The Presence of Active Brown Adipose Tissue Determines Cold-Induced Energy Expenditure and Oxylipin Profiles in Humans

Abstract

Background

Accumulating evidence links brown adipose tissue (BAT) to increased cold-induced energy expenditure (CIEE) and regulation of lipid metabolism in humans. BAT has also been proposed as a novel source for biologically active lipid mediators including polyunsaturated fatty acids (PUFAs) and oxylipins. However, little is known about cold-mediated differences in energy expenditure and various lipid species between individuals with detectable BAT positive (BAT_pos) and those without BAT negative (BAT_neg).

Methods

Here we investigated a unique cohort of matched BAT_pos and BAT_neg individuals identified by 18F-fluorodeoxyglucose positron emission tomography combined with computed tomography ([18F]-FDG PET/CT). BAT function, CIEE, and circulating oxylipins, were analyzed before and after short-term cold exposure using [18F]-FDG PET/CT, indirect calorimetry, and high-resolution mass spectrometry, respectively.

Results

We found that active BAT is the major determinant of CIEE since only BAT_pos individuals experienced significantly increased energy expenditure in response to cold. A single bout of moderate cold exposure resulted in the dissipation of an additional 20 kcal excess energy in BAT_pos but not in BAT_neg individuals. The presence of BAT was associated with a unique systemic PUFA and oxylipin profile characterized by increased levels of anti-inflammatory omega-3 fatty acids as well as cytochrome P450 products but decreased concentrations of some proinflammatory hydroxyeicosatetraenoic acids when compared with BAT_neg individuals. Notably, cold exposure raised circulating levels of various lipids, including the recently identified BAT-derived circulating factors (BATokines) DiHOME and 12-HEPE, only in BAT_pos individuals.

Conclusions

In summary, our data emphasize that BAT in humans is a major contributor toward cold-mediated energy dissipation and a critical organ in the regulation of the systemic lipid pool.

brown adipose tissue, thermogenesis, cold exposure, energy expenditure, lipid metabolism, oxylipins

Brown adipose tissue (BAT) is a thermogenic organ that dissipates chemical energy through heat production. Promotion of BAT function counteracts adiposity in numerous animal models and holds great promise as novel antiobesity concept in humans (1-5). Cold exposure is the most potent physiologic stimulus of BAT thermogenesis that results in increased energy expenditure in mice and humans. In humans, cold-activated BAT can be specifically detected by ¹⁸F-fluorodeoxyglucose positron emission tomography combined with computed tomography ([¹⁸F]-FDG PET/CT) (1, 2, 5-8). Cold-induced [¹⁸F]-FDG uptake in BAT has been shown to correlate positively with energy expenditure and negatively with adiposity in humans (7, 9-11). Indeed, BAT-dependent cold-induced thermogenesis significantly contributes to total energy expenditure, and chronic cold exposure studies demonstrated a decrease in body fat mass (9-13). However, the presence of BAT is very heterogeneous in the general population with large variations even in young healthy adults. For example, in a Japanese cohort of 260 healthy volunteers, only 125 (48%) participants had detectable BAT depots (BAT_pos) after cold exposure (14). As previously reported in a number of other studies, participants without detectable BAT (BAT_neg) were significantly older, had higher fat mass and higher expression of adiposity-related parameters such as hemoglobin A_1c, triglycerides, and cholesterol (14-16). However, very little is known about the differences in energy expenditure and lipid metabolites between age- and weight-matched BAT_pos and BAT_neg individuals and therefore the individual contribution of BAT toward systemic fuel metabolism. Studying such interindividual differences is of particular importance given accumulating evidence that cold-activated BAT also improves insulin sensitivity and glucose uptake and significantly alters lipid homeostasis (12, 17, 18). BAT activation promotes lipid mobilization from peripheral stores and oxidative disposal, presumably to accommodate the increased fuel needs of BAT mitochondrial activity (19, 20). In keeping with this notion are observations of elevated triglyceride-rich very low-density lipoprotein concentrations following short-term cold exposure, possibly as a consequence of increased hepatic output (21). Recent evidence suggests that activated BAT also represents a source of specific lipid species that may contribute to metabolic homoeostasis (19, 22, 23). These and other observations suggest that BAT may be associated with the individual risk for type 2 diabetes (24). Polyunsaturated fatty acids (PUFAs) as well as their oxidation products, known as eicosanoids, are the endogenous ligands for peroxisome proliferator-associated receptors (PPARs), which act as transcriptional regulators of glucose and lipid metabolism (25). Specific lipid species such as 12,13-dihydroxyoctadecenoic acid (DiHOME) have been described to be secreted from brown adipocytes upon exercise (23) and cold exposure (26) with effects on fatty acid release and metabolism, while circulating levels of the lysophosphatidylcholine-acyl C16:0 correlate with BAT activity in humans (27). The identification of specific lipid patterns associated with functional BAT contributes toward a better understanding of the importance of BAT for lipid metabolism and may help identify potential BAT biomarkers. In this study we investigated the differences in cold-induced energy expenditure (CIEE) and circulating oxylipin profiles between BAT_pos and BAT_neg individuals_. We found that a single bout of cold exposure significantly elevated energy expenditure for more than 2 hours after cessation of the cold challenge, whereas CIEE was absent in BAT_neg individuals_. The presence of active BAT was associated with an anti-inflammatory oxylipin/eicosanoid profile, and cold exposure increased systemic oxylipin levels only in BAT_pos but not BAT_neg individuals. We identified distinct PUFA and oxylipin clusters that were concomitantly regulated by the presence of BAT and cold exposure and were associated with CIEE.

Materials and Methods

Participants and study approval

We included 16 age- and body mass index (BMI)-matched healthy volunteers, all from the same geographical region (28) that were previously identified as BAT_pos or BAT_neg (n = 8/group) in a cross-sectional BAT imaging study (NCT02381483). While BAT_neg participants had no significant [¹⁸F]-FDG uptake at either room temperature (RT) or after cold exposure, BAT_pos participants had strong [¹⁸F]-FDG uptake in the typical cervical and thoracic BAT regions following cold exposure (Fig. 1A, B). For the initial imaging study, interested volunteers were recruited through printed advertisement. 16 age- and BMI-matched participants were invited for a follow-up study to analyze total CIEE after a single bout of cold exposure. All participants were screened for their medical history and underwent physical examination. Participants that fulfilled the inclusion criteria (28) were enrolled in the study after providing written informed consent. This study is registered at ClinicalTrials.gov (NCT02381483) and has been approved by the Ethics Committee of the Medical University of Vienna (no. 1032/2013). The study was conducted at the Division of Endocrinology and Metabolism, Department of Medicine III, Medical University of Vienna in accordance with the principles of the Declaration of Helsinki.

Figure 1.

BAT_neg individuals lack [18F]-FDG uptake only in BAT but not in other metabolic tissues. (A) Representative [18F]-FDG PET image of a BATpos participant before (left) and after (right) cold exposure. (B) Representative [18F]-FDG PET image of a BATneg participant before (left) and after (right) cold exposure. (C) [18F]-FDG uptake in BAT before and after cold exposure in BATpos and BATneg participants. (D) Changes in mean norepinephrine concentrations in BAT_pos (black line) and BAT_neg (interrupted gray line) before and after cold exposure. ***P < 0.001, **P < 0.01, *P < 0.05. [18F]-FDG PET/CT, 18F-fluorodeoxyglucose positron emission tomography combined with computed tomography; BAT, brown adipose tissue; neg, negative; pos, positive; SUV, standardized uptake value.

Study design

Study participants were examined at 3 separate study visits in the morning after an overnight fast (>10 hours) (28). At study visit 1, all participants underwent the first [¹⁸F]-FDG PET/CT scan at RT (23°C) to detect any basal BAT activity. Participants were instructed to wear comfortable clothing and stay warm until arriving in the hospital. At study visit 2 (no longer than 14 days after study visit 1), study participants underwent the second [¹⁸F]-FDG PET/CT scan, this time after cold exposure, to detect cold-induced BAT activity. A personalized cooling protocol (28) was applied using a water-perfused cooling vest (CoolShirt Systems, Stockbridge, Georgia, USA) similar to as previously described (29). The water temperature was kept slightly above the shivering temperature, and muscle activity was monitored by electromyography (OT Bioelettronica, Torino, Italy). After 90 minutes of cold exposure, [¹⁸F]-FDG was administered, and cold exposure was continued for another 60 minutes until PET/CT acquisition started between 10 am and 11am. Participants that had [¹⁸F]-FDG-positive BAT depots after cold exposure were deemed BAT positive, whereas participants that had no significant [¹⁸F]-FDG uptake at baseline and after cold exposure, respectively, were considered BAT negative. All PET/CT scans were performed during the cool season (fall to spring) to avoid any seasonal influence on BAT activity.

At study visit 3, eight BAT-positive and 8 BAT-negative participants underwent repeated indirect calorimetry examinations before, during, and after cold exposure to determine the magnitude and duration of CIEE after a single bout of cold exposure. All examinations started between 8 am and 9 am, and all participants were in fasting condition (>10 hours). After a 30-minute resting phase, the basal resting energy expenditure was determined by indirect calorimetry. Next, participants were cooled for 90 minutes using the same temperature settings as during study visit 2. During the final 30 minutes of the cooling period, another indirect calorimetry was performed. Thereafter, during rewarming, indirect calorimetry was repeated every hour until energy expenditure had returned to the baseline (resting) levels. Blood was taken before and after 90 minutes of cold exposure.

PET/CT scanning protocol

Each participant received a dose of 2.5 MBq kg^-1 body weight [¹⁸F]-FDG (at least 150 MBq and a maximum of 350 MBq). A combined PET/CT acquisition started 60 minutes after the administration of the radioactive isotope on a Siemens Biograph 64 True Point scanner (Siemens Healthcare Sector, Erlangen, Germany). First, a low dose CT scan (120 kV, 50 mAs) was performed for scatter and attenuation correction as well as for precise anatomical localization of the BAT depots, followed by PET acquisition in 3D mode of 3 minutes/bed position. Both PET and CT were acquired from the base of the skull to mid thigh.

PET/CT analysis

PET/CT scans were analyzed using Hermes Hybrid 3D Viewer (Hermes Medical Solutions, Stockholm, Sweden). The identification of the BAT depots was performed using a multithreshold model in accordance with the Brown Adipose Reporting Criteria in Imaging Studies (30). Briefly, the regions of interest were delineated in the axial fusion images. Only regions of interest located in regions with a CT radio density of –180 to –10 Hounsfield units and with a minimal standardized uptake value higher than 2 were considered for brown fat quantification.

Indirect calorimetry

The resting energy expenditure was measured by computerized open circuit indirect calorimetry (Quark RMR, Cosmed, Rome, Italy). Briefly, after a resting phase of 30 minutes, participants were studied for 30 minutes in a supine position in a quiet room with a constant temperature of about 23° to 24°C, or during cold exposure. The participants were instructed to stay motionless and awake during the entire study period. The results of the last 20 minutes of each experiment were averaged. CIEE was calculated as the percentage change in total energy expenditure between RT and cold exposure.

PUFA and oxylipin extraction

Blood was drawn in cooled ethylenediaminetetraacetic acid-coated tubes and was immediately centrifuged at 4000 rpm at 4°C. After centrifugation, the samples were frozen at –80°C until use in further experiments. Frozen plasma samples were thawed on ice. In order to precipitate proteins, 500 µL plasma were mixed with 2.0 mL ethanol (abs 99%; AustroAlco) including 10 to 100 nM of each internal standard (12S-, 15S- hydroxyeicosatetraenoic acids [HETE]-d8, 5-Oxo-ETE-d7,11.12- dihydroxyeicosatrienoic acid [DiHETrE]-d11, prostaglandin E (PGE)-d4, 20-HETE-d6; Cayman Europe, Tallinn, Estonia) and stored over night at –20°C. Samples were centrifuged (30 minutes, 5000 rpm, 9 deceleration, 7 acceleration, 4°C), and the supernatant transferred to a 15 mL Falcon tube. Ethanol was evaporated via vacuum centrifugation at 37°C until the original sample volume was restored. Samples were loaded on preconditioned 30 mg mL^-1 StrataX solid phase extraction columns (Phenomenex, Torrance, CA, USA), using Pasteur pipettes, washed with 2 mL mass spectrometry grade water, and lipids were eluted with 500 µL ice cold methanol (MeOH abs.; VWR International, Vienna, Austria) containing 2% formic acid (FA; Sigma-Aldrich). MeOH was evaporated using a nitrogen stream at RT, and samples were reconstituted in 150 µL of reconstitution buffer (H₂O/acetonitrile/MeOH + 0.2% FA - 65:31.5:3.5), containing another set of 10 to 100 nM of internal standards (5S-HETE-d8, 14.15-DiHETrE-d11, 8-iso-PGF2a-d4; Cayman Europe, Tallinn, Estonia).

Liquid chromatography with tandem mass spectrometry method

Analytes were separated using a Thermo Scientific Vanquish (ultra high-performance liquid chromatography) system and a Kinetex C18—column (2.6 μm C18 100 Å, LC Column 150 × 2.1 mm; Phenomenex). Applying a 20-minute gradient flow method, starting at 35% solvent B steadily increasing to 90% B (1-10 minutes), going up to 99% B in 0.5 minutes and held for 5 minutes. Solvent B was dropped to 35% in 0.5 min and the column equilibrated for 4 minutes. With an injection volume of 20 µL, flow rate was kept at 200 µL minute^-1, and column oven temperature was set to 40°C. Solvent A: 0.2% FA in H₂O, eluent B: ratio of acetonitrile to MeOH (90:10) containing 0.2% FA. Mass spectrometric analysis was performed with a Q Exactive HF Quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific, Austria), equipped with a heated electrospray ionization source for negative ionization. Mass spectra were recorded operating from 250 to 700 m/z at a resolution of 60 000 @ 200 m/z on MS1 (full scan) level. The 2 most abundant precursor ions were selected for fragmentation (higher energy collisional dissociation 24 normalized collision energy), preferably molecules from an inclusion list that contained 31 m/z values specific for PUFAs and oxylipins (28). MS2 (fragment ion scan) was operated at a resolution of 15 000 @ 200 m/z. For negative ionization, a spray voltage of 2.2 kV and a capillary temperature of 253°C were applied, with the sheath gas set to 46 and the auxiliary gas to 10 arbitrary units. For the liquid chromatography–mass spectrometry (LC-MS) data interpretation, raw files generated by the Q-Exactive Orbitrap were analyzed manually using Thermo Xcalibur 4.1.31.9 (Qual browser), comparing reference spectra from the Lipid Maps depository library from July 2018 (3). For peak integration, the software TraceFinder (version 4.1, Thermo Scientific, Austria) was used.

Statistics

Statistical analyses were performed in R version 3.5.2 and CRAN or Bioconductor R-packages as needed (31). P values were corrected for multiple testing according to the false discovery rate (FDR) method (also known as the Benjamini–Hochberg procedure (32), and FDR values <10% were considered as statistically significant. Analytes abundance values were log₂ transformed to get near normal distributions. Lipids with missing values in more than one-third of samples were removed (4/100), and the remaining missing values imputed with R-package mice version 3.4.0 using the predictive mean matching method (33). Differences of analytes were calculated in a regression model using Bioconductor package limma version 3.38.3 (34), and P values were corrected for multiple testing. Pair-matched information and the confounders sex and BMI (represented by the first principal component, PC1, explaining 82.6% of variance) were included into the limma model. Differences for the following contrasts were estimated: (a) all BAT_pos versus BAT_neg samples, (b) BAT_pos versus BAT_neg samples from RT controls, (c) cold exposure versus RT (paired), (d) cold exposure versus RT (paired) in BAT_pos and, (e) cold exposure versus RT (paired) in BAT_neg participants. Clinical parameters and laboratory values were also log₂ transformed, and the following contrasts were estimated (corrected also for PC1 comprising sex and BMI information): (a) cold exposure versus RT (paired) and (b) BAT_pos versus BAT_neg at RT.

To integrate lipid levels with clinical parameters and laboratory values, lipid abundance values were first used for graphical Gaussian modeling (GGM) using R-package GGMselect (0.1–12.1) (35). GGM, also known as covariance selection or concentration graph modeling, is used to build association networks from abundance data. The principle behind GGM is to use partial correlations as a measure of independence of any 2 analytes, which allows distinguishing direct from indirect interactions. The tuning parameter K for the penalty function was varied between 1 and 4 with 0.5 increments, and the function selectFast [family = c(“C01,” “LA,” “EW”)] was used for optimizing the model employing the C01 algorithm, the Lasso-And (LA), and the adaptive lasso (EW) families (Vignette of R-package GGmselect). To select the optimal K value, the GGM-obtained subclusters were analyzed together as analytes-set for significant associations with above defined contrasts employing the procedure “gene set analysis” with raw P values and regulation-directions (logFC-values) from the corresponding single analytes-limma analyses using the runGSA function of R-package piano version 1.22.0 (36). The K value with the highest relative number of significant analytes-sets (= GGM subclusters) across all contrasts (based on the absolute number of subclusters) was used for further analyses, that is, K = 2, yielding 20 subclusters containing between 2 and 8 lipid species.

We correlated significantly regulated lipids or lipid clusters with clinical parameters and laboratory values from cold exposure and RT respectively, by modeling the different comparisons (contrasts) with limma (corrected for the PC1, representing sex and BMI information). We used all significant (FDR < 10%) associations to build the final association networks.

Finally, these associations together with the subcluster information were plotted as rich annotated networks for each contrast. Node colors (green-red) represent the log₂ fold-change between groups according to corresponding contrast (large node meaning significant; small node meaning not significant). The edge color (yellow-blue) represents the correlation between oxylipins and clinical parameters or laboratory values (labeled with “c” if RT control correlations were described and not labeled if cold exposure correlations were described). Edges in red, green, and gray between oxylipins indicate the subcluster structure (ie, whether the oxylipins were connected by graphical Gaussian modeling). The line thickness indicates the statistical significance of the subclusters between different conditions (BAT_neg vs BAT_pos, RT vs cold exposure) considering all 5 contrasts (bold line: significant; thin line: not significant; red: positive correlation; green: negative correlation; and gray, direction undefined).

Prediction of lipid candidates using CFM-ID 3.0

This analysis strategy was used to identify DiHOME candidates in plasma samples. Peaks with an exact mass of 313.2384 and a fragmentation pattern similar to 9,10-DiHOME were subjected to CFM-ID 3.0 as the identification tool (37), searching for candidate molecules following the protocol based on MS2-spectra obtained with fragmentation energies of 10, 20, and 40 eV. This was performed for the following molecules: DiHOME_10.81, 313_10.40, 313_10.45, and 313_10.75. The CFM-ID 3.0 can search against either HMDB 4.0 (38), other lipid databases, or manually generated candidate lists. Using HMDB 4.0 suggested 9,10-DiHOME and 12,13-DiHOME as potential candidate molecules for the DiHOMEs listed above. However, HMDB 4.0 contains only 9,10-DiHOME and 12,13-DiHOME and no other DiHOMEs. We can exclude that DiHOME_10.81, 313_10.40, 313_10.45, 313_10.75 were 9,10-DiHOME or 12,13-DiHOME as we have assigned these molecules in our dataset to the retention times 8.56 and 8.40, respectively, based on the lipid standards. However, characteristic fragment ions strongly suggested that DiHOMEs_10.81, 313_10.40, 313_10.45, 313_10.75 were indeed DiHOMEs. Therefore, an extended candidate list was generated manually using SMILES code for all DiHOMEs available on LIPID MAPS (28). MS2 peak list files containing exact masses and intensity values were exported from Xcalibur 4.1.31.9 (Qual browser) for each DiHOME candidate peak, respectively. Compound identification criteria included ESI spectra in negative ion mode, Jaccard as the scoring function, and a mass tolerance of 20 ppm. To control for the reliability of this software, the experimental MS data obtained for 9,10-DiHOME were processed and resulted in the correct identification of 9,10-DiHOME (28).

Results

Identification of BAT_pos versus BAT_neg participants

In order to study the relevance of active BAT for CIEE and cold-regulated lipid species, we performed [¹⁸F]-FDG PET/CT scans in 30 healthy normal-weight individuals at RT and after 2.5 hours of moderate cold exposure. Using this approach, we identified 16 age- and BMI-matched participants (28) of whom 8 had detectable BAT [¹⁸F]-FDG uptake after cold exposure (BAT_pos) and 8 that did not have any significant BAT activity at RT or at cold-stimulated conditions (BAT_neg, Fig. 1A-C). Metabolic parameters such as circulating glucose, triglycerides, cholesterol, and BAT activity parameters were similar in both groups at RT and after cold exposure (28). The water temperatures of the cooling garment did not differ between BAT_pos and BAT_neg (28), whereas the cold-induced increase of circulating norepinephrine concentrations was even higher in BAT_neg than BAT_pos participants (Fig. 1D), emphasizing that insufficient cooling or lower sympathetic tone was not the reason for a lack of [¹⁸F]-FDG uptake in BAT_neg participants._. In contrast to BAT, [¹⁸F]-FDG uptake was similar between groups in other metabolically relevant tissues including liver, subcutaneous abdominal WAT, and skeletal muscle (pectoralis major and erector spinae) (28).

CIEE is increased in BAT_pos but not in BAT_neg individuals and reverses 150 minutes after discontinuation of cold exposure

To better understand the importance of the BAT status for CIEE, we quantified whole body energy expenditure in BAT_pos and BAT_neg participants before, during, and after cold exposure. At baseline, resting energy expenditure and the respiratory quotient were similar in both groups (Fig. 2A, B). However, CIEE increased significantly in BAT_pos reaching a maximum during the 90-minute cold exposure period. In contrast, energy expenditure was unaltered in BAT_neg in response to cold (Fig. 2C), suggesting that BAT status is the primary determinant of cold-mediated energy consumption. Given the lack of information about the duration of cold-induced BAT activity, we continued to examine energy expenditure after the discontinuation of cold exposure. After cold exposure, CIEE slowly declined to baseline levels within 150 minutes in BAT_pos individuals, whereas BAT_neg individuals did not experience any changes in energy expenditure during the entire test period (Fig. 2C). The incremental area under the curve of energy consumption during cooling and rewarming was approximately 20 kcal/day in BAT_pos and, therefore, significantly higher compared with BAT_neg (Fig. 2D).

Figure 2.

Cold-induced energy expenditure (CIEE) differs significantly between BAT_pos and BAT_neg individuals. (A) Respiratory quotient (RQ) and resting energy expenditure (REE) (B) of BAT_pos (black bars) and BAT_neg participants (gray bars) at room temperature. (C) Change of CIEE over time during and after the cessation of the cold stimulus in BAT_pos (black line) and BAT_neg (gray line) participants. CIEE was calculated as the percentual increase in REE after cold exposure as compared with baseline. (D) Integrated energy consumption following a single bout of 90-minute cold exposure. ***P < 0.001, **P < 0.01, *P < 0.05. BAT, brown adipose tissue; iAUC, incremental area under the curve; min, minute(s); neg, negative; pos, positive; REE, RQ,

BAT_pos have a distinct systemic PUFA and oxylipin profile compared with BAT_neg individuals

Next we investigated whether (a) the presence of BAT defined the systemic lipid pool in humans and (b) cold exposure altered the PUFA and oxylipin concentrations distinctly in BAT_pos and BAT_neg. Therefore, blood plasma was collected from all 16 participants before and after a 90-minute cold exposure period for subsequent lipid extraction and analysis by LC-MS/MS. Molecules were identified by exact mass, MS2 fragmentation pattern, and chromatographic retention time. A panel of 31 commercially available PUFA and oxylipin standards was included in the analyses, supporting the exact assignment of the corresponding molecules in the plasma samples (28). However, an additional 67 molecules were identified as oxylipins without more detailed structural specifications (28). These molecules were annotated according to their mass and retention time, that is, 319_10.31 corresponds to a molecule with 319 Da and a retention time of 10.31 minutes.

The abundance of detected plasma PUFAs and oxylipins was very heterogeneous, covering a concentration range of more than 3 orders of magnitude (28). Various molecules including arachidonic acid, 14,15-DiHETrE, and 11-HEPE showed less interindividual variation than others such as 5-HETE, 11-HETE, and (15d-) PGJ2.

A principle component analysis indicated that the general variance of analytes between individuals was larger than the variance between groups (28). For further statistical analysis, we thus performed analytes-set analyses and a coassociation network analysis as described in Materials and Methods.

This analysis strategy identified 4 coassociation clusters that were distinctly regulated between BAT_pos and BAT_neg participants (Fig. 3A). The first cluster included 20-HETE and 2 DiHETrEs, all of them significantly upregulated in BAT_pos individuals and all representing known cytochrome P450 products (39). Therefore, we termed this cluster 1, P450 products. The second cluster was also increased in BAT_pos participants and included the ω3 fatty acids stearidonic acid, eicosapentaenoic acid (EPA), (40) and was thus designated as ω3 precursors. The third cluster was decreased in BAT_pos versus BAT_neg individuals and was named cluster 3, HETEs, because it contained only HETEs formed by lipoxygenases (5, 12, and 15-isoform) or via free oxygen radicals (9 and 11 isoform). A fourth cluster comprising mainly PUFAs was also distinctly regulated between BAT_pos and BAT_neg participants; however, this analytical strategy could not distinguish between down- or upregulation.

Figure 3.

Systemic PUFAs and oxylipins differ between BAT_pos and BAT_neg and correlate with clinical BAT parameters. Association networks of (A) various lipids in BAT_pos versus BAT_neg participants at room temperature and of (B) various lipid species and clinical parameters in BAT_pos versus BAT_neg participants before cold exposure. Gray ellipses indicate significantly regulated lipid clusters. Node colors “green-red” represent log2 fold-change of individual lipids between groups (red, high in BAT_pos). Colored molecule names indicate significant differences between BAT_pos and BAT_neg. The color code yellow-blue represents the correlation between various lipids and clinical BAT parameters or laboratory values (red, positive correlation; green, negative correlation). BAT, brown adipose tissue; FA, fatty acid; HETE, hydroxyeicosatetraenoic acid; neg, negative; pos, positive; PUFA, polyunsaturated fatty acid.

Next, we included clinical BAT parameters such as CIEE, mean standardized uptake value, and BAT volume in the model and found that cluster 1, containing the cytochrome P450 products 20-HETE and 11,12- and 14,15-DiHETrE, was significantly correlated with CIEE in BAT_pos participants. The PPAR-γ ligand 15-deoxy-Δ12,14-prostaglandin J2 as well as 315_9.55 and 313_12.63 were positively associated with BAT volume and total lesion glycolysis, a marker of BAT activity (Fig. 3B).

To further address the effect of the BAT status on PUFA and oxylipin profiles, we performed a comparative analysis of such circulating lipids in BAT_pos and BAT_neg individuals at RT. By using a linear model analysis corrected for sex and BMI, with subsequent correction for multiple testing, and applying a threshold of an FDR < 10%, we identified 36 lipids that were significantly different between BAT_pos and BAT_neg participants at baseline, including 19 upregulated and 17 downregulated molecules (Table 1, Fig. 3A). The lipid molecule with the highest relative expression in BAT_pos compared with BAT_neg participants was DiHOME_10.81, which belongs to a class of BAT-derived lipokines (BATokines) previously associated with increased brown adipocyte and skeletal muscle fatty acid uptake (23, 26). The matching molecular mass and isotope pattern, as well as characteristic fragment ions at 295.2 and 251.2 Da, strongly suggest that the molecule that we identified as DiHOME_10.81 indeed belongs to the family of DiHOMEs. Previously, 12,13-DiHOME has been reported as an oxylipin highly expressed in BAT (26). However, DiHOME_10.81 is distinct from 12,13-DiHOME, which we also identified here using a lipid standard with the retention time of 8.4 minutes. Despite thorough analysis of the mass spectrometry data with the specialized software CMF-ID 3.0 and due to the lack of commercially available standards, the exact molecular characterization of DiHOME_10.81 and other candidate molecules of the DiHOME family was not yet possible (28). Other known molecules that were significantly increased in BAT_pos compared to BAT_neg included the anti-inflammatory EPA and stearidonic acid and the recently identified BATokine 12-HEPE, which improve glucose metabolism in vitro and in vivo (40). In contrast, a number of proinflammatory arachidonic acid metabolites such as 5-oxo-eicosatetraenoic acid and a number of HETEs were decreased in BAT_pos individuals (Table 1).

Table 1.

Differences in the abundance of various lipid species between BAT_pos and BAT_neg individuals. Log2 fold change (logFC) and false discovery rates (FDR) for comparisons between BAT_pos and BAT_neg at baseline (before cold exposure)

ID	LogFC	FDR
DiHOME_10.81	8.25	0.016
335_9.86	2.76	0.015
EPA	2.01	0.007
317_8.69	1.81	0.046
Stearidonic acid_11.63	1.41	0.085
3.HOTrE_9.39	1.41	0.007
HpODE_10.04	1.41	0.007
319_11.07	1.38	0.010
12-HEPE	1.37	0.007
265_9.08	1.24	0.029
HETE_9.57	0.99	0.012
313_10.75	0.94	0.029
6trans-LTB4_7.98	0.93	0.086
11,12-DiHETrE	0.78	0.025
HpODE_7.71	0.78	0.077
HHTrE_9.01	0.77	0.098
14,15-DiHETrE	0.77	0.015
335_9.77	0.73	0.085
20-HETE	0.58	0.094
313_10.45	–0.71	0.098
15S-HETrE_10.67	–1.05	0.055
13,14-dihydro-15-keto-PGF2a_6.3	–1.40	0.077
Linoleic acid	–1.65	0.012
9/10-DiHOME	–2.10	0.002
15-HETE	–2.18	0.001
313_12.27	–2.19	0.015
343_10.07	–2.30	0.009
319_10.31	–2.88	0.001
12-HETE	–2.89	0.001
metPGE2_6.35	–2.91	0.067
5-HETE	–2.94	0.007
11-HETE	–3.10	0.002
9-HETE	–3.11	0.001
15S-HETrE_10.77	–3.28	0.019
14-HDoHE and 10-HDoHE	–3.75	0.010
5-Oxo-ETE	–3.75	0.094

Abbreviations: 6 metPGE2, 6 met prostaglandin E2; 6-trans-LTB, 6-trans-leucotriene B; 13,14-dihydro-15-keto-PGF2a, Prostaglandin F; DiHETrE, dihydroxyeicosatrienoic acid; HDoHE, hydroxy-docosahexaenoic acid; HETre, hydroxy-eicosatrienoic acid; HHTrE, hydroxy-heptadecatrienoic acid; HOTre, hydroxy-octadecatrienoic acid; HpODE, hydroperoxy-octadecadienoic acid; ID, identification; neg, negative; pos, positive.

Cold exposure significantly increases PUFA and oxylipin concentrations only in BAT_pos but not in BAT_neg individuals

The direct comparative analysis of cold-mediated changes in lipid species revealed 11 molecules that were significantly altered by cold exposure. All of them were cold-induced but only in BAT_pos and not in BAT_neg individuals (Table 2). Particularly, PUFAs such as docosahexaenoic acid (DHA) and docosapentaenoic acid (DPA) as well as members of cluster 2 ω3 precursors were elevated in BAT_pos individuals in response to cold exposure (Fig. 4A,B, Table 2) Fig. 4 depicts the unprocessed signal intensity normalized area under the curve (nAUC) values from the LC-MS experiments, whereas Table 2 lists the results obtained after data processing, including correction for sex and BMI. This list also included the ω3 oxylipin 12-HEPE, which was not only higher in BAT_pos compared with BAT_neg individuals (Table 1) but significantly increased with cold exposure, supporting the recent notion of 12-HEPE as a cold-regulated BAT-derived lipokine (40). In addition, 12-HEPE was one of only 8 lipids that were elevated in BAT_pos at RT and increased even further in response to cold exposure (Fig. 5) as would be expected from a typical BATokine. Other lipids in this group were the ω3 PUFAs and cluster 3 members EPA and stearidonic acid (Fig. 5). Interestingly, some molecules including the PPAR-γ agonist PGJ2 and its metabolite 15d-PGJ2 showed oppositional regulation; they were nonsignificantly increased in BAT_pos but reduced in BAT_neg after cold exposure (28).

Table 2.

Significant cold-induced changes in circulating fatty acids and oxylipins occurred only in BAT_pos but not BAT_neg individuals. LogFC and FDR values for comparisons before versus after cold exposure in BAT_pos and BAT_neg, respectively

	BATpos		BATneg
ID	LogFC	FDR	LogFC	FDR
12-HETE	0.77	0.023	0.41	0.212
Oleic acid	0.72	0.092	0.12	0.874
DPA	0.72	0.092	0.65	0.187
305_13.04	0.67	0.096	0.23	0.692
Stearidonic acid_11.63	0.63	0.096	0.38	0.336
DPA_12.97	0.54	0.092	0.29	0.336
Linoleic acid	0.50	0.096	0.36	0.267
12-HEPE	0.49	0.058	0.18	0.425
DHA	0.46	0.092	0.46	0.187
AA	0.45	0.096	0.44	0.187
319_11.45	0.32	0.092	0.26	0.212

Abbreviations: BAT, brown adipose tissue; DHA, docosahexaenoic acid; DPA, docosapentaenoic acid; FDR, false discovery rate; HEPE, hydroxyeicosapentaenoic acid; HETE, hydroxyeicosatetraenoic acid; ID, identification; LogFC, log2 fold change; neg, negative; pos, positive.

Figure 4.

Cold exposure raises PUFA and oxylipin levels only in BAT_pos individuals. A) Cold-induced changes in the abundance (nAUC) of AA, 12-HETE, DPA, and oleic acid in BAT_pos and BAT_neg, respectively. Unprocessed abundance values (nAUC) determined from LC-MS experiments are depicted. Fold-change values may thus slightly differ from processed and corrected values reported in Tables 1 and 2; Significance levels were used as given in Tables 1 and 2 and are corrected for age, sex, pairedness, cold exposure, and BAT status (see Material and Methods). * FDR < 10%, ** FDR < 5%. B) Selected association networks showing distinct effects of cold exposure on lipid clusters in BAT_pos and BAT_neg individuals. Red molecules were cold induced; red edges indicate lipid clusters that were increased after cold exposure. AA, arachidonic acid; BAT, brown adipose tissue; DPA, docosapentaenoic acid; FDR, false discovery rate; HEPE, hydroxyeicosapentaenoic acid; HETE, hydroxyeicosatetraenoic acid; neg, negative; pos, positive; PUFA, polyunsaturated fatty acid.

Figure 5.

Potential BAT-derived lipids are elevated in BAT_pos and after cold exposure. Venn diagram showing the number of overlapping lipids that were significantly increased (“up in BAT_pos”) or decreased (“down in BAT_pos”) in BAT_pos compared with BAT_neg individuals as well as lipids that were significantly upregulated (“up with cold”) or downregulated (“down with cold”) in response to cold exposure in both cohorts. All comparisons were corrected for age, sex, and BAT status, the latter only for cold exposure comparisons. Eight lipids were significantly higher in BATpos and increased by cold exposure, particularly suggesting BAT as a potential source. BAT, brown adipose tissue; DiHETrE, dihydroxyeicosatrienoic acid; HEPE, hydroxyeicosapentaenoic acid; HETE, hydroxyeicosatetraenoic acid; neg, negative; pos, positive.

Discussion

Despite accumulating evidence that BAT activation is associated with increased energy expenditure, improved insulin sensitivity, and lipid clearance in humans (11, 12, 17-19), direct comparison of substrate metabolism and lipid species in matched individuals with and without active BAT depots has not been studied yet. Here we demonstrated that CIEE increased by 15% in BAT_pos, with no changes in BAT_neg participants, suggesting that BAT is the predominant contributor to CIEE (Fig. 2). This notion is not only supported by the absence of CIEE in BAT_neg individuals but also the lack of cold-induced FDG uptake in skeletal muscle in the entire study cohort (28). Numerous studies have previously shown an association between BAT activation and increased energy expenditure in humans (7, 9-11), however it had remained unclear how long this effect persisted once BAT was activated. We showed here that energy expenditure was elevated for approximately 150 minutes after discontinuation of the cold challenge and thus culminates in an additional energy loss of 20 kcal/day in BAT_pos, which is in line with a previous report studying oxidative metabolism in BAT using oxygen-15(¹⁵O)/PET imaging (41). Considering that 20 kcal/day is the amount of excess energy consumption conveyed by a single 90-minute bout of cold exposure, chronic cold adaption may have more pronounced effects on energy expenditure given that repeated cold exposure over 6 weeks increased BAT activity by 150%. In addition, chronic cold exposure as well as weight loss resulted in the emergence of newly recruited BAT in individuals with previously no detectable BAT (11, 42) emphasizing that therapeutic interventions targeting BAT activation may not be limited to individuals with the presence of BAT before treatment.

Recently, human BAT has been established as a metabolic tissue that takes up and burns lipids but also represents a source of lipid production by releasing so-called lipokines into the circulation (19, 22). Lipokines act locally or systemically and control a number of metabolic processes such as glucose homeostasis and inflammation.

Oxylipins, including eicosanoids are a complex class of lipid-derived signaling molecules that govern inflammatory, anti-inflammatory, and metabolic processes (26, 40, 43, 44). In addition, eicosanoids represent the most important physiological ligands for PPARs, which are critical in regulating carbohydrate and lipid metabolism (45). BAT has recently been identified as a source of oxylipin production in humans (26, 40). Using high-resolution mass spectrometry analysis (46) we found that circulating PUFA and oxylipin profiles were distinct in BAT_pos and BAT_neg individuals and that cold exposure significantly altered plasma lipid composition only in BAT_pos, strongly suggesting that BAT is a significant contributor to the systemic oxylipin pool in humans. Indeed, the BATokine 12,13-DiHOME has recently been demonstrated to be induced in mice and humans upon cold exposure. In addition, 12,13-DiHOME administration in mice stimulated BAT fuel uptake, triglyceride clearance, and conferred protection against cold stress (23, 26). We detected 12,13-DiHOME in all participants and found it increased in BAT_neg and BAT_pos individuals upon cold exposure, reaching statistical significance when combining both groups (logFC 1.36, FDR 6.57 %). Notably, the potential lipokine that was most strongly upregulated in BAT_pos compared with BAT_neg in our study cohort also belonged to the DiHOME family (designated here as DiHOME_10.81, Table 1 (28), suggesting that DiHOMEs are a class of lipids regulated by BAT with a potential for therapeutic targeting. In keeping with this notion, 12,13-DiHOME was inversely associated with obesity and related metabolic complications in a recent cross-sectional study with more 2200 participants (47). Like DiHOMEs, DiHETrEs are another group of cytochrome P450 oxidase products that were significantly higher in BAT_pos versus BAT_neg participants (cluster 1 in Fig. 3A). Notably this was the only lipid cluster that correlated positively with CIEE in BAT_pos (Fig. 3B), directly linking specific systemic oxylipin availability with BAT thermogenesis. Indeed, DiHETrEs are known agonists of the nuclear receptor PPAR-α (39), a master regulator of fatty acid oxidation and lipid catabolism. These observations also suggest that BAT may be associated with higher P450 enzymatic activity resulting in the formation of PPAR-α agonists and thus affecting systemic lipid metabolism.

The presence of BAT was also characterized by increased concentrations of ω3 fatty acids and their precursor molecules such as stearidonic acid, EPA, and 12-HEPE (cluster 2). In accordance with previous reports (48, 49), plasma levels of PUFAs such as DPA and DHA and their precursors were elevated in response to cold exposure only in BAT_pos participants (Table 2, Fig. 4). In mice, DPA supplementation had a comparable favorable effect on cardiometabolic parameters as does treatment with DHA (50). Supplementation with the ω3 fatty acids EPA and DHA reduces cardiovascular risk in a dose-dependent manner in large randomized trials (51). In addition, EPA administration has been linked to BAT thermogenesis as well as WAT browning in vitro and in mice (49, 52, 53). One possible mediator of this effect is the G-protein–coupled free fatty acid receptor GPR120, which has been shown to induce mitochondrial fission and thermogenic gene expression via fibroblast growth factor 21 (54-56). Stearidonic acid, which was more abundant in BAT_pos at RT as well as after cold exposure, has been shown to suppress lipopolysaccharide-induced inflammation in murine adipocyte stem cells (57). Most recently, 12-HEPE has been identified as a potential novel cold-induced BATokine that promotes glucose uptake in adipocytes and skeletal muscle through activation of an insulin-like intracellular signaling pathway (40). These findings not only suggest BAT as an important source for PUFA release as previously suggested (58) but could hint toward potential anti-inflammatory and positive metabolic effects mediated by BAT-derived ω3 fatty acids (59). In contrast, a number of proinflammatory HETEs including 5-, 11-, 12-, and 15-HETE were significantly reduced in BAT_pos compared with BAT_neg (cluster 3 in Fig. 3A, Table 1). All of them have been associated with obesity-related complications (60, 61). Particularly 12-HETE has been suggested to be involved in the pathogenesis of type 2 diabetes by triggering insulin resistance and β-cell dysfunction (60, 62-64). Both, 12-HETE and 12-HEPE are products of 12-lipoxygenase (12-LOX) and rise with cold exposure, which is in line with the recent finding of a cold-mediated increase in 12-LOX activity in BAT (40). However, the observation that 12-HETE levels were generally higher in BAT_neg versus BAT_pos individuals (Fig. 4B), suggests that other tissues than BAT may also contribute toward systemic 12-HETE production. Particularly, WAT may represent a potential source for oxylipins as previously reported (60). In addition, WAT depots are significantly larger than BAT depots, and WAT is also known to respond to cold stimuli given that adrenergic receptors are abundantly expressed in white adipocytes (65). However, 12-HEPE was 1 of only 8 molecules that were elevated in BAT_pos at baseline and increased even further after cold exposure (Fig. 5), suggesting that BAT is indeed a significant source of 12-HEPE production. This further demonstrates that 12-LOX activity has both detrimental as well as positive effects in adipocyte biology. Many of the identified lipids are part of the same biosynthesis pathways (eg, the LOX pathway) (28).

A limitation of this study is the relatively small sample size due to the age- and BMI-matched design and the expensive technology used here. Also, the radioactive exposure associated with [¹⁸F]-FDG PET/CT scans would hardly justify the screening of a larger number of individuals given that these were healthy volunteers. Despite the small sample size and the exploratory nature of the study, we found a robust regulation of several lipid species by cold exposure and BAT status that is in accordance with previous reports of cold-induced BAT-derived oxylipins (26, 40). However, it cannot be excluded that other tissues responsive to cold stimulation contribute to the observed phenotype. The fact that very distinct lipid clusters are regulated concomitantly by BAT status and cold exposure strongly suggests that BAT is indeed an important organ orchestrating the systemic PUFA/oxylipin pool in a coordinated manner and therefore warrants further investigation.

In summary, cold exposure and the presence of active BAT were associated with a distinct PUFA and oxylipin profile with potentially anti-inflammatory and metabolically favorable characteristics (eg, increased ω3 fatty acids and endogenous PPAR ligands). These data add to the growing body of evidence that BAT may affect metabolic processes beyond thermogenesis, including lipid metabolism and inflammation, and may communicate with other metabolic tissues through BATokines.

Abbreviations

Abbreviations
 
• [¹⁸F]-FDG PET/CT

  18F-fluorodeoxyglucose positron emission tomography combined with computed tomography

 
• 12-LOX

  12-lipoxygenase

 
• 15d-PGJ2

  15-deoxy-Δ12,14-prostaglandin J2

 
• 5-oxo-ETE

  5-oxo-eicosatetraenoic acid

 
• BAT

  brown adipose tissue

 
• BATokines

  BAT-derived circulating factors

 
• BMI

  body mass index

 
• CIEE

  cold-induced energy expenditure

 
• DHA

  docosahexaenoic acid

 
• DiHETrE

  dihydroxyeicosatrienoic acid

 
• DiHOME

  dihydroxyoctadecenoic acid

 
• DPA

  docosapentaenoic acid

 
• EPA

  eicosapentaenoic acid

 
• FA

  formic acid

 
• FDR

  false discovery rate

 
• GGM

  graphical Gaussian modeling

 
• HETE

  hydroxyeicosatetraenoic acid

 
• HMDB

  human metabolome database

 
• LC

  liquid chromatography

 
• MBq

  megabecquerel

 
• MeOH

  methanol

 
• MS

  mass spectrometry

 
• P450

  cytochrome P450

 
• PC1

  first principal component

 
• PPAR-α

  peroxisome proliferator-activated receptor alpha

 
• PPAR-γ

  peroxisome proliferator-activated receptor gamma

 
• PUFAs

  polyunsaturated fatty acids

 
• REE

  resting energy expenditure

 
• RQ

  respiratory quotient

 
• WAT

  white adipose tissue

Acknowledgments

Financial Support: This work was supported by the Vienna Science and Technology Fund, LS12-059, Austrian Science Fund, P 27391 and the Austrian Diabetes Association Research Fund all to F.W.K.

Clinical Trial Information: ClinicalTrials.gov. registration no. NCT02381483 (registered March 6, 2015).

Author Contributions: All authors listed on this manuscript researched and/or analyzed data, contributed intellectually, and revised the manuscript.

Additional Information

The authors have nothing to disclose.

Disclosure Summary: The authors have nothing to disclose.

Data Availability: The datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.

Author notes

These authors contributed equally.