p38α blocks brown adipose tissue thermogenesis through p38δ inhibition

Adipose tissue has emerged as an important regulator of whole-body metabolism, and its capacity to dissipate energy in the form of heat has acquired a special relevance in recent years as potential treatment for obesity. In this context, the p38MAPK pathway has arisen as a key player in the thermogenic program because it is required for the activation of brown adipose tissue (BAT) thermogenesis and participates also in the transformation of white adipose tissue (WAT) into BAT-like depot called beige/brite tissue. Here, using mice that are deficient in p38α specifically in adipose tissue (p38αFab-KO), we unexpectedly found that lack of p38α protected against high-fat diet (HFD)-induced obesity. We also showed that p38αFab-KO mice presented higher energy expenditure due to increased BAT thermogenesis. Mechanistically, we found that lack of p38α resulted in the activation of the related protein kinase family member p38δ. Our results showed that p38δ is activated in BAT by cold exposure, and lack of this kinase specifically in adipose tissue (p38δ Fab-KO) resulted in overweight together with reduced energy expenditure and lower body and skin surface temperature in the BAT region. These observations indicate that p38α probably blocks BAT thermogenesis through p38δ inhibition. Consistent with the results obtained in animals, p38α was reduced in visceral and subcutaneous adipose tissue of subjects with obesity and was inversely correlated with body mass index (BMI). Altogether, we have elucidated a mechanism implicated in physiological BAT activation that has potential clinical implications for the treatment of obesity and related diseases such as diabetes.


Author summary
Accumulation of fat in adipose tissue is essential to store energy and insulate the body; however, excessive body fat leads to obesity. Of the 2 existing types of adipose tissue, white adipose tissue (WAT) stores energy, whereas brown adipose tissue (BAT) can produce heat. Activation of BAT and transformation of WAT into brown-like 'brite/beige' adipocytes have recently emerged as novel strategies against obesity. The uncoupling protein 1 (UCP1) is a hallmark of BAT and is responsible for triggering these 2 processes under the regulation of the p38 MAP kinase (p38MAPK) pathway, but the underlying mechanisms remain unknown. Here, we have analysed this process in detail and demonstrate that a protein kinase called p38α directly correlates with UCP1 levels in human adipose tissue, while it inversely correlates with body mass index (BMI). We find that mice lacking p38α in adipose tissue are protected against diet-induced obesity due to increased body temperature. In addition, another p38 family member, p38δ, is activated in these adipocytes lacking p38α and reduces their thermogenic capacity. Our results suggest that these 2 members of the p38 family have opposite roles in controlling thermogenesis.
Obesity is a serious worldwide health problem, associated with a higher risk of life-threatening diseases [2], that has had a dramatic increase in prevalence [1]. As the main organ for fat storage, adipose tissue has a fundamental role in metabolism [3]. Whereas white adipose tissue (WAT) stores energy in the form of triglycerides and releases free fatty acids on demand, brown adipose tissue (BAT) burns fat to maintain the temperature in a process called nonshivering thermogenesis [4]. Classically, it was assumed that in adult humans BAT played a minor role in energy metabolism. However, recent findings have indicated that this tissue can be modulated by several stimuli presenting lower activity in individuals with obesity [5][6][7]. Additionally, under certain stimuli, WAT can increase its thermogenic capacity in a process called browning [8][9][10][11]. This remodelling of WAT has acquired special interest because it has important therapeutic implications in the treatment of obesity [12,13].
The p38MAPK pathway is activated during browning, and it has been suggested that this drives adipose tissue remodelling [14,15]. There are 4 p38 isoforms: p38α, p38β, p38γ, and p38δ, all of which are activated by stress stimuli in a cell-dependent manner, controlling cellular fate [16][17][18][19][20]. It has been extensively described that p38MAPK triggers browning and BAT activation through the transcription of uncoupling protein 1 (UCP1) via cAMP response element-binding (CREB), activating transcription factor 2 (ATF2), and peroxisome proliferatoractivated receptor gamma coactivator 1α (PGC1α) activation. In fact, β-adrenergic stimulation and other browning agents stimulate the p38MAPK cascade, promoting thermogenesis [18,[21][22][23]. Although most of these studies assumed that the phenotype is driven by p38α, the specific role of the isoform p38α and other p38 isoforms in the development and transformation of adipose tissue has not been elucidated yet using genetically modified mouse models.
Using conditional animals for p38α (p38α Fab-KO ), unexpectedly, we found that deletion of this kinase in adipose tissue protected animals against high-fat diet (HFD)-induced obesity together with increased energy expenditure followed by higher BAT thermogenesis. Lack of p38α in BAT resulted in higher activation of p38δ. In agreement with this, conditional deletion of p38δ in adipose tissue led to obesity, with higher body weight and reduced energy expenditure due to a lower body and skin surface temperature in the BAT region. Besides, lack of p38α in inguinal fat (iWAT) increased p38γ activation and UCP1 expression. Our results indicate that p38α controls p38δ activation in BAT, regulating thermogenesis and energy expenditure. In contrast, in WAT, p38α would have opposite effects depending on the fat depot, blocking browning through inhibition of p38γ in iWAT and promoting browning in epididymal fat (eWAT). Thus, these findings challenge the classical view of p38α as an activator of BAT thermogenesis. These studies provided important insights into p38δ and p38α function in BAT regulation that could have therapeutic implications to efficiently fight obesity.

Results
p38α has emerged as one of the main player that could activate the thermogenic capacity of adipose tissue. Because the thermogenesis of adipose tissue is reduced in obesity [6,7,21], we wondered whether expression of this kinase changes in human WAT during obesity. Using 2 cohorts for visceral fat and subcutaneous fat (sWAT) of adult patients with 80 and 170 samples, respectively, we found that the expression of p38α (Mapk14) in visceral fat and sWAT from individuals with obesity was reduced compared with those without obesity (Fig 1A and  1D). In fact, mRNA levels of Mapk14 in visceral fat inversely correlated with body mass index (BMI) (Fig 1B). It has been suggested that p38α in WAT activates browning by triggering the expression of UCP1 [18], the main protein responsible for adipose tissue thermogenic capacity [22]. In visceral fat and sWAT from individuals with obesity and those without obesity, we found that expression of Mapk14 correlated positively with the levels of Ucp1 (Fig 1C and 1E). This correlation reinforced the idea that p38α in visceral fat and sWAT controls the levels of UCP1 and could regulate browning in humans.
Then, we evaluated the function of p38α in adipose tissue using conditional mice (p38α Fab-KO ), which lacked p38α in WAT and BAT (S1 These data suggest that lack of p38α might protect against type 2 diabetes. Moreover, we evaluated whether lack of p38α affects adipogenesis, browning, and metabolism in eWAT and BAT. BAT from p38α Fab-KO mice presented an increase of Cidea, a marker of browning, together with higher expression of glycolytic and β oxidation genes (S3 Fig).
To further evaluate the role of p38α in adipose tissue, mice were fed an HFD, and we observed that p38α Fab-KO mice were completely protected from diet-induced obesity because their weight was identical to the weight of the control animals in ND (Fig 2A). This reduced weight gain was in line with lower fat mass ( Fig 2B) and reduced weight of the different fat depots, including eWAT, sWAT, iWAT, pWAT, and BAT (S4A Fig). Moreover, liver weight was also reduced in agreement with protection against HFD-induced liver steatosis in p38α Fab-KO mice (Fig 2C and S4A Fig). The protection against HFD-induced obesity was associated with reduced fasted and fed hyperglycaemia in p38α Fab-KO mice, with no differences in triglyceridemia ( Fig 2D and S4E Fig). In addition, p38α Fab-KO mice were protected against HFDinduced glucose intolerance even when glucose dose was adjusted to lean mass (Fig 2E, S4B  Fig.) and insulin resistance as shown by the reduced glucose levels during the insulin tolerance test (ITT) (Fig 2E). HFD-induced obesity was associated with liver insulin resistance and reduced insulin-stimulated Akt phosphorylation in livers from HFD-fed Fab-Cre mice (S4C Fig). Evaluation of insulin sensitivity in several tissues indicated that HFD-fed p38α Fab-KO Correlation between mRNA levels of Mapk14 (p38α) and BMI (r 2 = −0,365; p = 0.001) or (C) Ucp1 in visceral fat (r 2 = 0.316; p = 0.007). The mRNA levels of Mapk14 (p38α) and Ucp1 were determined by qRT-PCR (n = 71). (D) mRNA levels of Mapk14 (p38α) in sWAT from lean individuals and individuals with obesity. mRNA expression was normalised to the amount of Gapdh mRNA. (E) Correlation between mRNA levels of Mapk14 (p38α) and Ucp1 in sWAT (r 2 = 0.320; p < 0.0001). Graph correlating mRNA Mapk14 and log mRNA Ucp1 is also shown. The mRNA levels of Mapk14 (p38α) and Ucp1 were determined by qRT-PCR (n = 168). See also S1 Data. Linear relationships between variables were tested using Pearson's correlation coefficient. BMI, body mass index; qRT-PCR, quantitative real-time polymerase chain reaction; UCP1, uncoupling protein 1. Fab-Cre and p38α Fab-KO male (8-10-wk-old) mice fed an HFD over 8 weeks. Data are presented as the increase above initial weight (left panel) or as total weight comparing mice fed an HDF with mice fed an ND (right panel). HFDinduced weight gain was significantly higher in Fab-Cre than p38α Fab-KO mice (mean ± SEM; Fab-Cre HFD n = 10 mice; p38α Fab-KO HFD n = 11 mice; Fab-Cre ND n = 9 mice; p38α Fab-KO ND n = 8 mice). (B) NMR analysis of fat mass in p38α Fab-KO and Fab-Cre mice after 8 weeks of HFD (mean ± SEM; Fab-Cre n = 10 mice; p38α Fab-KO n = 8 mice). (C) Representative haematoxylin-eosin and oil red O staining of liver sections (Fab-Cre n = 6 mice; p38α Fab-KO n = 6 mice; and 3 pictures from each mouse). Scale bar: 50 μm. (D) Fasting and fed blood glucose in Fab-Cre and p38α Fab-KO mice fed the HFD (8 weeks) (mean ± SEM; Fab-Cre n = 10 mice; p38α Fab-KO n = 11 mice). (E) GTT and ITT in Fab-Cre and p38α Fab-KO mice fed the HFD for 8 weeks. Mice were fasted overnight (for GTT) or 1 hour (for ITT), and blood glucose concentration was measured in mice given intraperitoneal injections of glucose (1 g/kg of total body weight) or insulin (0.75 U/kg of total body weight) (mean ± SEM; Fab-Cre n = 10 mice; p38α Fab-KO n = 11 mice). (F) Immunohistochemistry of eWAT sections using anti-GLUT4 (green), anti-Cav-1 (red) antibodies, and the nuclear dye DAPI (blue). Location of GLUT4 was analysed in mice treated without or with insulin (1.5 IU/kg) for 15 minutes after overnight fasting. Scale bar: 20 μm. (G) Representative haematoxylin-eosin BAT and eWAT sections (Fab-Cre n = 6 mice presented higher insulin-induced phosphorylation of Akt at Thr308 and Ser473 than HFD-fed Fab-Cre mice in liver and muscle but not in eWAT nor BAT (S4D Fig). Furthermore, we observed a slight increase of insulin-stimulated GLUT4 translocation in eWAT ( Fig 2F). Together, these results demonstrate that p38α Fab-KO mice are protected against diet-induced obesity and diabetes.
Histological analysis showed that interscapular BAT depot from HFD-fed p38α Fab-KO mice had small multilocular adipocytes (Fig 2G), whereas in eWAT, we observed a slight decrease of adipocyte size (Fig 2G), which correlates to reduced cell size in BAT and WAT adipocytes from HFD-fed p38α Fab-KO with respect to HFD-fed Fab-Cre (S5A and S5C Fig). Then, we evaluated HFD-induced WAT adipocyte expansion by bromodeoxyuridine (BrdU) staining [23], observing reduced expansion in p38α Fab-KO ( Fig 3A). However, no differences in Ki67 staining were observed after HFD in WAT or BAT adipocytes (S5A and S5C Fig).
To further investigate the mechanism by which lack of p38α in adipose tissue could protect against HFD-induced obesity, we evaluated whole-body metabolism using metabolic cages. HFD-fed p38α Fab-KO mice showed a significant increase in whole-body energy expenditure analysed by ANCOVA, with no changes in food intake or respiratory exchange ratio ( Fig 3B). These data are consistent with the observation that HFD-fed p38α Fab-KO mice have higher skin temperature in the region of BAT compared with Fab-Cre mice ( Fig 3C). Western blot analysis of BAT indicated that HFD-fed p38α Fab-KO mice presented a slight increase of UCP1 expression associated with higher AMPK and Creb phosphorylation (Fig 3D and 3E). In addition, higher expression of UCP1 levels was observed in iWAT from HFD-fed p38α Fab-KO mice (S5B and S7A Figs), suggesting an increased browning of this adipose depot. In contrast with the up-regulated UCP1 levels in iWAT, analysis of eWAT by western blot and immunohistochemistry showed that HFD-fed p38α Fab-KO mice have reduced UCP1 levels in this tissue (S6 and S7B Figs). These results are in agreement with the results found in human visceral fat ( Fig 1C) suggesting that, in visceral fat, p38α directly correlates with UCP1.
In vitro-differentiated brown adipocytes from p38α Fab-KO mice confirmed a key role of this kinase inhibiting browning in a cell-autonomous manner because several browning markers (UCP1, PGC1b, Cidea, Cox7a1, Cox7a2, and Cox8b) were up-regulated in p38α Fab-KO brown adipocytes (S8A Fig). In concordance with the results observed in the BAT tissue, glycolytic genes were also up-regulated, while many lipogenic genes that correlated with the lower triglyceride content in p38α Fab-KO brown adipocytes were down-regulated (S8B, S8C, S8D and To further confirm the autonomous role of p38α in BAT, we crossed p38α loxP mice with UCP1-Cre mice [24], which express Cre recombinase specifically in the interscapular brown fat at room temperature, generating p38α UCP1-KO mice. In agreement with our previous results, these mice were protected against HFD-induced obesity and presented lower fat mass and increased temperature. mice; p38α Fab-KO n = 6 mice; and 3 pictures from each mouse). Scale bar: 50 μm. Ã p < 0.05, ÃÃÃ p < 0.001 Fab-Cre versus p38α Fab-KO . '&&' indicates p < 0.01, '&&&' indicates p < 0.001 Fab-Cre ND versus Fab-Cre HFD (2-way ANOVA coupled with Bonferroni's post-tests or t test or Welch's test when variances were different). See also S1 Data. BAT, brown adipose tissue; Cav-1, caveolin-1; eWAT, epididymal fat; GLUT4, glucose transporter type 4; GTT, glucose tolerance test; HFD, high-fat diet; ITT, insulin tolerance test; ND, normal-chow diet; WAT, white adipose tissue.
Our data at 23˚C demonstrated that lack of p38α resulted in increased whole-body energy expenditure due to the activation of BAT and iWAT thermogenesis. At this temperature, BAT is already fully differentiated; because it is complicated to detect an even higher level of UCP1, genetic modifications that up-regulate UCP1 levels cannot be easily detected [25]. For this reason, we therefore evaluated p38α Fab-KO phenotype in thermoneutrality (30˚C) because it has been suggested to be more similar to the human situation [25]. At 30˚C, p38α Fab-KO mice were also protected against HFD-induced obesity ( Fig 5A) and presented lower body fat mass and increased BAT thermogenesis (Fig 5B and 5C), indicating that, even at temperatures at which BAT is impeded, these mice maintain BAT activation. In fact, UCP1 expression was much higher in BAT from p38α Fab-KO than in the control Fab-Cre mice at 30˚C (Fig 5D). In addition, p38α Fab-KO were also protected from HFD-induced diabetes at thermoneutrality ( Fig  5E and 5F). Together, these data confirm that lack of p38α protects against HFD-induced obesity and diabetes due to an activation of BAT thermogenesis.
To gain insight into the molecular mechanism that might account for increased UCP1 levels and thermogenic capacity, we studied the signalling in the different adipose tissue depots. The p38MAPK pathway has been shown to trigger BAT activation in several models [18,[26][27][28]. Additionally, it has been found that p38α can inhibit the other p38 isoforms by a negative feedback loop that blocks the activation of the upstream kinases of this pathway [29]. Therefore, we evaluated the expression and phosphorylation state of the other p38s, with a phospho-p38 antibody that recognises all p38 isoforms [30]. Using adipocytes lacking p38γ/δ, we confirmed that p38α/β run around 38 kDa, while p38γ/δ run higher-around 41 kDa-allowing us to distinguish the phosphorylation of these kinases (S9A Fig . Activation of p38δ in BAT was diminished when mice were maintained at 30˚C ( Fig  6A), suggesting that this p38 isoform might activate BAT thermogenesis. To further evaluate this hypothesis, mice lacking p38δ in adipose tissue (p38δ Fab-KO ) were generated. In agreement with the importance of this kinase in BAT activation, p38δ Fab-KO mice fed with ND presented higher body weight, associated with increased fat mass and weight of all fat depots (Fig 6C and  6D and S10A Fig). In concordance, p38δ Fab-KO presented reduced energy expenditure, wholebody temperature, and decreased BAT thermogenesis (Fig 6E and 6F) as well as lower Quantification of positive BrdU nuclei is showed in the middle panel. (B) Comparison of energy balance between HFD-fed Fab-Cre and p38α Fab-KO mice. HFD-fed mice were examined in a metabolic cage over a 2-day period to measure FI, respiratory exchange, and EE. FI and EE (left) over 2 days were corrected by lean mass. EE expressed as ANCOVA analysis (middle panel) and hour by hour over 48-h period (right panel) are also shown (mean ± SEM; Fab-Cre n = 10 mice; p38α Fab-KO n = 8 mice). (C) Body (mean ± SEM; Fab-Cre n = 20 mice; p38α Fab-KO n = 18 mice) and skin temperature of surrounding interscapular BAT (mean ± SEM; Fab-Cre n = 10 mice; p38α Fab-KO n = 7 mice). Lower panels show representative infrared thermal images. (D) Immunoblot analysis of UCP1 levels and Creb and AMPK phosphorylation in lysates from BAT. Quantification is shown in the lower panel. (E) Immunohistochemistry staining of UCP1 after 8 weeks of HFD in BAT. Scale bar: 50 μm. Statistically significant differences between Fab-Cre and p38α Fab-KO mice are indicated: ÃÃ p < 0.01 (t test or Welch's test when variances were different). See also S1 Data. AMPK, 5' adenosine monophosphate-activated protein kinase; BAT, brown adipose tissue; BrdU, bromodeoxyuridine; Creb, cAMP response element-binding; EE, energy expenditure; eWAT, epididymal fat; FI, food intake; HFD, high-fat diet; IR temperature, infrared temperature; UCP1, uncoupling protein 1; WAT, white adipose tissue.  p38δ is activated in BAT upon cold exposure and in adipocytes after stimulation with the thyroid hormone T3 or norepinephrine (NE) (Fig 6G and 6H), suggesting that this p38 isoform might activate BAT thermogenesis. In fact, at 4˚C, p38δ Fab-KO mice have lower body and skin temperature in the BAT region ( Fig 6I). Moreover, HFD-fed p38δ Fab-KO mice were more obese with higher fat mass and weight of all fat depots (S11A-S11C Fig). This increased adiposity correlated with lower BAT thermogenesis and lower UCP1, Ppargc1a, and Cidea levels in BAT (S11D-S11F Fig).
Our data indicated that p38δ was triggering thermogenesis because in vitro-differentiated brown adipocytes lacking p38δ have reduced expression of important genes implicated in BAT thermogenesis (Ppargc1b, Ppargc1a, Cidea, and Cox8b) and a slight decrease of Ucp1 and Cox7a1 supporting the cell-autonomous effect of p38δ in BAT thermogenesis ( Fig 7A), with no differences in amount of mitochondrial DNA (Fig 7B and 7C).
Therefore, we evaluated respiration profiles in brown adipocytes lacking p38α and p38δ. Brown adipocytes lacking p38α presented higher leak respiration after isoproterenol (ISO) or NE treatment ( Fig 7D). However, this augmented respiration capacity induced by NE or ISO was diminished when p38δ was chemically inhibited by BIRB796, a known inhibitor p38δ [31], as well as in p38δ-deficient brown adipocytes (Fig 7E and 7F), supporting the important role of this kinase in brown adipocyte activation.
In conclusion, we demonstrated that p38α in BAT inhibits p38δ activation, which in turn regulates BAT thermogenesis, energy expenditure, and body weight. We demonstrated that p38α and p38δ have opposite roles in BAT: whereas p38α inhibits BAT thermogenesis, p38δ induces it upon several physiological stimuli (Fig 8).

Discussion
Adipose tissue has become an important target for the treatment of obesity, not only because its dysfunction could be responsible for diabetes development but also because increasing BAT thermogenesis and/or browning of WAT could lead to new therapeutic approaches against obesity [32,33]. In this scenario, p38MAPK signalling has been proposed to be a key activator of these processes. Consequently, there is an increasing interest to understand the function of this pathway in the regulation of adipose tissue metabolism, remodelling, and browning.
We also demonstrated that p38δ is activated in BAT by 3 stimuli widely known to activate this tissue: cold exposure, NE, and thyroid hormone treatment [35,36], whereas its phosphorylation is reduced under thermoneutrality conditions. In addition, p38δ expression in BAT was reduced in obese mice, while this down-regulation was ablated in p38α Fab-KO mice, suggesting that activation of p38δ in p38α Fab-KO mice is responsible for the protection against diet-induced obesity observed in these mice. Indeed, inhibition of p38δ in p38α Fab-KO brown adipocytes abolished the increased respiratory capacity induced by β3-adrenergic stimuli. In agreement with the role of p38δ-promoting thermogenesis, mice lacking this kinase in adipose tissue developed overweight, even in ND, and showed decreased whole-body energy expenditure associated with lower temperature and reduced BAT activation. Moreover, we confirmed the cell-autonomous role of p38δ inducing browning using differentiated adipocytes.
Our results were completely unexpected because the p38MAPK pathway has been shown to trigger BAT activation in several models [18,[26][27][28], and-until now-it was thought that the only implicated family member was p38α. Moreover, we have recently found that hyperactivation of p38α in MKK6-deficient animals induces browning of eWAT [37]. These finding might indicate opposite effects of p38α in eWAT versus iWAT or BAT. While p38α would activate browning in eWAT-increasing energy expenditure-it would prevent it in iWAT, and it would block thermogenesis through the negative regulation of p38δ in BAT. In agreement with this hypothesis, we observed reduced levels of UCP1 in epididymal fat lacking p38α. In fact, our data from human samples indicated that the p38α mRNA levels in visceral fat directly correlates with UCP1 expression and inversely correlates with the BMI, suggesting that p38α triggers visceral fat browning. We also found that p38α in sWAT inversely correlates with UCP1. This is in accordance with results observed in mouse models, in which we found a decrease of all p38s after HFD in all fat depots. However, the levels of UCP1 expression in these human fat depots is quite low judging by the low Ct obtained (higher than 29), and evaluation of UCP1 protein expression in human fat depots would be necessary. Moreover, further studies to determinate the expression of p38 family members and upstream kinases in other human fat depots would help us to understand the role of these kinases in human adipocytes.
It has been proposed that p38α induces adipogenesis [38][39][40]. However, using genetically modified animals, we showed here that lack of p38α in preadipocytes did not affect their differentiation to adipocytes, nor did it affect changes in the differentiation markers evaluated in the major fat depots. This capacity of cells lacking p38α to still differentiate to adipocytes could be due to the hyperactivation of the other members of the family: p38γ and p38δ. In fact, it has been shown that p38 isoforms can compensate for each other [30]. Here, we demonstrated the cell-autonomous and opposite effects of 2 p38 isoforms in adipocytes, p38α and p38δ. The cell-specific actions of p38α in each fat depot could be explained by the specific expression pattern of p38 family members-p38α being the main isoform expressed in eWAT, whereas p38δ n = 10 mice; p38α Fab-KO n = 6 mice). Lower panels show representative infrared thermal images. (D) Immunoblot analysis of UCP1 protein levels in lysates from BAT. Quantification is shown in the lower panel. (E) Fasting and fed blood glucose in Fab-Cre and p38α Fab-KO mice fed the HFD at 30˚C (mean ± SEM; Fab-Cre n = 10 mice; p38α Fab-KO n = 6 mice). (F) GTT in HFD-fed Fab-Cre and p38α Fab-KO at 30˚C. Blood glucose concentration was measured in mice given intraperitoneal injections of glucose (1 g/kg of total body weight) (mean ± SEM; Fab-Cre n = 10 mice; p38α Fab-KO n = 6 mice). Statistically significant differences between Fab-Cre and p38α Fab-KO mice are indicated: Ã p < 0.05; ÃÃ p < 0.01; ÃÃÃ p < 0.001 (t test or Welch's test when variances were different). See also S1 Data. BAT, brown adipose tissue; eWAT, epididymal fat; GTT, glucose tolerance test; HFD, high-fat diet; IR temperature, infrared temperature; iWAT, inguinal fat; pWAT, perirenal fat; sWAT, subcutaneous fat; UCP1, uncoupling protein 1; WAT, white adipose tissue. https://doi.org/10.1371/journal.pbio.2004455.g005 Regulation of thermogenesis by p38 signalling We also evaluated the controversial role of p38α in GLUT4 translocation [41][42][43]. Under ND, insulin-induced GLUT4 translocation was the same in both control and p38α Fab-KO mice. However, p38α Fab-KO mice maintained the insulin-induced translocation after the HFD, perhaps due to the fact that these animals did not gain weight and were protected against dietinduced insulin resistance. In fact, our data suggest that these mice are more glucose tolerant using a dose of glucose based on their total body weight.
Due to the potential clinical implications of these results, it would be necessary to further evaluate the function of each p38 family member in browning to better understand how this pathway controls adipose tissue metabolism.
In summary, we have demonstrated that p38α and p38δ in adipose tissue have opposite roles: p38α negatively regulates BAT thermogenesis, energy expenditure, and body weight, while p38δ induces thermogenesis in BAT in response to several physiological stimuli. These results have potential clinical implications because inhibition of p38α or activation of p38δ might be of therapeutic interest against obesity.

Ethics statement
This population study was approved by the Ethics Committee of the University Hospital of Salamanca and the Carlos III (CEI PI 09_2017-v3) with the all subjects providing written informed consent to undergo visceral fat biopsy under direct vision during surgery. Data were collected on demographic information (age, sex, and ethnicity), anthropomorphic measurements (BMI), smoking and alcohol history, coexisting medical conditions, and medication use.

Study population and sample collection
For the analysis of visceral fat, the study population included 71 patients (58 adult patients with BMI !35), while for the analysis of sWAT, the study population included 170 patients mice after 8 weeks of ND measured by NMR (mean ± SEM; Fab-Cre n = 6 mice; p38δ Fab-KO n = 5 mice). (E) Comparison of energy balance between ND-fed Fab-Cre and p38δ Fab-KO mice. ND-fed mice were examined in a metabolic cage over a 3-day period to measure FI and EE. FI (upper left panel; mean ± SEM; Fab-Cre n = 12 mice; p38δ Fab-KO n = 10 mice) and EE (upper right panel; mean ± SEM; Fab-Cre n = 6 mice; p38δ Fab-KO n = 6 mice) over 2 days were corrected by lean mass. EE expressed as ANCOVA analysis (lower left panel; mean ± SEM; Fab-Cre n = 9 mice; p38δ Fab-KO n = 12 mice) and hour by hour over a 48-hour period (lower right panel; mean ± SEM; Fab-Cre n = 12 mice; p38δ Fab-KO n = 12 mice) are also shown. (F) Body temperature of ND-fed Fab-Cre and p38δ Fab-KO mice (mean ± SEM; Fab-Cre n = 9 mice; p38δ Fab-KO n = 11 mice). Skin temperature surrounding interscapular BAT in ND-fed Fab-Cre and p38δ Fab-KO . Right panels show representative infrared thermal images (mean ± SEM; Fab-Cre n = 10 mice; p38δ Fab-KO n = 12 mice). (G) Adipocytes differentiated from interscapular BAT were stimulated with 100 nM T3 for 48 hours. Immunoprecipitation from cell lysates of p38δ were evaluated by immunoblot with antibodies against phospho-p38 and p38δ. Adipocytes differentiated from sWAT were stimulated with 1 μM NE for 1 hour, and p38 phosphorylation was analysed by immunoblot. (H) Control mice (C57BL/6) were exposed to cold (4˚C) for the indicated time, and phosphorylation of the different p38s in BAT was evaluated by immunoblot (n = 5 for each group; representative blot presented). (I) Body temperature of ND-fed Fab-Cre and p38δ Fab-KO mice exposed to cold (4˚C) for 1 hour (mean ± SEM; Fab-Cre n = 10 mice; p38δ Fab-KO n = 8 mice). Skin temperature surrounding interscapular BAT in ND-fed Fab-Cre and p38δ Fab-KO after 1 hour of cold exposure. Right panels show representative infrared thermal images (mean ± SEM; Fab-Cre n = 9 mice; p38δ Fab-KO n = 8 mice). Ã p < 0.05; ÃÃ p < 0.01; ÃÃÃ p < 0.001 (t test). See also S1 Data. BAT, brown adipose tissue; Creb, cAMP response element-binding; EE, energy expenditure; FI, food intake; HFD, high-fat diet; IR temperature, infrared temperature; ND, normal-chow diet; NE, norepinephrine; NMR, nuclear magnetic resonance; qRT-PCR, quantitative real-time polymerase chain reaction; sWAT, subcutaneous fat.  Patients were excluded if they had a history of alcohol use disorders or excessive alcohol consumption (>30 g/day in men and >20 g/day in women) or had chronic hepatitis C or B. Control subjects (n = 13 for visceral fat study; n = 30 for sWAT study) were recruited among patients who underwent laparoscopic cholecystectomy for gallstone disease. Before surgery, fasting venous blood samples were collected for measuring complete cell blood count, total bilirubin, aspartate aminotransferase (AST), alanine aminotransferase (ALT), total cholesterol, high-density lipoprotein, low-density lipoprotein, triglycerides, creatinine, glucose, and albumin (S1 and S2 Tables).

Animal models
Mice with a germ-line mutation in Mapk14 (p38α) and Mapk13 (p38δ) have been reported before [44,45]. These animals were crossed with Tg (Fabp4-cre)1Rev/J [46] line or B6.FVB-Tg (Ucp1-cre)1Evdr/J [24] on the C57BL/6J background (Jackson Laboratory) to generate the mice lacking p38α or p38δ in adipose tissue (both WAT and BAT or just in BAT, respectively). All mice were maintained on a C57BL/6J background (back-crossed 10 generations). Genotype was confirmed by PCR analysis of genomic DNA. Mice were fed with an ND or an HFD, Research Diets Inc.) for 8 weeks ad libitum. For fat expansion measurement, mice were treated with BrdU (0.4 mg/ml; Sigma) in the drinking water (water was refreshed every 3 days) during the first week of a 6-week HFD. For temperature experiments, mice were housed at 30˚C for 8 weeks while feeding an HFD in case of thermoneutrality analysis. Mice were exposed to 4˚C for 1 hour, 1 day, or 1 week in case of cold adaptation studies.

Cell culture
Immortalised and primary brown preadipocytes from WT, Fab-Cre, p38α Fab-KO , and p38δ-KO mice were differentiated to brown adipocytes in 10% FCS medium supplemented with 20 nM insulin, 1 nM T3, 125 μM indomethacin, 2 μg/ml dexamethasone, and 50 mM IBMX for 48 hours and maintained with 20 nM of insulin and 1 nM of T3 for 8 days. For some experiments, cultures were incubated with 100 nM T3 for 48 hours before extraction.

Analysis of mitochondrial function
Primary brown preadipocytes were plated and differentiated in gelatin-coated (0.1%) 96 seahorse plates. MitoStress oxygen consumption rate (OCR) was assessed in XF medium containing 25 mM glucose, 2 mM L-glutamine, and 1 mM sodium pyruvate using a XF-96 Extracellular Flux Analyzers (Seahorse Bioscience, Agilent Technologies). Cells were stimulated with following drugs: NE or ISO, oligomycin, FCCP, and antimycin A plus rotenone (1 μM finally; all from Sigma Aldrich). The protocol for the all drugs followed a 3-minute mix, 2-minute wait, and 3-minute measure cycle that was repeated 3 times. After the analysis, data were induced OCR. (F) OCR induced by NE and ISO in differentiated brown adipocytes from Fab-Cre and p38α Fab-KO mice was abolished by pretreatment with BIRB796 (10 μM) for 1 hour (mean ± SEM; Fab-Cre n = 6 or p38α Fab-KO n = 7 wells treated with NE; and Fab-Cre n = 7 or p38α Fab-KO n = 8 wells treated with ISO). See also S1 Data. BAT, brown adipose tissue; FCCP, carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone; ISO, isoproterenol; NE, norepinephrine; OCR, oxygen consumption rate; qRT-PCR, quantitative real-time polymerase chain reaction; WT, wild-type. https://doi.org/10.1371/journal.pbio.2004455.g007 Regulation of thermogenesis by p38 signalling Fig 8. Regulation of browning and BAT activation by p38 pathway. Graphical abstract summarising the role of p38 isoforms in adipose tissue. In eWAT, p38α activates browning through the phosphorylation of Creb and ATF2 increasing UCP1 expression. In iWAT and BAT, p38α activation inhibits p38γ and p38δ and in consequence reduces browning and BAT activation, respectively, by down-regulation of UCP1. ATF2, activating transcription factor 2; BAT, brown adipose tissue; Creb, cAMP response element-binding; eWAT, epididymal fat; iWAT, inguinal fat; UCP1, uncoupling protein 1. normalised to protein level assessed by Bradford quantification. Basal Respiration Capacity (OCR basal − OCR nonmitochondrial) and oxygen consumption in response to NE (OCR NE − OCR basal) or ISO (OCR ISO-OCR basal) were calculated. For some experiments, cultures were pretreated with 10 μM BIRB796 for 1 hour.
For the immunoprecipitation assay, cell extracts were incubated with 4 μg of anti-p38 delta coupled with protein-G-Sepharose. After an overnight incubation at 4˚C, the captured proteins were centrifuged at 10,000 g, the supernatants discarded, and the beads washed 4 times in lysis buffer. Beads were boiled for 5 minutes at 95˚C in 10 μl sample buffer. The antibodies employed were anti-phospho p38 and anti-p38δ (Santa Cruz, sc7585). Immune complexes were detected by enhanced chemiluminescence (NEN).

GTT
Overnight-starved mice were injected intraperitoneally with 1 g/kg of body weight of glucose, and blood glucose levels were quantified with an Ascensia Breeze 2 glucose meter at 0, 15, 30, 60, 90, and 120 minutes post injection. Alternatively, GTT was performed injecting intraperitoneally 1 g/kg of lean mass of glucose.

ITT
ITT was performed by injecting intraperitoneally 0.75 IU/kg of insulin at mice starved for 1 hour and detecting blood glucose levels with a glucometer at 0, 15, 30, 60, 90, and 120 minutes post injection.

Indirect calorimetry system
Energy expenditure, respiratory exchange, and food intake were quantified using the indirect calorimetry system (TSE LabMaster, TSE Systems, Germany) for 3 days.

Temperature
Body temperature was detected by a rectal thermometer (AZ 8851 K/J/T Handheld Digital Thermometer-Single, AZ Instruments Corp., Taiwan).
BAT-adjacent interscapular temperature was quantified by thermographic images using a FLIR T430sc Infrared Camera (FLIR Systems, Inc., Wilsonville, OR) and analysed through FlirIR software.

Nuclear magnetic resonance analysis
Body, fat, and lean mass were quantified by nuclear magnetic resonance (Whole Body Composition Analyzer; EchoMRI, Houston, TX) and analysed by ImageJ software.

Triglyceride measurement
Blood triglyceride content was quantified using a Dimension RxL Max analyser (Siemens). For triglyceride analysis in cells, brown adipocyte cultures were lysed in isopropanol, centrifuged at 10,000 g for 15 minutes at 4˚C, and triglycerides were detected in the supernatant with a commercial kit (Sigma).

DNA isolation
Brown adipocyte cells were scraped in PBS and pellet lysed in TNES buffer supplemented with Proteinase K (20 mg/ml) overnight at 55˚C. Reaction was stopped with sodium chloride 6 M and samples centrifuged 5 minutes at 13,000 g. DNA was precipitated in supernatants with 100% ethanol and washed with 70% ethanol. After drying, DNA was resuspended in DNase free water, quantified, and analysed by RT-PCR. Mitochondrial DNA was detected using primers for COII and nuclear DNA, using primers for Sdh1 (S3 Table).

qRT-PCR
RNA 500ng-extracted with RNeasy Plus Mini kit (Quiagen) following manufacturer instructions-was transcribed to cDNA, and qRT-PCR was performed using Fast Sybr Green probe (Applied Biosystems) and the appropriated primers in the 7900 Fast Real Time thermocycler (Applied Biosystems). Relative mRNA expression was normalised to Gapdh mRNA measured in each sample. Primers used are listed in S3 Table. Histology staining Fresh livers, brown, and epididymal white fat were fixed with formalin 10%, included in paraffin, and cut in 5 μm slides followed by a haematoxylin-eosin staining.
Fat droplets were detected by oil red staining (0.7% in propylenglycol) in 8 mm slides included in OCT compound (Tissue-Tek) and in differentiated brown and white adipocytes.