Phospholipids of human serum. Lipids 1966, 1, (2), 89–97. [DOI] [PubMed] [Google Scholar]
43.Massart J; Zierath J. R. J. T. i. E.; Metabolism, Role of diacylglycerol kinases in glucose and energy homeostasis. 2019, 30, (9), 603–617. [DOI] [PubMed] [Google Scholar]
44.Boni LT; Rando R. R. J. J. o. B. C., The nature of protein kinase C activation by physically defined phospholipid vesicles and diacylglycerols. 1985, 260, (19), 10819–10825. [PubMed] [Google Scholar]
45.Van Meer G; Voelker DR; Feigenson GW, Membrane lipids: where they are and how they behave. Nature reviews Molecular cell biology 2008, 9, (2), 112–124. [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Walter A; Korth U; Hilgert M, et al. , Glycerophosphocholine is elevated in cerebrospinal fluid of Alzheimer patients. Neurobiology of aging 2004, 25, (10), 1299–1303. [DOI] [PubMed] [Google Scholar]
47.Pinz I; Robich MP; Ryzhov S, et al. , A novel lipidomic approach to understand human diabetic heart disease. The FASEB Journal 2017, 31, (1_supplement), 883.19–883.19. [Google Scholar]
48.Alshehry ZH; Mundra PA; Barlow CK, et al. , Plasma lipidomic profiles improve on traditional risk factors for the prediction of cardiovascular events in type 2 diabetes mellitus. 2016, 134, (21), 1637–1650. [DOI] [PubMed] [Google Scholar]
49.Fahy E; Subramaniam S; Brown HA, et al. , A comprehensive classification system for lipids. European journal of lipid science and technology 2005, 107, (5), 337–364. [DOI] [PubMed] [Google Scholar]
50.Futerman AH; Hannun YA, The complex life of simple sphingolipids. EMBO reports 2004, 5, (8), 777–782. [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Kuchta AM; Kelly PM; Stanton C, et al. , Milk fat globule membrane–a source of polar lipids for colon health? A review. International journal of dairy technology 2012, 65, (3), 315–333. [Google Scholar]
52.Monk JM; Wu W; Lepp D, et al. , Navy bean supplemented high-fat diet improves intestinal health, epithelial barrier integrity and critical aspects of the obese inflammatory phenotype. The Journal of nutritional biochemistry 2019, 70, 91–104. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials