Impaired Metabolic Flexibility to Fasting is Associated with Increased Ad Libitum Energy Intake in Healthy Adults

Abstract

Objective:

We investigated how changes in 24-h respiratory exchange ratio (RER) and substrate oxidation during fasting compared an energy balance condition influence subsequent ad libitum food intake.

Methods:

Forty-four healthy, weight-stable volunteers (30 male, 14 female; age 39.3±11.0 years; BMI 31.7±8.3 kg/m²) underwent 24-h energy expenditure measurements in a respiratory chamber during energy balance (50% carbohydrate, 30% fat, 20% protein) and 24-h fasting. Immediately after each chamber stay, participants were allowed 24-h ad libitum food intake from computerized vending machines.

Results:

24-h RER decreased by 9.4% (95% CI −10.4 to −8.5; p<.0001) during fasting compared to energy balance, reflecting a decrease in carbohydrate oxidation (−2.6±0.8 MJ/day; p<.0001) and an increase in lipid oxidation (2.3±0.9 MJ/day; p<.0001). Changes in 24-h RER and carbohydrate oxidation in response to fasting were correlated with the subsequent energy intake such that smaller decreases in fasting 24-h RER and carbohydrate oxidation, but not lipid oxidation, were associated with greater energy intake after fasting (r=0.31, p=0.04; r=0.40, p=0.007; r=−0.27, p=0.07, respectively).

Conclusions:

Impaired metabolic flexibility to fasting, reflected by an inability to transition away from carbohydrate oxidation, is linked with increased energy intake.

Introduction:

Identification of individuals prone to overeat may facilitate prevention and treatment of obesity and its complications. Fat-free mass and energy expenditure (EE) are important determinants of energy intake (1, 2, 3). During periods of food deprivation, animals have evolved survival mechanisms that include energy expenditure reduction (4) and flexibility to switch mitochondrial fuel sources from glucose to fatty acids and ketones (5). In the fasting state, reductions in insulin decrease transport of glucose into tissues such as adipose and skeletal muscle while glucagon stimulates glycogenolysis to produce glucose to sustain adequate delivery of glucose to key glucose dependent tissues such as brain, nerves, red blood cells, renal medulla, and bone marrow. Metabolic flexibility is the ability to change the dominant source of fuel in response to food availability (6). There is evidence to suggest that being metabolically inflexible (inability to match the rates of oxidation to the macronutrients ingested) may be predictive of weight gain in the long term (7). The mechanism by which this occurs is yet to be established and likely differs based on flexibility (or lack thereof) to different dietary conditions. Metabolic flexibility during transition from the fed to the fasting state may be measured in several ways including by changes in respiratory exchange ratio (RER) (8, 9). Higher 24-h RER, reflecting a tendency towards carbohydrate oxidation (CARBOX) over lipid oxidation (LIPOX) (10), is associated with increased ad libitum food intake (3, 11) and the propensity to gain weight and fat mass (12). Current literature shows that an impaired metabolic flexibility to overfeeding diets predicts weight gain at one year (7). The evidence linking 24-h RER and energy intake has been generated when RER was measured in energy balance conditions (3, 11) or an effect on energy intake has been inferred based on gain of total weight and adiposity in free living conditions (12, 13). However, it is unclear whether changes in 24-h RER as a result of metabolic stressors such as fasting, influences subsequent food intake when food is available.

In the present study, we sought to determine whether 36 hours of fasting would induce a compensatory increase in energy intake after fasting and whether the ability to adapt one’s energy metabolism in response to this deprivation (i.e., metabolic flexibility), as reflected by changes in 24-h RER and EE by whole-room indirect calorimetry, would be associated with subsequent ad libitum energy intake measured objectively in the laboratory. We hypothesized that fasting would induce a compensatory increase in energy intake, with a reduced ability to decrease 24-h RER and CARBOX, and increase LIPOX during fasting being associated with greater subsequent ad libitum energy intake. We also hypothesized that both during energy balance and changes in 24-h EE during fasting would be associated with food intake. In exploratory analyses, we also measured hormones that have been shown to influence food intake and energy metabolism such as GLP-1 and ghrelin (14, 15, 16), along with hormones that have been shown to be stimulated under caloric deprivation and may have implications in metabolic flexibility such as FGF-21 (17). Additionally, we measured urine catecholamine excretion as a surrogate for symptathetic nervous system which has been shown to influence substrate utilization (18, 19) and thus may be implicated in metabolic flexibility.

Methods:

Research Design and Methods

This study (ClinicalTrials.gov identifier NCT02939404) was approved by the Institutional Review Board of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). Other findings from this protocol unrelated to main aim of the current study were previously reported (20). All participants provided written informed consent prior to beginning the study. Eligible volunteers were between 18 and 55 years of age and weight stable by self-report. Participants were determined to be healthy by medical history, physical examination, and routine lab tests (e.g., no chronic illnesses). Female participants were premenopausal. Complete inclusion and exclusion criteria are described in Supplemental Methods. After screening, enrolled volunteers were admitted to the NIDDK clinical research unit in Phoenix, Arizona, and placed on a daily weight-maintaining diet (50% carbohydrate, 30% fat, and 20% protein; food quotient 0.87) throughout the admission period, with total daily intake calculated using unit-specific equations based on weight and sex (21). Body weight was maintained by adjusting energy intake as needed (22). After at least 3 days of the weight-maintaining diet, participants had a 75-g oral glucose tolerance test and withdrawn from the study if they had diabetes based on American Diabetes Association criteria for fasting and 2-h plasma glucose concentrations (23). Subsequently, participants completed two 24-h sessions inside a whole-room indirect calorimeter, the first during energy balance and the second during fasting. On the day of the energy balance chamber, standardized meals were provided at 07:00AM (breakfast), 11:00AM (lunch), 4:00PM (dinner), and 7:00 PM (snack). Foods were as follows: a) breakfast - cheese, eggs, tortilla, orange juice, and decaffeinated coffee, b) lunch and dinner – beef, rice, butter, bread, banana, cookie, and green beans, and c) snack – turkey breast, bread, butter, peach, orange juice, mustard. Food intake–related hunger ratings were assessed during (6:30PM), and before and after each chamber stay (before breakfast for the energy balance chamber), using a visual analog scale (VAS). Participants rated their hunger scores ranging from 0 (“not at all”) to 100 (“extremely”). Participants were given ad libitum access to food in a vending machine paradigm for 3 days after the energy balance chamber and for 1 day after fasting chamber. To make the time periods comparable, only the first day of intake after the energy balance chamber was used to assess changes in ad libitum intake. A list of food items that were made available for participants during the vending periods are shown in Table S1. An overview of the study procedures is shown in Figure S1. Of 109 volunteers screened, 64 met eligibility criteria. Of the 64 enrolled participants, 44 completed both chambers and had measurements of ad libitum energy intake and were included in the current analysis (Figure S2).

Whole-Room Indirect Calorimetry

The 24-h EE and RER were assessed in an open-circuit, whole-room indirect calorimeter as previously described (24, 25). Briefly, volunteers entered the calorimeter at 08:00AM after having fasted overnight and eating breakfast (specifically during energy balance chamber) at 07:00AM. They remained therein for ~23 hours and 30 minutes during which 3 meals were provided (only during energy balance chamber) at 11:00AM, 4:00PM, and 7:00 PM. While in the calorimeter, participants were able to move freely around the chamber but were asked to refrain from exercise (e.g., push-ups). The mean ambient temperature while in the calorimeter was 23.4°C ± 0.23°C [mean ± standard deviation (SD)]. Prescribed energy intake during the equicaloric assessment was calculated using unit-specific formulas as previously described and was ~20% less compared with the weight-maintaining diet outside the calorimeter to account for limited physical activity inside the chamber (26). During the fasting chamber, no food was provided after the previous day’s 07:00PM meal, but participants were provided with 4 liters of water to self-hydrate ad libitum. O₂ consumption and CO₂ production during each respiratory chamber were measured every minute, averaged, and extrapolated to 24 h, and used to calculate 24-h EE and 24-h RER (24). Urine was collected over the 24 h to measure urea nitrogen excretion to calculate substrate oxidation rates [CARBOX, LIPOX, and protein oxidation (PROTOX)] as previously described (26, 27) and to measure urine catecholamine excretion rates. Fasting plasma samples were collected before breakfast and within the hour of entering and exiting the respiratory chamber. Plasma was frozen at −70°C for later measurements.

Ad Libitum Food Intake as Measured by Vending Machine System

The measurement of ad libitum food intake by an automated food-selection system was previously described, validated, and tested for reproducibility (28). Participants were asked to complete a food preference questionnaire which consisted of a list of 80 food items presented in random order. Individuals were asked to assign each food item a hedonic rating with a 9-point Likert scale (1=dislike, 5=neutral, 9=like extremely). An automated food-selection system made up of a refrigerated vending machine (model 3007; U-Select-It) included 40 items for breakfast, lunch, and dinner as previously described (28). The 40 food items were made available to participants consisted of foods that the participants rated as intermediate (between 4 and 8) on the food preference questionnare.

Each volunteer was assigned to a single vending machine, had unrestricted access to the machine for 23.5 h/d for 3 days after energy balance chamber and 1 day after fasting chamber. Participants were instructed to eat only in the vending room, whatever they wished whenever they desired and to return the unconsumed food portions to the metabolic kitchen for calculation of actual energy consumed obtained using a food database. Daily energy, protein, fat, and carbohydrate intakes were calculated from the actual weights of food and condiments consumed with the use of the Food Processor SQL Edition (ESHA, version 10.0.0; ESHA Research, Salem, OR) modified to reflect the nutrient content of specific food items as indicated by the manufacturer.

Analytical Measurements

Plasma glucose concentrations were measured using the glucose oxidase method (Analox Instruments, Lunenburg, MA). Plasma insulin concentrations were measured with an automated immunoenzymometric assay (Tosoh Bioscience Inc., Tessenderlo, Belgium). FGF-21 was measured using the human FGF-21 quantikine ELISA kits from R&D Systems (Minneapolis, MN; intra-assay CV: 2.51%, and inter-assay CV: 5.2%). Active GLP-1 were measured using immunoassay kits from Meso Scale Discovery (Rockville, MD; intra- and inter- assay CV were 3.63% and 8.53%). Total Ghrelin was measured using ELISA kits from EMD-Millipore (Billerica, MA) and the inter and intra assay % CV were 4.11 and 8.63, respectively. Free fatty acids were measured using a kit from Fujifilm-Wako Diagnostics (Mountain View, CA), and the intra- and inter-assay CV was 4.4% and 5.8% respectively. Urine with added hydrochloric acid was stored at −70° C until analysis. Urinary epinephrine, norepinephrine, and dopamine were measured by liquid chromatography-tandem mass spectrometry (Mayo Clinic Laboratories, Rochester, MN).

Statistical Analysis

This analysis assessed a pre-specified hypothesis of an ongoing clinical trial (ClinicalTrials.gov identifier NCT02939404). Statistical analysis was performed using the SAS Statistical Software package (SAS Enterprise Guide version 7.15; SAS Institute, Cary, NC). Data were expressed as mean ± SD with range values (min/max) or mean with 95% confidence limits. Insulin distribution was skewed and was log₁₀-transformed before analyses to meet the assumptions of parametric tests. The associations presented were quantified by the Pearson correlation coefficients ^®. General Linear Models were used to quantify the beta coefficients between measures of energy metabolism and energy intake. Changes (Δ) in plasma concentrations of hormones and free fatty acids were calculated as post-chamber minus pre-chamber concentrations. Changes (Δ) in 24-h urinary catecholamine excretion were calculated as fasting chamber minus energy balance chamber. Relative overeating was calculated in two ways: 1) ad libitum energy intake after fasting minus intake after energy balance (Δ) and 2) ad libitum energy intake after fasting minus 24-h energy expenditure during the energy balance chamber. Metabolic flexibility was calculated as changes in energy expenditure measures (e.g., RER, CARBOX, LIPOX, PROTOX) during fasting minus the energy balance chamber. One sample t-test was used to compare if any given delta (Δ) was significantly different than 0. Independent samples t -test was used to examine sex differences. Statistical significance was defined as p < 0.05 (two-tailed).

Results

Characteristics of the 30 male and 14 female participants are shown in Table 1. A wide range of BMI was represented in the cohort (range 17.6 to 48.6 kg/m², mean 31.7 ± 8.3 kg/m²). There were no differences in 24-h RER during the energy balance chamber by sex or race (p=0.2 and p=0.6 respectively). There was a negative correlation between 24-h RER during the energy balance chamber and fat mass (r=−0.38 p=0.01), fat free mass (r=−0.32 p=0.03) but not age or sex (both p>0.1). Energy balance and fasting measures of 24-h RER, CARBOX and LIPOX were positively associated with each other (r=0.52, p=0.0003; r=0.34, p=0.02; r=0.73; all p<0.0001 respectively; Figure S3).

Table 1.

Characteristics of the study cohort.

	Whole cohort n= 44	Males n=30	Females n=14
Age (y)	39.3±11.0 (19.6, 59.7)	41.8±10.4 (21.0, 59.7)	33.8±10.8 (19.6, 52.0) *
Race/ethnicity (n)	B 6, W 4, H 4, N 27, O 3	B 5, W 4, H 4, N 16, O 1	B 1, N 11, O 2
BMI (kg/m2)	31.7±8.3 (17.6, 48.6)	29.2±7.5 (17.6, 46.5)	37.3±7.5 (23.4, 48.6) *
Height (cm)	171.2±8.5 (153.5, 188.0)	174.4±7.4 (163.0, 188.0)	163.8±5.7 (153.5, 173.0)
Weight (kg)	92.1±22.4 (61.2, 145.5)	88.4±20.8 (61.2, 139.2)	98.7±24.5 (61.4, 145.5)
Fat mass (kg)	34.7±18.1 (9.1, 79.4)	28.1±14.5 (9.1, 64.1)	49.9±16.4 (21.6, 79.4)
Fat-free mass (kg)	57.4±9.0 (39.7, 75.0)	60.2±7.9 (46.7, 75.0)	51.0±8.4 (39.7, 66.0)
Body fat (%)	35.6±11.8 (14.9, 54.6)	30.0±9.1 (14.9, 46.1)	48.3±6.2 (35.3, 54.6) *
Fasting glucose (mg/dL)	91.2±7.2 (77.0, 113.0)	90.3±7.1 (77.0, 105.0)	92.9±7.4 (81.0, 113.0)
2-h glucose (mg/dL)	123.2±31.5 (50.0, 183.0)	120.1±33.4 (50.0, 183.0)	129.6±27.3 (90.0, 183.0)
Fasting insulin (pmol/L)	96.0±84.0 (15.0, 337.8)	75.6±72.0 (15.0, 334.8)	139.2±93.0 (32.4, 337.8)

B, Black; W, White; H, Hispanic; N, Native American; O, Other

Data are presented as mean ± SD (minimum, maximum) unless otherwise indicated.

^*p-values for differences between males and females by independent samples t-test.

Ad Libitum Food Intake Following Energy Balance and Fasting

Self-reported feelings of hunger before, during and after the energy balance and fasting chamber stays are presented in Figure S4. During the fasting chamber, there was an increase in hunger after 24 hours of fasting compared with 12 hours of fasting. However, no additional increase in hunger was observed at 36 hours of fasting. Mean ad libitum energy intake was 16.4 ± 5.6 MJ/day (range: 6.2 to 35.6 MJ/day) for the day following the energy balance chamber and 16.7 ± 5.5 MJ/day (range: 5.2 to 34.7 MJ/day) for the day following the fasting chamber. There was no difference in ad libitum energy intake from after energy balance condition to after fasting condition (Figure 1A) though there was wide variability in individual responses (+0.3 ± 3.8 MJ/day, range −7 to 10.5 MJ/day, p=0.6). Energy intake above 24-h EE was observed during the vending days both after energy balance (7.2 ± 5.6 MJ/day, range −2.4 to 26.4 MJ/day, p<0.001) and after fasting condition (7.6 ± 5.6 MJ/day, range −6.1 to 25.6 MJ/day, p<0.0001; Figure 1B]

Figure 1.

(A) Differences in ad libitum energy intake assessed with a vending machine paradigm after energy balance and fasting. (B) Ad libitum energy intake after the fasting chamber above 24-h energy expenditure during energy balance. One sample t-test used to assess whether means are different from 0. Mean change and 95% confidence intervals are shown.

Changes in Energy Metabolism from Energy Balance to Fasting

Twenty-four hour RER decreased from energy balance to the fasting condition (−0.08 ± 0.02; range −0.13 to −0.01; p<0.0001; Figure 2A), reflected by an increase in LIPOX (2.3 ± 0.9 MJ/day; range 0.02 to 4.8 MJ/day; p<0.0001; Figure 2C) and a decrease in CARBOX (−2.6 ± 0.8 MJ/day; range −4.7 to −0.7 MJ/day; p<0.0001; Figure 2C) (Table S2). PROTOX decreased from energy balance to fasting condition (−0.2 ± 0.3 MJ/day; range −1.2 to 0.7 MJ/day; p<0.0001; Figure 2C) (Table S2). Twenty-four-hour EE also decreased from energy balance to the fasting condition (−0.6 ± 0.4 MJ/day; range −1.5 to 0.4 MJ/day; p<0.0001; Figure 2B).

Figure 2.

(A) Respiratory exchange ratio (RER) during energy balance and fasting. (B) Energy expenditure during energy balance and fasting. (C) Carbohydrate, lipid, and protein oxidation rates during energy balance and fasting. One sample t-test was used to assess whether means are different from 0. Asterisk (*) indicates p <0.05.

Energy Metabolism and Ad Libitum Energy Intake

Unadjusted and adjusted (age, sex, fat mass, fat free mass) 24-h RER, 24-h EE, and 24-h substrate oxidation rates (i.e., CARBOX, LIPOX, and PROTOX) during the energy balance chamber were not associated with ad libitum energy intake after this chamber (all p>0.2). Twenty-four-hour RER during fasting showed a positive correlation with ad libitum energy intake after fasting but did not reach statistical significance; (r=0.28, p=0.06, Figure S5A), and was not associated after adjusting for age, sex, fat mass and fat free mass (r=0.15, p=0.37). Ad libitum energy intake after fasting was associated with CARBOX during the fasting chamber (r=0.31, p=0.03, Figure S5B) but was not associated with LIPOX or PROTOX (r=−0.13, p =0.39; r=0.23, p=0.13; Figure S5C-D). The association between CARBOX and ad libitum energy intake was no longer significant after adjusting for age, sex, fat mass, and fat-free mass (r=0.18, p=0.28).

Metabolic Flexibility to Fasting and Ad Libitum Energy Intake

The decreases in 24-h RER and CARBOX from energy balance to the fasting condition were associated with changes in ad libitum energy intake (r =0.31, p =0.04, Figure 3A; r =0.40 p =0.007 Figure 3B). For every 0.01 decrease in 24-h RER from energy balance to the fasting condition, ad libitum energy intake was on average 0.4 MJ/day lower the following day (95% CI: 0.01 to 0.8 MJ/day, p<0.001). Similarly, for every 1 MJ/day decrease in CARBOX from energy balance to fasting condition, energy intake was 1.75 MJ/day lower (95% CI: 0.5 to 3 MJ/day, p<0.001). Similar results were obtained when these results were adjusted for deviations in energy balance during the energy balance chamber stay (partial r=0.32 p=0.03 and partial r=0.39 p=0.01 for changes in 24-RER and CARBOX and changes in ad libitum energy intake correlation respectively.) The increases in LIPOX from energy balance to fasting were not associated with changes in ad libitum energy intake. (r=−0.27 p=0.07 Figure 3C). Similar results were obtained when energy intake was expressed in excess of 24-h EE after the fasting chamber (i.e., ad libitum energy intake after fasting minus 24-h EE during energy balance chamber), which was positively correlated with fasting-induced decreases in 24-h RER (r = 0.33, p = 0.03 Figure 3D) and CARBOX (r =0.38, p =0.01 Figure 3E), and was negatively correlated with increases in LIPOX (r = −0.38, p =0.01 Figure 3F).

Figure 3.

Correlations between change in ad libitum energy intake and changes in (A) RER, (B) carbohydrate oxidation, and (C) lipid oxidation from energy balance to fasting. Correlations between energy intake after fasting above 24-h energy expenditure during energy balance and changes in (D) RER, (E) carbohydrate oxidation, and (F) lipid oxidation from energy balance to fasting. The strength of association was quantified by the Pearson correlation coefficient.

There was a positive relationship between 24-h RER during energy balance chamber stay and the deviations from energy balance during the energy balance chamber stay (r=0.46 p=0.001 Figure S6A). Changes in 24-h RER from energy balance to fasting chamber stay were not associated with energy balance that was achieved during the energy balance chamber stay (p=0.4, Figure S6B).

Decreases in 24-h EE from energy balance to fasting was not associated with change in ad libitum energy intake (p=0.96) or energy intake above 24-h EE (p =0.6) after fasting (Figure S7A-B).

Investigations into Hormonal Determinants of Metabolic Flexibility and Ad Libitum Energy Intake

On average, there were decreases in both plasma glucose and insulin after fasting (Table 2). Plasma GLP-1 and FGF-21 concentrations decreased with fasting (Table 2). Plasma ghrelin concentrations did not change after fasting condition. Free fatty acid levels were increased after fasting condition (Table 2). There was an increase in urinary epinephrine excretion and decrease in norepinephrine from energy balance to fasting (Table 3). Plasma concentrations of beta-hydroxybutyrate and lactate but not pyruvate showed an increase after fasting (Table S3).

Table 2.

Changes in plasma concentrations of glucose, insulin, GLP-1, FGF-21, ghrelin, and free fatty acids following dietary interventions

	Pre-chamber	Post-chamber	Change (Δ)	Percent Change (%)	p-value
Energy balance
Glucose (mg/dL)	93.7 ± 6.3 (82.0, 110.0)	93.6 ± 5.8 (80.0, 110.0)	−0.1 ± 4.0 (−10.0, 7.0)	−0.1 (−1.5, 1.2)	0.82
Insulin (pmol/L)	92.4 ± 81.0 (13.2, 361.8)	91.2 ± 77.4 (16.2, 324.0)	−1.2 ± 49.8 (−210.0, 150.0)	−1.8 (−18.8, 15.2)	0.42
GLP-1 (pg/ml)	1.5 ± 1.4 (0.1, 7.8)	1.9 ± 2.0 (0.1, 8.4)	0.3 ± 2.2 (−4.7, 7.0)	20.6 (−25.5, 66.7)	0.37
FGF-21(pg/ml)	136.6 ± 90.4 (2.9, 325.0)	129.0 ± 86.8 (3.0, 348.4)	−7.6 ± 44.1 (−91.0, 135.6)	−5.6 (−16.1, 4.9)	0.28
Ghrelin (pg/ml)	162.1 ± 88.6 (39.8, 431.0)	159.7± 88.2 (35.3, 456.6)	−2.4 ± 42.8 (−149.4, 141.4)	−1.5 (−10.1, 7.0)	0.72
FFA (mEq/L)	0.3 ± 0.1 (0.1, 0.7)	0.4 ± 0.2 (0.2, 0.8)	0.1 ± 0.1 (−0.4, 0.3)	17.8 (5.5, 30.1)	0.005
Fasting
Glucose (mg/dL)	92.0 ± 7.0 (79.0, 112.0)	87.8 ± 8.9 (70.0, 118.0)	−4.2 ± 9.0 (−21.0, 31.0)	−4.6 (−7.6, −1.5)	0.004
Insulin (pmol/L)	93.6 ± 69.0 (14.4, 313.8)	67.8 ± 64.2 (8.4, 282.6)	−25.2 ± 44.4 (−126.0, 132.0)	−24.0 (−38.8, −9.1)	0.003
GLP-1 (pg/ml)	2.1 ± 1.6 (0.5, 7.0)	1.3 ± 1.5 (0.1, 8.6)	−0.8 ± 1.8 (−6.8, 2.6)	−39.8 (−68.0, −11.5)	0.007
FGF-21 (pg/ml)	201.7 ± 129.4 (3.2, 445.7)	130.2 ± 85.3 (15, 344.4)	−71.5 ± 100.9 (−319.4, 65.1)	−35.4 (−51.7, −19.2)	<0.0001
Ghrelin (pg/ml)	168.5 ± 72.9 (59.7, 342.3)	158.3 ± 7.8 (48.3, 379.2)	−10.2 ± 49.7 (−107.9, 109.8)	−6.0 (−15.6, 3.5)	0.20
FFA (mEq/L)	0.3 ± 0.2 (0.1, 1.2)	0.8 ± 0.4 (0.1, 1.8)	0.5 ± 0.4 (−1.0, 1.4)	163.9 (119.8, 208.1)	<0.0001

FFA, free fatty acids. Data are presented as mean ± SD (minimum, maximum), except for percentage changes that are reported as mean with 95% CI. Significant changes (p < 0.05) from zero by one sample t-test are reported in bold for absolute change.

Table 3.

Changes in 24-h urine catecholamines from energy balance to fasting

Catecholamine	Energy Balance	Fasting	Absolute Change	Percent Change (%)	p-value
Dopamine (mcg/24 h)	358.4 ± 107.6 (155.0, 693.0)	302.5 ± 85.4 (34.0, 485.0)	−55.9 ± 112.0 (−403.0, 209.0)	−15.6 (−25.9, −.3)	0.0039
Epinephrine (mcg/24 h)	3.5 ± 2.3 (0.9, 11.0)	5.2 ± 2.5 (2.0, 12.0)	1.7 ± 2.5 (−6.8, 7.5)	47.4 (23.6, 71.3)	0.0003
Norepinephrine (mcg/24 h)	37.3 ± 17.2 (12.0, 88.0)	31.5 ± 11.9 (12.0, 60.0)	−5.8 ± 11.5 (−39.0, 13.0)	−15.7 (−25.8, −5.6)	0.003

Data are presented as mean ± SD (minimum, maximum), except for percentage changes that are reported as mean with 95% CI. Significant changes (p < 0.05) from zero by one sample t-test are reported in bold for absolute change.

Pre-chamber fasting insulin was negatively correlated with 24-h RER during fasting (r=−0.35, p=0.02, Figure 4A) but not during energy balance (r=−0.30, p=0.06). This was reflected by a positive correlation between fasting insulin and lipid oxidation during the fasting conditions (r=0.57, p<0.0001, Figure 4B) but not carbohydrate oxidation (p=0.6). Pre-chamber, post-chamber, and the changes in plasma glucose concentrations during energy balance and fasting were not associated with measures of energy metabolism, including metabolic flexibility (ΔRER, ΔCARBOX, ΔLIPOX, all p>0.5)

Figure 4.

Correlations between fasting insulin and (A) respiratory exchange ratio (RER) during fasting and (B) lipid oxidation rate during fasting. Correlations between urine norepinephrine excretion during fasting and RER during fasting (C) and lipid oxidation rate during fasting (D). The strength of association was quantified by the Pearson correlation coefficient.

Pre-chamber plasma concentrations of ghrelin were associated with 24-h RER during fasting (r=0.32, p=0.04, Figure S8A) and LIPOX (r= −0.44, p=0.004). Pre-chamber plasma concentrations of free fatty acids (Figure S8B), GLP-1 and FGF21 were not associated with 24-h RER during fasting (all p>0.06). None of these hormonal measurements were associated with changes in measures of metabolic flexibility, namely ΔRER, ΔCARBOX and ΔLIPOX (all p >0.13). Post-chamber hormone measurements the morning of the vending day were not associated with ad libitum energy intake after energy balance or fasting (all p>0.2).

Urinary norepinephrine but not epinephrine (all p>0.3) was associated with 24-h RER (r=−0.44, p=0.005, Figure 4C) and LIPOX (r=0.56, p=0.0002, Figure 4D) but not CARBOX (p=0.15) during fasting conditions. Changes in epinephrine and norepinephrine concentrations were not associated with changes in measures of metabolic flexibility (ΔRER, ΔCARBOX, ΔLIPOX, all p>0.09. None of the urinary catecholamine measurements were associated with ad libitum energy intake (all p>0.2).

Plasma concentrations of lactate, pyruvate, and beta-hydroxybutyrate were not associated with measures of energy metabolism (including metabolic flexibility) or energy intake (all p>0.1).

Discussion:

This study evaluated whether fasting would influence subsequent eating behavior over the following 24-h and whether acute metabolic adaptations (e.g., in RER and EE) to food deprivation would affect subsequent ad libitum food intake. Although ad libitum energy intake did not increase on average after fasting compared to eucaloric conditions, there was a wide range of interindividual variability in energy intake regardless of whether measured as absolute intake or as overeating above 24-h EE. The current study found that a decreased ability to reduce 24-h RER during fasting was associated with a subsequent greater energy intake and further demonstrated that a greater reliance on carbohydrate over lipid oxidation influenced energy intake.

The literature on compensatory increase in energy intake as a response to short term fasting or negative energy balance is conflicting. Some studies showed an increase in ad libitum intake (29, 30), while others showed no change (31, 32). As others suggest, a longer time period of fasting (>24 hours) may be required to see an increase in ad libitum intake after fasting conditon compared to after energy balance (33). In the current study, participants last ate at 0700PM the day before the fasting chamber and thus were fasting for 36-h before the 24-h ad libitum intake period. Despite the 36-h fasting, an increase in ad libitum energy intake was not observed. Thus, 36 hours may still be insufficient to elicit an increase in ad libitum intake or monitoring of subsequent intake over a longer period (>24 hours) may be required to see a compensatory increase in food intake.

Although there was a wide range of responses in 24-h RER, CARBOX, and LIPOX within our cohort after fasting, there remained a strong intraindividual preference for fuel oxidation, such that individuals who tended to rely more on a specific substrate for substrate oxidation rates during the energy balance condition also demonstrated this preference during fasting. The determinants of the changes in 24-h RER and hence metabolic flexibility in different conditions are largely unknown. Besides the strong intraindividual consistency of 24-h RER between dietary conditions, the greatest determinant of macronutrient oxidation and thus 24-h RER is the macronutrient profile of the ingested food (e.g., food quotient) (27). However, less is known about the determinants of 24-h RER during a fast where food intake is not present.

Cross-sectional studies have shown that a higher 24-h RER during energy balance conditions, reflecting a higher carbohydrate-to-fat oxidation, is associated with future weight gain (13, 34, 35). Animals and human ancestors evolved in environments with intermittent food availability. Thus, food deprivation is a physiologically relevant perturbation. As expected, there was a decrease in 24-h RER and CARBOX and an increase in LIPOX during short-term fasting when compared to the energy balance condition. Consistent with prior evidence that higher 24-h RER and specifically higher CARBOX during energy balance predicts greater food intake (3, 35), the current study shows that during acute energy deficit, the metabolic inflexibility to caloric deprivation is also a determinant of increased food intake which might contribute to weight gain. This RER association was mainly due to CARBOX and not LIPOX.

The current study found that a reduced ability to switch away from carbohydrate oxidation was associated with increased ad libitum energy intake. One proposed mechanism for this may involve regulation of glycogen stores as described by Flatt (36), which proposes depletion and replenishment of liver glycogen stores as a major driver of food intake. Early in a fast (e.g., several hours), blood glucose is maintained in part by glycogenolysis in liver and muscle. As fasting continues, glucoeneogenesis and energy expenditure are supported by non-carbohydrate sources (e.g., triglycerides and ketones), reflected in a decrease in CARBOX and 24-h RER and an increase in LIPOX (37, 38). Given the small size of glycogen reserves relative to its turnover and the physiological importance of maintaining minumum blood glucose concentrations, there may be evolutionary adaptations to maintain glycogen levels through feedback signals to the brain driving hunger to increase ad libitum food and carbohydrate intake (36). The evolution of these signals may be driven by the “selfish brain” that is reliant upon glucose as substrate and accounts for about one-half of glucose utilization in the entire body (39). While reviews of more recent studies that investigated the depletion of glycogen stores and its effects on subsequent intake could not find such an effect (40). It may be important to note however, that the method by which glycogen depletion was achieved in these studies was not fasting, but rather low carbohydrate overfeeding that was paired with exercise (40). The reported effects of low carbohydrate feeding (41) and exercise (42) on appetite may confound their effects on glycogen mediated appetite regulation. In the current study, given the conflicting results within the literature, it may be possible that individuals who were unable to switch from CARBOX to LIPOX may have depleted their liver glycogen stores faster than those who were able to switch succesfully, leading to increased food intake (43).

It has been proposed that signals linking glycogen to the brain occur via the autonomic nervous system (36). Attenuated sympathetic tone is independently associated with higher 24-h RER during energy balance (attributed to decreased capacity for lipid oxidation) (19) and long-term propensity to gain weight (44). Consistent with this, we found that urine norepineprhine excretion was negatively correlated with 24-h RER during both energy balance and fasting conditions and further associated with LIPOX. However, urine norepinephrine was not associated with subsequent ad libitum food intake, indicating other factors mediate the association between 24-h RER and energy intake. Other potential signals that may explain energy intake were also investigated including FGF-21, GLP-1, ghrelin, GLP-1, free fatty acids, insulin, and ketones but were not associated with eating behavior in this paradigm.

The strengths of this study include a period of weight stabilization prior to baseline measurement of energy expenditure and 24-h RER and sufficient duration of fasting (36 hours) to induce flexibility in energy metabolism. In addition, a number of hormones that are known to regulate energy metabolism and food intake were measured. Although the vending machine paradigm leads to overconsumption on average, it is a highly reproducible (28), objective measure of ad libitum energy intake, previously shown to predict weight gain (45). This study has several limitations. First, there was lack of an objective measurement of glycogen stores. Second, only one day of ad libitum food intake after fasting was completed and this fasting intervention was done after the measurement of ad libitum food intake after the energy balance stay due to a lack of randomization. It is unclear if eating behaviour would change beyond the initial 24-hour ad libitum food intake period and whether results would have differed if the fasting chamber stay preceeded the stay in the energy balance chamber. Third, although the relationship between RER and weight gain has been shown in both Native American and other populations, the majority of our participants were Native Americans which may limit the generalizability to other populations. Finally, participants chose to overfeed on the vending machine paradigm for unclear reasons. Although participants were allowed activities on the inpatient unit (e.g., watching television, playing board games, making crafts), we speculate that participants chose to overfeed due to boredom as these activities may differ from their daily activities. The novelty of the vending machine paradigm for our participants may be another contributing factor; it is unclear if the degree of overfeeding would have differed if another ad libitum food intake paradigm (e.g., ad libitum test meal with constant composition) was used. A decline in the novelty of the vending machine paradigm may have prevented an additional increase in food intake after fasting chamber stay above the ad libitum intake after the energy balance stay.

In conclusion, this study showed that an impaired metabolic flexibility to fasting, defined as the ability to switch substrates for fuel, is associated with subsequent greater ad libitum energy intake. The association was primarily due to an inability to decrease CARBOX, as opposed to increasing LIPOX, during prolonged fasting. Metabolic inflexibility in response to short-term food deprivation may be implicated in greater ad libitum energy intake. Thus, this study raises the possibility that interventions targeting fuel selection by making individuals more metabolically flexible to energy deficit may help to prevent or treat obesity through pathways involving food intake behavior.

What is already known about this subject?

• Higher 24-h respiratory exchange ratio (RER) during eucaloric conditions is associated with ad libitum food intake and weight gain.

• An impaired metabolic flexibility to high fat overfeeding as assessed by changes in 24-h RER is associated with gain in body weight and fat mass.

What are the new findings in your manuscript?

• Metabolic inflexibility to fasting assessed by changes in 24-h RER and carbohydrate oxidation is associated with higher ad libitum energy intake after fasting.

How might your results change the direction of research or the focus of clinical practice?

• Future research should explore the mechanisms for interindividual differences in metabolic inflexibility.

• Metabolic inflexibility to energy deficits may be a target to prevent or treat obesity through pathways involving food intake.

ABBREVIATIONS

CARBOX

carbohydrate oxidation

EE

energy expenditure

FFA

free fatty acid

FGF21

fibroblast growth factor 21

GLP-1

glucagon-like peptide-1

LIPOX

lipid oxidation

NIDDK

National Institute of Diabetes and Digestive and Kidney Diseases

PROTOX

protein oxidation

RER

respiratory exchange ratio

Data Sharing:

Data described in the manuscript, code book, and analytic code will be made available upon request pending application and approval.