Reduced brown adipose tissue activity during cold exposure is a Metabolic feature of the human thrifty phenotype

Abstract

Background

We recently demonstrated that thrifty subjects, characterized by a greater decrease in 24h energy expenditure (24hEE) during short-term fasting, have less capacity for cold-induced thermogenesis (CIT) during 24h of mild cold exposure.

Objective

As cold-induced brown adipose tissue activation (CIBA) is a determinant of CIT, we sought to investigate whether thrifty individuals also have reduced CIBA.

Methods

Twenty-four healthy subjects (age: 29.8±9.5y, body fat: 27.3±12.4%, 63% male) were admitted to our clinical research unit and underwent two 24hEE assessments in a whole-room indirect calorimeter during energy balance and fasting conditions at thermoneutrality to quantify their degree of thriftiness. Positron emission tomography/computed tomography scans were performed after exposure to 16°C for 2 hours to quantify peak CIBA.

Results

A greater decrease in 24hEE during fasting was associated with lower peak CIBA (r=0.50, p=0.01), such that a 100 kcal/day greater reduction in 24hEE related to an average 3.2 g/mL lower peak CIBA.

Conclusion

Our results indicate that reduced CIBA is a metabolic trait of the thrifty phenotype which might explain reduced CIT capacity and greater predisposition toward weight gain in individuals with a thrifty metabolism.

Introduction

The human thrifty phenotype hypothesis, which originally emerged from a concept that ascribes the high prevalence of type 2 diabetes in western societies to thrifty genes[1, 2], presupposes that lower 24-hour energy expenditure (24hEE) during famine preserves body mass and promotes survival. Our group showed in different clinical trials that individuals with a thriftier metabolism (defined by a greater decrease in 24hEE during acute fasting conditions) lose less weight during highly-controlled sustained caloric restriction[3] while gaining more weight during prolonged overfeeding[4] and in free-living conditions[5].

The underlying biological mediators of thriftiness are unclear. One factor might be the activity of brown adipose tissue (BAT), a putative contributor to whole-body EE[6-8]. In contrast to white adipose tissue, the main function of BAT is not energy storage but heat generation through energy dissipation to maintain core body temperature during cold exposure[9]. Active BAT is present in adult humans and its activity is impaired in individuals with obesity, suggesting that it might play a role in thriftiness and weight gain susceptibility[6, 10]. BAT activity is mainly stimulated via cold exposure[6, 11, 12]. Greater cold-induced BAT activation (CIBA) has been associated with greater cold-induced thermogenesis (CIT) in humans[13-17] although some studies argue that BAT does not have enough mass to contribute meaningfully to EE[18, 19].

We recently demonstrated that thriftier individuals show reduced CIT during 24h of mild cold exposure at 19°C[20], thus we hypothesized that these individuals would also have less CIBA compared to more spendthrift individuals. To test our hypothesis, we assessed the relationship between the decrease in 24hEE during 24h of fasting (as a reliable, quantitative measure of thriftiness) and CIBA obtained from a ¹⁸F-fluorodeoxglucose (18F-FDG) positron emission tomography/computed tomography (PET-CT) scan after 2 hours of 16°C mild cold exposure in 24 healthy individuals.

Material and Methods

Subjects

The analysis was performed using data from an inpatient study (clinicaltrial.gov identifier NCT00523627) aimed to identify associations between BAT activity during mild cold exposure and short-term EE responses to 24-h fasting and overfeeding. From 2009 to 2016, 24 individuals aged 18-51 years completed the study, had valid EE assessments during 24-h energy balance and fasting conditions, and underwent PET-CT scans to quantify cold-induced brown adipose tissue activation (CIBA) (Figure 1). All participants signed a written informed consent prior to admission. The study was approved by the Institutional Review Board of the National Institute of Diabetes and Digestive and Kidney Diseases at the National Institutes of Health.

Figure 1. Study flow chart.

From 2009 to 2016, a total of 24 individuals aged 18-51 years completed the clinical trial and had a valid 24-h energy expenditure assessment both during energy balance and fasting conditions and also participated in the PET-CT substudy which assessed cold-induced brown adipose tissue activation.

Study volunteers were weight stable (less than 10% variation in body weight) for at least 6 months before admission and were absent of medical diseases based on medical history, physical examination, and fasting blood tests. Upon admission to the clinical research unit, individuals were fed a weight maintaining diet (WMD; 50% carbohydrate, 30% fat, and 20% protein) using unit-specific equations based on gender and weight[21]. On the second day of admission, dualenergy X-ray absorptiometry (DXA) (DPX-1, Lunar Corp, Madison, Wisconsin, USA) was used to assess body composition with fat mass (FM) and fat free mass (FFM) were calculated from the percentage body fat (%fat) from the DXA scan. As a final inclusion criterion, an OGTT was performed after 3 days of WMD and only subjects with normal glucose regulation continued the study[22]. Body weight was recorded daily and was maintained within 1% of the admission weight by adjusting the energy intake of the WMD before any dietary interventions were implemented[23]. The WMD was consumed on all days when 24hEE was not measured.

Energy expenditure measurements

The assessment of 24hEE and respiratory quotient (RQ) (an index of substrate oxidation) inside the whole-room indirect calorimeter was performed as previously described[24]. Briefly, after an overnight fast and following breakfast (7:00AM) subjects entered the calorimeter around 8:00AM. The air temperature was maintained constant at 23.6±1.3°C by an air conditioning system and volunteers remained in the calorimeter for 23.25 hours, during which the average CO₂ production and O₂ consumption per minute were calculated and the 24hEE was derived using the Lusk equation [25]. The 24-hour RQ was calculated as the ratio of 24-hour CO₂ production to 24-hour O₂ consumption. Carbohydrate and lipid oxidation rates were derived from the 24-h RQ after accounting for protein oxidation, which was estimated from measurement of 24-h urinary nitrogen excretion[26].

To closely achieve the state of energy balance during the 24hEE assessment in the whole-room calorimeter, 24hEE was measured twice during energy balance conditions. The first assessment was obtained while subjects resided for 24h in the calorimeter with total energy intake calculated using a unit-specific formula to achieve 24h energy balance in the confined environment of the calorimeter[24]. Secondly, after two days, subjects had the second assessment inside the calorimeter when the total energy intake was equal to the measured 24hEE value obtained from the first 24h assessment, for precise determination of 24EE during energy balance inside the calorimeter (24h energy balance at first 24hEE assessment=145±220 kcal/day, or 8.2±11.3%, 24h energy balance at second 24hEE assessment=9±80 kcal/day, or 0.7±3.9%). Following the assessment of energy balance and after at least 3 days on the WMD, 24hEE was assessed inside the calorimeter during 24h of fasting when no food was provided, and subjects were instructed to keep themselves hydrated by drinking water and/or no-calorie no-caffeine beverages.

PET-CT Imaging

Study participants underwent an ¹⁸F-fluorodeoxglucose (18F-FDG) PET-CT scan after a day on a WMD and following an overnight fast. Prior to the scan, volunteers were exposed to 16°C ambient temperature for 2 hours in the morning while wearing standardized clothing (shorts, t-shirt and flip-flops, ~0.3 clo) inside the calorimeter. All possible measures were taken to avoid shivering and, if shivering occurred (by self-report), the volunteer was temporarily removed from the calorimeter for 5 minutes.

PET-CT scan was done 1 hour after injection of ¹⁸F-FDG into the antecubital vein (mean dose 14.7±0.3 mCi). PET and CT images by the Reveal 16 High Rez (CTI Molecular Imaging, Knoxville, TN) were reconstructed into image voxels of 1.95 mm × 1.95 mm × 4.00 mm (PET) and of 0.98 mm × 0.98 mm × 3.75 mm (CT) and uploaded into ImageJ[27] for image processing. The field of view was head-to-hip. The PET-CT Viewer plug-in with features customized for BAT quantification was used in each of the subsequent analyses with specific CT density ranges used to identify fat (−190 to −10 Hounsfield units (HU)) from air and other tissues[28]. The ¹⁸F-FDG uptake (g/mL) in each PET image voxel was quantified as an SUV initially normalized to the individual’s lean body mass to total body mass ratio, i.e. the BAT SUV lower threshold was calculated based on the individual’s percentage of lean body mass (1.2 g/mL / FFM / body weight) according to BARCIST guidelines [28]. The BAT SUV upper threshold was set to 100 g/mL[29]. We applied ROIs to quantify whole-body BAT based on instructions by Kim et al.[29]. Briefly, one ROI was created on each axial slice, avoiding regions that were not metabolically active fat, to minimize false positive detection. ROI selection began at the slice corresponding to vertebra C3 and continued inferiorly until the beginning of the abdominal region (around T12) was reached. For the regional BAT quantification, we segmented the detected BAT voxels in each subject into previously defined depots[29]. The BAT voxels of all depots were summed to calculate total body BAT volume. We then calculated two measures of CIBA according to BARCIST 1.0 criteria[28]: mean CIBA was calculated as the average SUV derived from all depots with active BAT (i.e., above the predefined SUV lower threshold) while peak CIBA was determined as the highest average SUV in a 1 mL spherical volume of whole-body BAT. The quantification of BAT by analysis of PET-CT scans was first obtained by one observer (T.H.) and then blindly confirmed by an independent observer (K.L.V): BAT volume: CV=4.7±13.3%, mean SUV: CV=0.7±1.1%, peak SUV: CV=0%)

Analytical measurements

Blood for the measurement of total cholesterol, HDL-C, triglycerides, uric acid, GGT, ALT, and AST was collected in the morning of admission after an overnight fast and analyzed at the same day in a local laboratory (Phoenix Indian Medical Center, Phoenix, AZ).

Statistical analysis

The primary aim of this study was to assess the relationship between fasting – energy balance 24hEE (as a quantitative measure of metabolic thriftiness) and peak SUV (as a quantitative measure of CIBA) after 2-h mild cold exposure at 16°C. Power calculations performed prior to data analyses found that a sample size of 24 subjects achieved a power>0.80 (2-sided alpha=0.05) to detect a minimum expected correlation of r=0.54 between changes in 24hEE during acute fasting and CIBA. We further performed secondary (exploratory) analyses to assess the relationship between CIBA and metabolic flexibility during short-term fasting (i.e. changes in RQ, lipid oxidation rate, and FFA concentrations), which further characterizes the human metabolic phenotype[30].

Statistical analysis was performed using the SAS statistical software package (SAS Enterprise Guide Version 7.15; SAS Institute, Cary, NC). Unless otherwise specified, data were expressed as mean±SD or mean with 95% confidence interval (CI). A p-value <0.05 was considered statistically significant. Low-density lipoprotein cholesterol (LDL-C) was calculated based on the Friedewald formula[31].

Multivariate linear regression analysis was used to calculate adjusted values of 24hEE and its components controlling for FFM and FM as covariates. Specifically, the residual values (observed minus predicted values) of 24hEE and its components obtained from regression models were calculated and adjusted values were then derived by adding the average unadjusted 24hEE value calculated in the whole cohort to the residual values obtained by regression analysis. The Pearson’s correlation coefficient was used to quantify associations between continuous variables (e.g., adjusted EE values and CIBA measures). Spearman nonparametric correlation analyses were also performed as sensitivity analyses to account for influential cases that may have inflated Pearson’s correlation values.

For graphical purposes and ease of interpretation, the aforementioned analyses using continuous EE data were followed up with confirmatory analyses using distinct metabolic groups. In these confirmatory analyses, subjects were arbitrarily classified as thrifty/spendthrift based on a lower/higher-than-median decrease in 24hEE during 24-h fasting from energy balance conditions, respectively, as done previously[3, 32]. Paired t-test was then used to evaluate the difference (Δ) in EE measures between fasting vs. energy balance conditions while unpaired t-test was used to compare EE and BAT measurements between metabolic groups.

Results

Clinical characteristics for the study group including metabolic measurements during energy balance and 24-h fasting conditions at low level of energy turnover are presented in Table 1. In this study, we compared two equally-sized metabolic groups (thrifty/spendthrift) identified by the median value of the decrease in 24hEE during fasting, which defines the extent of thriftiness in a quantitative (continuous) fashion[33]. By using the median value as cut-off for the extent of thriftiness, the thrifty group had a greater fasting-induced 24hEE decrease by 146 kcal/day (CI: −92, −200, p<0.0001) compared to spendthrift subjects. As shown previously[33], the greater decrease in 24hEE in thrifty subjects was due to higher 24hEE during energy balance conditions after adjustment for FFM and FM (2175±511 [thrifty] vs. 1747±371 kcal/day [spendthrift], p=0.03) as the adjusted 24hEE was similar between groups during fasting conditions (p=0.37) (Supplemental Figure 1 supplemental material can be found at [34]). There were no further demographic, anthropometric, or metabolic differences between groups, except for differences in CIBA which are discussed below in detail.

Table 1.

Clinical characteristics of the study cohort.

	TOTAL n = 24	THRIFTY n = 12	SPENDTHRIFT n = 12	p Value
Male (%)	15 (62.5)	8 (66.7)	7 (58.3)	0.69
Ethnicity, n (%)				0.30
African-American	8 (33.3)	3 (25.0)	5 (41.7)
Caucasian	6 (25.0)	2 (16.7)	4 (33.3)
Native American	8 (33.3)	2 (16.7)	0 (0.0)
Hispanic	2 (8.3)	5 (41.7)	3 (25.0)
Age (years)	29.8 ± 9.5 (18.2, 51.4)	31.1 ± 9.6 (19.6, 51.4)	28.5 ± 9.7 (18.2, 47.3)	0.52
Height (cm)	170.7 ± 7.6 (156.5, 186.5)	170.7 ± 7.4 (157.5, 182)	170.7 ± 8.1 (156.5, 186.5)	0.98
Body composition measures
Body weight (kg)	75.3 ± 17 (47.5, 107.5)	77.5 ± 17 (54.1, 103.5)	73.1 ± 17.5 (47.5, 107.5)	0.54
BMI (kg/m2)	25.8 ± 5.5 (17.7, 39.2)	26.5 ± 5 (18.3, 33.4)	25.1 ± 6 (17.7, 39.2)	0.54
Body fat (%)	27.3 ± 12.4 (6.9, 53.8)	27.1 ± 9.5 (6.9, 40.3)	27.6 ± 15.1 (7.3, 53.8)	0.93
FM (kg)	21.5 ± 13.5 (4.9, 54.3)	21.4 ± 10 (5.9, 34)	21.7 ± 16.7 (4.9, 54.3)	0.96
FFM (kg)	53.7 ± 11.5 (33.9, 79.4)	56.1 ± 12.7 (39.5, 79.4)	51.4 ± 10.1 (33.9, 65.9)	0.33
Blood tests at baseline
Total cholesterol (mg/dL)	154.9 ± 33.3 (99, 259)	150.3 ± 27.6 (107, 193)	159.4 ± 38.9 (99, 259)	0.52
LDL-C, calculated (mg/dL)	85 ± 35 (30, 194)	79 ± 31 (30, 142)	92 ± 40 (31, 194)	0.38
HDL-C (mg/dL)	56 ± 12.7 (31, 82)	56.6 ± 14.1 (31, 82)	55.3 ± 11.7 (34, 75)	0.82
Triglycerides (mg/dL)	68.5 ± 44.5 (26, 189)	75.3 ± 46.3 (26, 166)	61.7 ± 43.5 (26, 189)	0.46
Uric acid (mg/dL)	4.5 ± 1.5 (2.6, 8.1)	4.6 ± 1.5 (2.6, 8.1)	4.5 ± 1.6 (2.6, 7.1)	0.80
GGT (U/L)	24.2 ± 14.8 (3, 67)	22.6 ± 12.3 (5, 51)	25.8 ± 17.4 (3, 67)	0.60
ALT (U/L)	39.5 ± 14.8 (16, 84)	39.4 ± 13 (20, 61)	39.5 ± 17 (16, 84)	0.99
AST (U/L)	21.8 ± 8 (11, 41)	21.5 ± 7.6 (12, 41)	22.1 ± 8.8 (11, 38)	0.86
Energy expenditure measures during 24h energy balance
RQ (ratio)	0.86 ± 0.03 (0.77, 0.91)	0.85 ± 0.02 (0.82, 0.9)	0.87 ± 0.04 (0.77, 0.91)	0.32
24hEE (kcal/day)	1961 ± 358 (1427, 2732)	2098 ± 404 (1427, 2732)	1825 ± 255 (1456, 2358)	0.06
Adjusted 24hEE (kcal/day)	1961 ± 488 (1064, 3033)	2175 ± 511 (1306, 3033)	1747 ± 371 (1064, 2490)	0.03
SPA (% of time)	5.1 ± 3.6 (0.6, 15.2)	6 ± 4.5 (0.6, 15.2)	4.3 ± 2.4 (0.7, 8.3)	0.28
24h Energy intake (kcal/day)	1970 ± 341 (1512, 2709)	2088 ± 392 (1512, 2709)	1853 ± 244 (1529, 2335)	0.09
Energy balance (%)	0.7 ± 3.9 (−10.5, 7.2)	−0.2 ± 4.3 (−10.5, 6)	1.7 ± 3.5 (−2.9, 7.2)	0.24
Energy expenditure measures during 24h fasting
RQ (ratio)	0.79 ± 0.03 (0.74, 0.88)	0.79 ± 0.03 (0.75, 0.84)	0.79 ± 0.03 (0.74, 0.88)	0.74
24hEE (kcal/day)	1809 ± 300 (1287, 2390)	1872 ± 352 (1287, 2390)	1746 ± 235 (1426, 2233)	0.31
Adjusted 24hEE (kcal/day)	1809 ± 405 (1110, 2579)	1885 ± 458 (1110, 2579)	1733 ± 346 (1148, 2424)	0.37
SPA (% of time)	4.6 ± 3 (0.4, 10.9)	5.3 ± 3.4 (0.4, 10.9)	3.9 ± 2.4 (0.8, 9.3)	0.25
Δ 24EE during fasting (kcal/day)	−152 ± 97 (−405, −18)	−225 ± 80 (−405, −140)	−79 ± 42 (−128, −18)	<0.0001

Categorial data are presented as absolute number with percentage in parentheses. Continuous data are presented as mean ± SD with minimum and maximum in parentheses. Classification of subjects into the spendthrift or thrifty metabolic group was based on the median value of the decrease in 24hEE from energy balance during fasting (−134 kcal/day), such that thrifty subjects were those with a greater–than-median decrease in 24hEE. Significance between thrifty and spendthrift metabolic groups was determined by unpaired Student’s t-test for numerical data and by χ² test for categorical data.

Multivariate regression analysis was used to calculate adjusted values of 24hEE controlling for FFM and FM. The residual values (observed minus predicted values) of 24hEE obtained from these regression models were considered as the unexplained variability in 24hEE and were added to the average 24hEE value calculated in the whole cohort.

LDL-C was calculated based on the Friedewald formula[31].

24hEE, 24-h energy expenditure; ALT, alanine aminotransferase; AST, aspartate transaminase; BMI, body mass index; FM, fat mass; FFM, fat-free mass; GGT, gamma-glutamyl transferase; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; RQ, respiratory quotient; SPA, spontaneous physical activity.

Associations between BAT measures and the short-term change in 24hEE during acute fasting

There were no associations between peak CIBA and adjusted 24hEE neither during energy balance nor during fasting conditions (both p≥0.09, Figure 2A). In Figure 2A, the fasting-induced decrease in adjusted 24hEE in each individual is denoted with red arrows (longer arrows indicate a greater decrease in adjusted 24hEE during fasting compared to energy balance conditions). Individuals with a greater decrease in adjusted 24hEE during fasting (i.e. longer arrows ≙ thriftier) were more prevalent on the left side of the graph, indicating an association between peak CIBA and the fasting-induced decrease in 24hEE. This was confirmed by the positive correlation between these two quantitative variables (r=0.50, p=0.01, Figure 2B), that is, a greater decrease in 24hEE by 100 kcal/day during 24h fasting was associated with a lower peak CIBA by an average of 3.2 g/mL (CI: −5.6 to −0.8). Similar results were obtained when considering both the percentage change in 24hEE (r=0.52, p=0.009, Supplemental Figure 2A[34]) and the change in adjusted 24hEE (r=0.41, p=0.05, Supplemental Figure 2B[34]) during 24h fasting. There was no seasonal effect on peak CIBA (p=0.21). Additional adjustment for age and sex did not change the results (partial r=0.47, p=0.03). The clinical characteristics of thrifty and spendthrift groups stratified by sex are shown in Supplemental Table 1. Adjustment for percentage deviation from energy balance at baseline did not change the results (partial r=0.47, p=0.03). Similar results were also obtained when considering mean CIBA SUV while BAT volume was not associated with the decrease in 24hEE during 24h fasting (data not shown). There were no associations between spontaneous physical activity (SPA) and either peak CIBA or the decrease in 24hEE from energy balance during 24-h fasting (both p>0.17).

Figure 2. Greater decrease in 24hEE during 24-h fasting at thermoneutrality (thrifty phenotype) is associated with lower peak CIBA.

(A) Association between peak CIBA and adjusted 24hEE during 24-h energy balance (black dots) and fasting (white dots). The fasting-induced decrease in adjusted 24hEE in each individual is denoted with red arrows (longer arrows indicate a greater decrease in 24hEE while shorter arrows indicate a smaller decrease in 24hEE). (B) Positive relationship between peak CIBA and the decrease in 24hEE during fasting (similar results were obtained using the percentage decrease in 24hEE and the decrease in adjusted 24hEE during fasting, see Supplemental Figure 2). (C) Comparison of peak CIBA between thrifty vs. spendthrift individuals.

Individual changes (Δ) in 24hEE were calculated as the difference in 24hEE values between the fasting condition minus the energy balance condition. The Pearson’s correlation coefficient was used to quantify associations between continuous variables. Spearman nonparametric correlation analyses were also performed as sensitivity analyses to account for influential cases that may have inflated Pearson correlation values and similar results were obtained (panel B: Spearman rho=0.42, p=0.04).

Individuals were categorized as thrifty or spendthrift based on the median value (−134 kcal/day) of the difference in 24hEE between energy balance and fasting conditions. Unpaired t-test was used to evaluate between-group differences in energy expenditure measures.

Multivariate regression analysis was used to calculate adjusted values of 24hEE during both conditions controlling for fat-free mass and fat mass. The residual values (observed minus predicted values) of 24hEE obtained from these regression models were considered as the unexplained variability in 24hEE and were added to the average 24hEE value calculated in the whole cohort.

24hEE, 24-hour energy expenditure; CIBA, cold-induced brown adipose tissue activation.

For illustrative purposes and to confirm our results obtained using continuous 24hEE data, we also performed a group-wise comparison of thrifty vs. spendthrift subjects as arbitrarily defined by the median decrease (−134 kcal/day) in 24hEE from energy balance during 24-h fasting. In these confirmatory analyses, on average spendthrift individuals had a 64% higher peak CIBA (+5.7 g/mL, CI: +0.9, +10.4, p=0.02, Figure 2C; Table 2) compared to thrifty individuals. In sensitivity analyses, similar results were obtained when defining thrifty vs spendthrift individuals based on the cut-off of −162 kcal/day in 24hEE decrease during fasting as reported in a larger cohort[33].

Table 2.

Measures of cold-induced brown adipose tissue activation

	TOTAL n = 24	THRIFTY n = 12	SPENDTHRIFT n = 12	p Value
BAT volume (mL)	265 ± 128 (29, 553)	247 ± 121 (57, 404)	283 ± 138 (29, 553)	0.50
Mean CIBA (g/mL)	3.1 ± 0.9 (2, 6.2)	2.8 ± 0.4 (2, 3.2)	3.4 ± 1.1 (2, 6.2)	0.06
Peak CIBA (g/mL)	11.8 ± 6.2 (3.3, 28.7)	8.9 ± 2.8 (3.3, 13.6)	14.6 ± 7.4 (4.4, 28.7)	0.02
Max CIBA (g/mL)	15.3 ± 8.8 (3.9, 41.6)	11.2 ± 3.2 (3.9, 17.6)	19.5 ± 10.8 (5.2, 41.6)	0.02

Data are presented as mean ± SD with minimum and maximum in parentheses. Classification of subjects into the spendthrift or thrifty metabolic group was based on the median value of the decrease in 24hEE from energy balance during fasting (−134 kcal/day), such that thrifty subjects were those with a greater–than-median decrease in 24hEE. Significance between thrifty and spendthrift metabolic groups was determined by unpaired Student’s t-test for numerical data.

24hEE, 24-h energy expenditure; BAT, brown adipose tissue; CIBA, cold-induced brown adipose tissue activation.

Further, we visually compared the PET-CT images from the extremes of the continuous data distribution, namely, the thriftiest vs. the most spendthrift subject based on the largest/smallest decrease in 24hEE during 24h fasting (Figure 3). It is visible that the thriftiest subject, who decreased 24hEE by 405 kcal/day during short-term fasting (Figure 3A), had a considerably lower peak CIBA compared to the most spendthrift subject (7.3 vs. 18.9 g/mL), who decreased 24hEE by only 18 kcal/day (Figure 3B). Two more PET/CT images of the second thriftiest and most spendthrift individuals as determined by the ranks of the decrease in 24hEE during 24-h fasting are shown in Supplemental Figure 3.

Figure 3. Comparison of the ¹⁸F-FDG PET-CT images of the two subjects with the greatest (“thriftiest”) and smallest (“most spendthrift”) decrease in 24hEE during acute fasting.

(A) Volunteer with the greatest decrease in 24hEE during fasting (thriftiest individual) in which only little visualization of CIBA is seen in PET (top) and fused coronal PET-CT (bottom) images along with a low peak CIBA of 7.3 (g/mL). (B) Volunteer with the smallest decrease in 24hEE during fasting (most spendthrift individual) in which PET (top) and fused coronal PET-CT (bottom) images show a large area of CIBA along with a high peak CIBA of 18.9 (g/mL).

24hEE, 24-hour energy expenditure; BAT, brown adipose tissue; CIBA, cold-induced brown adipose tissue activation.

There were no associations between detectable BAT volume and adjusted 24hEE during energy balance and fasting conditions at thermoneutrality (both p≥0.63, Supplemental Figure 4A[34]). We also assessed whether the decrease in adjusted 24hEE from energy balance during 24-h fasting was associated with detectable BAT volume. In Supplemental Figure 4A[34], individuals with a greater decrease in 24hEE during fasting (i.e. longer arrows ≙ thriftier) were evenly distributed on the graph indicating no association between detectable BAT volume and the decrease in adjusted 24hEE which was statistically confirmed when quantifying the correlation between variables (p=0.45, Supplemental Figure 4B[34]). In accordance with the results obtained using continuous data, there was no difference in detectable BAT volume between thrifty and spendthrift individuals (p=0.50, Table 2; Supplemental Figure 4C[34]).

Associations between BAT measures and metabolic flexibility during 24h fasting

We also performed exploratory analyses to determine whether changes in macronutrient preference during 24h fasting (another independent trait of metabolic phenotype) were associated with CIBA and BAT volume. A higher peak CIBA was related to greater decrease in RQ (r=−0.45, p=0.03, Figure 4A) and greater increase in lipid oxidation rate (r=0.47, p=0.02, Figure 4B) during 24h fasting. Similar results were obtained after adjustment of lipid oxidation rate for FM and FFM (r=0.44, p=0.03). A higher peak CIBA was also associated with greater increase in FFA during fasting (r=0.61, p=0.003, Figure 4C) and, accordingly, a higher FFA concentration following 24-h fasting (r=0.63, 0.002). Similar results were obtained when analyzing mean CIBA (data not shown). Conversely, detectable BAT volume was not associated with either the decrease in RQ or the increase in lipid oxidation and FFAs during/after 24h fasting (all p≥0.11, Supplemental Figures 5A-C[34]).

Figure 4. Lower peak CIBA is associated with reduced metabolic flexibility during 24h fasting.

Associations between peak CIBA and (A) the decrease in 24-h RQ, (B) the increase in lipid oxidation rate, and (C) the increase in plasma FFA concentration during/after 24h fasting. The Pearson’s correlation coefficient was used to quantify associations between continuous variables.

CIBA, cold-induced brown adipose tissue activation; FFAs, free fatty acids; RQ, respiratory quotient.

Determinants of CIBA and BAT volume

We further analyzed determinants of CIBA and BAT volume. A higher BMI was associated with lower peak CIBA (r=−0.50, p=0.01, Figure 5A) while no correlations were found with %FAT (p=0.35, Figure 5B) and age (p=0.31). Lower peak CIBA was associated with lower HDL-C (r=0.44, p=0.03, Figure 6C), higher triglycerides (r=−0.47, p=0.02, Figure 6D), and higher uric acid (r=−0.47, p=0.02, Figure 6E) but not with total cholesterol and LDL-C (both p>0.4, Figures 6A and 6B). The association between peak CIBA and HDL-C was still observed when controlling for PFAT, age, and sex in separate partial correlation analyses (all p≤0.05). Furthermore, lower peak CIBA was associated with higher GGT (r=−0.47, p=0.02, Figure 6F), ALT (r=−0.50, p=0.01, Figure 6G), and, to a lesser extent, AST (r=−0.37, p=0.07, Figure 6H). Similar results were obtained when considering mean CIBA (data not shown).

Figure 5. Higher BMI is associated with lower peak CIBA.

Associations between (A) BMI and (B) percentage body fat and peak CIBA. The Pearson’s correlation coefficient was used to quantify associations between continuous variables.

BMI, body mass index; CIBA, cold-induced brown adipose tissue activation.

Figure 6. Higher serum lipids and liver enzymes are associated with lower peak CIBA.

Associations between (A) total cholesterol, (B) LDL-C, (C) HDL-C, (D) serum triglycerides, (E) uric acid, (F) GGT, (G) ALT, and (H) AST and peak CIBA at baseline. The Pearson’s correlation coefficient was used to quantify associations between continuous variables.

ALT, alanine aminotransferase; AST, aspartate transaminase; CIBA, cold-induced brown adipose tissue activation; GGT, gamma-glutamyl transferase; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol.

A lower BAT volume was associated with higher BMI (r=−0.39, p=0.06, Supplemental Figure 6A[34]), %FAT (r=−0.43, p=0.04, Supplemental Figure 6B[34]), and ALT (r=−0.43, p=0.04, Supplemental Figure 7G[34]) whereas no associations were found for age (p=0.42), lipids, and uric acid (all p≥0.10, Supplemental Figures 7A-E[34]), and GGT and AST (both p≥0.18, Supplemental Figures 7F, 7H[34]). Brown adipose tissue volume was positively associated with peak CIBA (r=0.58, p=0.003, Supplemental Figure 8[34]).

Discussion

In 24 healthy individuals, we demonstrated that a thrifty metabolic phenotype, defined as a greater relative decrease in 24hEE during 24h fasting at thermoneutrality, was associated with less cold-induced brown adipose tissue activation (CIBA) after 2h of 16°C mild cold exposure. Furthermore, in exploratory analyses, we found that greater CIBA was related to a greater fasting-induced shift in macronutrient preference towards fats, reflected by a greater decrease in RQ and greater increases in lipid oxidation rate and circulating FFAs during short-term fasting conditions.

The thrifty phenotype is characterized by impaired cold-induced BAT activation

Brown adipose tissue is a metabolically active organ aiming to protect core body temperature during cold exposure through heat production. This process is mainly mediated by the high expression of tissue-specific uncoupling protein 1 (UCP1). While relatively large BAT depots are detectable in infants and young children, BAT mass declines throughout life reaching very low levels compared to white adipose tissue, usually ranging from 0-600g depending on sex, ethnicity, and body size[7, 10, 35, 36]. BAT is primarily localized in the cervical-supraclavicular region[10] although it is also detectable in the axillary, mediastinal, paraspinal, and abdominal regions[37, 38]. It is still unclear how much energy active human BAT can expend[39, 40]. Early research suggested that as little as 50g of highly active BAT can theoretically account for up to 20% of 24hEE in an adult human[41]; however, studies using ¹⁵O-labeled water to detect active BAT report much lower values of ~30 kcal/day[18, 19]. Further, BAT thermogenesis seems to depend on functional creatine uptake into BAT[42].

Previous research indicates that functional BAT is protective against diet-induced obesity in rodents[43-45]. In humans, BAT activity is 4-fold higher in lean compared to overweight individuals[6] and reduced CIBA has been associated with greater FM gain at follow-up [30, 46], supporting a role of BAT in determining the individual susceptibility to weight gain, which may not only be due to BAT energy consumption but also to “batokine” signaling[47, 48].

We recently demonstrated the existence of a thrifty phenotype (defined by a greater decrease in 24hEE during short-term fasting) which is more susceptible to weight gain during 6-week controlled low-protein overfeeding[4] and in free-living conditions[5], and more resistant to weight loss during 6-week controlled caloric restriction[3] – possibly through increased metabolic efficiency and energy conservation in situations of energy deprivation and energy excess[49]. We recently reported that thriftier individuals have less capacity for cold-induced thermogenesis (CIT) during 24h of mild cold exposure at 19°C[20], indicating that thermogenic mechanisms underlying the metabolic adaptation to short-term changes in energy intake and ambient temperature are mutual and subject-specific. In that previous study, individuals with less CIT capacity also had less increase in supraclavicular temperatures during cold exposure, which is indicative of less BAT activation[50]. However, in that previous study, we could not provide direct evidence that reduced BAT activity was responsible for lower CIT in thrifty individuals. Accordingly, in this present study we demonstrate that a greater decrease in 24hEE during 24h fasting (indicative of a thriftier phenotype) is associated with lower CIBA after 2h of mild cold exposure at 16°C, such that thrifty individuals have a ~40% lower peak CIBA compared to spendthrift individuals. These results extend the definition of this metabolic phenotype by including CIBA as a thermogenic mechanism to explain the reduced CIT capacity during 24h of mild cold exposure observed in thrifty individuals[20].

Importantly, BAT may influence weight gain susceptibility as it is not only activated during cold exposure but also by specific diets[18, 51, 52] and, as such, it may contribute to the adaptive component of diet-induced thermogenesis, i.e. adaptive thermogenesis, during feeding[41]. Based on these findings for feeding-induced BAT activation and assuming that the degree of BAT activation is similar in response to different “stressors” (i.e. cold and feeding), we might speculate that thrifty individuals may also experience a reduced BAT activation in response to excess energy intake, which could entail blunted overfeeding-induced adaptive thermogenesis and less energy dissipation in a setting of positive energy balance, ultimately explaining their greater weight gain susceptibility in the current obesogenic environment[49]. Furthermore, lower periprandial BAT activation has been linked to less postprandial satiety in rodents[52] providing another possible explanation for greater weight gain susceptibility of thrifty individuals, although the association between periprandial BAT activity and satiety still has to be proven in humans. Although SPA in the whole-room calorimeter did not correlate with peak CIBA and the decrease in 24hEE during 24-h fasting, we cannot rule out that physical activity levels in free-living conditions are a determinant of CIBA as shown by Dinas et al[53].

Our findings are independent from sex and age which is relevant since previous studies found that women and younger individuals possess more active BAT than men and older individuals[10, 54].

In this present study, higher BAT activity was also associated with improved lipid homeostasis, e.g., lower triglycerides and higher HDL-C levels, and with reduced markers of liver injury, as also shown in previous studies[8, 55, 56].

Exploratory analyses: Less cold-induced BAT activation is associated with metabolic inflexibility during acute fasting

In addition to the extent of metabolic adaptation to 24h fasting quantified by the changes in 24hEE, the shift in macronutrient preference for oxidation in response to acute dietary challenges with different macronutrient composition (i.e., metabolic flexibility[57-60]) constitute a further independent metabolic feature characterizing the susceptibility to future weight gain. Fuel preference is generally quantified by the respiratory quotient (RQ), that is, a lower RQ indicates relatively greater fat oxidation while a higher RQ indicates greater reliance on carbohydrate oxidation[61].

In our previous study, impaired metabolic flexibility during acute (24h) high-fat (60%) overfeeding, i.e. small decrease in RQ and reduced fat oxidation rate, was associated with greater weight gain after 1 year of follow-up[61]. Similarly, other studies found that a lower rate of nocturnal fat oxidation during 3-day overfeeding (indicative of impaired metabolic flexibility to fats) was associated with greater long-term weight gain[62, 63].

In this present study, we performed exploratory analyses to assess the association between metabolic flexibility during acute fasting and CIBA. Our current results consistently demonstrate that higher CIBA was associated with a greater decrease in RQ and a greater increase in lipid oxidation rate during 24h fasting, indicating a greater shift towards fat utilization (i.e. greater metabolic flexibility). These findings for macronutrient oxidation are supported by measurements of circulating metabolites, showing that higher CIBA was correlated with a greater fasting-induced increase in FFAs and a higher FFA concentration following 24h fasting. The ability to switch faster to fat oxidation during fasting, i.e. greater metabolic flexibility, may thus be a feature of individuals with higher CIBA. As BAT mainly relies on FFAs as energy source[64, 65], we hypothesize that the individual capability of utilizing fat as substrate for oxidation is consistent across different conditions, namely, during cold exposure to provide substrate for BAT as well as during prolonged fasting to compensate the lack of feeding. One might speculate that this preference for fat as fuel source also persists during conditions of overfeeding.

The sympathetic nervous system might play an important role in explaining the individual capacity to utilize fat in different conditions (cold exposure and acute fasting) as it controls BAT activity during cold exposure[66] and is likely involved in greater FFA release from white adipocytes during fasting conditions[67]. In conclusion, our exploratory findings indicate that metabolically inflexible individuals may be at increased risk for weight gain in part due to less capacity for BAT activation.

Limitations

Our study has limitations. First, our sample size was relatively small but is comparable with many previous studies studying the metabolic characteristics of BAT[6, 8, 10, 39, 68]. Second, although we did not measure FFA concentrations during or after cold exposure, the strong correlation between CIBA and fasting-induced FFA-release suggests that individuals with less/greater potential for BAT activation during cold exposure could be identified based on their FFA-response during fasting. Third, our current study cannot infer about causality between BAT and the decrease in 24hEE during fasting due to its cross-sectional nature. Lastly, we acknowledge that 24hEE was measured in conditions of reduced physical activity level inside the confined environment of a metabolic chamber, which may have influenced our results. However, we assessed SPA inside the whole-room calorimeter which is an indicator of free-living physical activity[69]. We did not have direct measures of physical activity levels in free-living conditions, nor did we assess muscle tone or physical fitness in our inpatient study, the latter being a potential determinant of BAT activity[53] and the thrifty phenotype as higher 24hEE during energy balance is associated with greater physical fitness[70].

Conclusion

Cold-induced BAT activation and fasting-induced variations in energy metabolism (changes in 24hEE and RQ) are related within an individual. That is, less BAT activation upon cold exposure (reflected by reduced CIT) is associated with decreased thermogenesis during fasting (thrifty phenotype). Therefore, cold-induced BAT activation may reflect the underlying metabolic phenotype (thrifty vs. spendthrift) informative of human predisposition to obesity. Lower CIBA may translate into other settings where BAT can be activated (i.e. food intake), resulting in reduced adaptive thermogenesis and further explaining the greater susceptibility to weight gain of thrifty individuals. Our findings have potential clinical implications as BAT is emerging to be a target for clinical interventions aimed to increase EE and to ultimately induce weight loss. Specifically, it might be speculated that thrifty individuals with reduced CIT may benefit from therapies that induce “browning” of white adipose tissue, thus increasing CIT and favoring weight loss in these individuals.

Highlights.

• Thrifty subjects are defined by a greater decrease in metabolic rate during fasting

• Thriftiness is associated with reduced cold-induced brown fat activation

• Less cold-induced brown fat activation is associated with less lipid oxidation

• Cold-induced brown fat activation may reflect the susceptibility for weight gain

Abbreviations:

18F-FDG

¹⁸F-fluorodeoxglucose

24hEE

24-h energy expenditure

ALT

alanine aminotransferase

AST

aspartate transaminase

BMI

body mass index

CIBA

cold-induced brown adipose tissue activation

CIT

cold-induced thermogenesis

FFA

free fatty acids

FM

fat mass

FFM

fat-free mass

GGT

gamma-glutamyl transferase

HDL-C

high-density lipoprotein cholesterol

LDL-C

low-density lipoprotein cholesterol

PET-CT

positron emission tomography/computed tomography

ROI

region of interest

RQ

respiratory quotient

SPA

spontaneous physical activity

SUV

standardized uptake value