Adiponectin and the mediation of HDL-cholesterol change with improved lifestyle: the Look AHEAD Study

Abstract

Adipose tissue dysfunction plays a key role in the development of the metabolic abnormalities characteristic of type 2 diabetes (T2DM) and participates actively in lipid metabolism. Adiponectin, found abundantly in circulation and a marker of adipose health, is decreased in obese persons with T2DM. We investigated whether the changes in adiponectin with an intensive lifestyle intervention (ILI) for weight loss could potentially mediate the increase in low HDL-cholesterol (HDL-C) with ILI. Adiponectin and its fractions were determined using an ELISA with selective protease treatment in 1,397 participants from Look AHEAD, a trial examining whether ILI will reduce cardiovascular events in overweight/obese subjects with T2DM when compared with a control arm, diabetes support and education (DSE). Multivariable regression and mediational analyses were performed for adiponectin and its high-molecular-weight (HMW) and non-HMW fractions. ILI increased baseline HDL-C by 9.7% and adiponectin by 11.9%; changes with DSE were 1.3% and 0.2%, respectively (P < 0.0001). In a model including changes in weight, fitness, triglycerides, and glucose control and that adjusted for demographics and medical history, adiponectin changes remained significantly associated with HDL-C change. Data supported the contribution of changes in both HMW- and non-HMW-adiponectin to the improvement in HDL-C with ILI

More than 1 in every 10 Americans has diabetes. In these individuals, cardiovascular disease (CVD) is the largest single cause of death and accounts for a significant decrease in life expectancy (1). Low HDL-cholesterol (HDL-C), a major CVD risk factor, is a characteristic component of diabetic dyslipidemia. The mechanisms that account for this abnormality are not completely understood. Adipose tissue plays a key role in the development of metabolic abnormalities that characterize type 2 diabetes (T2DM) (2, 3). It is involved in triglyceride regulation and participates actively in cholesterol (4) and phospholipid metabolism (5). Adiponectin (ADN), an adipose hormone found abundantly in circulation, is decreased in diabetic persons, particularly in men (6, 7), in whom lower levels predict the occurrence of cardiovascular events (8). A direct association between ADN and HDL-C has been observed in cross-sectional studies (7, 9). ADN undergoes extensive posttranslational modifications and assembles into isoforms of varying size and receptor affinity (10).

Reports from a few intervention trials in nondiabetic subjects suggest that weight loss, achieved through lifestyle behavior change or bariatric intervention, results in higher total ADN (11–13) and HDL-C levels (13). It has not been determined whether the effects of weight loss on HDL-C levels are mediated by ADN change or whether they are consequent to changes in other potential effectors modified with the intervention. In addition, given that ADN comprises a series of oligomeric/multimeric forms, it is of physiological and future pharmacological interest to determine which ADN fraction may account for the improvement in HDL-C. Both high-molecular-weight (HMW)-ADN and non-HMW-ADN fractions may participate in the modulation of HDL metabolism (14). Non-HMW-ADN is present in circulation in high concentrations (14), making it important to evaluate not only the role of changes in total ADN and HMW-ADN on HDL-C change, but also the effects of change in this less-studied ADN fraction.

We hypothesized that when compared with usual care, intensive lifestyle intervention (ILI) for weight loss would increase both HMW- and non-HMW-ADN and that their changes would potentially contribute to the increase in HDL-C with ILI. We posited that the effects of ILI on HDL-C could be mediated through ADN, lending support to the hypothesis that in obesity and diabetes, not only a reduction in adipose tissue mass, but also an improvement in adipose tissue function, is necessary for the improvements in HDL-C with ILI.

METHODS

Study design

We evaluated 1,397 Look AHEAD participants with biomarker and fitness data at baseline and 1 year. Look AHEAD is a clinical trial testing whether ILI will reduce CVD events in overweight/obese subjects with T2DM (15). The Look AHEAD study design, intervention, and participant characteristics have been previously described (15). Briefly, participants were randomized to ILI, aiming for a 7% weight loss from baseline, or to a diabetes support and education (DSE) arm, as control. ILI participants attended three group sessions and one individual monthly encounter during the initial 6 months, followed by two group sessions and one individual monthly appointment thereafter, in support of behavioral change to increase physical activity to 175 weekly minutes of moderate-intensity exercise and to reduce caloric intake. The activity program relied on at-home exercise, mostly brisk walking. The energy intake goal was 1,200–1,500 kcal/day if body weight was <114 kg and 1,500–1,800 kcal/day if weight was ≥114 kg. Liquid meal replacement for two daily meals was encouraged during the first 6 months of the study to help with portion control. DSE participants received three group health information sessions during the year. All participants continued care with their primary providers. The institutional review boards of the participating centers approved Look AHEAD and this ancillary study.

Laboratory, anthropometric, and fitness determinations

Plasma ADN and HMW-ADN levels were measured using a sandwich ELISA (American Laboratory Products Co.; Salem, NH) before and after selective protease pretreatment, respectively, as reported (16). Sensitivity of the assay was 0.019 ng/ml. Average intra- and interassay coefficients of variation were 2.3% and 9.6% for total ADN and 2.2% and 11.1% for HMW-ADN. Non-HMW-ADN was calculated by subtracting HMW-ADN from total ADN. HDL-C was measured enzymatically after precipitation of whole plasma with dextran sulfate-Mg2+ (17).

Determination of fitness (submaximal effort on a graded exercise stress test) and procedures for anthropometric measures, hemoglobin A1c (HbA1c), and lipids in Look AHEAD have been described previously (17).

Statistical analysis

Descriptive statistics included median and interquartile range for total ADN and its fractions and mean and standard deviation for HDL-C. Differences between the ILI and DSE arms in 1 year variable changes were evaluated using the two-sample t-test or the Wilcoxon rank sum test. Multivariable linear regression analyses tested the association of HDL-C change (outcome) and change in ADN and each of its fractions as independent variables. Changes in weight, fitness, glucose (HbA1c), and triglyceride control, either combined or within separate models, were also evaluated. All models were adjusted for baseline HDL-C level, demographics, clinic site, CVD history, diabetes duration, current smoking, hormone replacement in women, and treatment with statins, thiazolidinediones, and insulin. Treatment effect was examined with the use of an indicator for treatment group (“ILI versus DSE”). The significance of an ILI*gender interaction effect was evaluated in the full models to determine the need for stratified analyses by gender. Spearman's correlation coefficients were determined prior to construction of the regression models to exclude colinearity and to select the adiposity change measure (body mass index [BMI], waist, or weight) to include in the models. To test for the potential mediation of ADN and/or its fractions on 1 year HDL-C change with ILI, mediation analysis was performed as previously described (18) using the Baron and Kenny (19) approach and the Sobel (20) test to assess statistical significance of the potential mediators. Type I error rate was fixed at 0.05 and two-tailed for all analyses. Analyses were performed using SAS 9.2 (SAS Institute; Cary, NC).

RESULTS

Baseline characteristics

Participants were middle-aged, obese, and sedentary (Table 1). Baseline ADN, HMW-ADN, and non-HMW-ADN levels were low [geometric means (95% confidence interval) of 4.90 (4.76 to 5.04) and 1.89 (1.82 to 1.98) µg/ml for total ADN and HMW-ADN, respectively], when compared with those measured by the same laboratory and assay in nondiabetic adults (16) and by others with different methodology (6). We found no report on non-HMW-ADN levels in any other large study with which to compare our result. Total ADN, HMW-ADN, and HDL-C levels were higher in women than in men, as observed by others (16). Because Look AHEAD's age eligibility criteria changed after the first year of recruitment, subjects in this ancillary study were slightly younger and had less CVD when compared with the remaining cohort (21) (see supplementary Table I).

TABLE 1.

Baseline characteristics

	ILI (n = 732)	DSE (n = 665)
Age (years, mean [SD])	57.1 (7.1)	57.3 (7.3)
Gender (% females)	57	57
Race/ethnicity (% Caucasians)a	68	66
History of CVDb (%)	12	11
Current tobacco usec (%)	4	3
Statin therapy (%)	41	40
Thiazolidinedione therapy (%)	24	27
Insulin therapy (%)	15	14
Estrogen replacement in women c (%)	30	34
Duration of diabetesc (years)	6.56 (6.32)	6.57 (6.31)
Weight (kg)	102.85 (20.37)	101.33 (18.62)
BMI (kg/m )	36.43 (6.36)	36.04 (5.93)
Waist circumference (cm)	115.11 (14.8)	114.52 (14.08)
Fitness (submaximal, MET)	5.20 (1.49)	5.19 (1.54)
HbA1c (%)	7.32 (1.17)	7.34 (1.15)
LDL-C (mmol/l)	2.90 (0.68)	3.41 (0.85)
Triglycerides (mmol/l, median [IQR])	1.81 (1.25 to 2.61)	1.74 (1.20 to 2.47)
HDL-C (mmol/l)
Men	0.96 (0.23)	0.95 (0.23)
Women	1.46 (0.30)	1.19 (0.29)
ADN (µg/ml, median [IQR])
Men	4.04 (3.00 to 5.77)	4.30 (2.91 to 5.87)
Women	5.08 (3.73 to 7.30)	5.21 (3.86 to 7.76)
HMW-ADN (µg/ml, median [IQR])
Men	1.50 (0.93 to 2.62)	1.60 (0.97 to 2.69)
Women	2.14 (1.34 to 3.35)	2.21 (1.44 to 3.48)
Non-HMW-ADN (µg/ml, median [IQR])
Men	2.49 (1.95, 3.33)	2.55 (1.97, 3.31)
Women	2.87 (2.32, 3.88)	2.99 (2.35, 4.00)

Data shown as mean (SD) unless otherwise noted. IQR: interquartile range; MET: metabolic equivalent of task. To convert mmol/l to mg/dl, multiply by 38.67 for cholesterol and 88.57 for triglycerides.

^aOther races/ethnicities by treatment arm (ILI and DSE, respectively): African American (11% and 12%), Hispanic (10% and 10%), American Indian (7% and 8%), Other (4% and 4%).

^bSelf-reported history of myocardial infarction, stroke, transient ischemic attack, angioplasty/stent, coronary artery bypass graft, carotid endarterectomy, abdominal aortic aneurysm, or heart failure.

^cBy self-report.

1 Year changes in adiponectin and other metabolic variables of interest

ILI participants had significant 1 year improvements in adiposity, fitness, glucose, HDL-C, and triglyceride levels, but no significant changes in LDL-cholesterol, when compared with the DSE arm (Table 2), as observed in the overall cohort (17). Total ADN, HMW-ADN, and non-HMW-ADN increased significantly with ILI in men and women, when compared with DSE (Table 2). Overall ADN increases from baseline were geometric mean (95% confidence interval) 11.9 (–7.2% to 37.5%) and 0.2 (–15.6% to 20.1%), with ILI and DSE, respectively. Absolute and relative changes with ILI for total ADN, HMW-ADN, and non-HMW-ADN were greater in men than in women. Spearman's correlation coefficients (not shown) adjusted by age, gender, and race/ethnicity, showed that changes in total ADN and each of its fractions were correlated with changes in weight, BMI, and waist circumference (all P < 0.0001). Given that the correlations with change in weight and BMI for total ADN (–0.232, –0.229) in the overall group and for HMW-ADN (–0.274, –0.272) were somewhat greater than the correlations with change in waist circumference (–0.213 for total ADN and –0.246 for HMW-ADN change), we chose to include change in weight in our regression models. Correlation coefficients for non-HMW-ADN and alterations in weight, BMI, and waist circumference change were similar. The Spearman's partial correlation coefficient for changes in HMW-ADN and non-HMW-ADN was 0.45.

TABLE 2.

Changes in metabolic variables, HDL-C, and adiponectin by treatment arm

Variable a	ILI (n = 732)	DSE (n = 665)	Pc
Δ Weight (kg)	−8.7 (7.6)	−0.7 (5.1)	<0.0001
Δ Waist circumference (cm)	−7.4 (9.4)	−0.9 (8.2)	<0.0001
Δ Fitness (submaximal, MET)	1.0 (1.4)	0.2 (1.1)	<0.0001
Δ HbA1c (%)	−0.7 (1.0)	−0.2 (0.9)	<0.0001
Δ LDL-C (mmol/l)	−0.10 (0.68)	−0.11 (0.74)	0.76
Δ Triglycerides(mmol/L, median [IQR])	−0.23 (−0.78 to 0.17)	−0.06 (−0.49 to 0.30)	<0.0001
Δ HDL-C (mmol/l); %Δ
Overall	0.09 (0.18);9.7% (17.2%)	0.03 (0.17);4.0% (15.7%)	<0.0001
Men	0.10 (0.16);11.8% (17.1%)	0.04 (0.15);5.5% (16.6%)	<0.0001
Women	0.08 (0.19);8.0% (17.1%)	0.03 (0.18);2.9% (14.9%)	<0.0001
Δ ADN (µg/ml; % Δ)b
Men	0.9 (−0.1 to 1.8);23.6% (−2.2% to 45.5%)	0.1 (−0.6 to 0.9);2.8% (−14.6% to 21.8%)	<0.0001
Women	0.3 (−0.6 to 1.6);6.2% (−9.6% to 28.6%)	−0.1 (−1.0 to 0.9);−1.5% (−17.8% to 18.5%)	<0.0001
Δ HMW-ADN (µg/ml; % Δ)b
Men	0.44 (−0.03 to 1.16);35.90% (−1.27% to 76.64%)	0.05 (−0.35 to 0.48);3.29% (−18.86% to 31.93%)	<0.0001
Women	0.27 (−0.21 to 0.91);14.78% (−10.55% to 43.76%)	−0.03 (−0.44 to 0.46);−1.49% (−20.36% to 23.15%)	<0.0001
Δ Non-HMW ADN (μg/ml; % Δ)b
Men	0.35 (−0.21 to 0.88);21.40% (−15.04% to 65.65%)	0.02 (−0.51 to 0.58);1.82% (−30.71% to 35.16%)	<0.0001
Women	0.07 (−0.52 to 0.69);3.10% (−25.40% to 30.83%)	−0.08 (−0.64 to 0.58);−4.97% (−27.59% to 28.14%)	0.049

Abbreviations and conversion factors as in Table 1. Data are shown as mean (SD) unless specified otherwise.

^a1 Year change (Δ) from baseline expressed as follow-up minus baseline values.

^bMedian (IQR)

^cFor difference between ILI and DSE.

Association of changes in HDL-C with changes in ADN and its fractions

The regression models showed that total ADN and HMW-ADN changes with ILI were each associated with HDL-C change at 1 year (Table 3, Part I, Models B and C), and that, as hypothesized, the associations remained significant after adjusting for changes not only in weight, but also in fitness, HbA1c, and triglycerides (Table 3, Part II, Models E and F). Of interest, weight loss was no longer associated with HDL-C change in the full model with HMW-ADN change (Table 3, Part II, Model F, P = 0.068) and was only close to achieving statistical significance in the model with total ADN (Table 3, Part II, Model E, P = 0.053). As hypothesized, non-HMW-ADN change was also independently associated with HDL-C change (Table 3, Part I, Model D and Part II, Model G, P < 0.0001). Change in HMW-ADN and non-HMW-ADN explained to a similar extent the change in HDL-C with ILI (Table 3, Part I, Models C and D, R² = 0.122 and 0.116, respectively). In contrast to what was observed with HMW-ADN change (Table 3, Part II, Model F), change in weight remained significant in the full model with non-HMW-ADN (Table 3, Part II, Model G, P = 0.02), suggesting that non-HMW-ADN and HMW-ADN may influence HDL-C via different mechanisms. Furthermore, when changes in both HMW- and non-HMW-ADN were entered into the same model (Table 3, Part II, Model H), after excluding colinearity, they were each found to be independently associated with HDL-C change. The combined effect of changes in HMW-ADN and non-HMW-ADN on the variance in HDL-C change in the full model (Table 4, Part II, Model H, R² = 0.176) was identical to that accounted for by total ADN change (Table 4, Part II, Model E, R² = 0.176), confirming that total ADN change is an excellent indicator of the combined effect of the ADN fractions on HDL-C change. The HMW-ADN/total ADN ratio explained less of the variance in HDL-C change (R² = 0.104; full model as in Table 3, Part II, not shown) than did changes in total ADN, again pointing to the contribution of both ADN fractions to the HDL effect.

TABLE 3.

Association of HDL-C alterations with changes in ADN or its fractions

Change in HDL-C
Modela	B	SE	P	R
Part I. Basic Models
Model A				0.086
ILI v. DSE	2.331	0.356	<0.0001
Model B				0.129
ILI v. DSE	1.896	0.351	<0.0001
Δ ADN	0.518	0.063	<0.0001
Model C				0.122
ILI v. DSE	1.895	0.354	<0.0001
Δ HMW-ADN	0.759	0.102	<0.0001
Model D				0.116
ILI v. DSE	2.105	0.351	<0.0001
Δ Non-HMW-ADN	0.848	0.124	<0.0001
Part II. Full Models
Model E				0.176
ILI v. DSE	1.011	0.403	0.012
Δ ADN	0.442	0.063	<0.0001
Δ Weight	−0.056	0.029	0.053
Δ Fitness	0.195	0.150	0.19
Δ HbA1c	−0.158	0.189	0.40
Δ Triglycerides	−0.012	0.002	<0.0001
Model F				0.170
ILI v. DSE	1.028	0.405	0.011
Δ HMW-ADN	0.63	0.102	<0.0001
Δ Weight	−0.053	0.029	0.068
Δ Fitness	0.219	0.151	0.15
Δ HbA1c	−0.146	0.190	0.44
Δ Triglycerides	−0.012	0.002	<0.0001
Model G				0.169
ILI v. DSE	1.047	0.405	0.0098
Δ Non-HMW-ADN	0.740	0.121	<0.0001
Δ Weight	−0.067	0.029	0.0214
Δ Fitness	0.210	0.151	0.1635
Δ HbA1c	−0.237	0.189	0.2092
Δ Triglycerides	−0.012	0.002	<0.0001
Model H				0.176
ILI v. DSE	1.011	0.403	0.0123
Δ HMW-ADN	0.418	0.120	0.0005
Δ Non-HMW-ADN	0.472	0.143	0.0010
Δ Weight	−0.057	0.029	0.0516
Δ Fitness	0.194	0.150	0.1957
Δ HbA1c	−0.162	0.189	0.3942
Δ Triglycerides	−0.012	0.002	<0.0001

^aEach was analyzed independently and adjusted for baseline HDL-C, age, gender, race/ethnicity, clinic site, history of CVD (defined in Table 1), diabetes duration, current smoking, and use of insulin, thiazolidinediones, statins, and hormone replacement in women.

“ILI vs DSE” is an indicator for treatment group. An interaction term ILI* gender was evaluated in the full models (E–G) and was nonsignificant (P = 0.675 for total ADN, P = 0.746 for HMW-ADN, and P = 0.5983 for non-HMW-ADN). Sums of squares for Δ HMW-ADN and Δ non-HMW-ADN were 476.9 and 427.6 in Model H.

TABLE 4.

Summary of mediational analyses for alterations in adiponectin and its oligo/multimers as potential mediators of HDL-C change with ILI

Δ, Change. Regression parameters for all paths based on models adjusting for treatment group and clinic site. For Paths B, C, and C', models also adjust for baseline HDL-C, Δ HbA1c, Δ weight, Δ fitness, Δ triglycerides; age, gender, race/ethnicity, history of CVD (defined in Table 1), diabetes duration, current smoking; use of insulin, statins, thiazolidinediones, and hormone replacement. Paths B and C' include mediator in the model.

Effect of metabolic changes on HDL-C change with ILI

Of the change in metabolic variables, in addition to ADN change, only triglyceride change remained significant in all full models (P < 0.0001, Table 3, Part II). Improved fitness and glucose control were independently associated with HDL-C change in intermediate models without and with ADN change (see supplementary Table II), but not in the full models (Table 3, Part II).

Mediational effect of total ADN and its fractions on HDL-C change with ILI

ADN, HMW-ADN, and non-HMW-ADN changes differed by treatment arm (Table 2) and were independently associated with HDL-C change (Table 3), fulfilling criteria to be examined as potential mediators of HDL-C change with ILI (19). The measure of the direct effect of ILI on the potential mediators (change in ADN or ADN fraction) is depicted by Path A in Table 4, the relationship between potential mediators and HDL-C change by Path B, and the effect of ILI on HDL-C without the mediational effect is indicated by Path C. Path C’ tests the effect of ILI on HDL-C change, with each putative mediator (changes in ADN, HMW-ADN, and non-HMW-ADN) entered into the regression model separately. Mediational analyses confirmed that changes in total ADN and each ADN fraction (HMW-ADN and non-HMW-ADN) were potential mediators of the increase in HDL-C with ILI, independently of baseline HDL-C, demographics, medical history, and changes in weight, fitness, triglycerides, and HbA1c. The potential mediating effect of HMW-ADN on HDL-C change was greater than that of non-HMW-ADN, although both were significant.

DISCUSSION

As hypothesized, lifestyle changes associated with moderate weight loss significantly increased the low baseline HDL-C levels of obese diabetic individuals, when compared with usual care, with ADN changes potentially contributing to the rise. Furthermore, the increase in HDL-C with ILI was associated with and potentially mediated by both ADN fractions: HMW-ADN and the less-recognized and more-abundant non-HMW-ADN fraction. Our results show that the effects of ADN and each of its fractions on HDL-C levels occurred independently of weight change, in agreement with the paradigm that an improvement in adipose tissue function is necessary for the favorable metabolic effects of weight loss.

Experimental evidence supporting a role for ADN in the formation of HDL particles comes from work in ADN-deficient mice (22). In these animals, there are decreases in apoA-I levels and in the expression of ATP-binding cassette transporters in liver, suggesting that ADN deficiency may impair HDL assembly. Recent reports have also identified the presence of ATP-binding cassette transporters in adipocytes and have confirmed an active role for adipose tissue in the synthesis of HDL (4, 23). However, differences in HDL metabolism between rodents and humans (24) and in the posttranslational modifications of ADN, which have functional implications and vary by species (25), preclude the extrapolation of ADN and HDL-C mouse findings to humans.

A few clinical reports on weight-reducing interventions with lifestyle behavior changes have documented increases in ADN and HDL-C levels (11, 12) but did not investigate their association. Instead, their attention focused on the now well-known effects of ADN on insulin sensitivity and incident diabetes (11, 12, 26). A small study of short duration that also evaluated ADN, but lacked an exercise component, did not show an increase in HDL-C level (27). However, a correlation between change in ADN and change in HDL-C was reported by Baratta et al. (13) with surgical weight loss, but because of the nature of the intervention, no control group was included, making the observations of limited value. In addition, this is, to our knowledge, the first study to investigate the contribution of change in ADN fractions to HDL-C change. Previous oligomer/multimer studies have focused on HMW-ADN and its effects on insulin sensitivity (16, 26). We show that in addition to HMW-ADN, the more-abundant non-HMW-ADN fraction participates in the modulation of HDL-C levels. Finally, given the multiple metabolic changes that occur with weight loss, which on their own could influence HDL-C levels, confirmation of a potential mediating role for change in ADN and its fractions in HDL-C has been lacking. In this study, we applied specialized analytical approaches to gain insight, within the context of a randomized lifestyle intervention trial, on the potential mediation of ADN and its oligo/multimers in HDL-C change. The analyses controlled for changes not only in weight but also in fitness, triglyceride, and glucose control. Our findings show a significant and independent association between the rise in ADN with ILI and the increase in HDL-C levels and suggest that ADN, through changes in both its fractions, may potentially mediate the favorable effects of ILI on HDL-C levels. Alterations in non-HMW- and in HMW-ADN were each strongly associated with HDL-C change and independently mediated the hormone's lipid effect. These findings are in agreement with a cross-sectional study in 68 diabetic and nondiabetic subjects reporting that both non-HMW-ADN and HMW-ADN are associated with HDL particle size (14). Our work supports a complex role for ADN, with distinct but complementary functions for each fraction in HDL-C levels. We show that changes in the non-HMW-ADN fraction were independently associated with HDL-C change, not only after full metabolic adjustment but also when HMW-ADN was accounted for in the model. These results, together with the potential mediational effect on HDL-C found for each of the ADN fractions, imply that HMW-ADN and non-HMW-ADN are both important in the modulation of HDL-C levels and suggest the possibility that each may operate through different mechanisms.

ADN activates AMP-activated protein kinase and peroxisome proliferator-activated receptor α, increasing FFA oxidation, improving insulin sensitivity, and reducing VLDL synthesis and triglyceride levels (28). Although these processes may influence HDL-C levels, our data and the experimental evidence for an ADN effect on apoA-I and ATP-binding cassette transporter expression (4, 22, 23) support a direct, triglyceride-independent effect of ADN on HDL-C. Recently, it has been reported that the upstream effects of ADN are mediated by activation of ceramidase and its effects on the formation of sphingosine-1-phosphate (S1P) (29). Because HDL is a major carrier of sphingolipid in circulation, we speculate that change in ADN expression may influence not only HDL-C but also HDL function by altering HDL sphingolipid content. HDL containing S1P, but not HDL without S1P, has been found to have vasculoprotective effects (30).

Strengths of the study include its findings within the largest randomized lifestyle intervention trial for weight loss in obese diabetic persons and the use of specialized statistics to evaluate the independent effects of ADN and its fractions on HDL-C change. It must be noted that although our mediational analysis is suggestive of a mediational effect for ADN and its fractions on HDL-C change with ILI, it cannot establish causality. Furthermore, although ADN is an excellent marker of adipose tissue function, it is but one of the multiple bioactive substances made in adipose tissue. Likewise, HDL is a complex lipoprotein whose metabolism cannot be fully appreciated by looking at HDL-C alone.

In summary, we report that when compared with usual care, ILI exerted important changes on HMW-ADN and non-HMW-ADN levels, and that their combined effect, easily assessed by measuring total ADN change, may partially mediate the increases in HDL-C with ILI, independently of changes in triglyceride, weight, fitness, and glucose control. Our findings suggest an active role for adipose tissue function in the modulation of HDL metabolism. Studies in humans are needed to provide confirmatory evidence of a causal relationship between ADN and HDL-C change and to elucidate the underlying mechanisms. CVD outcome data from Look AHEAD will inform whether improved HDL-C levels with ILI will contribute to CVD event reduction in obese diabetic persons and investigate whether changes in ADN may be a potential contributor.