Effect of Calorie Restriction With or Without Exercise on Insulin Sensitivity, β-Cell Function, Fat Cell Size, and Ectopic Lipid in Overweight Subjects

Details of this study are reported elsewhere (18). Briefly, healthy male (25–50 years) and female (25–45 years) overweight participants (25 ≤ BMI ≤ 30 kg/m²) were recruited for the Comprehensive Assessment of the Long-term Effects of Reducing Intake of Energy (CALERIE) trial. Participants were excluded if they smoked; exercised more than twice per week; were pregnant, lactating, or postmenopausal; had a personal history of obesity (BMI never >32 kg/m²), cardiovascular disease, or diabetes; or regularly used medications (except birth control). The study was approved by the Pennington Biomedical Research Center Institutional Review Board and the CALERIE Data Safety Monitoring Board, and all subjects provided written informed consent.

To carefully determine individual energy requirements, total daily energy expenditure was measured by two 14-day measures of doubly labeled water, once while participants followed their usual diet at home and once while they were being provided with a weight maintenance diet by the metabolic kitchen. Patients were then admitted to the ward for 5 days of metabolic testing.

After baseline testing, 48 participants were randomly assigned to one of four groups for 6 months: control (healthy diet for weight maintenance), 25% calorie restriction of baseline energy requirements (CR), 12.5% calorie reduction + 12.5% increase in total energy expenditure by structured exercise (CREX), and low-calorie diet until a 15% reduction in body weight, followed by maintenance of the new lower body weight (LCD). Participants were provided with all food for the first 12 weeks and for weeks 22–24. Diets were based on American Heart Association recommendations (≤30% fat). For weeks 13–22, participants self-selected their diet based on their calorie targets. LCD participants were placed on 890 kcal/day (HealthOne, Health and Nutrition Technology, Carmel, CA). Once target weight loss (−15%) was achieved, LCD participants were refed to an energy level that maintained this new body weight. CREX participants increased their energy expenditure by 12.5% above resting by undergoing structured exercise (i.e., walking, running, or stationary cycling) 5 days per week. Participants were required to conduct three sessions per week under supervision. For unsupervised sessions, participants wore a portable heart rate monitor (Polar S-610, Polar Beat, Port Washington, NY) with heart rate and exercise duration recorded. For support, all participants attended weekly group meetings that were led by clinical psychology professionals.

All metabolic tests were performed during inpatient stays at baseline and month 6 following a 12-h overnight fast and at least 48 h after the last bout of exercise. Body fat was measured by dual-energy X-ray absorptiometry (QDA 4500A; Hologics, Bedford, MA) and multislice computed tomography scanning of the abdominal region (GE High Speed Plus; General Electric, Fairfield, CT) was performed to quantify abdominal fat compartments (19). Muscle and liver lipid stores were determined by proton magnetic resonance spectroscopy using point-resolved spectroscopy (20). Subcutaneous abdominal needle biopsies were performed, and FCS was determined by the Multisizer-3 counter (Beckman Coulter, Fullerton, CA) as previously described (21). Insulin sensitivity was determined by the insulin-modified frequently sampled intravenous glucose tolerance test (22,23). Baseline blood samples were drawn before 300 mg/kg body weight of glucose (50% dextrose; Hospira, Lake Forest, IL) was injected and 32 blood samples were collected over 180 min. At 20 min, a bolus injection of insulin (0.03 units/kg Humulin; Eli Lilly, Indianapolis, IN) was given. The S_i and acute insulin response to glucose (AIR_g) were calculated by the minimal model (23). Because of illness or problems with intravenous lines, four tests could not be analyzed at month 6. Glucose was analyzed using a Synchron CX7 (Beckman-Coulter, Brea, CA) and insulin was analyzed via immunoassay on the DPC 2000 (Diagnostic Product Corporation, Los Angeles, CA).

Data are expressed as means ± SE and the level of significance for all statistical tests was set at P < 0.05. SAS version 9.1 was used for analysis, and all analyses were performed by biostatisticians in the Pennington Biomedical Research Center Biostatistics Core. Pearson or Spearman rank order correlations were used where appropriate, and general linear regression was used to identify any interactions of the changes with sex. To assess the effect of the intervention among the four groups, the change from baseline to month 6 was computed, and an ANCOVA was performed with baseline values included in the model as covariates and adjusted with respect to Tukey-Kramer. Two subjects withdrew during the study; one was a control subject who withdrew for personal reasons and the other subject, who was following the LCD diet, was lost to follow-up. Data are therefore presented on 46 subjects.

Characteristics of the subjects at baseline are reported in Table 1. Subjects were generally in good health with fasting glucose, insulin, and blood pressure within recommended ranges; 30 Caucasians, 15 African Americans, and 1 Asian were examined.

At baseline, IHL and VAT were positively correlated (r = 0.57, P < 0.0001), and both fat depots were related to FCS (Fig. 1). IMCL in the soleus was not correlated with FCS, VAT, or IHL. S_i was significantly related to IHL (r = −0.31, P = 0.04) and FCS (r = −0.36, P = 0.01), but not VAT, DSAT, or IMCL. A forward stepwise regression analysis, showed that S_i was best explained by FCS (P < 0.01), whereas fat mass, IHL, VAT, and IMCL were not additional independent determinants. All of the above correlations were also statistically significant at month 6 (data not shown).

The impact of the intervention can be seen in Table 1 by comparing results at month 6 versus baseline.

Body weight was significantly reduced in the CR (−8 ± 1 kg, 10 ± 1%), CREX (−8 ± 1 kg, 10 ± 1%), and LCD (−11 ± 1 kg, 14 ± 1%) groups compared with the control group, and each intervention group had significant losses of fat mass (CR 24 ± 3%, CREX 25 ± 3%, and LCD 32 ± 3%) and fat-free mass (CR 5 ± 1%, CREX 3 ± 1%, and LCD 6 ± 1%). Similar reductions in VAT (CR 28 ± 4%, CREX 27 ± 3%, and LCD 36 ± 3%; P < 0.005), SAT (CR 26 ± 4%, CREX 28 ± 3%, and LCD 34 ± 3%; P < 0.005), and DSAT (CR 29 ± 5%, CREX 30 ± 3%, and LCD 37 ± 3%; P < 0.005) were observed in each intervention group. The intervention also induced a significant (P < 0.001) reduction in FCS (CR 19 ± 4%, CREX 26 ± 5%, and LCD 26 ± 4%). The changes in body composition and abdominal fat were not dependent on whether the caloric deficit was achieved by exercise and diet (CREX) or diet alone (CR and LCD). IHL was significantly (P < 0.01) reduced by the intervention but was not additionally influenced by exercise (CR 37 ± 10%, CREX 29 ± 15%, and LCD 40 ± 10%). The intervention did not change the percentage of IMCL in the soleus for either the CR (−8 ± 7%) or CREX groups (3 ± 11%); however, it tended to decrease in the LCD group (−12 ± 6, P = 0.07).

After the 6-month intervention, there was a significant improvement in S_i in the CREX (37 ± 18%, P < 0.01) and LCD groups (70 ± 34%, P < 0.04), which also tended to increase in the CR group (40 ± 20%, P = 0.08). The improvement in S_i was not different among the three intervention groups. Similarly, AIR_g was significantly decreased from baseline in each of the treatment groups (CR 29 ± 7%, CREX 30 ± 8%, and LCD 28 ± 9%; P < 0.01).

The decrease in FCS was correlated with the decrease in weight (r = 0.61, P < 0.001), percent fat (r = 0.61, P = 0.0001), VAT (r = 0.38, P < 0.01), SAT (r = 0.54, P = 0.0001), and fasting serum triglyceride levels (r = 0.44, P < 0.005). Similarly the decrease in IHL was related to the reduction in weight (r = 0.46, P = 0.001), percent fat (r = 0.37, P < 0.001), VAT (r = 0.49, P < 0.001), and fasting serum triglyceride levels (r = 0.60, P < 0.005). The improvement in S_i was related to the reduction in percent fat (r = −0.46, P = 0.002) (Fig. 2A), VAT (r = −0.51, P < 0.01) (Fig. 2B), SAT (r = −0.32, P < 0.05), and DSAT (r = −0.47, P < 0.01), but it was not related to the change in FCS (P = 0.11) (Fig. 2C), IHL (P = 0.59) (Fig. 2D), or IMCL (P = 0.52). A forward stepwise regression analysis showed that 25% of the improvement in S_i was attributed to the reduction in VAT (P = 0.002). The changes in IMCL, IHL, FCS, and the other abdominal fat depots were not additional independent determinants. These correlation analyses were repeated with the control group removed. The significance of the relationship between the changes in FCS and VAT and the changes in IHL and percent fat was lost, but no other relationships were affected.

In this study we examined the relationships between S_i and various indexes of body fat in overweight, glucose-tolerant subjects before and after calorie restriction. At baseline, we found that 1) fat deposition in liver was related to the accumulation of fat in the abdominal visceral area and to enlarged subcutaneous abdominal adipocytes and 2) increased FCS but not ectopic fat deposition in muscle and liver was independently associated with reduced insulin sensitivity. In response to 6 months of calorie restriction, we found that 1) weight, visceral fat, and FCS are reduced with improvements in S_i and reduced AIR_g and 2) fat deposition in liver but not muscle was reduced by the intervention, but the changes were not associated with improvements in S_i.

Whether ectopic lipid deposition in skeletal muscle and/or liver is related to total body fat is debated. Several studies have suggested that ectopic fat accumulation is independent of whole-body adiposity (16,24–30). However, other studies have noted that lipid accumulation in both muscle (27,31–33) and liver (34–37) increases as a function of obesity, providing that subjects with a wide range of adiposity are studied. In this study, we observed that lipid deposition in liver (but not muscle) was related to both total and abdominal adiposity. Specifically, our findings indicate that ectopic fat in the liver may be related to visceral fat stores. This relationship between liver lipid and visceral adiposity has been noted in some (34,38) but not all (29,30) studies. Most interestingly, we observed that liver lipid infiltration tended to be greater in overweight individuals who had enlarged adipocytes and increased visceral abdominal adiposity. Furthermore, visceral fat was related to FCS. These findings support the hypothesis that inadequate subcutaneous adipose stores result in lipid overflow into visceral fat and other nonadipose tissues (39). In this regard, visceral fat could be considered as a marker of ectopic fat.

At baseline and at month 6, large fat cells were also the strongest determinant of insulin resistance in these nondiabetic subjects. This finding prompts speculation that impaired adipogenesis may be the primary defect in insulin resistance, and the hypothesis is supported by findings that humans with partial or complete loss of adipose tissue are extremely insulin resistant (40), that surgical replacement of adipose stores in the fatless mouse restores insulin sensitivity (41), and that expression of Wnt signaling genes and adipogenic transcription factors are reduced in nondiabetic subjects with a family history of type 2 diabetes (42). Large fat cells have also been shown to have a different pattern of adipocytokine secretion than smaller fat cells (43), which may contribute to the strong association between large FCS and insulin sensitivity.

In contrast to previous studies (24,26–28,31,44,45), we observed that IMCL was not related to insulin sensitivity. Furthermore, IMCL was not related to adipocyte size. Our results are consistent with the hypothesis that IMCL stores alone are not sufficient to account for impaired insulin action (46–48). Liver lipid, on the other hand, was inversely related to insulin sensitivity. Liver lipid content has previously been reported to correlate with measures of whole-body insulin sensitivity in individuals with and without diabetes (30,34,35,38,49), but this relationship is difficult to explain mechanistically because most ingested (or infused) glucose is taken up by muscle. Theoretically, IHL is expected to correlate with reduced hepatic insulin sensitivity (impaired insulin suppression of glucose rate of appearance) and not necessarily with whole-body insulin action. However, the accumulation of hepatic triglyceride has been hypothesized to reduce insulin clearance and lead to peripheral insulin resistance via a downregulation of insulin receptors (34,50). Clearly, prospective human studies that define whether lipid accumulation in liver precedes insulin resistance would be of interest.

Contrary to some previous studies (51,52), we observed that diet alone or with exercise produced identical reductions in weight, fat mass, and abdominal fat mass. These conflicting results may be due to inaccurate calculations of the energy costs of the prescribed activity in those studies, which would lead to differences in energy deficits among groups. We also observed that FCS was reduced in response to an energy deficit, but we could not detect an additional effect of exercise. Our study was underpowered to detect differences in FCS among groups and our results contrast with the reports of You et al. (53), who found that whereas total body fat reduction was equivalent among groups, abdominal adipocyte size was reduced only with the combination of diet and exercise. The current study is also the first to simultaneously measure ectopic fat stores in both muscle and liver in response to a calorie restriction intervention. We found that the calorie restriction alone or with exercise did not affect IMCL in the soleus. These results are consistent with previous studies (6,14,15) and (together with the findings that IMCL was not independently related to S_i) suggest that IMCL accumulation alone is not likely to be a causal factor leading to acquired insulin-signaling defects in muscle. Many other factors, including lipid droplet size, location of lipid droplets relative to mitochondria, and muscle oxidative capacity, are all potential determinants of insulin resistance (15,48,54). An alternate hypothesis is that the capacity for lipid metabolism is an important mediator in the association between IMCL and insulin resistance. We observed, however, that IHL was sensitive to calorie restriction being reduced by an average 29–40% in the intervention groups. Caution must be exercised when interpreting these results because the study may have been underpowered to detect small differences in IHL among groups. The reduction in liver lipid levels is consistent with results of Tiikkainen et al. (16), who reported a 39–49% reduction in IHL with a simultaneous reduction in body mass of 8% in obese nondiabetic women. In addition, we also observed parallel reductions in IHL and abdominal visceral fat.

In summary, calorie restriction by diet alone or in conjunction with exercise leads to similar improvements in insulin sensitivity and reductions in β-cell sensitivity in overweight, glucose-tolerant subjects. The study also provides support for the hypothesis that the underlying pathologic cause of insulin resistance is related to abnormal partitioning of fat among adipose, hepatic, muscle, and pancreatic tissues, probably as a result of an inability to make new fat cells. However, the finding that IMCL was not responsive to weight loss (despite improvements in insulin sensitivity) suggests that intracellular fat accumulation is not a causal factor in insulin resistance in muscle. Overall, this study provides new evidence to suggest that impaired adipogenesis and increased liver lipid infiltration occur early in the pathogenesis of insulin resistance.

This work was supported by grants U01 AG20478 (to E.R.) and K01 DK062018 (to D.E.L.-M.). L.M.R. is supported by a Neil Hamilton-Fairley Training Fellowship awarded by the National Health and Medical Research Council of Australia (ID 349553).

The authors thank the remaining members of the Pennington CALERIE Research Team: James DeLany, Corby Martin, Julia Volaufova, Marlene Most, Lilian de Jonge, Tuong Nguyen, Frank Greenway, Emily York-Crow, Catherine Champagne, Brenda Dahmer, Andy Deutsch, Paula Geiselman, Jennifer Howard, Jana Ihrig, Michael Lefevre, Darlene Marquis, Connie Murla, Sabrina Yang, Robbie Durand, Sean Owens, Aimee Stewart, and Vanessa Tarver. Our gratitude is extended to the excellent staffs of the Inpatient Clinic and Metabolic Kitchen. Our thanks also go to Health and Nutrition Technology (Carmel, CA) for providing us with all of the HealthOne formula used in the study and to Edward J. Robarge for technical assistance with collection of the magnetic resonance spectroscopy data. Finally, our profound gratitude goes to all the volunteers who spent so much time participating in this very demanding research study.