The Effects of Macronutrient Intake on Total and High Molecular Weight Adiponectin: Results from the OMNI-Heart Trial

Abstract

Higher levels of the adipocyte-specific hormone adiponectin have been linked to increased HDL and lower insulin resistance. This study was conducted to determine the influence of macronutrient intake on adiponectin levels. One hundred and sixty-four pre- and stage-1 hypertensive adults participated in OMNI-Heart, a cross-over feeding study originally testing the effects of macronutrients on blood pressure. Participants underwent three 6-week feeding periods: one rich in carbohydrates (CARB), one rich in monounsaturated fat (MUFA), and one rich in protein (PROT), while maintaining body weight. Their median plasma high molecular weight (HMW) and total adiponectin levels were 2.3 and 8.2 μg/ml, respectively, resulting in an average of 27% HMW adiponectin. Both HMW and total adiponectin levels decreased after baseline while the percent HMW adiponectin remained unchanged. Between diets, the MUFA diet maintained a higher level of both HMW and total adiponectin level than either the CARB (HMW: +6.8%, p=0.02; total: +4.5%, p=0.001) or PROT (HMW: +8.4%, p=0.003; total: +5.6%, p<0.001) diets. Changes in total adiponectin levels were positively correlated to changes in HDL cholesterol irrespective of diets (Spearman r = 0.22–0.40). No correlation was found between changes in lipids, blood pressure, or insulin resistance (HOMA-IR). Macronutrient intake has effects on HMW and total adiponectin levels independent of weight loss. A diet rich in monounsaturated fat was associated with higher levels of total and HMW adiponectin in comparison to a carbohydrate- or protein- rich diet. Effects seen in adiponectin paralleled those found with HDL cholesterol.

The adipoctye-specific hormone, adiponectin, is associated with decreased risk of cardiovascular disease (CVD) and type 2 diabetes.(1) In large population-based studies, high levels of adiponectin have been consistently associated with reduced adiposity, dyslipidemia, insulin resistance, and inflammation.(2, 3).

Results from observational or weight loss studies have suggested that low carbohydrate(4) and high unsaturated fat diets might increase adiponectin.(5, 6) A cross-sectional analysis of 532 men found that higher carbohydrate intake was associated with lower adiponectin levels.(4) However, food intake was assessed by questionnaire which is subject to inaccurate recall. In weight loss studies that concurrently changed macronutrient composition, increases in unsaturated fat by fish oil(6) or by Mediterranean diet(5) appeared to increase adiponectin. Small intervention studies that investigated macronutrient effects independent of weight loss have been inconsistent.(7–9) These did not measure the high molecular weight (HMW) form of the hormone, previously found to the most bioactive.(10, 11) Hence, the effects of macronutrient intake on adiponectin are uncertain, especially in the setting of stable weight.

Understanding the effects of macronutrient intake on adiponectin levels independent of weight loss is important since weight loss remains difficult to achieve and maintain and because higher levels of adiponectin have been associated with reduced risk of CVD and diabetes.(1) Thus, the objective of this study was to test whether macronutrient intake, in the context of a weight stable, healthy diet, would change adiponectin levels and whether changes in adiponectin were associated with changes in cardiovascular risk factors. We nested our investigations in the Optimal Macro-Nutrient Intake Heart Trial (OMNI-Heart), a randomized, 3-period crossover feeding trial which previously showed that a healthy diet rich in protein could further decrease blood pressure and lipids when compared to a carbohydrate diet.(12)

Research Design and Methods

Study Population

OMNI-Heart recruited 164 adult participants (≥30 years old) that had pre- or stage 1-hypertension with mean SBP of 120–159 mmHg or DBP of 80–99 mmHg over 3 screening visits at one of two clinical centers (Baltimore and Boston). Institutional Review Boards at both centers approved the study. Each participant provided written informed consent.

Study characteristics have been previously reported. Major exclusion criteria were diabetes, cardiovascular disease, LDL over 5.70 mmol/l, fasting triglycerides over 8.48 mmol/l, weight over 350 lbs, and consumption of over 14 alcoholic beverages per week.(13) Of the 164 participants, exclusion of individuals who did not complete all 3 feeding periods and for outliers resulted in 161 baseline comparisons, 158 comparisons of analysis between carbohydrate and protein, 156 comparisons between unsaturated fat and protein, and 155 comparisons between carbohydrate and unsaturated fat for HMW adiponectin.

Dietary Intervention

In random order, participants consumed each of three healthy diets: 1) a diet rich in carbohydrates (CARB), 2) a diet rich in protein (PROT) and 3) a diet rich in unsaturated fat (MUFA). Each feeding period lasted six weeks. During each dietary period, participants ate their main weekday meal on site and were provided the rest of their food. 2–4 weeks wash-out periods where the participants ate their own food separated each feeding period. All diets were healthy with reduced saturated fat, sodium, and cholesterol and rich in vegetables, fruits, fiber, and other minerals at recommended levels. The distribution of the macronutrients and sources of each has been previously published.(14)

The three isocaloric diets differed by 10% kilocalories of macronutrient content each so that during the CARB diet participants consumed 58% of their calories in carbohydrates while this was reduced to 48% during the MUFA and PROT diets. Sample menus for each of the diets are available at http://www.omniheart.org/. Participants consumed 37% in fat during the MUFA diet. The increase in fat was predominantly in monounsaturated fat (+8% kcal) with a small increase in polyunsaturated fat (+2%), achieved through additional olive, safflower, and canola oils along with nuts and seeds. During the PROT diet, participants consumed 25% of their calories in protein. Two-thirds of the increased protein came from vegetable sources (i.e. seeds, legumes) with soy protein intake kept low (~7.3g/day). There was also an increase in low fat dairy and eggs consumed during protein diet. The type of carbohydrate for the three diets was kept similar as determined from the glycemic index. The micronutrient intake was also kept constant between the three periods as reflected from 24-hour urinary sodium, potassium and phosphorous concentrations previously published.(12)

Measurements and Adherence

Weight was measured at each weekday visit and effectively kept to a maximum of 2% difference by adjustment of caloric intake.(12) Mean energy intake (kcal/d) did not differ by diet and were: 2599 for CARB, 2558 for PROT, and 2564 for MUFA.(12) Mean weight (kg) by diet also did not differ and were: 86.6 for CARB, 86.7 PROT, and 86.6 MUFA.(12) Alcohol intake (oz/day) through daily diaries did not differ by diet and were: 1.9 for CARB, 2.1 for PROT, 2.0 for MUFA. Behavioral information (i.e. smoking status, alcohol intake, physical activity) was assessed by questionnaires at baseline. Systolic and diastolic blood pressure were measured by trained personnel blinded to diet, at multiple times during clinic visits and averaged together.(13)

Laboratory Methods

Blood after 8 hour fast was collected at baseline and at the end of each period. Plasma was separated and aliquoted for frozen storage (−70°C) and sent to the Core Laboratory for Clinical Studies (Washington University, St. Louis, MO) for measurement in 2005 for values of glucose, insulin, and lipids. Plasma glucose, analyzed by the hexokinase method (Roche, Indianapolis, IN), and insulin, analyzed by microparticle enzyme immunoassay (Abbott Imx analyzer, Abbott Park, IL), was used to calculate insulin resistance by the homeostasis model of assessment HOMA-IR (15). Total triglycerides, total cholesterol and HDL were measured by a Hitachi 917 analyzer. LDL was estimated by the Friedewald equation.(16)

HMW and total adiponectin were measured from frozen plasma in 2007 by the Core Laboratory of the Bayview General Clinical Research Center (Baltimore, MD). Samples from each participant were analyzed with baseline, feeding period (FP) 1, FP2, and FP3 samples located next to each other on the same well plate. Technicians were blinded to order of diet randomized to each participant. Quality control samples were included with each assay. Control values of adiponectin were within the range expected by commercial standards. Plasma HMW (Millipore ELISA, Bellerica, MA(17)) and total adiponectin (Linco ELISA, St. Charles, MO) had intra-assay coefficients of variation for 4.4% and 2.2%, respectively. Percent HMW was calculated from HMW divided by total adiponectin.

Statistical Methods

Continuous variables were examined for normality and outliers. Since HOMA-IR, HMW and total adiponectin were highly right skewed, their distributions were described as medians and 25^th and 75^th percentiles and log-transformed in analyses. Spearman correlations were used to test the associations at baseline and for changes between cardiovascular risk factors and adiponectin measures. Period and first order carry-over effects were tested using repeated measures analysis of variance. The regression model was performed with fixed effects and then random effects by methods as described in Jones & Kenward.(18) We tested for period by diet interaction by including an interaction term in the model. A carry-over term was created lagging exposure by one feeding period. Statistically significant carry-over effects were found to modify the association between macronutrient effects and total adiponectin. Analyses using generalized estimating equations (GEE) with exchangeable correlation yielded results that were not qualitatively different. Diet-to-diet comparisons are controlled for micronutrients and other factors by study design, thus these differences better reflect the macronutrient effect on adiponectin than baseline comparisons and represent our primary analysis. No adjustments were made for multiple testing. All analyses were carried out using Stata version 9.2 (College Station, Texas).

Results

Baseline Characteristics

Table 1 shows baseline characteristics among the 161 participants with adiponectin measured. Median total adiponectin was 8.2 μg/ml, and median HMW was 2.3 μg/ml. The participants had a mean %HMW adiponectin of 27%. We found a strong correlation between HMW and total adiponectin (Spearman r=0.89, p<0.01). This correlation did not differ by obesity status or dietary period. Percent HMW and total adiponectin were moderately correlated with each other (Spearman r=0.50). Levels of total and HMW adiponectin were significantly higher among women compared to men (p<0.001) and Caucasians compared to African Americans (p<0.001) even after adjusting for BMI.

Table 1.

Baseline characteristics of 161 OMNI-Heart participants with plasma HMW and total adiponectin measured

Variable	mean (SD)
Age, years	53.5 (11)
Male, n(%)	89 (55%)
Race, n(%)
Black	89 (55%)
White	64 (40%)
Other	8 (5%)
Weight, kg	87.6 (19)
Body Mass Index, kg/m2	30.3 (6)
Normal, BMI=<25	33 (21%)
Overweight, BMI 25–<30	55 (34%)
Obese, BMI >=30	73 (45%)
Current Smoking, n(%)	16 (10%)
Alcohol, oz/db	4.95 (8.2)
Systolic BP, mmHg	131 (9.3)
Diastolic BP, mmHg	77 (8.3)
Triglycerides, mmol/l a	1.1 (0.8–1.8)
Total Cholesterol, mmol/l	53 (9)
HDL, mmol/l	1.3 (0.4)
LDL, mmol/l	3.3 (0.8)
Insulin, pmol/l a	49 (35–90)
Glucose, mmol/l	5 (0.9)
HOMA-IR a	1.7 (1.0 – 2.9)
HMW adiponectin, μg/ml a	2.3 (1.3–3.7)
Total adiponectin, μg/ml a	8.1 (6.0–11.2)
Percent HMW adiponectin, %	27% (11)

^amedian (25^th – 75^th percentile)

^bamong n=160 with baseline FFQ data

Associations between Adiponectin and Cardiovascular Risk Factors at Baseline

Table 2 shows the cross-sectional spearman correlations between adiponectin and cardiovascular risk factors. Similar associations were found for all three measures of adiponectin. A direct correlation was found with HDL and inverse ones were found with triglycerides, insulin, glucose and HOMA-IR. Adjustment for age, gender, race, BMI, and smoking status decreased each of the correlations found but they remained significant. (Online Supplemental Table A1). No significant associations were seen with smoking status, physical activity, alcohol intake, caffeine, blood pressure, and total or LDL cholesterol in adjusted and unadjusted analyses.

Table 2.

Spearman Correlations Between Baseline CVD risk factors and Measures of Adiponectin

All Participants (n=161)	Total Adiponectin	HMW Adiponectin	Percent HMW Adiponectin
Age, years	0.28c	0.26b	0.16a
Alcohol, oz/day	−0.08	0.005	0.11
Caffeine, 100mg/d	−0.01	−0.02	−0.02
Vigorous Activity, times/week	0.03	0.02	0.02
BMI, kg/m2	−0.11a	−0.17a	−0.21b
SBP, mmgHg	0.05	0.09	0.09
DBP, mmHg	−0.09	−0.07	−0.04
Triglycerides, mmol/l	−0.33c	−0.32c	−0.22b
Total cholesterol, mmol/l	0.11	0.07	0.002
LDL cholesterol, mmol/l	0.05	0.02	−0.03
HDL cholesterol, mmol/l	0.41c	0.38c	0.26b
Insulin, pmol/l	−0.36c	−0.36c	−0.29c
Glucose, mmol/l	−0.25b	−0.25b	−0.20b
HOMA-IR	−0.36c	−0.36c	−0.29c

^ap<0.05,

^bp<0.01,

^cp<0.001

Changes in Adiponectin Compared to Baseline Levels

Compared to baseline, participants had lower levels of HMW and of total adiponectin at the end of each of the three feeding periods (Table 3). Percent HMW decreased approximately 1% from baseline for all three diets although this was not statistically significant for any of the diets (p>0.05).

Table 3.

Mean changes (95%Confidence Interval) in log HMW, log total, and %HMW adiponectin from baseline

	n	Log HMW	Log Total	% HMW
CARB – Baseline	159	−0.13 (−0.16 to −0.06) p<0.01	−0. 10 (−0.14 to −0.06) p<0.01	−1.0% (0.2 to −2.2%) p=0.11
MUFA – Baseline	156	−0.07 (−0.11 to 0.01) p=0.05	−0.06 (−0.09 to −0.02) p<0.01	−0.7% (0.6 to −2.1%) p=0.28
PROT – Baseline	159	−0.14 (−0.19 to −0.09) p<0.01	−0.11 (−0.14 to −0.08) p<0.01	−0.9% (0.4 to 2.3%) p=0.18

Effect of Macronutrient Composition on Adiponectin Levels

Although both total and HMW adiponectin decreased from baseline after each diet, there were significant differences in the level of decrease between diets. Of the three diets, a diet rich in unsaturated fat was beneficial in retaining high levels of HMW and total adiponectin. In comparison to the protein diet, the unsaturated fat diet achieved greater levels of HMW (+8.4%, p=0.005) and total adiponectin (+5.6%, p<0.001). (Figure 2) Similarly, the unsaturated fat diet had higher levels of HMW (+6.8%, p=0.02) and total (+4.5%, p=0.001) than the carbohydrate diet. In these comparisons, the protein diet had the lowest adiponectin levels although this did not differ significantly from the levels found from the carbohydrate diet. Lastly, even though significant differences were detected with the absolute measures of adiponectin, no difference was found with the percent HMW measure (p>0.7 for all tests). A similar pattern was evident in analyses stratified by race or gender.

Associations between Changes in Adiponectin and Changes in Cardiovascular Risk Factors

Changes in adiponectin from the three diets were then compared with changes in cardiovascular risk factors. Six different comparisons could be made (i.e. three baseline-to-diet or three diet-to-diet). In all such comparisons, changes in total adiponectin were significantly correlated with changes in HDL (with Spearman correlations of 0.20 to 0.37, all p<0.01). Changes in HDL remained strongly associated with changes in total adiponectin after adjustment for baseline total adiponectin and baseline HDL levels. Changes in other cardiovascular risk factors from the dietary intervention including lipids, blood pressure and insulin resistance, were not associated with changes in adiponectin.

Discussion

The results of our study demonstrate that in the context of a healthy diet and stable weight, modest changes in macronutrient intake (i.e. a 10% shift in calories) lead to changes in adiponectin levels. Moreover, the effects were observed for both HMW and total adiponectin, resulting in no significant change in the ratio of HMW to total adiponectin (percent HMW). Hence, changes in macronutrient intake do not preferentially change the molecular weight distribution of adiponectin. Across the three diets (one rich in carbohydrate, another rich in mono-unsaturated fat, and a third rich in protein), the diet rich in mono-unsaturated fat led to the largest increases in total and HMW adiponectin levels. Changes in adiponectin were strongly associated with changes in HDL but were not associated with changes in insulin resistance and other cardiovascular risk factors.

To our knowledge, this is the first study to demonstrate that an unsaturated fat diet is beneficial for total and HMW adiponectin even without weight loss compared to a high protein diet. Our findings of an unsaturated fat rich diet being beneficial in adiponectin levels compared to one that is high in carbohydrate are in agreement with other studies.(4, 5, 7) Unsaturated fatty acids are ligands for peroxisome proliferator-activator receptor gamma (PPARg)(19) whose activation has been directly found to increase adiponectin levels.(1) Thus, a high unsaturated fat rich diet could have direct effects on adiponectin levels. This direct effect has been also supported by a gene by environment interaction found between MUFA consumption and an adiponectin gene polymorphism on weight.(20)

Another mechanism by which adiponectin production may have changed could be through redistribution of fat deposits without weight loss. One recent 28-day isocaloric crossover study found that a low fat high carbohydrate diet led to a redistribution of fat from peripheral depots to central depots compared to a monounsaturated fat rich diet.(9) Although their investigations on fasting and postprandial plasma adiponectin were not conclusive, perhaps due to a smaller sample size (n=11), this observation of shifting fat deposition corresponds with our observation that a diet rich in carbohydrate led to larger decreases in adiponectin levels compared to a monounsaturated fat diet. They also found a decrease in peripheral adipose tissue adiponectin mRNA production postprandially after the high carbohydrate diet.(9) Although the authors concluded that the plasma levels did not correspond to their mRNA findings due to depot specific effects of adiponectin, we now provide evidence that a more global and not depot specific change in plasma adiponectin can occur.

The average changes in adiponectin documented in our study (i.e. often <0.5 μg/ml) are smaller than changes observed in other studies (e.g. +2.2 μg/ml in a longer term study of a Mediterranean diet with ≥10% weight loss (5)). Yet, the observed between-diet differences in both HMW and total adiponectin corresponded to differences between diets in HDL cholesterol, suggesting small differences in adiponectin are metabolically important. This connection with HDL is expected as it has been found that incubating hepatocytes with recombinant adiponectin can increase the production of HDL via increasing ApoA-I production.(21, 22) HDL levels also have stronger associations with adiponectin in human population studies than other cardiovascular risk factors(2, 23). Of the other lipids, LDL and total cholesterol has not been found to be consistently associated with adiponectin.

In comparison, the lack of association with changes in insulin resistance was surprising because of their associations cross-sectionally in the present study and others.(2, 23) Moreover, adiponectin has been found to act through increasing beta oxidation and decreasing hepatic glucose production to decrease insulin resistance.(1) Several reasons may explain the lack of corresponding change in HOMA-IR. First, the effect of macronutrient on both HOMA-IR and adiponectin without weight loss was small. This may have been due to short period in diet. Also, there is evidence that adiponectin may not play a large role in the early stages of insulin resistance, which may be applicable in our study because participants were non-diabetic and otherwise healthy except for raised blood pressure.(24)

Our finding of a decrease in adiponectin from baseline was intriguing. This was unlikely the result of systematic differences in lab assays as samples were not batched by feeding period. Theoretically it is possible that HMW adiponectin may dissociate into its lower molecular weight forms, thereby increasing total adiponectin over time of storage. Such changes have been observed due to temperature or pH changes which we do not believe our samples were exposed to. Although we do not have data using our own samples for the stability of the samples over long-term storage, another study using the same assay (Linco) for total adiponectin has shown that long-term storage (over 6, 12, 20, and 30 months) under frozen conditions (i.e. −30 degrees Celsius) does not systematically degrade samples.(25) In addition, since baseline samples were only stored for a couple months longer than FP1 samples (and 6–8 months for FP3 samples), we do not believe that degradation could explain the systematic decrease found from baseline.

Alternatively, the finding is real. The reduction in adiponectin from baseline could be related to the low alcohol intake of study participants. Some studies have demonstrated that alcohol increases adiponectin levels.(4, 26–28) However, analysis stratified by low versus high baseline alcohol consumption showed no differences in results (data not shown). Recent evidence into the central nervous system effects of adiponectin have revealed that it may be a starvation signal which triggers fat storage in the periphery. Our finding of decreased levels of adiponectin from baseline agrees with this hypothesis as the OMNI-Heart diet may have been more satiating than a free-living “Western” diet and its healthy content may have decreased the necessity for fat storage.(29) Nevertheless, this remains a hypothesis that requires further research. Regardless of the explanation for the decreases in adiponectin from baseline, these comparisons are pre- and post- randomization. The strongest aspect of this study is the randomized trial phase which directly compared the three diets.

Our study had several limitations. As with many cross-over trials, there are concerns with carry-over effects. We detected carryover effects by statistical modeling a carryover term using repeated linear models. However, statistical adjustment for this effect only reduced the difference in total adiponectin levels from carbohydrate versus protein feeding to non-significant levels but had otherwise no effect on the MUFA results with total adiponectin or with any of the results with HMW adiponectin. Despite the 2–4 week wash-out period, carryover effects may have occurred due to the process of adipose tissue remodeling. No measures of body composition were taken to corroborate this hypothesis. Another limitation is that multiple measures during each of the feeding periods were not assessed to perform a full mediation analysis where we would be able to determine whether changes in HDL occurred before or after changes in adiponectin levels. Lastly, the study was conducted among individuals who are hypertensive or pre-hypertensive. Their adiponectin levels, are slightly lower than the general population, even accounting for BMI, thus findings may not be applicable to all populations.

There were many strengths of our study as well. We measured the HMW form of adiponectin, the most bioactive form, which gave us more insight into the effects of macronutrients. OMNI-Heart was designed to look at effects of diet independent of weight loss. The study included a large sample of African American and Caucasian participants. Diets were designed so that they were feasible and relevant to the public - none of the macronutrient levels were extreme. Finally, the study was a randomized, tightly controlled feeding study with high adherence and retention rates. In conclusion, changes in macronutrient intake affect both HMW and total adiponectin levels independent of weight loss. A diet rich in unsaturated fat resulted in higher adiponectin levels in comparison to a carbohydrate- or protein- rich diet. Moreover, changes in HMW and total adiponectin were highly correlated with changes in HDL cholesterol but not insulin resistance and other cardiovascular risk factors. These findings show that adiponectin can be manipulated by changes in diet composition in the setting of stable weight. Although population-based studies have shown that higher levels of adiponectin levels are associated with reduced risk of diabetes and cardiovascular disease, it remains to be determined whether changes in adiponectin levels, possibly through macronutrient manipulation such as the OMNI diets, actually prevent cardiovascular disease.

Figure 1. Mean between diet differences (95% confidence interval) in log HMW and log total adiponectin.

Both HMW and total adiponectin levels were higher during the MUFA diet than the CARB or PROT diets, p<0.05. No significant differences were detected between CARB and PROT diets. Changes in log adiponectin levels were adjusted for period and carryover effects.