Plasma Endocannabinoids Are Independently Associated With the Metabolic Function of White Adipose Tissue

Abstract

Context

Little is known about the link between the endocannabinoid (EC) system and the in vivo metabolic function of white adipose tissue (WAT).

Objective

We aimed to evaluate whether ECs are linked to postprandial fatty acid metabolism and WAT metabolic function.

Methods

Men and women, with (IGT, n = 20) or without impaired glucose tolerance (NGT, n = 20) underwent meal testing with oral and intravenous stable isotope palmitate tracers and positron emission tomography with intravenous [¹¹C]-palmitate and oral [¹⁸F]-fluoro-thia-heptadecanoic acid to determine systemic and organ-specific dietary fatty acid (DFA) and nonesterified fatty acid (NEFA) metabolism and partitioning. We determined fasting and postprandial plasma levels of EC by ultra-high performance liquid chromatography–tandem mass spectrometry.

Results

All ECs of the 2-monoacylglycerol (2-MAG) family displayed a progressive postprandial increase up to 360 minutes after meal intake that was more pronounced in women with IGT. N-acylethanolamine (NAE) levels decreased between fasting and 180 minutes, followed by a return to preprandial values at 360 minutes and were also increased in women with IGT. Postprandial area under the curve (AUC) of palmitate appearance rate was significantly and independently associated with postprandial AUC of anandamide (AEA; P = .0003) and total energy expenditure (P = .0009). DFA storage in abdominal subcutaneous adipose tissue was positively predicted by fasting 2-arachidonoylglycerol (2-AG; P < .04).

Conclusion

EC levels of the NAE family independently follow plasma NEFA metabolism, whereas 2-MAG closely follow the spillover of triglyceride-rich lipoprotein intravascular lipolytic products. Whether these associations are causal requires further investigation.

Over the past 3 decades, the endocannabinoid (EC) system has been associated with the regulation of energy homeostasis and metabolism, and evidence suggests that it is involved in energy conservation, notably by promoting energy storage and fat accumulation through activation of cannabinoid-1 receptor by their respective endogenous ligands, the ECs N-arachidonoylethanolamine (anandamide or AEA) and 2-arachidonoylglycerol (2-AG) (1, 2). The EC system machinery is by nature ubiquitous and has been found to regulate metabolism via the central nervous system but also through peripheral tissues, including white adipose tissue (WAT) (3-5). In fact, accumulating evidence shows that activation of the EC system in WAT by cannabinoid-1 receptor stimulation increases lipogenesis and heightens lipoprotein lipase activity, promoting lipolysis of circulating triglyceride (TG)-rich lipoproteins and uptake and storage of dietary fatty acids (DFAs) in WAT (6-9). Mechanistic in vitro studies have highlighted that activation of the EC system in WAT during the postprandial state leads to impaired insulin signaling and loss of its antilipolytic action, impeding the mobilization of nonesterified fatty acids (NEFAs) and promoting fatty acid storage and WAT expansion and remodeling (10, 11). The EC system is therefore increasingly considered as a key modulator of WAT metabolic function (3). However, the relationship between EC and the metabolic functions of WAT has never been established in vivo in humans.

On the other hand, it has been repeatedly reported that circulating AEA and/or 2-AG increases in obesity and diabetes correlate with measures of obesity and metabolic syndromes such as body mass index (BMI), waist circumference, TGs, and WAT mass, making these compounds potential candidates as biomarkers of obesity and dysmetabolism (2, 5, 12). However, to our knowledge, circulating EC have never been investigated in individuals with impaired glucose tolerance (IGT) and the mechanisms leading to increased circulating EC in obesity remain unclear and still debated (13-15). Indeed, most studies have measured only AEA and 2-AG and have not determined other EC molecules such as 2-monoacylglycerols (2-MAGs) and related N-acylethanolamines (NAEs), which are also potential candidates as biomarkers of obesity and dysmetabolism. Furthermore, the very few studies that have investigated circulating EC were limited to fasting or 1 to 3 hours postprandially, offering a very limited understanding of circulating EC dynamics in the postprandial state, when DFA digestion and NEFA metabolism are known to extend up to 6 hours and beyond (16-20).

In the present study, we sought to determine whether ECs are associated with WAT metabolic function (ie, NEFA mobilization through lipolysis of intracellular TGs in WAT and storage of DFAs in WAT and partitioning of DFA to other tissues) in men and women with normal (n = 20) or IGT (n = 20) of similar age and BMI. We took advantage of this cohort of individuals who underwent comprehensive measurement of WAT metabolic function and organ-specific postprandial fatty acid metabolism using our state-of-the-art imaging and tracer methods. Comparison of fasting and 6-hour postprandial circulating EC levels between sex and GT status was also investigated as a secondary exploratory objective of our study. The hypothesis of the present study made a posteriori in the present cohort was that circulating ECs are increased in patients with IGT and linked to either DFA partitioning or NEFA mobilization. Using a multivariable regression modeling approach, we sought to determine whether any of these relationships were dependent on other factors known to potentially modulate WAT metabolism and circulating EC levels such as sex, GT status, insulin resistance, energy expenditure, and adiposity parameters (waist circumference, fat mass, and subcutaneous and visceral WAT volume).

Materials and Methods

Study Participants

Healthy adult volunteers including 21 White men (10 with NGT, 11 with IGT) and 19 White women (10 NGT, 9 IGT) were recruited. One of the individuals with IGT was subsequently found to have fasting plasma glucose over 7 mmol/L during one of the metabolic studies, but this participant was kept in the present analysis. All participants underwent the postprandial protocols described in detail in previous publications (NCT04088344; NCT02808182) (21-23). IGT was defined according to the Diabetes Canada guidelines (http://guidelines.diabetes.ca/cpg/chapter3). For NGT individuals, inclusion criteria were based on a glucose concentration of less than 7.8 mM after a 2-hour post 75-g oral glucose tolerance test and on a glycated hemoglobin A_1c of less than 5.8%. IGT and NGT participants were also matched for age (±10 years) and BMI (±5). A history of any dietary or severe past allergic reaction, severe conditions such as type 2 diabetes or other major illness, or participation in any research trial involving radiation exposure within the past 12 months were exclusion criteria. Participants taking drugs known to affect lipid or carbohydrate metabolism were also excluded, except those that could be safely interrupted prior to the metabolic studies. The study was approved by the ethics committee of the Centre de Recherche du CHUS and all participants gave written informed consent to participate in the study in line with the Declaration of Helsinki.

Study Design and Experimental Procedures

Three days prior to each of the four 6-hour postprandial protocols (A0, A1, B0, B1), participants were put on an isocaloric diet and were instructed to maintain their daily physical activities and avoid strenuous physical effort (21-23). In all 4 protocols, a standard liquid meal (400 mL for a total of 906 kcal, 33 g as fat, 34 g as proteins, and 101 g as carbohydrates) was consumed within 20 minutes at time 0. The results of protocols A1 and B1 to test nicotinic acid effect on postprandial metabolism are not reported here. On arrival at the test center the morning after a 12-hour overnight fast, body weight, height, and waist circumference were measured following standardized procedures, and lean and fat body mass was measured with a body composition analyzer (Model TBF 300A, Tanita Corporation of America Inc). For infusion of stable isotope tracers, a first intravenous catheter was placed in one forearm. For blood sampling, a second catheter was inserted in the distal vein of the other arm that was maintained in a heating pad (55 °C).

Administration of radioactive and isotopic tracer during the 6-hour postprandial protocols have been described previously (23). Briefly, in protocol A0, we administered [¹¹C]-palmitate 90 minutes after meal intake and performed dynamic positron emission tomography (PET) acquisition to measure cardiac and hepatic NEFA uptake, oxidation, and nonoxidative metabolism (23). In protocol B0, a constant intravenous infusion of [1,1,2,3,3-²H]-glycerol and [7,7,8,8-²H]-palmitate was first administered from time −60 before to 360 minutes after meal intake. This infusion was preceded by an intravenous bolus of [1,1,2,3,3-²H]-glycerol and NaH[¹³C]O₃ to prime the bicarbonate pool. In this protocol, the meal also contained [U-¹³C]-palmitate and 70 MBq of [¹⁸F]-fluoro-thia-heptadecanoic acid (FTHA) mixed into Intralipid 20% (Baxter) or into olive oil and incorporated into a gel capsule. Orally administered [¹⁸F]-FTHA was used with a whole-body PET acquisition at 360 minutes to determine lean organs and WAT DFA partitioning and 6-hour WAT DFA storage and cardiac and liver DFA uptake, a method that we established, validated, and published extensively (21, 24). The total radioactivity exposure to the participants was less than 20 mSv, and all tracers were tested for sterility and pyrogenicity. Indirect calorimetry (Vmax29n, Sensormedics) measures were taken for 10 minutes at fasting and every hour after the liquid meal to assess total energy expenditure, corrected for protein oxidation. Breath samples were collected at fasting and every hour to 360 minutes postprandial.

Blood Sampling and Processing

Venous blood was drawn every hour between times −60 and 360 minutes in BD Vacutainer tubes containing K₂-EDTA and ready-to-use aprotinin solution, gently inverted multiple times, and kept on ice a maximum of 10 minutes before processing. Processing consisted of centrifugation at 2000g for 10 minutes at 4 °C, and the plasma taken from the upper layer was aliquoted and stored at −80 °C in Eppendorf tubes for subsequent dosage and analysis. Plasma that had been unfrozen more than once were also excluded to avoid any artifacts that could affect EC concentrations, and when available, plasma aliquots that had not been unfrozen were prioritized.

Analysis of Circulating Endocannabinoids

All chemicals used in this work were of analytical grade. Water, methanol, acetonitrile (Optima grade for liquid chromatography/mass spectrometry), formic acid, and ammonium acetate were purchased from Fisher Scientific. The standards used in this work were of analytical grade with a purity up to 98% or more. The compounds 2-AG, AEA, N-palmitoylethanolamine (PEA), N-oleoylethanolamine (OEA), were purchased from Sigma Aldrich. The compounds 2-oleoylglycerol (2-OG), 2-linoleoylglycerol (2-LG), N-linoleoylethanolamine (LEA), N-stearoylethanolamine (SEA), N-docosahexaenoylethanolamine (DHEA), and arachidonic acid were purchased from Cedarlane. Deuterated compounds (AEA-d4, 2-AG-d5) used as internal standards were purchased from Cayman chemical. A 200 µL aliquot of plasma samples was fortified with 10 µL of internal standards (d4-AEA and d5-2-AG). The extraction of the analytes was performed using a solid phase extraction (Oasis HLB prime µElution 96-well plate, 3 mg Sorbent). After sample loading, the targeted compounds were eluted using 2 × 25 µL of acetonitrile:methanol (90:10, v/v). Prior to liquid chromatography–tandem mass spectrometry analysis, the eluates were filtered through a 0.22-µm polytetrafluoroethylene syringe filter and subsequently transferred to glass amber vials. EC analyses were performed on a Xevo TQ MS (Waters Corporation) equipped with a BEH C₁₈ column (50 mm × 2.7 mm,1.8 µm). The solvent flow rate was set to 0.4 mL·min⁻¹ and the column temperature kept at 40 °C. The sample volume injected was 3 µL. The mobile phase was 0.1% formic acid/water with 2 mM of ammonium acetate and 0.1% formic acid/acetonitrile. The mass spectrometry analysis was performed using a positive electrospray ionization source in Multiple-Reaction-Monitoring mode. The tandem mass spectrometry acquisition and data processing were performed with Masslynx 4.1 software from Waters Corporation. The method allowed the quantification of AEA, OEA, PEA, SEA, LEA, DHEA, as well as 2-MAG, including 2-AG, 2-OG, and 2-LG and the EC precursor arachidonic acid in 200-μL plasma samples. During peak integration of the quantitative ion, monoacylglycerol isomers at positions sn-1, 2, and 3 can sometimes be distinguished, but given their rapid interconversion and the preferential esterification of polyunsaturated fatty acids at the sn-2 position of phospholipids, 2-MAGs were quantified as the sum of the monounsaturated fatty acids and polyunsaturated fatty acid–derived 2-MAG isomers (25). The lower limit of quantification of all these assays ranged between 0.0079 and 0.0489 ng/mL.

Other Biochemical Measurements

Glucose, total NEFAs, TGs, insulin, chylomicron triglyceride [U-¹³C]-palmitate and plasma [U-¹³C]-palmitate, [7,7,8,8-²H]-palmitate, and [1,1,2,3,3-²H]-glycerol tracer-to-tracee ratios were determined and assayed as previously described (23). Insulin, C-peptide, and incretin hormone levels were determined by using Millipore (catalog No. HMHEMAG-34K, RRID:AB_2910198), while adiponectin level was determined by using Alpco Diagnostics (catalog No. 80-ADPHU-E01, RRID:AB_2892778). Total plasma oleate, palmitate, and linoleate were previously measured in plasma using gas chromatography–tandem mass spectronomy (21-23).

Calculations

The homeostatic model assessment of insulin resistance (HOMA-IR), Matsuda index, and insulin secretion rate were calculated as previously described (26, 27). Total postprandial plasma palmitate appearance rate (Ra_palmitate), dietary palmitate spillover rate (Ra_{palmitate spillover}), palmitate appearance rate from intracellular lipolysis ((Ra_{palmitate ICL}), plasma NEFA appearance rate ((Ra_NEFA), plasma glycerol appearance rate ((Ra_glycerol), fractional palmitate oxidation (FOx_palmitate), plasma palmitate oxidation (Ox_palmitate), net total fatty acid oxidation (FA_ox), and DFA partitioning were calculated as previously described (23). All areas under the curve (AUC) were calculated using the linear trapezoid method with GraphPad Prism version 9.2.0.

Statistical Analyses

All analyses were performed on untransformed data to simplify interpretation. Three-way analyses of variance (ANOVAs) were performed to evaluate the effect of sex, GT status, postprandial time, and the sex vs glucose tolerance status interaction on plasma concentrations of each of the ECs, arachidonic acid, and other fatty acid precursor levels. In addition, 2-way analyses of covariance (ANCOVAs) were conducted to determine the effect of sex and GT status and their interaction on 6-hour postprandial AUC of plasma EC levels after adjusting for postprandial AUC of plasma levels of their respective fatty acid precursors (ie, arachidonic acid, oleate, palmitate, and linoleate).

Univariate linear regression using Pearson correlations between WAT metabolic function or prespecified variables (insulin resistance, energy expenditure, waist circumference, fat mass, and subcutaneous and visceral WAT volumes) known or suspected a priori to influence WAT metabolic function and fasting or postprandial AUC of EC levels were performed. Stepwise multiple linear regression analysis was then performed using the forward-backward method between WAT metabolic function parameters as dependent variables, and all ECs showing reasonable association (r ≥ 0.2; P < .1) in univariate analysis as independent variables to identify whether EC levels were independently associated with WAT metabolic function parameters in models. Several iterations were also performed using the prespecified potential determinants of WAT metabolic function mentioned previously and forcing GT status and sex in the model to determine the optimal model showing significant independent association, based on maximum F value and R².

Because our data set includes an important number of variables (>90 variables) not a priori planned to be tested in the present study, factor analysis and principal component (PC) analysis were applied to determine relationship between fasting and postprandial EC levels and metabolic clusters. In addition to screen-plot and natural interpretation, parallel analysis with Monte-Carlo simulation with 7000 permutations was performed to determine the most statically likely number of PCs to extract. After direct Oblimin rotation with Kaiser normalization of extracted PCs, the pattern matrix of factor loadings was produced. The extracted PCs were named according to the variables that had high loadings and based on the existing literature. The variables that had communalities of less than 0.3 were excluded. Missing values (≤2 per variable) were replaced by mean of the standardized data and variables that missed more than 2 values were excluded. Z scores for each participant of the PCs were extracted for subsequent correlation and regression analysis.

The associations between these metabolic clusters and fasting and postprandial AUC of EC levels were first analyzed with univariate linear regression using Pearson correlations. Stepwise multiple linear regression analysis was then employed using the forward-backward method using all PCs showing statistically significant or reasonable trend association in univariate analysis (r ≥ 0.2; P < .1) to identify PCs that best predicted ECs in models and forcing GT status and sex to determine the optimal model showing significant independent association, based on maximum F value and R². Diagnosis of the models was evaluated using Shapiro-Wilk test, residual, quantile-quantile plot, homoscedasticity, and residual vs order plot to determine the solidity, equal variance, normality, and serial correlation of the models. All independent variables retained in the models had a variance inflation factor below 5. A 2-tailed P value less than .05 was considered statistically significant. All analyses were performed with SPSS software (IBM) or GraphPad Prism version 9.2.

Results

Participant Characteristics

The characteristics of each of the 4 groups are shown in Table 1. BMI tended to be higher in women with IGT than in those with NGT (P = .06). Men had significantly higher lean mass and waist circumference than women, while women had a significantly higher fat mass, fasting glucose, and NEFAs. Women with IGT were more insulin resistant (ie, higher HOMA-IR and adipose tissue insulin resistance [Adipo-IR]) and had significantly higher insulin secretion index and AUC insulin secretion rates than women with NGT. Men with IGT showed significantly higher fasting TGs than men with NGT. Insulin sensitivity (ie, Matsuda index) was significantly lower in men with IGT compared to men with NGT.

Table 1.

Characteristics of participants

	NGT male (N = 10)	IGT male (N = 11)	NGT female (N = 10)	IGT female (N = 9)	P sex	P status	P interaction
Age, y	60 ± 2	62 ± 3	62 ± 2	63 ± 3	NS	NS	NS
BMI	28.2 ± 0.93	31.4 ± 1.47	28.4 ± 1.74	32.5 ± 1.37	NS	.01	NS
Lean mass, %	72.0 ± 1.6	67.8 ± 2.7	60.9 ± 1.5	57.5 ± 1.8	<.0001	.06	NS
Fat mass, %	28.0 ± 1.6	32.2 ± 2.7	39.0 ± 1.5	43.9 ± 1.3	<.0001	.02	NS
Waist circumference, cm	101.3 ± 3.1	108.8 ± 4.1	89.1 ± 4.8	98.6 ± 3.4	.009	.04	NS
Fasting glucose, mM	4.95 ± 0.13	5.52 ± 0.15	4.96 ± 0.17	5.70 ± 0.41	NS	.006	NS
Fasting insulin, pM	77 ± 16	127 ± 19	89 ± 25	145 ± 30	NS	.02	NS
Fasting NEFA, μM	0.40 ± 0.02	0.44 ± 0.04	0.56 ± 0.05	0.57 ± 0.05	.0008	NS	NS
Fasting TGs, mM	0.92 ± 0.16	1.88 ± 0.25	1.21 ± 0.20	1.60 ± 0.26	NS	.004	NS
HOMA-IR	2.9 ± 0.6	5.3 ± 0.9	4.1 ± 1.2	7.1 ± 1.9	NS	.02	NS
Adipo-IR	29 ± 5	54 ± 10	47 ± 12	75 ± 11	.05	.01	NS
Disposition Index	255 ± 42	148 ± 16	155 ± 29	151 ± 37	NS	NS	NS
Matsuda Index	16.7 ± 2.5	8.3 ± 1.4	12.8 ± 1.8	8.0 ± 1.4	NS	.0009	NS
ISI	16.9 ± 1.7	19.7 ± 2.2	14.0 ± 1.3	24.1 ± 3.1	NS	.004	NS
AUC0-360 ISR	33 424 ± 3188	43 109 ± 4582	27 445 ± 2668	54 361 ± 6076	NS	.0001	.0495

Data are presented as mean ± SEM. P values are from 2-way analysis of variance.

Abbreviations: Adipo-IR, adipose tissue insulin resistance; AUC, area under the curve; BMI, body mass index; HOMA-IR, homeostatic model assessment of insulin resistance; IGT, impaired glucose tolerance; ISI, insulin secretion index; ISR, insulin secretion rate; NEFA, nonesterified fatty acid; NGT, normal glucose tolerance; NS, nonsignificant; TG, triglycerides.

Postprandial Plasma Glucose, Insulin, Nonesterified Fatty Acid Levels, and Appearance Rates and White Adipose Tissue Dietary Fatty Acid Partitioning

Postprandial plasma levels of glucose and insulin were statistically significantly increased in IGT participants, independently of sex (Fig. 1A and 1B; P < .0001). NEFA levels were higher in women vs men and tended to be higher in IGT vs NGT (Fig. 1C; P < .04 and P = .09, respectively), Ra_palmitate was higher in IGT vs NGT (Fig. 1D; P = .02), Ra_{palmitate spillover} was higher in men vs women (Fig. 1E; P = .04), and Ra_{palmitate ICL} was higher in IGT vs NGT (Fig. 1F; P = .04). DFA partitioning in total WAT was higher in women vs men and tended to be higher in IGT vs NGT (Fig. 1H; P = .04 and P = .06, respectively). DFA partitioning in subcutaneous abdominal adipose tissue was higher in women vs men (Fig. 1H; P = .004). DFA partitioning in subcutaneous thigh adipose tissue was not affected by sex or IGT status (Fig. 1I).

Figure 1.

Postprandial plasma levels of glucose, insulin, and NEFA levels and appearance rates. Plasma levels of A, glucose; B, insulin; C, NEFA levels; D, palmitate appearance rate; E, palmitate spillover appearance rate; and F, palmitate appearance rate from intracellular lipolysis; and G to I, total, abdominal, and femoral subcutaneous adipose tissue dietary fatty acid partitioning. NEFA, nonesterified fatty acids; Ra_palmitate, palmitate appearance rate; Ra_{palmitate ICL}, palmitate appearance rate from intracellular lipolysis; Ra_{palmitate spillover}, palmitate spillover appearance rate. P values are from 3-way analysis of variance or Fisher least significant difference post hoc test.

Postprandial Plasma Endocannabinoid Levels of the 2-Monoacylglycerol Family, Triglycerides, and Chylomicron-Triglycerides

Plasma concentrations of 2-AG (Fig. 2A) were higher in women vs men (P = .005) and in IGT vs NGT (P = .02) without statistically significant interaction (P = .47). 2-OG levels (Fig. 2B) were also higher in women vs men (P = .002) and in IGT vs NGT (P = .007) without statistically significant interaction (P = .12). 2-LG levels (Fig. 2C) were higher in IGT vs NGT (P < .05), essentially attributable to higher levels in women with IGT (interaction P = .006). Plasma total TGs (Fig. 2D) and plasma chylomicron-TGs (Fig. 2E) were both higher in IGT vs NGT (P < .0001). All 2-MAG, TGs, and chylomicron-TG levels significantly increased with postprandial time (time P ≤ .05), assuming a similar postprandial pattern and reaching maximum plasma levels 360 minutes after meal intake.

Figure 2.

Plasma endocannabinoid levels of the 2-monoacylglycerol family and plasma triglycerides (TG) and chylomicron-TG (CM-TG). Plasma levels of A, 2-AG; B, 2-OG; C, 2-LG; and D, total TG, and E, CM-TG. 2-AG, 2-arachidonoylglycerol; 2-LG, 2-linoleoylglycerol; 2-OG, 2-oleoylglycerol; CM, chylomicron; TG, triglycerides. P values are from 3-way analysis of variance.

Postprandial Plasma Endocannabinoid Levels of the N-Acylethanolamine Family and of Their Fatty Acid Precursors

Plasma concentrations of AEA (Fig. 3A) were higher in women vs men (P < .0001) and in IGT vs NGT (P < .05), due to higher levels in women with IGT (interaction P < .006). Plasma arachidonic acid levels (Fig. 3B) were not, however, significantly different between sexes and between GT status. Plasma OEA levels (Fig. 3C) were also higher in women vs men (P < .0001), but not in IGT vs NGT. Plasma oleic acid levels (Fig. 3D) were higher in women vs men (P = .02) and higher in IGT vs NGT (P = .03). Plasma LEA and PEA levels (Fig. 3E and 3G) were also higher in women vs men (P < .0001), but not in IGT vs NGT. Plasma linoleic acid levels (Fig. 3F) were higher in women vs men (P = .002) but not in IGT vs NGT. Plasma palmitic acid levels (Fig. 3H) were higher in women vs men (P = .02) and in IGT vs NGT (P = .003). DHEA levels (Fig. 3I) were selectively higher in women with IGT (sex P < .0001, glucose intolerance status P = .68, interaction P = .008). All NAE and NEFA precursors assumed similar postprandial change in concentration (all time P ≤ .0001), with a nadir reached at 180 minutes followed by an increase back toward fasting concentration levels at 360 minutes after meal intake.

Figure 3.

Plasma endocannabinoid levels of the N-acylethanolamine family and plasma nonesterified fatty acid precursors. Plasma levels of A, N-AEA and B, AA; C, N-OEA; and D, OA; E, N-LEA and F, LA; G, N-PEA; and H, PA. I, Plasma levels of N-DHEA. AA, arachidonic acid; AEA, N-arachidonoylethanolamine; DHEA, N-docosahexaenoylethanolamine; LA, linoleic acid; LEA, N-linoleoylethanolamine; OA, oleic acid; OEA, N-oleoylethanolamine; PA, palmitic acid; PEA, N-palmitoylethanolamine; Ra_palmitate, plasma palmitate appearance rate; SCAT, subcutaneous adipose tissue. P values are from 3-way analysis of variance.

Effects of Sex and Glucose Tolerance Status on 6-Hour Postprandial Area Under the Curve of Plasma Endocannabinoid Levels After Adjusting for Fatty Acid Precursors

After controlling for each fatty acid precursor using ANCOVA, all the statistically significant effects of sex and/or GT status described above using ANOVA remained significant (Supplementary Table S1 (28)), suggesting glucose tolerance- and sex-driven effects independent of any changes in EC precursors.

Associations Between Fasting and 6-Hour Postprandial Plasma Levels of Endocannabinoid and White Adipose Tissue Metabolic Function

Univariate linear regression using Pearson correlations revealed that postprandial AUC of Ra palmitate was strongly associated with postprandial AUC of AEA, OEA, PEA, LEA (Fig. 4A-4D) and, to a lesser extent, fasting AEA (r = 0.39; P = .02). Postprandial AUC of Ra NEFAs from WAT intracellular lipolysis was significantly associated with postprandial AUC of AEA (r = 0.35; P = .03), OEA (r = 0.34; P = .04), and LEA (r = 0.36; P = .03). Abdominal subcutaneous adipose tissue DFA partitioning was significantly correlated with fasting 2-AG (Fig. 4E), fasting (r = 0.37; P = .02), and postprandial AUC (r = 0.44; P = .005) of DHEA, and inversely with postprandial AUC of 2-LG (r = −0.36; P = .03). Femoral subcutaneous adipose tissue DFA partitioning was positively and significantly associated with postprandial AUC of 2-OG (Fig. 4F), AEA (Fig. 4G), OEA (Fig. 4H), LEA (r = 0.41; P = .01), PEA (Fig. 4I), SEA (r = 0.37; P = .02), and DHEA (r = 0.37; P = .02).

Figure 4.

Correlations between plasma EC and WAT metabolic function. Plasma palmitate appearance rate vs postprandial area under the curve of A, N-AEA; B, N-OEA; C, N-PEA; and D, N-LEA. E, Abdominal subcutaneous adipose tissue dietary fatty acid partitioning vs fasting 2-AG. Femoral subcutaneous adipose tissue dietary fatty acid partitioning vs postprandial area under the curve of F, 2-OG; G, N-AEA; H, N-OEA; and I, N-PEA. 2-AG, 2-arachidonoylglycerol; 2-OG, 2-oleoylglycerol; AEA, N-arachidonoylethanolamine; DFA, dietary fatty acids; EC, endocannabinoids; LEA, N-linoleoylethanolamine; OEA, N-oleoylethanolamine; PEA, N-palmitoylethanolamine; Ra_palmitate, plasma palmitate appearance rate; SCAT, subcutaneous adipose tissue; WAT, white adipose tissue.

Other factors such as fat mass, subcutaneous adipose tissue volume, and AUC of insulin secretion rate and 6-hour total energy expenditure were also determinants of WAT metabolic function parameters and served as covariates to evaluate independent associations between EC and adipose tissue metabolic function parameters in multivariate regression (Supplementary Table S2 (28)). In the best multivariate model, postprandial 6-hour AUC of Ra palmitate was positively associated with postprandial AUC of AEA and postprandial 6-hour total energy expenditure, independently of all other variables, including fat mass (Table 2). Similar results were obtained with models including OEA, PEA, and LEA, but OEA was dependent on fat mass. The best model for postprandial AUC of Ra palmitate from WAT intracellular lipolysis included postprandial 6-hour total energy expenditure, subcutaneous adipose tissue volume, and postprandial AUC of AEA (all positively associated), but postprandial AUC of AEA depended on fat mass. In the best multivariate model, abdominal subcutaneous adipose tissue DFA partitioning was positively predicted by female sex (fat mass dependent) and fasting 2-AG (GT status, fat mass, and subcutaneous adipose tissue volume dependent). Similar results were obtained with fasting 2-LG (negatively). Femoral subcutaneous adipose tissue DFA partitioning was positively associated with fat mass (subcutaneous adipose tissue volume dependent) and negatively associated with sex (postprandial AUC of insulin secretion rate and visceral adipose tissue volume dependent), and no EC was an independent predictor.

Table 2.

Multiple linear regression of palmitate flux and abdominal and femoral subcutaneous adipose tissue dietary fatty acid partitioning with fasting and postprandial area under the curve of plasma endocannabinoid levels

Predictor	β coefficient	t value	P	R 2	Adj. R2	Sig.
AUC0-360 Ra palmitate
Model				0.46	0.43	0.00002
Constant		3.450	.0015
AUC AEA	1921	3.991	.0003
AUC TEE	160.3	3.644	.0009
AUC0-360 Ra palmitate ICL
Model				0.68	0.64	0.0000004
Constant		5.257	.00001
AUC TEE	580.5	4.584	.00009
SCAT volume	4794	3.094	.0044
AUC AEA	4032	2.640	.0134
Abdominal DFA partitioning
Model				0.35	0.31	0.0005
Constant		6.151	<.0001
Sex, female	5.224	3.207	.0029
Fasting 2-AG	1.167	2.140	.0394
Femoral DFA partitioning
Model				0.40	0.43	0.00006
Constant		2.050	.0481
Fat mass, %	.5958	4.872	.00003
Sex, female	−4.595	2.266	.0230

Sig.: P value of the linear regression. For sex, female = 0, male = 1.

Abbreviations: 2-AG, 2-arachidonoylglycerol; Adj, adjusted; AEA, N-arachidonoylethanolamine; AUC, area under the curve; DFA, dietary fatty acid; ECs, endocannabinoids; ICL, intracellular lipolysis; Ra, appearance rate; SCAT, subcutaneous adipose tissue; TEE, 6-hour total energy expenditure.

Exploratory Principal Component Analysis Identifying the Association Between Plasma Endocannabinoid Levels and Metabolic Clusters

Parallel Monte-Carlo analysis simulation identified 3 major clusters of metabolic variables (Supplementary Table S3 (28)), which were designated as follow: PC1 (TGs, insulin secretion, and metabolic syndrome), PC2 (insulin resistance, energy expenditure, and DFA partitioning), and PC3 (NEFA metabolism and body composition). These 3 PCs accounted for 44% of the whole variance in the data set. PC1 had the highest eigenvalue (15%) and accounted for 20% of the variability in the data set. PC2 accounted for 13% of the variance, while PC3 represented 12% of the variance. As shown in Supplementary Tables S4 and S5 (28), PC1 was mainly associated with fasting 2-OG (r = 0.35; P = .04) and postprandial AUC of 2-OG (r = 0.43; P = .01). PC3 was strongly associated with all fasting levels and postprandial AUC of NAE (except DHEA) and tended to be associated with postprandial AUC of 2-OG (r = 0.32; P = .06). Multiple linear regressions were then performed using these 3 PCs together with sex and GT status to predict each of the EC fasting levels and postprandial AUCs showing reasonable trend in univariate analysis (Table 3). Fasting 2-OG was independently and positively predicted by PC1 and PC2 (R² = 0.35; P = .001) and similar associations, but in the negative direction, were obtained for fasting 2-LG (not shown). Fasting AEA and PEA were independently predicted by PC3 (not shown). Postprandial AUC of 2-OG was positively predicted by PC1 and sex (R² = 0.30; P = .004), independently of GT status. The postprandial AUC of AEA was positively predicted by PC3 (R² = 0.41; P < .0001), independently of sex and GT status. Similar results were obtained for postprandial AUCs of OEA, PEA, SEA, and LEA (Supplementary Table S6 (28)).

Table 3.

Multiple linear regression of post-prandial area under the curve and fasting EC plasma levels with principal components as independent variables, adjusting for sex and glucose tolerance status

Predictor	Standardized β-coefficient	t value	P	R 2	Adj. R 2	Sig.
AUC 2-OG
Model				0.30	0.25	0.004
Constant		2.233	.0329
PC1	.4500	3.480	.0015
Sex, female	.5687	2.228	.0333
AUC AEA
Model				0.41	0.40	<0.0001
Constant		1.455	.1555
PC3	.4585	4.756	<.0001
Fasting 2-OG
Model				0.35	0.31	0.001
Constant		1.072	.2919
PC1	.3517	3.327	.0023
PC2	.2599	2.459	.0197
Fasting AEA
Model				0.30	0.28	0.0008
Constant		0.7365	.4668
PC3	.5350	3.700	.0008

PC1 regroups triglycerides, insulin secretion and metabolic syndrome-related variables; PC2 regroups insulin resistance, energy expenditure and dietary fatty acid partitioning-related variables; PC3 regroups nonesterified fatty acid metabolism and body composition-related variables; β coefficient represents the standardized β coefficient.

Sig.: P value of the linear model. For sex, female = 0, male = 1.

Abbreviations: 2-OG, 2-oleoylglycerol; Adj, adjusted; AEA, N-arachidonoylethanolamine; AUC, area under the curve; PC, principal component.

Discussion

This study is the first to evaluate the postprandial plasma levels of 9 ECs during a 6-hour postprandial period and to investigate their relationship with WAT metabolic function in men and women, with NGT or IGT phenotyped with tracer and PET imaging protocols (19, 23). In the present study, we demonstrated that ECs of the 2-MAG family are linked to circulating TG metabolism and DFA storage in WAT whereas ECs of the NAE family are closely linked to plasma NEFA metabolism determined by intracellular TG lipolysis in WAT. We have thus established for the first time an independent relationship between circulating EC levels and the metabolic function of WAT in vivo in humans.

We found that all ECs of the 2-MAG family investigated in this study shared similar dynamics during the 6-hour postprandial period, with a gradual increase between 180 and 360 minutes postprandially, which was more pronounced in women with IGT. Postprandial circulating 2-MAG probably originate from the spillover of TG-rich lipoprotein intravascular lipolytic products under the action of lipoprotein lipase and, to a lesser extent, endothelial lipase into WAT (2, 29-31). In support of this interpretation, postprandial levels of 2-AG and 2-OG increased progressively up to 360 minutes after meal intake, following kinetics similar to that of DFA spillover (20). We also demonstrated strong correlations between fasting and postprandial levels of 2-OG with the PC1 that includes chylomicron-TGs and very low-density lipoprotein–TGs, consistent with the findings of others (16, 31, 32). We also found that fasting 2-AG levels entered the best models for predicting subcutaneous abdominal adipose tissue DFA storage.

On the other hand, all EC members of the NAE family showed very different postprandial dynamics compared with those of the 2-MAG, with a marked decline between fasting and 180 minutes, followed by a return to preprandial values at 360 minutes. This pattern follows the postprandial dynamics of NEFAs, suggesting that NAEs are simultaneously released into the circulation by WAT from intracellular TG lipolysis. It has previously been proposed that ECs may regulate WAT lipolysis (10, 11, 33). Previous human studies have established a relationship between circulating EC levels and abdominal obesity (31, 32, 34). For example, Blüher et al (32) showed that individuals with abdominal obesity exhibit higher circulating 2-AG levels than lean individuals. Di Marzo et al (16) showed that AEA decreased in healthy nonobese participants following an insulin infusion clamp, which was not observed in individuals with type 2 diabetes. D’Eon et al (35) demonstrated that treatment of insulin-sensitive adipocytes with insulin leads to a decrease in EC levels via an increase in EC catabolic enzymes (monoacylglycerol lipase [MAGL] and fatty acid amide hydrolase [FAAH]) and a decrease in anabolic enzymes N-acyl phosphatidylethanolamine-specific phospholipase D [NAPE-PLD], in contrast to insulin-resistant adipocytes, which are insensitive to the action of insulin on FAAH and MAGL expression. Abdominal obesity and adipose tissue insulin resistance are important predictors of increased postprandial NEFA onset (20, 23, 36, 37). Our results demonstrate that the association between abdominal obesity and insulin resistance with circulating NEA is likely mediated by intracellular lipolysis of WAT TGs.

Gatta-Cherifi et al (17) were the first to investigate the postprandial dynamics of plasma EC in humans, and found that postprandial levels of AEA, but not 2-AG, were higher in obese compared than in normal-weight participants. Similar results were also observed in a subsequent study (18). In contrast, Di Marzo et al (16) showed no variation in AEA or 2-AG after an oral glucose tolerance test in individuals with abdominal obesity, while postprandial AEA and 2-AG levels decreased in lean control individuals. However, these 3 previous studies were limited to a period of 1 to 3 hours after meal intake and the last one was performed with glucose, not with a mixed-meal challenge. The digestion and postprandial metabolism of DFA usually last for up to six hours and beyond (19, 20). Therefore, these previous studies were unable to study the complete postprandial dynamics of circulating fatty acid metabolism in relation to circulating EC levels. Differences in meal composition could also play a role in the differences observed between studies, given that different diets have the potential to modulate EC levels. For example, the Western diet has been associated with increased levels of 2-AG and AEA, independently of age, physical activity, BMI, waist circumference, and fat mass (37). One previous study showed a higher 2-hour postprandial plasma level of 2-AG, but not AEA, PEA, nor OEA, in response to a high-caloric and -fat meal in lean individuals (38).

We found that fasting and postprandial plasma levels of ECs were higher in women than in men, especially when associated with IGT status. Blüher et al (32) showed that healthy women naturally have higher fasting levels of AEA than healthy men. Another study has shown that plasma 2-AG was higher in postmenopausal, insulin-resistant obese women (39). Circulating 2-AG and/or AEA are increased in patients with obesity or type 2 diabetes, and these increases are associated with markers of obesity and metabolic syndrome such as BMI, waist circumference, visceral fat mass, insulin resistance, TGs, and NEFAs (18, 31, 32, 40, 41). Interestingly, we also demonstrated that women have significantly higher fat mass, fasting NEFAs, HOMA-IR, Adipo-IR, and abdominal subcutaneous adipose tissue DFA partitioning than men. Our multivariate analyses demonstrated that the reason for the increase in plasma NAE in women is likely the higher plasma NEFA appearance rate from WAT intracellular TG lipolysis relative to total body mass (42).

Our study is cross-sectional and therefore cannot establish a causal link between circulating ECs and in vivo WAT metabolic function. Although WAT metabolic function likely drives circulating EC levels, it cannot be excluded that ECs may exert modulating effects on in vivo WAT metabolism. Interventions aimed at modifying EC levels or WAT metabolic function independently, together with sophisticated phenotypic measurements of adipose tissue function, such as those we have established, may help to settle this issue in the future.

In conclusion, we found an independent relationship between circulating EC levels and the metabolic function of WAT in vivo in humans. NEAs are associated with NEFA release from WAT, while 2-MAGs are associated with postprandial metabolism of TG-rich lipoproteins and storage of DFA in WAT. Further studies are needed to reproduce the findings of this first study examining the relationship between circulating EC and WAT function and to determine the causality and direction of causality between these processes.

Abbreviations

2-AG

2-arachidonoylglycerol

2-LG

2-linoleoylglycerol

2-MAG

2-monoacylglycerol

2-OG

2-oleoylglycerol

Adipo-IR

adipose tissue insulin resistance

AEA

N-arachidonoylethanolamine

ANCOVA

analysis of covariance

ANOVA

analysis of variance

AUC

area under the curve

BMI

body mass index

DFA

dietary fatty acid

DHEA

N-docosahexaenoylethanolamine

ECs

endocannabinoids

HOMA-IR

homeostatic model assessment of insulin resistance

ICL

intracellular lipolysis

IGT

with impaired glucose tolerance

LEA

N-linoleoylethanolamine

NAEs

N-acylethanolamines

NEFA

nonesterified fatty acid

NGT

normal glucose tolerance

OEA

N-oleoylethanolamine

PC

principal component

PEA

N-palmitoylethanolamine

PET

positron emission tomography

Ra

appearance rate

Ra_palmitate

palmitate appearance rate

Ra_{palmitate ICL}

palmitate appearance rate from intracellular lipolysis

Ra_{palmitate spillover}

palmitate spillover appearance rate; SCAT, subcutaneous adipose tissue

SEA

N-stearoylethanolamine

TEE

6-hour total energy expenditure

TGs

triglycerides

WAT

white adipose tissue

Funding

This work was largely supported by funds from the Canadian Institutes of Health Research (MOP53094). A.C.C. holds the Canada Research Chair: Tier 1 in Molecular Imaging of Diabetes. P.S. is the holder of the Canada Research Chair: Tier 1 in Neurophysiopharmacology of Chronic Pain. This work was also made possible with the generous studentship to D.D. from the University of Sherbrooke.

Author Contributions

A.C.C. and D.D. conceived and designed research; D.D., C.N., M.F., L.H., and S.S. performed experiments; D.D., C.N., M.F., and A.C.C. analyzed data; D.D., C.N., and A.C.C. interpreted results of experiments; D.D. and C.N. prepared figures; D.D., C.N., and A.C.C. drafted manuscript; and all authors edited, revised, and approved the final version of the manuscript.

Disclosures

The authors have nothing to disclose.

Data Availability

The data sets generated during and/or analyzed during this study are available from the corresponding author on reasonable request.

Clinical Trial Information

Clinicaltrials.gov numbers NCT04088344 and NCT02808182 (registered on September 11, 2019 and on April 14, 2016 respectively).