Greater improvement in insulin sensitivity with weight loss during calorie restriction and exercise interventions may be related to improved coupling of β-oxidation and tricarboxylic acid cycle flux.
Abstract
Objectives:
The objective of the study was to evaluate whether serum concentrations of metabolic intermediates are related to adiposity and insulin sensitivity (Si) in overweight healthy subjects and compare changes in metabolic intermediates with similar weight loss achieved by diet only or diet plus exercise.
Design:
This was a randomized controlled trial.
Participants and Intervention:
The cross-sectional study included 46 (aged 36.8 ± 1.0 yr) overweight (body mass index 27.8 ± 0.7 kg/m2) subjects enrolled in a 6-month study of calorie restriction. To determine the effect of diet only or diet plus exercise on metabolic intermediates, 35 subjects were randomized to control (energy intake at 100% of energy requirements); CR (25% calorie restriction), or CR+EX: (12.5% CR plus 12.5% increase in energy expenditure by exercise).
Main Outcome Measures:
Serum concentrations of eight fatty acids, 15 amino acids, and 45 acylcarnitines (ACs) measured by targeted mass spectrometry.
Results:
In overweight subjects, the concentrations of C2 AC and long-chain ACs were positively associated with percent fat (R2 = 0.75, P = 0.0001) and Si (R2 = 0.12, P = 0.05). The percent fat (R2 = 0.77, P < 0.0001), abdominal visceral fat (R2 = 0.64, P < 0.0001), and intrahepatic fat (R2 = 0.30, P = 0.0002) were positively associated with fatty acid concentrations. There was a significant increase in an AC factor (comprised of C2 and several medium chain ACs) in the CR group (P = 0.01).
Conclusion:
In nonobese subjects, fasted serum ACs are associated with Si and fat mass. Despite similar weight loss, serum ACs increase with CR alone but not CR+EX. A greater improvement in Si with weight loss during CR+EX interventions may be related to improved coupling of β-oxidation and tricarboxylic acid cycle flux induced by exercise.
The Comprehensive Assessment of the Long-Term Effects of Reducing Intake of Energy (CALERIE; phase 1) was a study that examined the potential health benefits of caloric restriction (CR) in sedentary, nonobese, healthy individuals. Whereas the primary aim of the study was to determine the impact of CR on metabolic adaptation and on biomarkers of longevity, the secondary aims were to evaluate changes in risk factors for type 2 diabetes mellitus and cardiovascular disease. In the first phase of CALERIE, a 6-month study of 25% CR (with or without exercise) was conducted in overweight men and women at Pennington Biomedical Research Center (Baton Rouge, LA) (1). As reported, several metabolic improvements were observed including a lowering of the metabolic rate and a metabolic adaptation (1), improvements in biomarkers of longevity such as fasting insulin and body core temperature (1), an improvement in insulin sensitivity (Si) (2), and a lowering of 10-yr estimated cardiovascular disease risk (3).
Another objective was to better understand the metabolic pathways of a CR-induced improvement in insulin resistance in overweight individuals and explore possible mechanisms by which CR can prevent diseases of aging. Metabolic profiling by targeted mass spectrometry, termed metabolomics, is proving to be an important tool in human metabolic disease research (4). Targeted metabolic profiling has identified an association between branched-chain amino acids (AA) and insulin resistance in multiple human cohorts (5–7). It has also led to development of a novel hypothesis for obesity-induced insulin resistance in which increased availability of energy, particularly dietary fat and protein, overloads muscle mitochondria and increases β-oxidation, which in turn induces mitochondrial stress and intramitochondrial accumulation of incompletely oxidized lipids that can impair insulin action (4, 6, 8–10). In support of this hypothesis, an improvement in Si with diet-induced weight loss in obese subjects was accompanied by decreases in branched-chain AAs and related metabolites (11). No studies have yet compared the metabolic correlates of weight loss and increased Si induced by CR only or in conjunction with exercise.
In our cohort of nonobese individuals, 6 months of CR with or without aerobic exercise led to similar losses of body weight, fat mass, and abdominal visceral fat (12). In addition, participants in both groups had a significant increase in Si and a decrease of the acute insulin response to glucose [AIRg (increased β-cell function)] when compared with the control condition (2). We observed that the increase in Si was associated with losses of body weight, fat mass, and abdominal visceral fat but not with changes in ectopic lipid in skeletal muscle or liver.
There were three objectives of this study: first, to evaluate whether serum concentrations of metabolic intermediates measured by targeted mass spectrometry are related to adiposity and Si at baseline; second, to investigate whether calorie restriction or calorie restriction in conjunction with exercise led to unique changes in circulating metabolic intermediate concentrations; and finally to determine whether such changes in concentrations of metabolic intermediates with calorie restriction without or with exercise are associated with changes in body composition or Si.
Subjects and Methods
Subjects
The study was approved by the Pennington Biomedical Research Center Institutional Review Board and the Data Safety Monitoring Board of CALERIE, and subjects provided written informed consent before participating. Of 48 randomized subjects, 46 healthy, overweight [25 ≥ body mass index (BMI) < 30] men (25–50 yr) and premenopausal women (25–45 yr) completed the study. Details of the screening process and study population have been extensively described (1).
Study design
Participants were randomized into one of four groups for 24 wk: control, healthy weight maintenance based on American Heart Association step 1 diet (five males and six females); CR, 25% calorie restriction from baseline energy requirements (six males and six females); CR plus increase in energy expenditure by exercise (CR+EX), 12.5% calorie restriction and 12.5% increase in energy expenditure through structured aerobic exercise (five males and seven females); and low-calorie diet (890 kcal/d until 15% weight loss) and weight loss maintenance.
Dietary interventions
During 3 wk at baseline and wk 1–12 and 22–24 of the intervention, participants were provided with all meals, which were prepared by the metabolic kitchen at the center. Individualized energy intakes were prescribed from two consecutive measures of energy expenditure by doubly labeled water at baseline (13). For wk 13–22, participants consumed a closely monitored (by food records and daily body weight) self-selected diet. The diet composition was based on the American Heart Association guidelines, 30% calories from fat, 15% from protein, and 55% from carbohydrate.
Exercise prescription and compliance
CR and control participants were required to continue their usual pattern of physical activity, whereas participants in CR+EX increased their energy expenditure by 12.5% above baseline through structured aerobic exercise (i.e. walking, running, or stationary cycling), 5 d/wk according to an individualized exercise prescription (14). The target energy cost of the exercise sessions was calculated from the weekly energy expenditure target divided by 5 d/wk. Individual exercise prescriptions to meet target energy expenditure goals were calculated by measuring the oxygen cost (VMax 29 Series; SensorMedics, Yorba Linda, CA) during three individually prescribed levels of activity (i.e. walking at 3.0, 3.5, and 4.0 MPH), generating an energy cost equation from the workload vs. oxygen cost above rest, and assigning exercise duration according to target energy expenditure and self-selected workload. The energy equivalents were determined using the calculated food quotient of 4.89 kcal/liter of oxygen consumed.
To prevent skeletal muscle soreness and injury, the exercise load was progressively increased during the initial 6 wk and energy intake was adjusted so that the energy deficit equaled 25% of daily energy requirements. Following wk 6, participants were allowed to self-select exercise intensity (as long as their heart rate was within 65–90% of maximal heart rate), and exercise duration was adjusted accordingly to maintain the target energy expenditure. During the first 6 wk, all exercise sessions were conducted at the Pennington Biomedical Research Center Health and Fitness Center under supervision. For wk 7–24, at least three of the five weekly sessions were conducted at the center under supervision. A wireless heart rate monitor (Polar S-610; Polar Beat, Port Washington, NY) was used to record exercise duration and average heart rate during both supervised and unsupervised sessions. The target energy cost was maintained at 403 ± 63 kcal per session for women and 569 ± 118 kcal per session for men throughout the entire intervention, resulting in an average exercise duration of 53 ± 11 and 45 ± 14 min per session for women and men, respectively (14).
Clinical testing
All physiological testing was conducted during a 5-d stay in the inpatient unit at baseline and during wk 12 and 24 of the intervention (1). Following a 12-h fast and morning void, percent body fat was measured using dual-energy x-ray absorptiometry (Hologics QDR 4500A, Bedford, MA) and fat mass and fat-free mass calculated. Muscle and liver lipid stores were determined by proton magnetic resonance spectroscopy using point-resolved spectroscopy as previously detailed (2). Si and AIRg were determined by the insulin-modified frequently sampled iv glucose tolerance test (2).
Blood sample collection and metabolic profiling
On the fourth inpatient day, subjects underwent a 24-h blood-sampling protocol. For this analysis, two fasted plasma aliquots collected 30 min apart were pooled and used to measure concentrations of eight fatty acids (FAs), 15 AAs, and 45 acylcarnitines (ACs), as described previously (6, 8), by gas chromatography/mass spectrometry and tandem mass spectrometry. Insulin was analyzed via immunoassay on the DPC 2000 (Diagnostic Product Corp., Los Angeles, CA).
Data analysis
Principal components analysis (PCA) was used to reduce the dimensionality of the data. PCA was performed on each separate class of metabolites; AAs, FAs, and ACs. For our first objective, we determined whether metabolic intermediates might be associated with adiposity and Si in all 46 overweight individuals during weight maintenance at baseline. Data for each metabolite were logarithmically transformed [log10(metabolite + 1)] to approximate a normal distribution. Then PCA was performed using varimax rotation and components with an eigen value greater than 1 was retained. Individual metabolites with a component load of 0.5 or greater were considered as comprising that factor. These metabolites are reported for each factor (see Tables 1 and 3).
Table 1.
PCA for fasting metabolites at baseline
Components | Component loadings | Eigen value | Cumulative variance |
---|---|---|---|
FA factor 1 | |||
Palmitic acid | 0.95 | 5.74 | 0.72 |
Linoleic acid | 0.94 | ||
Oleic acid | 0.94 | ||
Palmitoleic acid | 0.90 | ||
Myristic acid | 0.88 | ||
α-Linolenic acid | 0.82 | ||
Stearic acid | 0.81 | ||
AA factor 1 | |||
Leucine/isoleucine | 0.92 | 6.67 | 0.45 |
Ornithine | 0.86 | ||
Phenylalanine | 0.85 | ||
Valine | 0.81 | ||
Methionine | 0.81 | ||
Aspartate/asparagine | 0.76 | ||
Histidine | 0.71 | ||
Glutamine/glutamate | 0.71 | ||
Tyrosine | 0.62 | ||
Alanine | 0.66 | ||
Proline | 0.57 | ||
AC factor 1 | |||
C8:1 | 0.65 | 14.04 | 0.31 |
C8 | 0.59 | ||
C6-DC | 0.50 | ||
C10:3 | 0.57 | ||
C10:2 | 0.51 | ||
C10:1 | 0.71 | ||
C10 | 0.73 | ||
C10-OH/C8-DC | 0.61 | ||
C12:1 | 0.76 | ||
C12 | 0.81 | ||
C12-OH/C10-DC | 0.64 | ||
C14:2 | 0.75 | ||
C14:1 | 0.86 | ||
C14:1-OH | 0.52 | ||
C20-OH/C18-DC | 0.53 | ||
C16:2 | 0.57 | ||
AC factor 2 | |||
C2 | 0.60 | 3.67 | 0.39 |
C16 | 0.69 | ||
C18:2 | 0.54 | ||
C18:1 | 0.68 | ||
C18 | 0.54 | ||
C18:1-OH | 0.56 | ||
C20 | 0.73 | ||
C16:2 | 0.54 | ||
C16:1 | 0.73 | ||
C16:1-OH/C14:1-DC | 0.56 | ||
C20:4 | 0.71 | ||
AC factor 3 | |||
C3 | 0.72 | 3.08 | 0.46 |
C4/Ci4 | 0.70 | ||
C5s | 0.75 | ||
C5-OH/C3-DC | 0.55 | ||
Ci4-DC/C4-DC | 0.62 | ||
C8:1 | 0.50 | ||
C5-DC | 0.51 | ||
C10:3 | 0.61 | ||
C10:2 | 0.63 |
PCA was performed separately for each metabolite class: FAs, ACs, and AAs. Key metabolites within each component (i.e. metabolites with component load ≥0.5) are presented.
Table 3.
PCA for changes in metabolic intermediates with CR
Components | Component loadings | Eigen value | Cumulative variance |
---|---|---|---|
FA change factor 1 | |||
Linoleic acid | 0.98 | 6.54 | 0.82 |
Oleic acid | 0.97 | ||
Myristic acid | 0.96 | ||
Palmitoleic acid | 0.94 | ||
Palmitic acid | 0.93 | ||
Stearic acid | 0.90 | ||
α-Linolenic acid | 0.89 | ||
Arachidonic acid | 0.60 | ||
AA change factor 1 | |||
Alanine | 0.84 | 6.1 | 0.40 |
Proline | 0.83 | ||
Methionine | 0.82 | ||
Tyrosine | 0.77 | ||
Histidine | 0.75 | ||
Arginine | 0.73 | ||
Leucine/isoleucine | 0.67 | ||
Valine | 0.65 | ||
Glycine | 0.64 | ||
Phenylalanine | 0.56 | ||
Citrulline | 0.52 | ||
AA change factor 2 | |||
Glutamine/glutamate | 0.87 | 2.0 | 0.54 |
Aspartate/asparagine | 0.76 | ||
Ornithine | 0.60 | ||
Phenylalanine | 0.56 | ||
AC change factor 1 | |||
C2 | 0.70 | 10.7 | 0.24 |
Ci4-DC/C4-DC | 0.51 | ||
C8 | 0.74 | ||
C6-DC | 0.59 | ||
C10:2 | 0.63 | ||
C10:1 | 0.71 | ||
C10 | 0.69 | ||
C10-OH/C8-DC | 0.57 | ||
C12:1 | 0.84 | ||
C12 | 0.86 | ||
C12-OH/C10-DC | 0.45 | ||
C14:2 | 0.73 | ||
C14:1 | 0.84 | ||
C14 | 0.47 | ||
C14:1-OH | 0.48 | ||
C14-OH/C12-DC | 0.45 | ||
C16 | 0.68 | ||
C18:2 | 0.63 | ||
C18:1 | 0.66 | ||
C16:2 | 0.57 | ||
C16:1 | 0.64 |
Using slopes of change for each metabolite, PCA was performed separately for each metabolite class: FAs, ACs, and AAs. Key metabolites within each component (i.e. metabolites with component load ≥0.5) are presented.
To determine the effects of calorie restriction on concentrations of metabolic intermediates, we elected to study only CR, CR+EX, and control (n = 35). The rationale for this approach was that these two groups had the same energy deficit (25% reduction from baseline), which resulted in similar weight loss and time course of weight loss. For each participant we computed a trajectory (or slope) for each metabolite over time (M0, M3, and M6). Individual trajectories were reduced within each metabolite class by PCA with varimax rotation, such that one PCA was performed for changes in FAs, changes in ACs, and changes in AAs. The numbers of components retained for each model was selected to obtain a solution that explained the greatest percent variance. ANOVA was then used to test for between-group differences in factor scores derived from the changes in metabolic intermediates over time.
To test whether factor scores derived from the changes with CR (trajectories) were associated with changes in adiposity and Si, the individual trajectories for adiposity, and Si variables over times (M0, M3, and M6) were computed and then used as dependent variables in the regression analyses. For each separate class of metabolites (FA, AC, or AA), we performed multiple linear regressions initially with full models, which included the factor scores retained from the PCA analysis, sex, age, and their interaction terms. If the interaction terms were not significant, these terms were removed and the reduced models were refitted. P < 0.05 was considered statistically significant. Data are reported as means ± sem, unless otherwise stated. SAS version 9.12 (SAS Institute, Cary, NC) was used for analysis.
Results
At baseline, 26 women (57%) and 20 men (43%) comprised the 46 subjects. Thirty subjects (65%) were Caucasian, 15 subjects (32%) were African-American, and one subject (3%) was Asian. The subjects were young (aged 38.5 ± 6.4 yr), overweight (BMI 27.7 ± 1.7, kg/m2) but in good health including a good Si 3.2 ± 1.4, 10−4 mU/liter−1 · min−1.
PCA at baseline in 46 nonobese subjects
Previous work had indicated that metabolic intermediates typically clustered into factors comprised of intermediates from the same class, such that FAs clustered into a single factor, and ACs clustered into factors mostly with other ACs (5). Given this, we chose to perform PCA separately for each metabolite class (Table 1). Of note, performing PCA separately or with all metabolites together produced similar results (data not shown). For FAs, we observed that a single factor comprised of seven FAs (of eight) explained 72% of the variance in FAs (Table 1). For AAs, a single-factor solution comprised 11 AAs (of 15) had an eigen value of 6.7 explained 45% of the variance in AA concentrations. When PCA was performed for ACs, one, two, three, four, and five factor models were evaluated. A three-factor model was selected for ACs, given that factors four and five models were dominated by a single AC. The three-factor solution had an eigen value of 3.1 (for 45 ACs) and explained 46% of the variance in fasting AC concentrations at baseline (Table 1).
The association of metabolic intermediates with adiposity and Si in nonobese men and women
After reducing the data into the three separate classes of metabolic intermediates by PCA, we sought to determine whether concentrations of fasting metabolic intermediates were related to adiposity (weight, percent fat, visceral fat, intrahepatic lipid, intramyocellular lipid, and abdominal fat cell size) or factors related to insulin action (Si, AIRg, and fasting insulin) at baseline. Results of the significant regression models are shown in Table 2.
Table 2.
Linear regression models for measures of body composition and insulin action controlling for sex and age
Model | Parameter estimate | sem | F | P |
---|---|---|---|---|
FA baseline factor | ||||
Percent fat: adjusted R2 = 0.77, P < 0.0001 | ||||
FA factor 1 | 1.488 | 0.582 | 6.541 | 0.01 |
Sex (male = 0) | 11.32 | 1.158 | 95.51 | 0 |
Abdominal visceral fat: adjusted R2 = 0.64, P < 0.0001 | ||||
FA factor 1 | 0.354 | 0.151 | 5.461 | 0.02 |
Sex (male = 0) | −2.66 | 0.301 | 78.04 | 0 |
Intrahepatic lipid: adjusted R2 = 0.30, P = 0.0002 | ||||
FA factor 1 | 0.14 | 0.053 | 6.932 | 0.01 |
Sex (male = 0) | −0.48 | 0.106 | 20.78 | 0 |
Fasting FFA: adjusted R2 = 0.72, P < 0.0001 | ||||
FA factor 1 | 0.159 | 0.018 | 82.23 | 0.000 |
AC baseline factors | ||||
Percent fat: adjusted R2 = 0.75, P < 0.0001 | ||||
AC factor 1 | 1.122 | 0.554 | 4.098 | 0.049 |
Sex (male = 0) | 12.97 | 1.454 | 79.57 | 0 |
Si: adjusted R2 = 0.12, P = 0.05 | ||||
AC factor 2 | 0.568 | 0.214 | 7.033 | 0.01 |
Fasting FFA: adjusted R2 = 0.20, P = 0.02 | ||||
AC factor 2 | 0.068 | 0.027 | 6.11 | 0.02 |
Sex (male = 0) | 0.148 | 0.065 | 5.213 | 0.03 |
Only significant predictors are shown.
Fatty acids
Fasting FA concentrations measured by targeted mass spectrometry were positively associated with serum FAs measured by ELISA (Table 2). The single FA factor (FA factor 1) was associated with measures of adiposity. With adjustments for sex and age, FA factor 1 (Table 2) was strongly associated with percent body fat (r2 = 0.77, P < 0.0001) and also visceral fat (r2 = 0.64, P < 0.0001). Interestingly, the FA factor was also associated with ectopic liver fat, explaining 30% of the variance.
Acylcarnitines
After controlling for sex and age (Table 2), AC factor 1, comprised predominantly of medium-chain ACs, was positively associated with whole percent body fat (r2 = 0.75, P < 0.0001). AC factor 2 comprised of C2 and long-chain ACs was also positively associated with Si (r2 = 0.12, P < 0.05) and with serum FA concentrations, (r2 = 0.20, P = 0.02). AC factor 3, which contains end products of AA metabolism, was not associated with measures of adiposity or Si in this nonobese group of men and women.
Amino acids
The single AA factor (AA factor 1) was not associated with any measure of adiposity. Additionally, in contrast to previous reports in three separate human cohorts, including Asian subjects with average BMI of 24 kg/m2 (5–7), the AA factor (containing leucine/isoleucine, valine, and phenylalanine) was not associated with Si at baseline. Furthermore, there was also no correlation between leucine/isoleucine and Si or valine and Si when examining the log-transformed raw data.
The association of calorie restriction or calorie restriction in conjunction with exercise with unique changes in circulating metabolic intermediate concentrations
As reported previously (12), the 25% energy deficit by CR or CR+EX resulted in equivalent losses of body weight (CR: −10 ± 1%; CR+EX: −10 ± 1%), fat mass (CR: −24 ± 3%; CR+EX: −25 ± 3%), and abdominal visceral (CR: −28 ± 4%; CR+EX: −27 ± 3%) and sc fat stores (CR: −26 ± 4%; CR+EX: −28 ± 3%). Si was improved in both CR and CR+EX but reached significance only in the CR+EX group (CR: 40 ± 20%; CR+EX: 66 ± 22%). Fasting insulin (CR: −28 ± 6%; CR+EX: −20 ± 12%) and insulin secretion (AIRg: CR: −29 ± 7%; CR+EX: −30 ± 8%), however, were decreased significantly in both treatment groups (2).
To determine the effect of a 25% energy deficit by CR or CR with exercise on serum metabolite species, we measured serum metabolic concentrations after 3 and 6 months. To simplify these time-series data, we calculated the changes in serum metabolite species from a linear regression of serum metabolite species over time (BL, M3, and M6). We rationalized these data based on the observation that the weight loss curves over this time period were roughly linear. The slopes for each metabolite were reduced further within each separate class of metabolites by PCA. Each significant factor solution and components are reported in Table 3.
For free fatty acid (FFA) trajectories (changes in FFAs over time), PCA identified that changes in FFA species could be condensed to a single factor (FA change factor 1), comprised of changes of eight FAs. This single factor had an eigen value of 6.5 and explained 82% of the variance in the serum FA species changes over time (Table 3).
When PCA was performed on the changes in AAs, one-, two-, three-, and four-factor models were evaluated. A two-factor model was selected because factors three and four contained only a single AA. Together the two-factor solution explained 54% of the variance in changes in AA concentrations over time (Table 3). The first AA factor (AA change factor 1) was comprised of alanine, proline, methionine, tyrosine, histidine, arginine, leucine/isoleucine, valine, glycine, phenylalanine, and citrulline and explained 40% of the variance in the time-series changes in AAs. The second AA factor (AA change factor 2) explained an additional 15% of the total variance and is comprised of glutamine/glutamate, aspartate/asparagine, phenylalanine, and ornithine. Phenylalanine had similar loadings for each factor.
For ACs, the PCA performed on the AC trajectories determined that a single AC factor explained 24% of the variance in the AC species changes (Table 3). This single AC factor (AC change factor 1) had an eigen value of 10.7 (for 45 ACs) and was comprised of AC (C2) and medium chain ACs (Table 3).
Using ANOVA, we tested between-group differences in each change factor (Fig. 1). First with three treatment groups in the model (CR, CR+EX, and control), we observed no significant treatment effect for FA change factor 1 (P = 0.80) or AA change factors 1 (P = 0.37) and 2 (P = 0.34). There was a significant treatment effect for AC change factor 1 (P = 0.05). The post hoc comparison indicated a significant difference between CR and control (P = 0.04). LS means indicated a significant increase in AC change factor 1 for CR but not for CR+EX, indicating an increase in C2 and medium-chain AC concentrations was unique to CR (Fig. 1).
Fig. 1.
Effect of CR alone (CR) and with exercise (CR+EX) on changes in serum concentrations of FAs, AAs, and ACs. Changes in individual metabolite concentrations within each class were reduced to common change factors by PCA. Four significant factor models were identified (FA change factor 1, AA change factor 1, AA change factor 2, and AC change factor 1), and treatment effects were tested by ANOVA. *, Significant between-group difference (P < 0.05).
To confirm the results of a differential effect between CR and CR+EX on serum ACs, we computed the changes from baseline to M3 and baseline to M6 for each AC species and performed a simple ANOVA to test for within- and between-treatment group effects. In support of the PCA, there was a significant treatment effect for several of the AC species comprised within AC change factor 1. C2 AC was significantly increased from baseline at M3 (P < 0.000) and M6 (P < 0.000) with CR. Pair-wise comparisons (with Tukey-Kramer adjustment) indicated a significant difference between CR and CR+EX (P = 0.003) and also CR and control (P < 0.000). Similarly, C14:1 (P = 0.001), C16 (P = 0.032), and C18:1 (P = 0.03) were significantly increased from baseline at M3 and M6 with CR and not with CR+EX. These findings confirm that C2 as well as medium-chain and long-chain (C16 and C18:1) ACs are increased uniquely with 6 months of CR and not with CR+EX.
Association of changes in concentrations of metabolic intermediates with calorie restriction without or with exercise with changes in body composition or Si
Although mean changes indicated an increase in AC concentrations over time with CR and a decrease with control, there was significant variation across the sample. To determine which change was associated with improvements in adiposity and insulin action, the changes in ACs (change factors described above) were related to the changes in Si, fasting insulin, and adiposity measures. We used multiple linear regressions (controlling for sex and age) to determine whether changes in metabolic intermediates were associated with changes in adiposity and measures of Si during the 6-month study.
Acylcarnitines
After controlling for sex and age, we found that the single AC change factor was inversely associated with the change in fasting insulin, accounting for 14% of the variance in this parameter (Table 4). Interestingly the results indicate that greater increases in fasting AC concentrations were associated with greater reductions in fasting insulin concentrations (Fig. 2). Consistent with the baseline analysis, the single AC change factor was also positively related to changes in serum FA concentrations, measured by ELISA, and accounted for 51% of the variance (P < 0.001).
Table 4.
Linear regression models for changes in metabolic intermediates and changes in body composition and insulin action with calorie restriction
Model | Parameter estimate | sem | F | P |
---|---|---|---|---|
AC change factor | ||||
Fasting insulin: adjusted R2 = 0.14, P = 0.03 | ||||
AC factor 1 | −0.257 | 0.102 | 6.295 | 0.02 |
Fasting FFA model: adjusted R2 = 0.51, P < 0.0001 | ||||
AC factor 1 | 0.026 | 0.006 | 18.18 | 0.00 |
FA change factor | ||||
Fasting FFA model: adjusted R2 = 0.51, P < 0.0001 | ||||
FA factor 1 | 0.025 | 0.006 | 17.514 | 0.00 |
Sex (male = 0) | 0.035 | 0.011 | 9.689 | 0 |
Models controlled for sex and age. Only significant predictors are shown.
Fig. 2.
Relationship between changes in fasting insulin and serum ACs (AC change factor 1) during 6 months of calorie restriction. Changes in ACs were reduced to a common change factor using PCA. Factor change scores were related to trajectories for fasting insulin adjusting for sex and age using linear regression.
Fatty acids
Importantly, as observed at baseline, the changes in fasting FA species measured by targeted mass spectrometry were positively associated with changes in serum FAs measured by ELISA (Table 4). This single FA change factor was not associated with changes in any measure of adiposity or Si.
Amino acids
Neither the AA change factor 1 or AA change factor 2 was associated with changes in any measure of adiposity or Si.
Discussion
In this analysis we: 1) evaluated whether measures of adiposity and Si in healthy but overweight men and women were correlated with fasting metabolic intermediates; 2) measured changes in fasting metabolic intermediates with a 25% energy deficit by CR only or CR in conjunction with exercise, and 3) determined whether changes in adiposity and Si with 6 months of CR were associated with changes in fasting metabolic intermediates.
After reducing the dimensionality of the data by PCA within each separate class of metabolites, we found that in 46 overweight men and women, whole-body adiposity (percent fat by dual energy x-ray absorptiometry), abdominal visceral fat and ectopic fat in the liver were positively associated with FA concentrations (FFA factor 1). Additionally, concentrations of C2 AC and long-chain ACs (AC factor 2) were positively associated with Si and whole-body adiposity. In response to a 25% energy deficit by diet only or diet with exercise, we found no changes in FA (FA change factor 1) or AA concentrations (AA change factor 1 and factor 2) despite significant reductions in body weight, adiposity, and improvements in Si. We did, however, observe a significant treatment effect for AC change factor 1. Serum AC concentrations were increased by 6 months of CR but were not changed in CR+EX or control (Fig. 1). The post hoc tests indicated that AC change factor 1 was significantly different between CR and control. Furthermore, ANOVA performed on each AC species found that 6 months of CR resulted in a significant increase in C2, C14:1, C16, and C18:1 ACs. There was no significant change in serum AC concentration with 6 months of CR+EX or a weight-maintaining controlled diet (control). AC factor 1 therefore appears to be a factor that changes uniquely with CR and not with CR+EX and is not related to other markers of metabolic adaptation to CR other than fasting insulin, a biomarker of aging.
Products of incomplete FA oxidation in the mitochondria such as ACs are elevated in obesity (6) and type 2 diabetes (15) and are positively associated with insulin resistance (5, 6), even after adjusting for BMI (7). What is not currently known is whether weight loss or weight loss-associated improvements in Si is coupled with changes in lipid-derived CAs. The current study therefore provides new findings with regard to the effects of CR on metabolic intermediates. We compared the effects of a 25% energy deficit achieved by dietary restriction alone or dietary restriction in conjunction with exercise on body composition and various physiological outcomes including Si (2), energy expenditure (1, 13), and mitochondrial function (16). Changes in body weight, whole-body fat mass, and abdominal visceral fat were comparable between the CR and CR+EX groups. One of the few differences between the two groups was change in Si. As measured by frequently sampled iv glucose tolerance test, Si was significantly increased with CR+EX by 66% after 6 months and by only 40% with CR (not significant). Given the association between AC concentrations and insulin resistance in larger cross-sectional studies, a reasonable hypothesis is that CR+EX would produce more complete FA oxidation, leading to less production of AC in serum. Indeed, we observed reduced accumulation of ACs with CR+EX, suggesting that the addition of exercise to CR improves the coupling between β-oxidation and the tricarboxylic acid cycle.
Our results agree with previous studies demonstrating that in the presence of work or physical activity, the accumulation of lipid-derived ACs is attenuated in rodents fed a high-fat diet (17). The ability of trained muscle to fully oxidize lipid substrate is thought to be related to the increase in peroxisome proliferator-activated receptor-γ coactivator 1α (PGC1α) mRNA expression and probably mitochondrial density. Both exercise and myocytes treated with recombinant adenovirus encoding PCG1α are shown to trigger a downstream activation of genes involved in tricarboxylic acid cycle and electron transport chain (17). The end result is a tighter coupling of between β-oxidation and tricarboxylic acid cycle (18).
In our subjects, PGC1α was increased in CR and CR+EX in parallel with an increase in mitochondrial DNA content (16). Given that there was a significant change in serum ACs (AC change factor) in the CR group and that the CR group had an increase in C2 concentrations (Supplemental Table 1, published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org), we speculated that CR leads to a more complete β-oxidation in the fasted state over time without a proportionate increase in tricarboxylic acid cycle flux. Interestingly, the enzyme activity of the key mitochondrial enzymes involved in β-oxidation (β-hydroxyacyl-coenzyme A), the tricarboxylic acid cycle (citrate synthase), or the electron transport chain (cytochrome C oxidase II) was not changed in either group (16). Therefore, an increase in the mitochondrial biogenesis or changes in allosteric regulation of enzymes may be necessary in diet and exercise interventions to enhance complete oxidation of lipids and improve insulin action. However, it should be noted that weight loss-mediated improvements in Si in obese individuals can occur independently of changes in skeletal muscle fat oxidation and mitochondrial size and function (19, 20).
Previously in a group of overweight to obese dyslipidemic men and women, we showed that increased concentrations of large neutral AAs were independently associated with insulin resistance (Si and disposition index) (5). Furthermore, previous studies identified that a distinct metabolic signature defined by branched chain AAs, other AA species, and C3 and C5 ACs (by-products of isoleucine, leucine, and valine catabolism) was significantly elevated in obese individuals compared with lean individuals (6). Independent of sex, age, and race, this branched chain AA-related metabolic factor was positively associated with insulin resistance as measured by homeostasis model assessment (HOMA) (6, 7). In our study, the single AA factor was not associated with any measure of insulin action. Furthermore, AC factor 3, in which fasting C3 and C5 AC concentrations were heavily loaded, was also not associated with insulin action. In agreement with our factor analysis, we did not find any relationships between fasting leucine/isoleucine or valine concentrations (log transformed data), key AAs believed to be important in skeletal muscle insulin resistance, and insulin resistance (5, 6), Si, or fasting insulin.
A major difference between the current study and foregoing studies by our group linking branched chain AAs and related metabolites with insulin resistance is that subjects in the current study were not insulin resistant at baseline. Thus, in our prior comparison of obese, insulin-resistant subjects with lean, insulin-sensitive controls, the former group had a HOMA score of 5.73, and the latter group a HOMA score of 2.51. The mean HOMA score for all subjects of the current study at baseline was 2.35. Moreover, among a subset of lean and obese subjects in which Si was measured in our prior study, Si was 2.12 in obese subjects and 4.44 in lean controls; in the current study, the Si at baseline for all subjects was 2.8 (2). In our prior study, the relationship between AAs and insulin resistance did not hold in analysis of the lean subjects alone, probably due to the small variability in Si in this group (6). We believe that the small variance in Si among the current subjects also explains the lack of association. We also note that PCA analysis performed in our prior study was applied to the entire metabolite data set, resulting a PCA factor comprised of a mixture of AAs and ACs, whereas in the current study, we chose to perform PCA separately on AAs and ACs. Finally, there is also the issue of statistical power with only 46 subjects in our cross-sectional analysis. With this limited number of subjects, we cannot be confident that the current finding of no association between AA concentrations and Si would hold true in a larger sample of overweight subjects. Nevertheless, with these caveats in mind, our study does show that when analyzing small but significant changes in Si in healthy subjects in response to CR or CR+EX, such changes are associated with changes in ACs but not AAs.
It is also important to recognize that baseline testing in this study was conducted after all subjects consumed a calorie-controlled isoenergetic weight-maintaining diet (fluctuations in daily weight <250 g) for 14 d. The foods that were prepared by the metabolic kitchen provided only 30% calories from fat, 15% calories from protein, and 55% calories from carbohydrate. Thus, the current study may have reduced intraindividual variability in fat and protein intake by imposing a standardized diet in our healthy overweight (not obese) participants. If the notion that an oversupply of dietary fat and protein is responsible at least in part for incomplete mitochondrial fat oxidation (6, 9, 10), then it is likely that the mitochondria were not overloaded in the fasted state in our subjects. According to our metabolic kitchen records, the percent calories from protein and fat ranged from 15–20 and 28–37%, respectively. It was interesting when comparing these data with those previously published in lean and obese subjects (6) that the median concentration of leucine/isoleucine, valine, and C3 and C5 in our overweight subjects (Supplemental Table 1) fell between those of lean and obese subjects.
In summary, we found that in healthy overweight individuals consuming a weight-maintenance, macronutrient-controlled diet, serum ACs and FA species in the fasted state are related to measures of adiposity and insulin action. In response to a 25% energy deficit by diet only or diet and structured exercise, CR alone produced an increase in serum AC concentrations, whereas with CR and exercise, the increase in serum AC concentrations is attenuated. A greater improvement in Si with diet and exercise interventions may therefore be related to an exercise-induced coupling of β-oxidation and tricarboxylic acid cycle flux.
Supplementary Material
Acknowledgments
This work was supported by Grant U01 AG20478 (to E.R.) and in part by Grant 1P30 DK072476 (to E.R.) and by Grant P30AG028716 (principal investigator, H. J. Cohen). L.M.R. is supported by Grant K99 HD060762. K.M.H. was supported by the American College of Rheumatology Research and Education Foundation/Association of Specialty Professors Junior Career Development Award in Geriatric Medicine funded by Atlantic Philanthropies, the American College of Rheumatology Research and Education Foundation, the John A. Hartford Foundation, and the Association of Specialty Professors as well as National Institutes of Health/National Institute of Arthritis and Musculoskeletal and Skin Diseases Grant K23AR054904.
Disclosure Summary: The authors have nothing to disclose related to this study.
Footnotes
- AA
- Amino acid
- AC
- acylcarnitine
- AIRg
- acute insulin response to glucose
- BMI
- body mass index
- CALERIE
- Comprehensive Assessment of the Long-Term Effects of Reducing Intake of Energy
- CR
- caloric restriction
- CR+EX
- CR plus increase in energy expenditure by exercise
- FA
- fatty acid
- FFA
- free fatty acid
- HOMA
- homeostasis model assessment
- PCA
- principal components analysis
- PGC1α
- peroxisome proliferator-activated receptor-γ coactivator 1α
- Si
- insulin sensitivity.
References
- 1. Heilbronn LK, de Jonge L, Frisard MI, DeLany JP, Larson-Meyer DE, Rood J, Nguyen T, Martin CK, Volaufova J, Most MM, Greenway FL, Smith SR, Deutsch WA, Williamson DA, Ravussin E. 2006. Effect of 6-month calorie restriction on biomarkers of longevity, metabolic adaptation, and oxidative stress in overweight individuals: a randomized controlled trial. JAMA 295:1539–1548 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Larson-Meyer DE, Heilbronn LK, Redman LM, Newcomer BR, Frisard MI, Anton S, Smith SR, Alfonso A, Ravussin E. 2006. Effect of calorie restriction with or without exercise on insulin sensitivity, β-cell function, fat cell size, and ectopic lipid in overweight subjects. Diabetes Care 29:1337–1344 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Lefevre M, Redman LM, Heilbronn LK, Smith JV, Martin CK, Rood JC, Greenway FL, Williamson DA, Smith SR, Ravussin E. 2009. Caloric restriction alone and with exercise improves CVD risk in healthy non-obese individuals. Atherosclerosis 203:206–213 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Bain JR, Stevens RD, Wenner BR, Ilkayeva O, Muoio DM, Newgard CB. 2009. Metabolomics applied to diabetes research: moving from information to knowledge. Diabetes 58:2429–2443 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Huffman KM, Shah SH, Stevens RD, Bain JR, Muehlbauer M, Slentz CA, Tanner CJ, Kuchibhatla M, Houmard JA, Newgard CB, Kraus WE. 2009. Relationships between circulating metabolic intermediates and insulin action in overweight to obese, inactive men and women. Diabetes Care 32:1678–1683 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Newgard CB, An J, Bain JR, Muehlbauer MJ, Stevens RD, Lien LF, Haqq AM, Shah SH, Arlotto M, Slentz CA, Rochon J, Gallup D, Ilkayeva O, Wenner BR, Yancy WS, Jr, Eisenson H, Musante G, Surwit RS, Millington DS, Butler MD, Svetkey LP. 2009. A branched-chain amino acid-related metabolic signature that differentiates obese and lean humans and contributes to insulin resistance. Cell Metab 9:311–326 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Tai ES, Tan ML, Stevens RD, Low YL, Muehlbauer MJ, Goh DL, Ilkayeva OR, Wenner BR, Bain JR, Lee JJ, Lim SC, Khoo CM, Shah SH, Newgard CB. 2010. Insulin resistance is associated with a metabolic profile of altered protein metabolism in Chinese and Asian-Indian men. Diabetologia 53:757–767 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. An J, Muoio DM, Shiota M, Fujimoto Y, Cline GW, Shulman GI, Koves TR, Stevens R, Millington D, Newgard CB. 2004. Hepatic expression of malonyl-CoA decarboxylase reverses muscle, liver and whole-animal insulin resistance. Nat Med 10:268–274 [DOI] [PubMed] [Google Scholar]
- 9. Koves TR, Ussher JR, Noland RC, Slentz D, Mosedale M, Ilkayeva O, Bain J, Stevens R, Dyck JR, Newgard CB, Lopaschuk GD, Muoio DM. 2008. Mitochondrial overload and incomplete fatty acid oxidation contribute to skeletal muscle insulin resistance. Cell Metab 7:45–56 [DOI] [PubMed] [Google Scholar]
- 10. Noland RC, Koves TR, Seiler SE, Lum H, Lust RM, Ilkayeva O, Stevens RD, Hegardt FG, Muoio DM. 2009. Carnitine insufficiency caused by aging and overnutrition compromises mitochondrial performance and metabolic control. J Biol Chem 284:22840–22852 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Lien LF, Haqq AM, Arlotto M, Slentz CA, Muehlbauer MJ, McMahon RL, Rochon J, Gallup D, Bain JR, Ilkayeva O, Wenner BR, Stevens RD, Millington DS, Muoio DM, Butler MD, Newgard CB, Svetkey LP. 2009. The STEDMAN project: biophysical, biochemical and metabolic effects of a behavioral weight loss intervention during weight loss, maintenance, and regain. OMICS 13:21–35 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Redman LM, Heilbronn LK, Martin CK, Alfonso A, Smith SR, Ravussin E. 2007. Effect of calorie restriction with or without exercise on body composition and fat distribution. J Clin Endocrinol Metab 92:865–872 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Redman LM, Heilbronn LK, Martin CK, de Jonge L, Williamson DA, Delany JP, Ravussin E. 2009. Metabolic and behavioral compensations in response to caloric restriction: implications for the maintenance of weight loss. PLoS One 4:e4377. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Larson-Meyer DE, Redman L, Heilbronn LK, Martin CK, Ravussin E. 2010. Caloric restriction with or without exercise: the fitness versus fatness debate. Med Sci Sports Exerc 42:152–159 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Adams SH, Hoppel CL, Lok KH, Zhao L, Wong SW, Minkler PE, Hwang DH, Newman JW, Garvey WT. 2009. Plasma acylcarnitine profiles suggest incomplete long-chain fatty acid β-oxidation and altered tricarboxylic acid cycle activity in type 2 diabetic African-American women. J Nutr 139:1073–1081 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Civitarese AE, Carling S, Heilbronn LK, Hulver MH, Ukropcova B, Deutsch WA, Smith SR, Ravussin E. 2007. Calorie restriction increases muscle mitochondrial biogenesis in healthy humans. PLoS Med 4:e76. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Koves TR, Li P, An J, Akimoto T, Slentz D, Ilkayeva O, Dohm GL, Yan Z, Newgard CB, Muoio DM. 2005. Peroxisome proliferator-activated receptor-γ co-activator 1α-mediated metabolic remodeling of skeletal myocytes mimics exercise training and reverses lipid-induced mitochondrial inefficiency. J Biol Chem 280:33588–33598 [DOI] [PubMed] [Google Scholar]
- 18. Muoio DM, Koves TR. 2007. Skeletal muscle adaptation to fatty acid depends on coordinated actions of the PPARs and PGC1α: implications for metabolic disease. Appl Physiol Nutr Metab 32:874–883 [DOI] [PubMed] [Google Scholar]
- 19. Berggren JR, Boyle KE, Chapman WH, Houmard JA. 2008. Skeletal muscle lipid oxidation and obesity: influence of weight loss and exercise. Am J Physiol Endocrinol Metab 294:E726–E732 [DOI] [PubMed] [Google Scholar]
- 20. Toledo FG, Menshikova EV, Azuma K, Radiková Z, Kelley CA, Ritov VB, Kelley DE. 2008. Mitochondrial capacity in skeletal muscle is not stimulated by weight loss despite increases in insulin action and decreases in intramyocellular lipid content. Diabetes 57:987–994 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.