145 young, sedentary adults were randomly assigned to either: (i) a control group (no exercise, hereinafter CON), (ii) a moderate-intensity exercise group (hereinafter MOD-EX), or (iii) a vigorous-intensity exercise group (hereinafter VIG-EX). Both exercise groups performed 150 min/week of endurance training, distributed over 3–4 sessions/week, plus 80 min per week of resistance exercise over two sessions/week (Supplementary Fig. S1). The primary endpoints were BAT volume, 18F-FDG uptake and mean radiodensity, as measured by a static 18F-FDG positron emission tomography/computed tomography (PET/CT) scan following 2 h of personalized cold exposure. The secondary endpoints included body weight and composition, cardiometabolic risk factors (i.e., blood pressure, fasting serum glucose, insulin, triacylglycerols, cholesterol, liver enzymes and C-reactive protein), and physical fitness (i.e., muscular strength and cardiorespiratory capacity).

Study participants

Out of 1127 initially interested individuals, 523 attended one of 76 scheduled information meetings. 126 of the latter individuals declined to participate, and 397 completed a pre-screening survey. Among these, 296 were eligible for screening (Fig. 1). When contacted to schedule their screening visit, 102 declined to participate. The remaining 194 participants provided informed consent to be included in the study, and began baseline assessment; however, 49 did not finish this phase (39 declined to participate, 8 had an abnormal exercise electrocardiogram, and 2 met a medical exclusion criterion). The first participant was enrolled on October 5th, 2015 and the last participant was enrolled on November 7th, 2016. Thus, 145 individuals were finally randomized into the three experimental groups. During the 24-week intervention period, another 38 participants dropped out of the study; 107 participants therefore completed it. Among these, three participants from the MOD-EX group, and two participants from the VIG-EX group, were excluded from the main analyses for attending <70% of the training sessions. Additionally, five participants (MOD-EX; n = 2 and VIG-EX; n = 3) were excluded from the main analyses due to problems in collecting the primary outcome data (the cooling vest or chiller involved suffered technical problems during testing, or the 18F-FDG tracer was improperly injected). The baseline characteristics of the remaining 97 participants (65% women, CON; n = 35, MOD-EX; n = 31, and VIG-EX; n = 31) were similar across the three experimental groups (all P ≥ 0.059, Table 1 and Supplementary Table S1).

Fig. 1: Participant enrollment in the ACTIBATE study. BMI body mass index, CON control group, MOD-EX moderate-intensity exercise group, VIG-EX vigorous-intensity exercise group, ECG electrocardiogram. Full size image

Table 1 Baseline characteristics of the study participants Full size table

No evidence of brown adipose tissue volume, 18F-fluorodeoxyglucose uptake, or mean radiodensity changes after 24 weeks of supervised exercise training

Neither the MOD-EX nor VIG-EX regimen significantly altered BAT volume (Δ [post-baseline values]: Δ CON = −22.2 ± 52.6 mL (mean ± standard deviation); Δ MOD-EX = −15.5 ± 62.1 mL; Δ VIG-EX = −6.8 ± 66.4 mL; p = 0.771, Fig. 2a). Nor did they alter 18F-FDG uptake (standardized uptake value [SUV] mean: Δ CON = −0.4 ± 0.7; Δ MOD-EX = −0.4 ± 1.0; Δ VIG-EX = −0.5 ± 1.1; P = 1.000, Fig. 2b; SUVpeak: Δ CON = −2.6 ± 3.1; Δ MOD-EX = −1.2 ± 4.8; Δ VIG-EX = −2.2 ± 5.1: P = 0.476, Fig. 2c). In addition, neither BAT volume nor 18F-FDG uptake in different BAT depots (i.e., laterocervical, supraclavicular, mediastinum or paravertebral areas; all P > 0.1, Fig. 2e) were significantly modified. BAT mean radiodensity (a proxy of fat content30 also remained unchanged (Δ CON = −0.04 ± 6.91 Hounsfield Units [HU]; Δ MOD-EX = 2.2 ± 3.5 HU; Δ VIG-EX = −2.7 ± 10.7 HU; P = 0.441, Fig. 2d). We repeated the analyses using mixed-effects regression models and the absence of evidence of BAT activation after supervised exercise training persisted (Supplementary Fig. S2).

Fig. 2: Effect of the 24-week supervised exercise intervention on brown adipose tissue (BAT) volume, 18F-fluorodeoxyglucose (18F-FDG) uptake (Standardized uptake Value [SUV] mean and peak), and mean radiodensity. a Total BAT volume. b Total BAT SUVmean. c Total BAT SUVpeak. d Total BAT mean radiodensity (CON n = 25; MOD-EX n = 17; VIG-EX n = 16). E Regional BAT volume, SUVmean, and SUVpeak. SUV values are shown relative to lean mass. Δ was calculated as post-intervention minus baseline values for every outcome. P values are from analyses of covariance adjusting for baseline values (n = 97). In panel e, all P values ≥0.1. CON control group, HU Hounsfield Units, MOD-EX Moderate-intensity exercise group, VIG-EX Vigorous-intensity exercise group. Bars represent mean and standard deviation. Source data are provided as a Source Data file. Full size image

Previous clinical studies have shown that BAT volume and 18F-FDG uptake in response to mild-cold and thermoneutral exposures are different in women and men31,32, which was also confirmed in this cohort33. However, the absence of evidence of BAT activation after exercise training was recorded for both sexes when analyzed separately (all P > 0.7). Only participants who attended ≥70% of the training sessions were included in the main analyses (Supplementary Fig. S3), although the results did not change when sensitivity analyses were performed including participants whose attendance was <70% and ≥85%.

Following the recommendations of the Brown Adipose Reporting Criteria in Imaging STudies (BARCIST 1.034) report, BAT was defined as those imaging voxels with a radiodensity between −190 and −10 HU, and a SUV higher than an individualized threshold (1.2/[lean mass/body mass]). Both our research group and others have reported there to be inter-study heterogeneity regarding the criteria used for BAT quantification in 18F-FDG PET/CT scans (i.e., radiodensity range and SUV threshold)35,36,37. Our group also showed that the most frequently used criteria for measuring BAT provide non-comparable estimates of BAT volume and 18F-FDG uptake35. Hence, the use of different criteria might sometimes allow the effect of an intervention to be detected. Accordingly, in the present work, sensitivity analyses were performed using alternative criteria to quantify BAT (i.e., radiodensity: −250/−50 HU and SUV threshold >2.0; and radiodensity: −180/−10 HU and SUV threshold > 1.5). The lack of evidence of BAT activation persisted (all P > 0.5).

In agreement with the present results, a previous study29 showed that six weeks of moderate resistance training did not modify BAT 18F-FDG uptake, although a slight reduction in BAT volume was seen in the laterocervical region in 11 men. However, the latter study did not include a control group, limiting the conclusions that can be drawn. BAT volume and 18F-FDG uptake are strongly influenced by outdoor temperature, with activation greater in colder months38,39,40,41,42,43,44,45,46,47; therefore, not including a control group leaves it impossible to know if any changes in BAT volume and/or 18F-FDG uptake were the result of seasonal temperature variation or a consequence of the intervention. The present results clearly support this notion, since the Δ outdoor ambient temperature was negatively associated with Δ BAT volume, Δ 18F-FDG uptake and Δ BAT mean radiodensity (all R2 > 0.09 and P < 0.001, Fig. 3a–d) in all three groups. These correlations remained unaltered when BAT-related outcomes were calculated for individual BAT depots separately (all r ≤ −0.207 and P ≤ 0.043, Fig. 3e). At baseline, we found that outdoor ambient temperature at which the PET/CT scans were performed negatively correlated with BAT-related outcomes (all r ≥ −0.364 and P < 0.001, Supplementary Table S2), although these significant correlations disappeared when post-intervention values were used instead (Supplementary Table S2). After observing these correlations, we performed the main analyses adding the outdoor ambient temperature as covariate and the results did not change (all P ≥ 0.182; Supplementary Table S3). In addition, we found weak to moderate correlations (range from r = 0.515 to r = 0.872) between baseline and post-intervention BAT-related outcomes, which suggests that our protocol is reproducible despite the presence of large seasonal variation (Supplementary Table S4).

Fig. 3: Associations between the Δ outdoor ambient temperature and Δ (post-intervention minus baseline value) brown adipose tissue (BAT)-related outcomes. a Total BAT volume. b Total BAT standardized uptake value (SUV) mean. c Total BAT SUVpeak. d Total BAT mean radiodensity (CON n = 25; MOD-EX n = 17; VIG-EX n = 16). e Regional BAT volume, SUVmean, and SUVpeak. SUV values are shown relative to lean mass. P and β values are obtained from linear regression analyses. β non-standardized coefficients, BAT brown adipose tissue, CON Control group, HU Hounsfield units, MOD-EX moderate-intensity exercise group, R2 explained variance, SUV standardized uptake value, VIG-EX vigorous-intensity exercise group, WAT white adipose tissue. Full size image

The present BAT assessments were made after a personalized 2 h cold exposure, i.e., based on the participants’ shivering threshold - a proxy of cold tolerance16. While this method is widely used15,33,34 it has been suggested that it might not be entirely adequate. Shivering patterns are variable among individuals, making the detection of shivering onset difficult to be determined objectively48,49. Moreover, exercise has been reported to increase cold tolerance50, and could therefore alter the detection of shivering onset during post-intervention personalized cold exposure, introducing a bias regarding the effect of exercise on BAT recruitment/activation. In the present work, however, the shivering threshold (assessed via the cooling vest water temperature at the onset of shivering [see Methods]) was not modified in any of the groups after the intervention (P = 0.154, Supplementary Table S1).

The lack of evidence of BAT volume, 18F-FDG uptake or mean radiodensity modification after exercise training is in line with the results of cross-sectional studies that show no association of BAT-related outcomes with objectively measured free-living physical activity51 or muscular or cardiorespiratory fitness52. Case-control studies have revealed, however, that endurance-trained athletes show lower BAT volumes and 18F-FDG uptakes than their sedentary counterparts53,54. It is thus conceivable that the high volume of exercise training undertaken by endurance-trained athletes reduces BAT volume and 18F-FDG uptake as an adaptation to high activity energy expenditure levels, while the more moderate exercise training regimens implemented in the present study do not. This hypothesis concurs with the constrained total energy expenditure model, which suggests that low levels of physical activity do not inhibit energy-consuming physiological processes such as BAT thermogenesis55, whereas higher levels of physical activity do in order to maintain total daily energy expenditure within a homeostatic range56,57.

Although 18F-FDG PET/CT scanning is still the most widely used technique for the in vivo assessment of human BAT, the use of 18F-FDG has important limitations58. Cold exposure increases BAT 18F-FDG uptake, but the amount of 18F-FDG taken up does not necessarily mimic the BAT’s thermogenic activity, since most glucose taken up by BAT does not fuel mitochondrial oxidative capacity59,60. Accordingly, insulin infusions increase human BAT 18F-FDG uptake without increasing BAT perfusion, suggesting that the increase in 18F-FDG uptake is not paralleled by increased BAT thermogenic activity61. 18F-FDG is taken up by BAT through insulin-independent (glucose transporter 1, GLUT-1) and insulin-dependent (GLUT-4) mechanisms62. Consequently, BAT insulin sensitivity is likely to partially bias the assessment of BAT volume and activity, when performing 18F-FDG uptake quantification63. Any intervention that changes BAT insulin sensitivity might therefore result in a different 18F-FDG uptake, regardless of the change in BAT thermogenic activity. Indeed, exercise is known to increase insulin sensitivity in many peripheral tissues64, so it is plausible that it modifies BAT insulin sensitivity too. Moreover, exercise training increases mitochondrial oxidative capacity in skeletal muscle65, and it might likewise increase BAT oxidative metabolism rather than BAT 18F-FDG uptake. BAT oxidative metabolism, as well as the uptake of fatty acids in human BAT, has been recently quantified by 11C-acetate, 15O-oxygen and 18F-fluoro-thiaheptadecanoic acid (18FTHA) PET in response to cold exposure in humans48,66,67,68. Thus, it would be interesting to ascertain in future studies if assessing BAT metabolism by methods other than glucose uptake would unveil whether exercise can activate/recruit BAT.

Even if a better PET radiotracer were available, PET/CT imaging would still be affected by important limitations when evaluating the effect of exercise on BAT recruitment/activation. It has been suggested that the human BAT detectable by PET/CT scanning in the laterocervical and supraclavicular areas is composed of a combination of brown and beige-like adipocytes69,70,71,72,73. Most preclinical studies have shown that exercise induces WAT browning23,24, although these results might be just applicable to room temperature environments (i.e., below thermoneutrality74,75). Thus, exercise might not increase the volume or activity of the brown-like depots detectable by PET/CT scanning, but may still induce the browning of small clusters of adipocytes, that are not detected by this technique. If this is the case, the limited resolution of PET imaging (usually 4 mm) would not allow the appearance of beige cells within the WAT to be captured58. The effect of exercise on BAT volume/activity and WAT browning needs to be further explored as better methods become available.

The present finding that exercise seems not to alter BAT volume nor activity in human adults should also be tested in new studies involving different types of exercise program and populations. It is well known that different exercise types (endurance vs. resistance vs. a combination of both) and intensities induce different physiological adaptations76. The present exercise regimens combined resistance and endurance exercises, but it cannot be determined from the results whether the resistance or endurance components on their own might have different effects on BAT recruitment/activation. Interestingly, we found that outdoor ambient temperature influenced BAT volume, 18F-FDG uptake and radiodensity, which might be masking the possible effect of exercise on BAT recruitment/activity. Thus, other study designs and experimental settings, with longer durations, and taking into account the potential effect of outdoor ambient temperature, should be implemented for addressing this scientific question. Finally, BAT presence and activity are reported to be blunted by aging6,13,77,78,79 and metabolic disease46. Since the present work involved relatively healthy young sedentary adults, the findings made may not apply to older or metabolically compromised individuals.

Twenty-four weeks of supervised exercise training reduces adiposity and increases muscular and cardiorespiratory fitness

Exercise reduced both total body fat (Δ CON = −0.3 ± 3.5 kg; Δ MOD-EX = −2.1 ± 2.3 kg; Δ VIG-EX = −2.7 ± 3.8 kg; P = 0.021) and visceral fat mass (Δ CON = −8.6 ± 90.3 g; Δ MOD-EX = −44.5 ± 57.9 g; Δ VIG-EX = −66.7 ± 91.3 g; P = 0.021), although post-hoc comparisons detected differences only between the CON and VIG-EX groups (Fig. 4a and Supplementary Table S1). As expected, exercise tended to increase lean mass in both intervention groups, although only a trend was seen with respect to the control group (P = 0.213, Fig. 4a and Supplementary Table S1). No evidence of a reduction in body weight (P = 0.283, Fig. 4a), or improvements on blood pressure, fasting serum concentrations of glucose, insulin, liver enzymes (i.e., γ-GT, ALT and ALP), cholesterol (i.e., HDL-C and LDL-C), triacylglycerols, and C-reactive protein (all P ≥ 0.092, Fig. 4b and Supplementary Table S1) after the exercise intervention were observed. These findings concur with those of previous exercise intervention studies in relatively healthy individuals80,81. The lack of evidence of an improvement in cardiometabolic risk factors after exercise training may be related to the young age (22 ± 2 years old) and the relatively healthy status of the present participants, whose cardiometabolic risk markers were usually within normal ranges.

Fig. 4: Effect of the 24-week supervised exercise intervention on secondary endpoints. a Body weight and composition parameters, b cardiometabolic risk factors, and c physical fitness parameters. Δ was calculated as post-intervention minus baseline value for every outcome. Serum concentrations were log 10 transformed. P values are from analyses of covariance (ANCOVAs), adjusting for baseline values. * and † indicate significant differences between pairs after Bonferroni correction. BP Blood pressure, CON Control group, HDL-C high-density lipoprotein cholesterol, MOD-EX Moderate-intensity exercise group, RM repetition maximum, VAT visceral adipose tissue, VO 2 oxygen consumption, VIG-EX vigorous-intensity exercise group. Bars represent the mean and standard deviation. Source data are provided as a Source Data file. Full size image

The intervention enhanced lower body muscular strength as measured via the repetition maximum (RM) leg press test (Δ CON = 3.7 ± 30.8 kg; Δ MOD-EX = 20.5 ± 31.1 kg; Δ VIG-EX = 26.3 ± 37.7 kg; P = 0.049, Fig. 4c, and Supplementary Table S1), whereas a lack of effect was observed on RM bench press or handgrip strength results after the exercise intervention (all P ≥ 0.993, Supplementary Table S1). In addition, the intervention increased the time to exhaustion in a maximum effort test (Δ CON = 21.8 ± 124.0 s; Δ MOD-EX = 144.2 ± 109.8 sec; Δ VIG-EX = 142.2 ± 96.2 s; P < 0.001, Fig. 4c) and VO 2 peak (Δ CON = 28 ± 371 mL/min; Δ MOD-EX = 305 ± 367 mL/min; Δ VIG-EX = 285 ± 376 mL/min; P = 0.006, Fig. 4c and Supplementary Table S1), although no significant differences were seen between the exercise groups. Only those participants who attended ≥70% of the total training sessions were included in the main analyses (Supplementary Fig. S3), but the results remained unaltered when other attendance criteria were taken into account (<70%, or ≥85%).

Exercise-induced changes in adiposity, muscular, and cardiorespiratory fitness do not correlate with any changes in brown adipose tissue

Preclinical studies have shown that, during exercise, BAT secretes signaling molecules that favor skeletal muscle function82,83,84. Thus, it may be hypothesized that individuals with higher BAT volumes might benefit to a greater extent from exercise. However, no significant correlations were detected between baseline values, or any change in BAT-related outcomes, and any exercise-induced changes in body composition, cardiometabolic risk profile, muscular strength or cardiorespiratory capacity (all P > 0.1, Supplementary Table S5). Although significant correlations were found between changes in total plasma cholesterol and BAT-related outcomes in the CON and MOD-EX groups, these did not persist after false discovery rate correction (all q values ≥0.54; Supplementary Table S5). These data agree with the findings of a previous study in which no relationship was seen between exercise-induced reduction in adiposity or improvement in the lipoprotein profile, and exercise-induced changes in BAT 18F-FDG uptake29. These observations suggest that the cardiometabolic improvements induced by the exercise intervention were not mediated by BAT activation, at least as quantified by 18F-FDG PET-CT.

In conclusion, there was no evidence of changes in BAT-related outcomes after 24-week supervised exercise intervention combining resistance and endurance training at different intensities in young sedentary adults. The exercise intervention reduced adiposity and enhanced muscular and cardiorespiratory fitness to a comparable extent in both exercise groups, but seemed not to modify other cardiometabolic risk factors. These exercise-induced changes in adiposity, muscular and cardiorespiratory fitness were not correlated with any individual changes in BAT-related outcomes, suggesting that the observed exercise-induced benefits are independent of BAT in such young sedentary adults.