Funding agencies: This research was supported by the Canadian Institutes of Health Research (T 0602145.02), and an operating grant from genome Quebec (Complications Associated with obesity; CAO). Rémi Rabasa-Lhoret holds a scholarship from FRSQ (Fonds de Recherche en Santé du Québec and is the recipient of the J-A DeSéve chair in clinical research). Siham Yasari is currently employed by the Canadian Institutes of Health Research and she declares no conflict of interests. Vincenzo Di Marzo and Fabiana Piscitelli are recipients of an NIH grant (DA-009789).
Circulating endocannabinoids in insulin sensitive vs. Insulin resistant obese postmenopausal women. A MONET group study
Version of Record online: 19 OCT 2013
© 2013 The Obesity Society
Volume 22, Issue 1, pages 211–216, January 2014
How to Cite
Abdulnour, J., Yasari, S., Rabasa-Lhoret, R., Faraj, May., Petrosino, S., Piscitelli, F., Prud' homme, D. and Di Marzo, V. (2014), Circulating endocannabinoids in insulin sensitive vs. Insulin resistant obese postmenopausal women. A MONET group study. Obesity, 22: 211–216. doi: 10.1002/oby.20498
Disclosure: The authors declare no conflict of interest.
Author Contributions: Joseph Abdulnour, Siham Yasari, Remi Rabasa-Lhoret, Vincenzo Di Marzo and Denis Prud'homme played a role in the conception/design as well as the acquisition of data. Stefania Petrosino and Fabiana Piscitelli carried out the endocannabinoid measurements. All authors were involved in writing and interpreting the paper and gave final approval of the submitted and published versions.
- Issue online: 11 JAN 2014
- Version of Record online: 19 OCT 2013
- Accepted manuscript online: 24 APR 2013 02:54PM EST
- Manuscript Accepted: 28 MAR 2013
- Manuscript Received:
To measure the circulating levels of endocannabinoids and related molecules at fasting, after acute hyperinsulinemia and after weight loss in insulin sensitive vs. insulin resistant obese postmenopausal women.
Design and Methods
The sample consisted of 30 obese postmenopausal women (age: 58.9 ± 5.2 yrs; BMI: 32.9 ± 3.6 kg/m2). Subjects underwent a 3-hour hyperinsulinaemic-euglycaemic clamp (HEC) (glucose disposal rate (M-value): 10.7 ± 3.3 mg min−1 kg−1 FFM) and 6-month weight loss intervention. Participants were classified as insulin sensitive obese (ISO) or insulin resistant obese (IRO) based on a predefined cutoff. Plasma levels of the endocannabinoids, anandamide (AEA), 2-arachidonoylglycerol (2-AG), and of the AEA-related compounds, palmitoylethanolamide (PEA) and oleoylethanolamide (OEA), were measured by liquid chromatography-mass spectrometry.
IRO presented higher levels of 2-AG (P < 0.05) independently of the HEC and weight loss, whereas the HEC had an independent inhibitory effect on AEA, PEA, and OEA levels (P < 0.05) in both groups. Furthermore, there was an independent stimulatory effect of weight loss only on PEA levels in both groups (P < 0.05).
This study is the first to show that higher circulating levels of the endocannabinoid 2-AG are found in IRO compared to ISO postmenopausal women, and that weight loss is associated with an increase in PEA, a PPAR-α ligand.
Obesity is known to be associated with an increased risk of insulin resistance, type 2 diabetes and cardiovascular disease (CVD) . However, two subgroups of obese subjects have been characterized. The first of these groups is the “insulin sensitive obese” (ISO) group; this phenotype presents a favorable metabolic profile and is not associated with an increased risk for type 2 diabetes mellitus (T2D) or CVD despite the presence of obesity [2, 3]. On the other hand, also prevalent is the insulin resistant obese (IRO) phenotype, in which subjects are more at risk of developing metabolic disorders than ISO subjects . It has been previously hypothesized that visceral fat accumulation, high birth weight and larger adipose cell size may favor the development of the IRO phenotype . However, it is still uncertain whether any of these factors are responsible for the metabolic differences between the two obesity phenotypes.
Obesity is also a condition in which “over-activation” of the endocannabinoid (EC) system occurs, characterized by higher circulating levels of endogenous ligands of cannabinoid receptors (endocannabinoids, ECs) [5-8]. ECs are endogenous lipids that are involved in the stimulation of food intake and fat accumulation in white adipocytes, as well as in hepatic and skeletal muscle insulin resistance, and insulin release from pancreatic islets mostly by acting on type-1 (CB1) cannabinoid receptors [9, 10]. Activation of CB1 receptors by ECs is associated with insulin resistance . This is believed to be, among other things, secondary to the failure of insulin-resistant adipocytes to regulate EC metabolism in response to insulin stimulation , suggesting that blocking the EC system may improve the cardiometabolic profile . The role of the EC system has not been studied in the ISO vs. IRO subjects. Although there has been interest in identifying the mechanisms characterizing the IRO and ISO phenotypes, it has never been tested whether or not EC overactivity is associated with any one of these obesity phenotypes .
In search for a possible explanation for the metabolic differences between ISO and IRO in the fasting state and in response to acute insulin infusion or weight loss, we investigated in the present study the differences in circulating levels of two key ECs, anandamide (AEA) and 2-arachidonoylglycerol (2-AG), in two sub-groups of obese individuals representing the ISO or IRO phenotype before and after weight loss or insulin infusion. We also measured the levels of two endogenous agonists of peroxisome proliferator-activated receptor-α (PPAR-α), palmitoylethanolamide (PEA) and oleoylethanolamide (OEA), which are metabolically related to AEA .
Based on previous data on circulating EC, PEA and OEA levels in subjects with varying degree of insulin sensitivity and obesity [8, 15, 16], we hypothesized that ISO subjects have lower fasting concentrations of plasma EC and a higher insulin-mediated suppression of plasma EC. Moreover, we hypothesized that, in response to weight loss, IRO women would exhibit a decrease and ISO women an increase in their circulating EC levels.
This is a post-hoc analysis of a sub-group of 30 obese postmenopausal women, matched for age (58.9 ± 5.2 years) and % fat mass (45.1 ± 4.0), enrolled in of the studies of the Montreal Ottawa New Emerging Team (MONET), 6-month weight loss project . The study was approved by the University of Montreal ethics committee.
Women were eligible to participate if they met the following criteria: 1) BMI ≥ 27 kg/m2; 2) cessation of menstruation for more than 1 year and a follicle-stimulating hormone level ≥ 30 U/l; 3) < 2 hours per week of reported structured exercise; 4) nonsmokers; 5) low to moderate alcohol consumption (reported < 2 drinks/day); 6) free of known inflammatory disease including diagnosis of diseases such as arthritis (polyarthritis, rheumatoid, lupus) and acute infections (medical questionnaire and physical examination); and 7) no use of hormone replacement therapy (pharmacy list). On physical examination or biological testing, all participants had no history or evidence of: 1) CVD, peripheral vascular disease, or stroke (detailed medical evaluation and, if suspected, further investigations); 2) diabetes (fasting plasma glucose < 7.0 mmol/L and 2-h 75-g oral glucose tolerance test < 11.0 mmol/L); 3) orthopaedic limitations such as chronic tendinitis, bursitis, back pain, etc. therefore no physical activity restrictions; and 4) medication such as beta, alpha blocker, nitrates, digitalis that could affect cardiovascular function and/or metabolism.
The study began at 07:30 h after a 12-h overnight fast following the procedure described by DeFronzo et al. . An antecubital vein was cannulated, by a trained research nurse under the supervision of a medical doctor, for the infusion of 20% dextrose and insulin (Actrapid®, Novo-Nordisk, Toronto, Canada). The other arm was cannulated for sampling of blood. Three basal samples of plasma glucoseand insulin were taken over 40 min. Then, insulin infusion was initiated at a rate of 75 mU·m−2·min−1 for 180 min. Plasma glucose was measured every 10 min with a glucose analyzer (Beckman Instruments, Fullerton, CA) and maintained at fasting levels with a variable infusion rate of 20% dextrose. Glucose disposal was calculated as the mean rate of glucose infusion measured during the last 30 min of the hyperinsulinaemic-euglycaemic clamp (HEC) (steady state) and was expressed in mg min−1 kg−1 FFM. Based on a previously published cutoff proposed by our group using similar techniques in a comparable group of patients , participants were classified according to their insulin sensitivity, using the glucose disposal rate mg min−1 kg−1 fat free mass (FFM) (M-values). Women with M-values greater than 12.0 mg min−1 kg−1 FFM were classified as the ISO group, whereas women with values below 9.5 mg min−1 kg−1FFM were classified as IRO. The homeostasis model assessment (HOMA-IR) was estimated using the following equation: HOMA-IR = (fasting glucose × fasting insulin)/22.5.
Blood samples were collected at times 0, 60, 160, 170, and 180 min during the HEC. Blood samples were centrifuged at 3900 x g for 10 min at 4°C and kept at −80°C° until further analyses were achieved. Fasting serum samples were collected for total cholesterol, high-density lipoprotein (HDL)-cholesterol, low-density lipoprotein (LDL)-cholesterol, triglyceride (TG), and insulin measurements. Serum blood glucose and the lipid profile were analyzed on the day of collection, whereas insulin samples were kept at −80°C until they were analyzed. Analyses were done on the COBAS INTEGRA 400 (Roche Diagnostic, Montreal, Canada) analyzer for total cholesterol (TC), HDL-C, and TG. TC, HDL-C, and TG were used in the Friedewald formula  to calculate LDL-C concentration (all subjects had a TG level of less than 4.52 mmol/l). Serum insulin levels were determined by RIA (Medicorp, Montreal, Canada). All samples were measured in duplicate.
The extraction, purification and quantification of AEA, 2-AG, OEA and PEA in serum was performed as described previously in the laboratory of one of the authors (Di Marzo) [15, 16, 21]. After addition of the respective deuterated internal standards to the plasma, repeated (three times) lipid extraction of the latter with chloroform:methanol (2:1 by vol), and prepurification of the lipid extracts on silica mini-columns eluted with chloroform:methanol (9:1 by vol), chromatographic fractions containing the compounds were subjected to isotope dilution liquid chromatography-atmospheric pressure chemical ionization-mass spectrometric analysis as described previously . The amounts of the compounds were calculated by isotope dilution and expressed as pmol/ml of blood. The ECs, OEA and PEA were measured at baseline, the last 30 min of the HEC (steady state) and after 6 months of weight loss. All measures were preformed in duplicate.
Body weight, lean body mass (LBM), and fat mass (FM) were evaluated by dual-energy X-ray absorptiometry (DXA) (General Electric Lunar Corporation version 6.10.019, Madison, WI, USA) as previously described . Standing height was measured using a wall stadiometer (Perspective Enterprises, MI, USA). Body mass index (BMI) was calculated as follows: BMI = body weight (kg)/height (m2). Waist circumference was measured with a nonextendable linear tape measure at the mid-distance between the lowest rib and the iliac crest. Calibration of the DEXA was executed daily with a standard phantom before each test and the intraclass coefficient correlation for test–retest for FM and LBM was 0.99 (n = 18) . Body composition measures were performed by a certified research assistant.
A GE High Speed Advantage CT scanner (General Electric Medical Systems, Milwaukee, WI) was used to measure visceral adipose tissue. CT scans were performed at the university hospital during schedule time dedicated for research by a certified radiologist. The subjects were examined in the supine position with both arms stretched above their head. The position of the scan was established at the L4 through L5 vertebral disk using a scout image of the body . Visceral fat area was quantified by delineating the intra-abdominal cavity at the internal most aspect of the abdominal and oblique muscle walls surrounding the cavity and the posterior aspect of the vertebral body. The cross-sectional areas of fat were highlighted and computed using an attenuation range of −190 to −30 Hounsfield Units (HU) . Test–retest measures of the different body fat distribution indices on 10 CT scans yielded a mean absolute difference of ±1% .
Study participants entered into a medically supervised weight loss program for 6 months aimed at reducing body weight by 10%. Subjects were either on calorie restriction only or on calorie restriction + resistance exercise as previously described . Diet prescriptions ranged from 1100 to 1800 kcal/day. Daily energy intake during weight loss was based on the baseline resting metabolic rate multiplied by an activity factor of 1.4, which corresponds to a sedentary state. Both the calorie restriction and the calorie restriction + resistance exercise were merged because there were no significant differences between the groups on all ECs variables and to increase the sample size.
Data are expressed as means ± SD. We verified the normality of the distribution of variables using descriptive analysis (skewness, kurtosis and normality plots) and found no significant deviation from normality. Baseline and 6 months postweight loss differences between various parameters of ISO and IRO individuals were evaluated by repeated measures ANOVA. One-way ANOVAs were performed to determine if significant differences are observed in plasma EC levels between obesity phenotypes (ISO vs. IRO), before and after the HEC and weight loss intervention. Pearson correlations were performed between EC levels and variables of interest. Repeated measures ANOVA were performed to determine the main effect of two groups (ISO vs. IRO) for: 1) baseline and last 30 min of HEC; 2) baseline and after 6 month weight loss on ECs. SPSS 17.0 for Windows (SPSS, Chicago, IL) was used to perform statistical analyses, a P-value of less than 0.05 was considered statistically significant.
At baseline the women's mean age was 58.9 ± 5.2 years, with a BMI of 32.9 ± 3.6 kg/m2. When both obesity phenotypes were compared, IRO women showed significantly higher body weight, BMI, LBM, visceral fat, waist circumference, and TG compared to ISO women of same age and FM. By design, the IRO group presented significantly higher insulin resistance values as defined by glucose disposal rate (M-value) and HOMA-IR score (P < 0.001). After weight loss, the women presented lower body weight, BMI, FM, visceral fat, and waist circumference. Also, we observed a significant interaction for LBM, insulin resistance values and total cholesterol. According to the M-value, 80% of the IRO group remained insulin resistant, whereas 20% improved their insulin sensitivity. On the other hand, 38% of the ISO group remained insulin sensitive, whereas 62% became insulin resistant (Table 1).
|Baseline||6 Months postweight loss||Repeated measure ANOVA P-value|
|Variables||ISO (N = 15)||IRO (N = 15)||ISO (N = 15)||IRO (N =15)||Within subjects||Between subject||Interaction|
|Age (y)||58.9 ± 5.4||59.0 ± 5.1|
|Weight (kg)||81.7 ± 8.4||91.8 ± 13.1||76.8 ± 6.9||84.8 ± 13.5||<0.001||0.026||0.237|
|BMI (kg/m2)||31.5 ± 3.0||34.3 ± 3.7||29.6 ± 2.4||31.7 ± 4.1||<0.001||0.044||0.278|
|Lean body mass (kg)||41.7 ± 2.7||48.1 ± 6.6||42.0 ± 2.7||45.8 ± 5.0||0.025||0.004||0.004|
|Fat mass (kg)||37.4 ± 6.6||41.3 ± 8.3||32.4 ± 5.4||36.6 ± 10.4||<0.001||0.163||0.813|
|Visceral fat (cm2)||166.4 ± 41.5||234.8 ± 67.2||143.0 ± 34.0||207.1 ± 61.4||<0.001||0.001||0.729|
|Waist circumference (cm)||100.8 ± 7.4||109.0 ± 8.0||97.6 ± 9.0||104.4 ± 11.5||0.007||0.021||0.587|
|M-value (mg min−1 kg−1 FFM)||13.7 ± 1.6||7.76 ± 1.3||11.8 ± 1.4||10.1 ± 2.5||0.372||<0.001||<0.001|
|Insulin (mU/L)||12.6 ± 3.8||21.4 ± 9.4||12.0 ± 4.2||17.3 ± 7.8||0.010||0.005||0.047|
|HOMA-IR||2.81 ± 0.9||5.16 ± 2.2||2.80 ± 0.96||4.04 ± 1.82||0.015||0.003||0.026|
|Total cholesterol (mmol/L)||5.13 ± 1.1||5.49 ± 1.0||5.56 ± 1.10||5.32 ± 1.05||0.299||0.877||0.016|
|HDL-C (mmol/L)||1.47 ± 0.4||1.38 ± 0.3||1.53 ± 0.35||1.28 ± 0.24||0.681||0.154||0.076|
|TG (mmol/L)||1.30 ± 0.5||2.09 ± 1.0||1.28 ± 0.49||1.90 ± 1.11||0.321||0.021||0.419|
|LDL-C (mmol/L)||3.07 ± 1.0||3.15 ± 0.9||3.45 ± 0.91||3.16 ± 0.92||0.074||0.752||0.090|
ECs and related mediators
IRO women displayed significantly higher 2-AG (P < 0.001) and PEA (P < 0.039) concentrations as compared to ISO women at baseline. After the HEC, the IRO women still displayed significantly higher 2-AG levels (4.83 ± 1.42 vs. 3.09 ± 1.22 (pmol/ml); P = 0.001); however, they presented lower AEA levels (0.98 ± 0.31 vs. 1.34 ± 0.37 (pmol/ml); P = 0.008) compared to ISO women. Finally, after weight loss the IRO group showed only lower AEA levels (2.67 ± 0.77 vs. 3.60 ± 1.28 (pmol/ml); P = 0.023) compared to the ISO group.
Association between insulin sensitivity and EC levels
Correlative analysis revealed that baseline M-value was significantly correlated with 2-AG (r = −0.54; P = 0.002) and PEA (r = −0.50; P < 0.005) at baseline; with AEA (r = 0.42; P = 0.021) and 2-AG (r = −0.57; P = 0.001) after the HEC; and with AEA (r = 0.40; P = 0.028) postweight loss. Baseline visceral fat was significantly correlated with baseline 2-AG (r = 0.42; P = 0.023), post-HEC 2-AG (r = 0.46; P = 0.010), and OEA (r = 0.38; P = 0.038). LBM was significantly correlated with AEA (r = −0.37; P = 0.043) and 2-AG (r = 0.42; P = 0.020) at baseline.
Independent and interaction effects of the HEC and weight loss on EC levels
Repeated measures ANOVA analysis was performed to determine the main effect of the HEC and between obesity phenotypes (ISO vs. IRO). The HEC had a significant main effect on AEA, PEA, and OEA (P < 0.001), as the plasma concentrations of all three compounds decreased after the HEC in both ISO and IRO women. On the other hand, when comparing obesity phenotype (ISO vs. IRO), there was a significant main effect on 2-AG (P < 0.001), which was higher in IRO women (Figure 1). No significant interaction was observed.
Further repeated measures ANOVA revealed that weight loss had a main effect on PEA, which significantly increased after weight loss in both obesity phenotypes (P = 0.028). After adjusting for the type of weight loss [diet (N = 21) vs. diet + resistance exercise (N = 9)] the effect was no longer significant. Thus again, the obesity phenotype had a significant main effect on AEA (P = 0.044) and 2-AG (P = 0.002), presenting lower AEA and higher 2-AG levels in IRO women compared to ISO women (Figure 2). No significant interaction was observed.
The present study is the first to investigate the differences in the “tone”of the EC system in a context of insulin resistance (IRO) vs. insulin sensitivity (ISO), in obese postmenopausal women in the fasting and insulin-stimulated state, before and after weight loss. The results presented here suggest that there is a difference in EC, especially 2-AG, levels between the two obesity phenotypes.
It has been previously shown that insulin resistance is associated with an increase in the concentrations of ECs, especially those of 2-AG, in the adipose tissue . This is in agreement with our present finding of higher plasma 2-AG levels in fasting IRO women. Increased local tissue concentrations of 2-AG, and subsequent elevation of CB1 receptor activity, are suggested to reduce glucose uptake in skeletal muscle and increase abdominal adiposity and free fatty acid flow from adipose tissue to the liver, thus increasing the risk of insulin resistance [23, 24]. To test the inhibitory acute effect of insulin on circulating EC levels [8, 12], we also investigated here the effect of a HEC and, altogether, the results obtained suggest that insulin resistance is associated with higher peripheral levels of 2-AG, but not with impaired reduction of AEA levels by insulin infusion. Thus, 2-AG, but not AEA, levels may serve as a discriminating marker between the ISO and IRO phenotypes. HEC was reported to reduce plasma AEA, but not 2-AG levels, also in a previous study in both lean and obese men, although plasma levels of OEA and PEA were not measured. Therefore, although type 2 diabetes is associated with higher fasting plasma levels of both AEA and 2-AG , acute hyperinsulinemia has an inhibitory effect only on circulating AEA, in both insulin sensitive and resistant subjects. This inhibitory effect also affects the two AEA-related molecules, OEA and PEA which are PPAR-α activators  with biosynthetic and inactivating processes similar to those of AEA  and plasma concentrations usually correlating with those of this EC [15, 21].
After 6 months of weight loss, we did not observe any statistically significant change in the plasma EC levels of either IRO or ISO women, although the difference in 2-AG levels between the two phenotypes became smaller. Based on the above discussion, and in view of the concurrent observation of overall increased and reduced insulin sensitivity in the former and latter group, respectively, this lack of intra-group changes was unexpected particularly for 2-AG. Furthermore, a previous study reported that an admittedly stronger weight loss caused by 1-year lifestyle intervention in abdominally obese men, and leading to an 8 cm reduction in waist circumference, was accompanied by decreased plasma 2-AG levels correlating with improved insulin sensitivity . It is possible that the weight-loss-induced changes in insulin sensitivity in the two present groups were not sufficient to determine a corresponding change in basal 2-AG levels. Thus, weight loss without strong improvement of insulin sensitivity and/or other metabolic disturbances may not be sufficient to produce a decrease in plasma 2-AG levels. As to plasma AEA levels, although following weight loss they became lower in IRO than in ISO women, we did not really expect major intragroup changes based on the fact that before weight loss no difference between the two phenotypes had been observed. Furthermore, the aforementioned previous study in obese men  showed that even with a higher weight loss and improvement of insulin sensitivity, plasma AEA levels are only decreased to a small extent.
Weight loss also had an impact on plasma PEA, yet, after adjusting for the type of weight loss intervention, the effect was lost. This could suggest that the type of weight loss intervention (diet vs. exercise vs. combined) might have an impact on plasma PEA levels. However, further studies are needed to understand the impact of weight loss and type of intervention on PEA, which is thought to be an anti-inflammatory marker and to reduce cardiovascular risk [5, 25, 27]. Nevertheless, the present finding of a negative correlation between PEA levels and M-values may represent an adaptive mechanism through which, when insulin sensitivity improves, there is less need to produce a PPAR-α ligand and anti-inflammatory mediator. Conversely, this finding is unlikely to suggest that increased M-values and insulin sensitivity are caused by decreased PEA levels.
Future studies will need to address the issue of what other factors (e.g., hormones other than insulin, like ghrelin, cholecystokinin, and leptin, which are all known to control EC tone [28, 29], or branches of insulin signaling pathways that are less impaired during insulin resistance) underlie, in concert with weight loss and insulin sensitivity/resistance, the higher levels of 2-AG in IRO women as well as the response of AEA, PEA, and OEA to insulin infusion. Also, comparing the activity of CB1 receptors in peripheral tissues, such as skeletal muscles and liver, from both ISO and IRO women could be useful in understanding the differences in the two phenotypes. Indeed, CB1 receptor activation was shown to inhibit the translocation of glucose transporter 4, and to have a negative effect on genes regulating insulin sensitivity in skeletal muscles [30, 31].
The present study has several limitations. First, the ISO and IRO cohorts included only overweight or obese nondiabetic sedentary postmenopausal women. Therefore, our findings are limited to this gender, hormonal status, and obesity level. Secondly, the cohort is composed of 30 matched subjects, and, although this design minimizes confounding factors, a larger sample size might be necessary to increase the statistical power of the present evaluations. Thirdly, a comparison between our study and others is difficult because of differences in populations and study designs. Despite these limitations, however, our results are supported by the use of gold standard techniques for the evaluation of insulin sensitivity and EC levels in a well-characterized cohort.
In conclusion, the present study is the first to show that the EC, 2-AG, is associated with insulin resistance in obese postmenopausal women. Therefore, 2-AG constitutes an interesting surrogate marker for insulin resistance in this population. The response, or lack thereof, of plasma EC, PEA and OEA levels to acute insulin after HEC was similar in both obesity phenotypes. However, weight loss, which is known to have beneficial effects on lipid metabolism and insulin sensitivity, induced opposite effects in the two phenotypes, with IRO subjects showing improvement as compared to ISO subjects, and no strong overall change in EC, PEA and OEA levels.
Within subjects: significant effect between baseline and after weight loss. Between subjects: significant effect between ISO vs. IRO groups.
Abbreviations: BMI, body mass index; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; Mvalue, Glucose disposal rate; NS, nonsignificant; TG, triglycerides.