Both proliferation and lipogenesis of brown adipocytes contribute to postnatal brown adipose tissue growth in mice

Brown adipose tissue (BAT) is the primary non-shivering thermogenesis organ in mammals, which plays essential roles in maintaining the body temperature of infants. Although the development of BAT during embryogenesis has been well addressed in rodents, how BAT grows after birth remains unknown. Using mouse interscapular BAT (iBAT) as an example, we studied the cellular and molecular mechanisms that regulate postnatal BAT growth. By analyzing the developmental dynamics of brown adipocytes (BAs), we found that BAs size enlargement partially accounts for iBAT growth. By investigating the BAs cell cycle activities, we confirmed the presence of proliferative BAs in the neonatal mice. Two weeks after birth, most of the BAs exit cell cycle, and the further expansion of the BAT was mainly due to lipogenesis-mediated BAs volume increase. Microscopy and fluorescence-activated cell sorting analyses suggest that most BAs are mononuclear and diploid. Based on the developmental dynamics of brown adipocytes, we propose that the murine iBAT has two different growth phases between birth and weaning: increase of BAs size and number in the first two weeks, and BAs size enlargement thereafter. In summary, our data demonstrate that both lipogenesis and proliferation of BAs contribute to postnatal iBAT growth in mice.

In this study, we investigated the growth of iBAT in postnatal mice by dissecting the developmental dynamics of the BAs. Our data indicate that the iBAT has two different growth phases between birth and weaning: the iBAT grows by increasing brown adipocytes (BAs) number and size in the first two weeks after birth (first growth phase), and expands through BAs hypertrophic growth thereafter (second growth phase). Furthermore, our data suggest that a small fraction of BAs are proliferative in the first growth phase. In summary, this study extends our understanding of postnatal iBAT growth, and suggests the presence of proliferating BAs in newborns.

Results
Lipogenesis of BAs contributes to iBAT growth. In mice, the iBAT primarily grows by increasing BAs number during embryogenesis 18 . We checked the postpartum iBAT growth dynamics, which showed that the iBAT grew rapidly in the first 30 days after birth and its weight increased around 11 times from postnatal day 1 (P1) to P30 (Fig. 1A). The iBAT weight and body weight ratio decreased in the first 8 days after birth but increased between P8 and P30 (Fig. 1B,C), indicating that the cellular mechanisms controlling iBAT growth might vary at different developmental stages. Generally, the adipose tissue has two different growth mechanisms: lipogenesis growth (caused by adipocyte enlargement) and proliferation growth (due to adipocyte number increase) 19 . To understand the cellular mechanisms controlling postnatal iBAT development, we first examined the BAs size at different postnatal stages: postnatal day 1, 4, 8, 12, 15 and 30. MCT1 (Monocarboxylate transporter 1) is a protein enriched in the BAs plasma membrane 20 . We used MCT1 antibody to visualize the cell borders of BAs 21 . From P1 to P4, the BAs cross-sectional area increased around 1.8 times (from 54.29 ± 7.447 µm 2 to 90.66 ± 5.85 µm 2 ). Compared to P4, the BAs cross-sectional area did not change significantly at P8 (103.2 ± 8.64) and P12 (109.1 ± 15.39 µm 2 ). From P12 to P15 and from P15 to P30 (135.1 ± 9.24 vs 200 ± 25.02 µm 2 ), the BAs cross-sectional area increased significantly in both stages (Fig. 1D,E). If hypertrophic growth of BAs was the only factor controlling iBAT expansion, the fold changes of BAs volume and iBAT weight (A, C and E) One-way ANOVA post-hoc Tukey's multiple comparisons test, *P < 0.05; ***P < 0.001. (F) Folds change comparison between iBAT weight and brown adipocyte volume. iBAT weight and brown adipocytes volume at different ages were normalized to the corresponding values of P1, respectively. Student t test, *P < 0.05; **P < 0.01. N = 4-6.

Scientific Reports
| (2020) 10:20335 | https://doi.org/10.1038/s41598-020-77362-x www.nature.com/scientificreports/ should be similar at each growth stages. We assumed that the BAs were spherical and calculated the BAs volume fold changes with the following equation: V n / V 1 = (A n / A 1 ) 3/2 . V n indicates the BA volume at different growth stages; V 1 indicates the BA volume at P1. A n represents the average BA cross-sectional area at different postnatal days; A 1 represents the average BA cross-sectional area at P1. Compared to P1, the BA volume and iBAT weight had similar fold increases at P4 and P8 (Fig. 1F). At P12, the increase folds of BAs volume and iBAT weight started to diverge, with BAs volume and iBAT weight increasing 2.8 and 4 folds, respectively. The growth difference of BA volume and iBAT weight continued through P15 and P30. At P30, the BAs volume and iBAT weight were 7 and 11 times higher than that of the P1, respectively (Fig. 1F). These data indicate that lipogenesis of BAs is not the only cause of iBAT expansion, and suggest that increase of BAs number may also contribute to iBAT growth.
Lipogenesis is associated with BAs volume enlargement. In sheep, the increase of perirenal BAT weight and lipid deposition coincides between P7 and P30, but the protein content of BAT remains constant during this growth period 22 , indicating that lipogenesis is highly associated with BAT growth. We then asked the question whether lipogenesis was the main driver of BAs hypertrophic growth. BODIPY 493/503 (BODI) is a well-defined lipophilic fluorescent probe for staining neutral lipids 23 . By double staining the iBAT sections with BODI and MCT1, we were able to assess the individual BA's lipid contents: BODI staining visualized the typical multilocular lipid droplets of BAs, and MCT1 immunofluorescence signals outlined the cell borders ( Fig. 2A).
The density and size of the typical multilocular lipid droplets in the BAs was low at P1, increased at P8 and P15, and was highest at P30 ( Fig. 2A). To quantify the BAs' lipid contents, we measured the cell BODI fluorescence intensity (CBFI). Compared to the CBFI at P1, the CBFI at P8, P15, and P30 increased 5.2, 7.5 and 14 times, respectively (Fig. 2B). We further measured the triglycerides (TG) contents of different ages iBAT. Consistent with the BODI staining results, from birth to weaning, the TG contents of iBAT increased with the age substantially (Fig. 2C). These data suggest that formation of lipid droplets is highly associated with BAs growth. BAs increase their lipid deposition through fatty acid uptake or de novo fatty acids synthesis 24 . To understand the temporal regulation of lipogenesis, we interrogated the expression dynamics of several crucial lipogenesis genes that control lipid uptake, transport and synthesis.
Lipoprotein lipase (LPL) is a key enzyme that hydrolyses the triacylglycerol components of circulating lipoproteins 25 , and the Lpl gene was highly expressed in BAT 26 . Cluster of differentiation 36 (CD36) is a transmembrane protein that transports fatty acids from the extracellular environment into the cell. Inside the cell, fatty acid binding proteins (FABPs), such as FABP3 27 and FABP4 (also known as aP2) 28 , bind and transport the free fatty acids to the mitochondria for oxidation or to the lipid droplets for storage. From P1 to P8, the expression of Lpl and Cd36 did not change much; from P8 to P12, both genes decreased significantly; from P15 to P30, Lpl increased substantially, and Cd36 showed out a mild increase (Fig. 2D). The expression of Fabp3 was low at P1, increased significantly from P1 to P4, and increased substantially from P15 to P30 (Fig. 2E). The expression of Fabp4 was constant from P1 to P15, but increased significantly from P15 to P30 (Fig. 2E). The de novo fatty acids synthesis process is oscillated by multiple key enzymes. We tested the expression of two crucial lipid synthesis genes: acetyl-CoA carboxylase 1 (Acc1) and fatty acid synthetase (Fasn) 29 . As crucial de novo fatty acids synthesis genes, both Acc1 and Fasn had low expression from P1 to P15 but increased robustly from P15 to P30 (Fig. 2F). The expression dynamics of these genes suggest that the BAs primarily increase their lipid deposition through fatty acids uptake in the first two weeks after birth, and that the de novo fatty acids synthesis in BAs happens after P15.
The lipid deposition is controlled by both lipogenesis and lipolysis. Our data suggest that the postnatal iBAT development process favors lipogenesis over lipolysis. To test this notion, we measured the expression of several lipolysis related genes, including Patatin like Phospholipase Domain Containing 2 (Pnpla2, also known as Atgl), Acyl-CoA Oxidase 1 (Acox1), and Carnitine palmitoyltransferase 1b (Cp1tb). Pnpla2 encoded protein catalyzes the initial step of triglyceride hydrolysis 30 . The protein products of Acox1 and Cpt1b are rate limiting enzymes of fatty acid (FA) β-oxidation, which catalyzes the first step of FA β-oxidation and facilitates the transfer of FA from cytoplasm to mitochondria, respectively 31,32 . During the postnatal iBAT growth, the expression of Pnpla2 did not change substantially; Acox1 and Cpt1 decreased with age ( Fig. 2G). We further examined the expression of Pparg and Pgc1a, two master regulators of adipose tissue development. In the four measured time points, the expression of Pgc1a was low at P1, robustly increased at P8, and decreased thereafter (Fig. 2H). The mRNA levels of Pparg did not change significantly between different ages iBAT, but the PPARγ protein level increased with age (Fig. 2H,I).
Together these data suggest that, during postnatal iBAT growth, lipid droplet deposition is the primary reason of BAs volume increase, and indicate that the lipogenesis and FA oxidation gene expression programs increase and decrease with age, respectively.
BAs have DNA synthesis activities in the first two weeks after birth. Our data suggest that lipogenesis of BAs is not the only cause for iBAT expansion. We suspected that proliferation of BAs might also contribute to iBAT growth. To test this hypothesis, we analyzed the BAs' uptake rate of 5-ethynyl-2′-deoxyuridine (EdU), a nucleotide analog that labels cells passing through S phase. Six time points were included in this study: P1, P4, P8, P12, P15 and P30. Uncoupling protein 1 (UCP1) is a well-known marker of BAs. We used UCP1 antibody to label brown adipocytes. In the iBAT, EdU uptake was detected at P1, P4, P8, P12, but not at P15 and P30 (Fig. 3A). High magnification microscopy imaging showed that EdU positive cells contained both BAs (UCP1 postive) and non-BAs (UCP1 negative) (Suppl. Figure 1A). The BAs EdU uptake rate was low at P1 (1.072 ± 0.489%), increased five times at P4 (4.977 ± 0.911%), reached to a peak at P8 (13.24 ± 1.574%), and decreased to a lower level at P12 (1.57 ± 0.43%). EdU uptake was rarely detected at P15 and P30 ( Fig. 2A,B). These data strongly suggest that a fraction of BAs have DNA synthesis activity in the first two weeks after birth.

Scientific Reports
| (2020) 10:20335 | https://doi.org/10.1038/s41598-020-77362-x www.nature.com/scientificreports/ To further test whether BAs have any changes in the cell cycle activity in the first two weeks after birth, we measured the expression of two sets cell cycle related genes. The first gene set includes three cell cycle genes: Cdk1, CycB1 and CycA2, which are required for executing the cell cycle. The second gene set has three cell cycle inhibitor genes: Rb1, a G1/S phase check point gene 33 ; Cdkn1a and Cdkn1b, two cyclin dependent kinase inhibitor genes 34 . In line with the iBAT EdU incorporation data, the expression of the three cell cycle genes peaked at P8, dramatically dropped at P12 and remained low thereafter (Fig. 3C). Although the overall expression dynamics of the three cell cycle inhibitor genes were distinct from each other, one common feature existed: they all had lower expression at P8 and higher expression at P30 (Fig. 3D). The expression patterns of these cell cycle related genes suggest the presence of proliferative BAs in the neonatal mice.   www.nature.com/scientificreports/ In the neonatal pups, besides originating from the preexisting BAs, the newly formed BAs could originate from two other cell populations: the adipocyte progenitor cells 35 , and dedifferentiated adipocytes 36 . If the newly formed BAs were derived from progenitor cells, we should expect a high expression of progenitor cell marker genes around one week after birth. If the BAs need to undergo dedifferentiation before they proliferate, we should detect a decrease of BAs marker gene expression at P8. Dlk1, Prdm16 and Pdgfra are three genes highly expressed in progenitor cells 37,38 . Our data showed that the expression of Prdm16 peaked at P4 but was relatively constant from P8 to P30 (Fig. 3E). The expression of Dlk1 decreased with age, and the expression of Pdgfra kept relatively constant from P1 to P30 (Fig. 3E). Our data showed that Ucp1 had a peak expression at P8 (Suppl. Figure 1B), suggesting that the BAs might not undergo dedifferentiation at this age. These gene expression data suggest that the high expression of cell cycle genes in the neonatal mice is neither due to enrichment of adipocyte progenitor cells nor due to BAs dedifferentiation, and further indicate the presence of proliferative BAs.
A subset of BAs are proliferative in the first week after birth. Our data suggest that the proliferation of existing BAs might be a reason for the proliferation growth of iBAT. Histone H3 phosphorylated on serine 10 (pH3) is a marker of M phase of the cell cycle. To further confirm the presence of proliferative BAs in neonatal mouse pups, we examined the iBAT for pH3 signals. In both P4 and P8 iBAT, pH3 positive (pH3 + ) BAs were detectable (Fig. 4A). Consistent with the EdU uptake data (Fig. 3B), P8 iBAT had more pH3 + BAs than P4 iBAT (Fig. 4B).
As an alternative strategy, we used fluorescence activated cell sorting (FACS) to analyze the cell cycle activity of BAs at P8. Rosa 26 tdTomato is a reporter mouse line, which expresses strong tdTomato (tdTom) fluorescence protein following Cre-mediated recombination 39 . Ucp1::Cre transgenic mice express Cre recombinase in the BAs 40 . We crossed the Ucp1::Cre mouse line with the Rosa 26 tdTomato mice to genetically label BAs with tdTom (Fig. 4C). After dissociating the P8 iBAT following a published protocol 41 , we used tdTom + cells for cell cycle analysis. The presence of typical raspberry shape tdTom + BAs confirmed the success of our iBAT dissociation procedure (Fig. 4D). We then ran FACS with the dissociated BAs, and gated tdTom + BAs for cell cycle analysis (Fig. 4E). Published data have shown that dissociation of adipose tissue generates free lipid droplets which can be counted as FACS events 41 . During our FACS analysis, we noticed that most of the recorded events were tdTom negative (Fig. 4E). However, we were not sure how many of these tdTom negative events were cells or free lipid droplets. To focus on analyzing the brown adipocytes, we only gated for cells with high tdTom fluorescence (Fig. 4E), and then measured the DNA contents of these gated cells. Although most BAs were at G1 phase (94.83 ± 0.88%), a small fraction of BAs undergoing DNA synthesis (S phase, 2.375 ± 0.72%) and mitosis (G2/M phase, 1.325 ± 0.35%) were detected (Fig. 4F).
Together, these data suggest that a subset of BAs are proliferative in the neonatal mice.
BAs are primarily mononuclear and diploid. Karyokinesis and cytokinesis are two related but different processes, and karyokinesis without cytokinesis may lead to polyploidy (more than two copies of DNA) or multinuclearity (more than one nucleus). Although we confirmed the presence of mitotic BAs in the neonatal mice, it was not clear whether these cells would complete the cell cycle and yield out new daughter cells. If the mitotic BAs in the neonates fail to complete karyokinesis or cytokinesis, we should be able to detect polyploidy or multinuclearity in the cell-cycle-quiescent BAs of young mice. In the iBAT of P15 or P30 mice, no more DNA synthesis activity was detected (Fig. 3A), suggesting that most of the BAs have already exited the cell cycle two weeks after birth. Therefore, we examined the BAs from P15 and P30 mice for polyploidy and nuclearity examination. Dissociated BAs from Ucp1::Cre;Rosa 26 tdTomato mice were used for the following studies. We first used microscopy method to examine the BAs nuclearity. At P15, 3 out of 550 tdTom + cells were binuclear, which accounted for 0.55% of the examined tdTom + cells (Fig. 5A, 5B). At P30, we detected 3 binuclear BAs out of 331 tdTom + cells (Fig. 5A, B, Suppl. Figure 2B). However, in 153 and P1 and 251 P8 brown adipocytes, we did not detect binuclear cells (Fig. 5B, Suppl. Figure 2A). We further checked the ploidy of the isolated tdTom + cells with FACS. In the gated P30 tdTom + cells (Fig. 5C), 96 ± 2.7% were diploid and 1.47 ± 1% were tetraploid (Fig. 5D), respectively. These data demonstrate that the majority of BAs from mouse iBAT contain a single diploid nucleus.
Together with the EdU incorporation data, these results further indicate that some BAs are proliferative in the first two weeks after birth, and suggest that DNA synthesis can be used as a marker for testing BAs proliferation.

Discussion
In this study, using neonatal and juvenile mice, we interrogated the cellular and molecular mechanisms that control iBAT growth. Our data demonstrate that both lipogenesis and proliferation of BAs account for iBAT growth in the first two weeks after birth, and that lipogenesis but not proliferation of BAs contributes to iBAT expansion between two weeks after birth and weaning (summarized in Fig. 5E).
Published data showed that impaired cBAT growth had a negative effect on adult energy homeostasis 16,42 , suggesting that increasing the mass of cBAT may be an effective way to prevent obesity. Therefore, it is essential to delineate the cellular mechanisms controlling cBAT growth. As a basic question, it is not clear how the mass of cBAT changes after birth. Theoretically, the cBAT expands depends on the increase of brown adipocyte number and volume. The number of brown adipocytes may increase through two different mechanisms: proliferation of the already exist brown adipocytes and differentiation of the adipocyte progenitor cells.
In this study, by interrogating the iBAT growth in the neonatal mice, we expanded the perceptiveness of cBAT development. Our data suggest that both proliferation and lipogenesis of BAs contribute to postnatal iBAT growth, and that lipogenesis mainly accounts for BA hypertrophic growth. Based on our new data, we propose the presence of two different iBAT growth phases in early life rodents: 1) naive growth phase, and 2) Scientific Reports | (2020) 10:20335 | https://doi.org/10.1038/s41598-020-77362-x www.nature.com/scientificreports/ maturation phase. In the naive growth phase, increase of BAs number and size coincides; in the maturation phase, lipogenesis-driven BAs hypertrophic growth contributes to the iBAT mass expansion (Fig. 5E). Our data support the notion that a fraction of embryo-derived BAs are proliferative in the iBAT naive growth phase, which at least partially accounts for the postnatal growth of iBAT. However, lineage tracing data are required to quantify how many newly formed BAs are derived from existing BAs. Based on the published data, two distinct opinions exist regarding to the proliferation capability of terminal differentiated BAs. One category of data suggests that terminal differentiated BAs do not proliferate, and that the newly formed BAs in the adult iBAT mostly originate from PDGFRα + progenitor cells 35 . The other category of data supports the view that differentiated BAs are proliferative. For example, overexpression of a cell cycle inhibitor p27 (encoded by Cdkn1b) in adipocytes largely reduces the iBAT depot size without affecting BAs size 43 . The controversy of these two tires observations can now be reconciled with our new hypothesis: the overexpression of p27 suppresses the proliferation of BAs in the naive growth stage, which results in the paucity of iBAT in the adult mice. In the adult mice, embryo-derived BAs are no longer proliferative and newly formed BAs are exclusively derived from activated adipocyte progenitor cells.
Lipogenesis and lipolysis are two biochemistry processes determining the amount of lipid droplets in adipocytes 44 . Our data show that BA lipogenesis happens throughout the two postnatal growth phases, which is evidenced by the increase of multilocular lipid droplets in each individual BAs and the increase of TG levels in the www.nature.com/scientificreports/ iBAT. Nevertheless, the underlying mechanism may be different. Compared with the genes at P30, the expression levels of lipid synthesis genes (such as Acc1 and Fasn) are constant and much lower in the naive growth phase (Fig. 2). Meanwhile, the expression of Cd36, a crucial fatty acid transporter gene, is much higher in the naive growth phase than in the mature phase. These gene expression data suggest that the brown adipocytes lipogenesis in naive growth phase mainly depends on extracellular lipid uptake, and that de novo lipid synthesis becomes the primary lipogenesis mechanism in maturation phase. This hypothesis is consistent with the fact that the pups mainly consume milk in the first two weeks, which contains much more crude fat than the regular chow diet 45 . Together, these data suggest that increased lipogenesis is the main reason of brown adipocytes volume increase. Precocial species are born with a mature hypothalamic-pituitary-adrenal (HPA) axis after a long gestation period, and therefore are able to rapidly turn on non-shivering thermogenesis after birth 7 . Altricial species are born with immature HPA axis after a short gestation period, and their pups mainly maintain body temperature by huddling together instead of switching on non-shivering thermogenesis. Large animals such as sheep and www.nature.com/scientificreports/ humans are precocial, and rodents such as rats and mice are altricial 17 . In humans, the cBAT depots are abundant in children under 10 years old 10 , and higher brown-like BAT composition in infants is associated with lower body fat gain in early life 13 . The current study shows that the proliferation of BAs happens in the first two weeks after birth in mice, which corresponds to the first one and a half years in human 46 . Our observation raised the possibility that a cBAT naive growth phase might exist in human infants, and that this naive growth process might be essential for preparing fully functional BAT for childhood. It is also possible that precocial and altricial species share different BAT growth mechanisms, and that proliferative naive BAs only exist in altricial neonates.
To have a definite answer, future works are required to examine the BAs proliferation in the neonates of precocial species such as sheep.
One of the major findings of this study is that a fraction of brown adipocytes are proliferative in the iBAT naive growth phase. It will be interesting to decipher the molecular mechanisms that control classical brown adipocytes proliferation during this growth period, which will be insightful for developing new therapeutic strategies of generating more thermogenic brown adipocytes. Beta-adrenergic signaling is one of the pathways that regulates brown adipocyte proliferation 47,48 ; however, because the HPA axis is not mature in the new born mouse pups, it is unlikely that the proliferation of classical brown adipocytes is driven by beta-adrenergic signaling in the iBAT naive growth phase. Growth hormone (GH) and Insulin growth factor (IGF) signaling pathways are essential for postnatal development 49,50 , and disruption of IGF but not GH signaling in the adipose tissue results in severe iBAT paucity 51,52 , suggesting that IGF signaling may regulate the classical brown adipocytes proliferation during the iBAT naive growth phase. Future studies are secured to assess whether manipulating the IGF signaling affects the proliferation of classical brown adipocytes.

Material and Methods
Mice. All animal procedures were approved by the Institutional Animal Care and Use Committee of Masonic Medical Research Institute. All experiments were performed in accordance with relevant guidelines and regulations. C57BL/6 J background UCP1::Cre 40 and Rosa 26 tdTomato mouse lines 39 were reported previously, and were purchased from the Jackson Laboratory. Otherwise mentioned, C57BL/6 J background wild type mice were used throughout this study. All mice were kept at room temperature ( 22 ± 1 °C ) with a 12 h light/dark cycle. Pups were weaned at 28 days of age.
Histology and immunostaining. Interscapular brown adipose tissue (iBAT) was fixed in 4% PFA. iBAT was cryoprotected with 30% sucrose and embedded in OCT. 10 μm sections were used for immunofluorescence staining. Antibody sources were listed in Supplementary Table 1. EdU was administered subcutaneously at 5 µg/ gram body weight 24 h before iBAT collection. EdU was detected with Click-iT chemistry (Invitrogen). After primary and secondary antibody staining, sections were treated with Sudan Black B (0.5% in 70% ethanol) for 5 min to reduce background autofluorescence 53 . Imaging was performed on a Zeiss Confocal LSM 700.
Cell size and fluorescence intensity measurements. The BAs's borders were visualized with MCT1 antibody, and the BAs' sizes were measured with Fiji software. For measuring BA's lipid contents, iBAT sections were double stained with MCT1 antibody and BODIPY 493/503. The cellular BODIPY fluorescence intensity (CBFI) was quantified based on an online protocol (https ://theol b.readt hedoc s.io/en/lates t/imagi ng/measu ringcell-fluor escen ce-using -image j.html).

BAs isolation.
BAs isolations were carried out following previous published protocol 41 with minor modifications. Briefly, iBAT from P15 or P30 mouse pups was digested in a 20 ml flask, which was equipped with a short metal bar and placed on a magnetic stirrer. The digestion was carried out in a 36 °C incubator. The digestion buffer contained 3.5 mg/ml Dispase II, 10 mM CaCl 2 and 1 mg/ml Collagenase II in PBS solution. 12 min after digestion, the cell suspension was transferred into a 50 ml falcon tube and filtered through a 70 µm strainer. Fresh digestion solution was added into the flask for two more round digestions. Filtered cells were incubated in PBS containing 3% BSA and 2 mM EDTA. The floated cells were fixed with 2% PFA at room temperature for 15 min. Adipocytes were subsequently stained for microscopy or FACS analysis with Hoechst to visualize nuclei.

Measurement of iBAT triglycerides (TGs).
Lipid from iBAT of different ages mice were extracted following a published protocol 54 . Briefly, iBAT was suspended in phosphate-buffered saline (PBS) buffer at a raito of 25 µl PBS/mg tissue, and homogenized with a glass dounce homogenizer. After centrifuge, 100 µl supernate solution was transfered into a new tube. To extract the lipid, 300 µl chloroform and 150 µl methanol was added into the supernate. The mixed solution was centrifuged to separate the aqueous phase from the organic phase (containing lipid), and the latter was transfered into a new ependorf tube for air dry. The TG concentration was determined with L-type triglyceride M Assay kit (Wako diagnostic), and normalized to the iBAT weight.

FACS analysis.
Fixed and Hoechst stained BAs were filtered through 40-μm cell strainers. BAs not labeled with tdTom reporter and Hoechst were used as negative control to establish the flow cytometer settings. FACS was performed on FACSAria III (BD). The flow cytometric data were analyzed with DIVA software (BD).
Gene expression. Total RNA was isolated using the Trizol reagent. For quantitative reverse transcription PCR (qRT-PCR), RNA was reverse transcribed (Superscript III) and specific transcripts were measured using SYBR Green chemistry and normalized to 36B4. Primer sequences were provided in Supplementary www.nature.com/scientificreports/ Statistics analysis. Values were expressed as mean ± SEM. Student's t-test or ANOVA with Tukey's honestly significant difference post-hoc test was used to test for statistical significance involving two or more than two groups, respectively.