Metabolically Active Human Brown Adipose Tissue Derived Stem Cells



Brown adipose tissue (BAT) plays a key role in the evolutionarily conserved mechanisms underlying energy homeostasis in mammals. It is characterized by fat vacuoles 5–10 µm in diameter and expression of uncoupling protein one, central to the regulation of thermogenesis. In the human newborn, BAT depots are typically grouped around the vasculature and solid organs. These depots maintain body temperature during cold exposure by warming the blood before its distribution to the periphery. They also ensure an optimal temperature for biochemical reactions within solid organs. BAT had been thought to involute throughout childhood and adolescence. Recent studies, however, have confirmed the presence of active BAT in adult humans with depots residing in cervical, supraclavicular, mediastinal, paravertebral, and suprarenal regions. While human pluripotent stem cells have been differentiated into functional brown adipocytes in vitro and brown adipocyte progenitor cells have been identified in murine skeletal muscle and white adipose tissue, multipotent metabolically active BAT-derived stem cells from a single depot have not been identified in adult humans to date. Here, we demonstrate a clonogenic population of metabolically active BAT stem cells residing in adult humans that can: (a) be expanded in vitro; (b) exhibit multilineage differentiation potential; and (c) functionally differentiate into metabolically active brown adipocytes. Our study defines a new target stem cell population that can be activated to restore energy homeostasis in vivo for the treatment of obesity and related metabolic disorders. Stem Cells 2014;32:572–581


Obesity is a worldwide epidemic with serious health consequences. In the United States, 35.7% of adults and 16.9% of children and adolescents are obese [1]. Obesity is a risk factor for insulin resistance, metabolic syndrome, and cardiovascular disease [2]. In the United States and other Western countries, obesity is driving a parallel epidemic of type 2 diabetes. In 2007, annual health spending attributed to diabetes alone was $174 billion, a cost that is projected to continue to rise dramatically with increasing rates of obesity [3].

The impact of increasing obesity prevalence and its deleterious sequelae has intensified research into obesity therapies. In particular, brown adipose tissue (BAT) has received increased attention over the past several years as a metabolic target. BAT is typically found in large quantities in newborns, decreasing with age to localized depots in adults. Recent studies have confirmed the presence of active BAT containing both classical brown and beige adipocytes in adult humans, with depots residing in the cervical, supraclavicular, mediastinal, paravertebral, and suprarenal regions [4-6]. Unlike white adipose tissue (WAT), which stores and accumulates fat, BAT metabolizes fat, generates heat, and increases overall metabolism. It also has a unique morphology, characterized by fat vacuoles of 5–10 µm in diameter, and possessing unique expression of uncoupling protein one (UCP1) central to the regulation of thermogenesis [7]. A recent report has demonstrated that in response to cold exposure, adult human BAT consumes more glucose per gram than any other tissue [8]. BAT is also purported to counteract weight gain by increasing whole-body energy intake [9]. BAT, therefore, may represent a novel target to activate and restore energy homeostasis in vivo for obesity treatment and related metabolic disorders [10].

Adipose derived stem cells are prevalent in WAT [11] and have been used in several applications, including tissue regeneration and immunomodulation [12]. Here, we describe our discovery and characterization of stem cells derived from human BAT that differ distinctly from stem cells derived from white adipose tissue. We demonstrate that these brown adipose derived stem cells (referred to as BADSCs) derived from a single depot are: (a) metabolically active; (b) can be expanded in vitro; (c) exhibit multilineage differentiation potential; and (d) can be functionally differentiated into metabolically active brown adipocytes. We have tested the functionality of these BADSCs and their derived brown adipocytes both in vitro and in vivo. We have also studied their potential as a cell-based therapy.

Materials and Methods

Mediastinal Adipose Tissue Procurement

Subjects were referred for elective coronary artery bypass grafting surgery at the University of Utah. Mediastinal adipose tissues were obtained from 54 patients (44 males and 10 females). Mean patient age was 72.4 ± 12 years (range 28–84 years). Institutional review board (IRB) numbers assigned to the study were 35,241 and 35,242.

Biopsy Procedure

In the anterior mediastinum, adipose tissue was removed from the right to left pleura and up to the innominate vein in the standard fashion in order to access the pericardium. A portion of the tissue was fixed in formalin for immunohistochemical confirmation of UCP1 expression and another portion was used for primary cell isolation.

Derivation of Primary Human BADSCs

Under aseptic conditions, the mediastinal adipose tissue was placed into a 50 mL conical tube containing Hanks' balanced saline solution (HBSS, Hyclone, Rochester, and washed by shaking vigorously for 5–10 seconds. This was repeated three times to remove erythrocytes and leukocytes. The tissue was then cut into 3 mm pieces, washed three times in Dulbecco's phosphate buffered saline (DPBS, -calcium, -magnesium, Life Technologies, Carlsbad, and plated onto a six-well dish. Dulbecco's modified Eagle's medium (DMEM) low glucose, containing 10% XcytePlus (iBiologics, Phoenix), 1× Glutamax, and 1× minimal essential medium (MEM)-nonessential amino acids (Life Technologies,, was used to derive primary cells from explants. Primary cultures were then plated onto 225-cm2 flasks and further expanded. At 70% confluence, cells were detached using TrypLE (Life Technologies) and frozen. Derivation of clonal cell lines via limiting dilution was performed as previously described [13].

Cell Growth Kinetics

To generate a growth curve and determine population doublings (PD), cells were cultured in 25-cm2 flasks, harvested, counted, and replated when they reached 80% confluence. Cultures were terminated when cells failed to double after 2 weeks in culture. PDs were calculated as previously described [14].

Brown Adipogenic Cell Differentiation

Cells were plated in triplicate in six-well dishes at a density of 50,000 cells per well. At 90% confluence, cells were differentiated using DMEM low glucose (Life Technologies) containing dexamethasone (5 µM), insulin (0.5 µg mL−1), isobutylmethylxanthine (0.5 mM), rosiglitazone (1 µM), T3 (1 nM), recombinant fibronectin type III domain containing 5 (FNDC5) (20 nM) and 10% fetal bovine serum (FBS) as previously described [15].

Fatty Acid Uptake Analysis

Analysis began with the replacement of growth media with HBSS buffer with 20 mM HEPES and 0.2% fatty acid-free bovine serum albumin (BSA). Cells were placed in the incubator for 1.5 h, and quencher based technology (QBT) fatty acid uptake (Molecular Devices LLC, Sunnyvale, media was added to the wells, and fluorescence was analyzed every minute using a Bio-Tek Synergy HT instrument.

Cellular Respiration and Glycolysis Analysis

The oxygen consumption rate (OCR) was performed using a Seahorse Bioscience XF-24 Analyzer (Seahorse Bioscience, North Billerica, by replacing cell growth media with XF assay media and incubating in a CO2-free chamber for 1 hour. The XF Cell Mito Stress Test simultaneously analyzed basal respiration, ATP turnover, proton leak, spare respiratory capacity, and glycolysis.

Mouse Model

Care of animals was in accordance with institutional guidelines. Eight-week-old nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice, NOD.CB17-Prkdcscid/J (Jackson Laboratories, Bar Harbor, ME, were implanted with scaffolds for 5 weeks and BADSCs for 7 weeks. Additional details on mouse model can be found in Supporting Information Material.

Cell and Scaffold Culture

Scaffolds (1 mm × 7 mm with 250 µm pores) were made by lyophilization as described in Supporting Information Data. Passage-4 BADSCs were harvested and resuspended in DPBS (-calcium, -magnesium) at a concentration of 106 cells per 100 µL for in vitro analysis, injection, and scaffold seeding. This concentration yielded 0.167 × 106 cells per scaffold, and six scaffolds were implanted per mouse for a total dose of 106 cells per mouse. Scaffolds used as implants were placed in brown fat differentiation media and cultured for 7 days with media changes at days 3 and 6. To confirm viability within three-dimensional (3D) scaffolds, cells were cultured in control media for 2 days and labeled with Calcein AM and propidium iodide (PI) according to manufacturer's instructions. For differentiation analysis, cells within scaffolds were placed directly into differentiation media and cultured for 21 days, with media changes every 3 days.

Passaging of Lipid-Filled Cells

Cells were differentiated into brown fat as described above. On day 21, media was removed, cells were washed 2× in DPBS, TrypLE was added, and cells were placed in a 37°C incubator for 1 hour to fully detach cells. Cells were counted using a hemacytometer. Trypan blue was used to determine cell viability. Cells were centrifuged first at 200 rcf for 5 minutes and then 300 rcf for 5 minutes to reduce shear stress experienced by the large lipid-filled cells. The supernatant was transferred to a new well of a six-well plate and allowed to settle for 1 hour prior to imaging. The pellet was resuspended in 2 mL of brown fat differentiation media and added to a new well of a six-well plate and allowed to adhere overnight prior to imaging with an Olympus IX51microscope and DP72 CCD camera.

Standard Methods

Standard procedures were implemented for flow cytometry, Western blot analysis, transmission electron microscopy, scanning electron microscopy (SEM), confocal imaging, histology, and differentiation into adipo-, chondro-, and osteo-lineages and are described in Supporting Information Materials. Antibodies used for all cell-specific markers were supplied by Molecular Probes, Eugene, OR (


All data are expressed as the mean ± SEM. The significance of differences was determined by the use of an unpaired, two-tailed Student's t test. AUCs represent the areas under the curves of plotted weekly time points collected over the time course of the study. AUCs were determined with GraphPad Prism software 5.0.


Multilocular UCP1-Positive Cells Are Prominent in Mediastinal Adipose Depots in Humans

It has been observed that mediastinal adipose depots may contain some BAT even in adults [16, 17]. Consistent with these studies, we obtained mediastinal adipose tissue biopsies from lean subjects (body mass index (BMI) under 25) undergoing elective coronary artery bypass grafting surgery (Fig. 1A). Subjects were mainly men in their early 70s and metabolically healthy (Fig. 1A). To confirm that adipose tissues biopsied were consistently BAT, immunohistochemical staining for UCP1 was performed. Microscopic examination of the mediastinal adipose tissue demonstrated distinctive multilocular lipid morphology, as opposed to a unilocular morphology typical of a subcutaneous depot (Fig. 1B, 1C, 1E). More importantly, the mediastinal adipose tissue expressed UCP1 (Fig. 1D), which is characteristic of BAT [16].

Figure 1.

Mediastinal adipose tissue characterization. (A): Patient demographics. Data are represented as means ± SD. (B): Biopsy of human mediastinal adipose depot from an adult. (C): Hematoxylin and eosin stain of human mediastinal adipose depot. (D): Immunohistochemical staining of uncoupling protein one (UCP1) using primary monoclonal anti-UCP1 antibody at 1:1,000 dilution of mediastinal adipose tissue. (E): Immunohistochemical staining of UCP1 using primary monoclonal anti-UCP1 antibody at 1:1,000 dilution of subcutaneous white adipose tissue. (F): Cell surface marker profile of BADSCs versus WADSCs. Abbreviations: BADSC, brown adipose derived stem cell; BMI, body mass index; CD, cluster of differentiation; HbA1c, glcosylated hemoglobin; HDL, high-density lipoprotein; HLA, human leukocyte antigen; LDL, low-density lipoprotein; SSEA, stage-specific embryonic antigen; TG, triacylglycerol; TMEM, transmembrane protein; WADSC, white adipose derived stem cell.

Derivation of Primary Human BADSCs

We recently identified a stem cell population within mediastinal adipose tissue exhibiting multilineage differentiation potential. In this study, we sought to isolate a stem cell population with functional brown adipose differentiation capacity. We first derived primary stem cells using a xeno-free explant method. After 7 days in culture, fibroblast-like cells could be observed. Primary cells were successfully derived from all 54 patients, with 100% derivation efficiency. These cells were further expanded to passage 1 and frozen until expression of UCP1 by immunohistochemistry was confirmed in the original mediastinal tissue.

Cloning and Characterization of Human BADSCs

Once UCP1 expression was confirmed in the mediastinal adipose tissue depots, the corresponding frozen primary cells were thawed for cloning into 96-well plates. Cloning efficiency (measured as the percentage of seeded cells) averaged 5% (data not shown). Clone BADSC18 was chosen for further characterization due to its higher expression of UCP1. Growth kinetics of clonal BADSCs demonstrated a cell population that could be expanded for greater than 20 passages (data not shown). Karyotyping at passage 10 showed normal diploid cells without chromosomal aberrations (Supporting Information Fig. S1). To further characterize the clonal human BADSCs, we performed flow cytometry analysis. The cells expressed cluster of differentiation 90 (CD90), CD166, CD44, stage-specific embryonic antigen-4, CD73, CD105, transmembrane protein 26 (TMEM26), and CD137 and were negative for the hematopoietic markers CD34, CD45 and human leukocyte antigen DR (Fig. 1F). Cells were negative for CD31 and had a dimly positive population of CD146 (Supporting Information Fig. S2). In order to rule out the possibility that clonal cell populations derived from mediastinal adipose depots could be white adipose derived stem cells, we isolated mesenchymal stem cells (MSCs) from human subcutaneous tissue using the same explant and cloning protocols as that used for deriving the mediastinal BADSCs. Subcutaneous WAT cells did not have multilocular lipid morphology and did not express UCP1 (Fig. 1D). WAT-derived stem cells expressed similar but not identical cell surface markers as those found on BADSCs. In particular, BADSCs had a notably higher expression of TMEM26 and CD137 (Fig. 1F). A recent report identified TMEM26 and CD137 as unique markers found in beige adipose tissue [18]. In addition, we found that BADSCs expressed CD9 and CD63 (data not shown), both of which are part of the tetraspanin family involved in cell motility and correlated with exosome production [19]. CD9 has been found to play a role in the proliferation and proangiogenic action of adipose derived stem cells [20]. The BADSCs also had a unique gene expression profile compared to the white adipose-derived stem cells, most notably the higher expression of genes associated with BAT such as peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α), PR domain containing 16 (PRDM16), CAMP responsive element binding protein one (CREB1), and uncoupling protein 1 (UCP1) (Fig. 2A). These data demonstrate that the BADSCs isolated from the mediastinal adipose depot are a unique population distinct from stem cells derived from subcutaneous white adipose depots.

Figure 2.

In vitro differentiation of BADSCs. (A): Gene expression profile comparing undifferentiated BADSCs to undifferentiated WADSCs derived from subcutaneous adipose tissue. Genes in red are associated with brown fat phenotype. (B): Gene expression profile comparing undifferentiated BADSCs to differentiated brown adipocytes. Biological replicates performed in triplicate from a single clone were used for gene expression profile. (C): Transmission electron microscopy of 21-day brown adipocyte differentiation induced with fibronectin type III domain containing five (FNDC5) demonstrate multiocular intracytoplasmic lipid vacuoles and mitochondria (arrows). (D): Alizarian red staining of brown adipose derived stem cells induced to undergo osteogenesis. (E): Alcian blue staining of brown adipose derived stem cells directionally differentiated into chondrocytes. (F): Fatty acid binding protein four immunocytochemistry of BADSCs induced to undergo white adipogenesis. (G): Undifferentiated BADSCs. (H): Western blot 21 days post-FNDC5 induction. Lane 1, brown adipose derived stem cells directionally differentiated into brown adipocytes. Lane 2, non-FNDC5 cells. Abbreviations: BADSC, brown adipose derived stem cell; INSR, insulin resistance; UCP1, uncoupling protein one; WADSC, white adipose derived stem cell.

Multilineage Differentiation of Mediastinal BADSCs

MSCs have the ability to differentiate into cell types of mesodermal lineage [21]. In order to investigate the ability of mediastinal BADSCs to differentiate into multiple cell types, passage 2 cells were directionally differentiated into osteogenic, chondrogenic, white adipogenic, and brown adipogenic lineages. When induced to differentiate under osteogenic promoting conditions, the cells formed a mineralized matrix as confirmed by alizarian red staining (Fig. 2D). Using a pellet culture system [22], cells were induced to undergo chondrogenesis. Chondrogenic differentiation was confirmed by alcian blue staining for sulfated proteoglycans on induced cell pellets (Fig. 2E). In order to test the ability of BADSCs to undergo adipogenic differentiation, cells were induced to undergo both white adipogenesis (Fig. 2F) and brown adipogenesis (Fig. 2C). BADSCs induced to undergo brown adipogenesis expressed genes associated with the mature phenotype, such as CREB1, type II iodothyronine deiodinase, UCP1, PRDM16, and insulin receptor, compared to undifferentiated cells (Fig. 2B). Western blots demonstrated expression of UCP1 in directionally differentiated cells (Fig. 2H). In addition, we also induced BADSCs to undergo white adipogenesis and compared resulting gene expression profiles to BADSCs induced to undergo brown adipogenesis. Cells not directionally differentiated with FNDC5 had a lower gene expression profile of brown fat associated genes (Supporting Information Fig. S3). Cells not induced to differentiate retained their normal morphology (Fig. 2G). Our results confirm previous reports that FNDC5 is a potent inducer of UCP1 and brown fat adipogenesis [15].

BADSCs Differentiated into Brown Adipocytes Demonstrate Mature Functional Properties

In addition to exhibiting the gene profile of brown adipocytes and expressing UCP1, we sought to determine whether differentiated BADSCs were metabolically active. Therefore, OCR and fatty acid uptake were analyzed in BADSCs previously differentiated for 7 and 21 days post-FNDC5 induction. Undifferentiated BADSCs (day 0) exhibited low basal oxygen consumption and minimal uncoupling, as evidenced by the efficient inhibition of OCR after exposure to oligomycin (Fig. 3A). In contrast, at 21 days post-FNDC5 induction, BADSC oxygen consumption was increased at baseline and was less sensitive to oligomycin, suggesting enhanced uncoupling (Fig. 3A). Furthermore, maximal oxidative capacity (after trifluorocarbonylcyanide phenylhydrazone) was significantly increased at day 21 post-FNDC5 compared to day 0 in differentiated BADSCs. Additionally, fatty acid uptake capacity shown in Figure 3B for BADSCs induced for 21 days was highest compared to control, noninduced cells. Taken in aggregate, these results confirm that multipotent BADSCs induced to differentiate into brown adipocytes exhibit the mature functional properties of these cells, including increased mitochondrial activity, an important functional characteristic of BAT.

Figure 3.

In vitro mitochondrial functional assay of differentiated BADSCs. (A): Functional mitochondrial respiration assay of BADSCs differentiated into brown adipocytes at 0, 7, 14, and 21 days postdifferentiation, showing increased respiration in BADSCs with increased differentiation time. (B): Fatty acid uptake of brown fat differentiated BADSCs at 0, 7, 14, and 21 days postdifferentiation. Abbreviations: BADSC, brown adipose derived stem cell; FCCP, trifluorocarbonylcyanide phenylhydrazone; FNDC5, fibronectin type III domain containing five; OCR, oxygen consumption rate.

Development and Characterization of BADSC Delivery Systems for Therapeutic Applications

Adipose-derived stem cells with brown adipogenic potential would represent a new modality for the treatment of obesity and related metabolic disorders. To explore this possibility, we first sought to develop effective cell-based delivery systems.

Lipid-Filled Cells Are Unable to Be Pelleted upon Centrifugation

We investigated whether directionally differentiated lipid-filled brown fat-like cells could be passaged and concentrated for potential injection. BADSCs were differentiated into brown fat-like cells with high lipid content (Fig. 4A). After 21 days, cells with lipid-filled vesicles were passaged but were unable to be pelleted after centrifugation and only cells with small or no lipids adhered upon plating (Fig. 4B, 4C).

Figure 4.

In vitro culture of brown adipose derived stem cells (BADSCs). Representative bright-field images of differentiated cells (A) prior to passaging, (B) postpassaging and centrifugation adherent, and (C) floating in the supernatant. Representative confocal images of live BADSCs stained with Calcein AM on the (D) surface and (E) through the center of the three-dimensional (3D) scaffolds; dead cells stained with propidium iodide on the (F) surface and (G) through the center. (H, I): Representative scanning electron micrographs of BADSCs on scaffolds in a (H) nondifferentiated state and (I) 21 days directionally differentiated toward brown adipocytes on scaffolds—secondary and backscattered electrons overlaid to show variation in composition (orange represents lipids). (J, K): Representative confocal images of (J) Actin (red) and 4′,6-diamidino-2-phenylindole (DAPI) (blue) staining, and (K) lipid (fatty acid binding protein 4, FABP4, green) and DAPI (blue) staining of brown fat-differentiated BADSCs in 3D scaffolds. (L): Representative bright-field image of a histology section of differentiated brown fat cells (arrow) within a scaffold (*) after 21 days. n = 3 separate scaffolds or three separate wells during 2D culture.

Culture and Differentiation of BADSCs on 3D Porous Extracellular Matrix-Derived Scaffolds

In order to determine the feasibility of culturing and differentiating BADSCs in a 3D culture system, undifferentiated BADSCs were seeded and differentiated on porous extracellular matrix-derived scaffolds fabricated from human subcutaneous fat depots. Undifferentiated cells were seeded at low density and cultured for 2 days. To ensure sufficient viability within the 3D construct, cells were labeled with Calcein AM (Fig. 4D, F) and PI (Fig. 4E, 4G). A large percentage of viable cells were seen both on the outside and in the center of the scaffolds (Fig. 4D, 4E). SEM analysis of undifferentiated BADSCs revealed highly adherent cells (Fig. 4H).

Undifferentiated BADSCs were also induced to undergo brown fat differentiation for 21 days. SEM analysis of differentiated cells demonstrated multilocular lipid formation (Fig. 4I), indicative of brown adipose differentiation. These differentiated cells were evenly distributed throughout the scaffold as seen by 4′,6-diamidino-2-phenylindole and actin staining (Fig. 4J). Differentiated BADSCs also revealed an even distribution of lipid content as demonstrated by fatty acid binding protein four staining (Fig. 4K) and showed cells with high lipid content even within the center of the scaffolds (Fig. 4L).

Undifferentiated Brown Fat Stem Cells Reduce Body Weight and Blood Glucose Levels in NOD-SCID Mice on a High Fat Diet

To analyze their in vivo functionality, the BADSCs were injected in mice fed a high fat diet. We first injected undifferentiated BADSCs, unsupported by a scaffold, subcutaneously. Over the course of 7 weeks postinjection, the mice injected with the unsupported and undifferentiated BADSCs experienced a slight decrease in blood glucose levels but no change in body weight compared to the saline control animals (Supporting Information Fig. S4). Control mice fed normal low fat chow were also monitored and compared to mice fed high-fat chow to show increases in weight and glucose and during this study (Supporting Information Fig. S5).

Following completion of the unsupported BADSCs injection study group, we implanted 3D adipose extracellular matrix scaffolds seeded with differentiated (for 7 days) BADSCs into a separate group of mice. The cells within the scaffolds maintained homogeneous distribution throughout the thickness of the scaffold after 5 weeks of implantation (Fig. 5A). Importantly, the scaffolds were found to contain viable human-derived cells as evidenced by vimentin labeling (Fig. 5B). SEM analysis of the explanted scaffolds showed substantial lipid content within the range of 5–10 µm, consistent with the size of brown fat lipid (Fig. 5C, 5D). Several brown fat genes were also detected within the explanted scaffolds (Supporting Information Fig. S6). Compared to the saline controls, high-fat fed NOD-SCID mice implanted with differentiated brown fat stem cells within the scaffolds experienced a significant reduction in blood glucose levels and body weight (Supporting Information Fig. S7).

Figure 5.

Brown adipose derived stem cells-seeded scaffolds explanted from mice after 5 weeks. Histology images of (A) H&E stained sections, revealing homogenous cell distribution throughout center of scaffolds and (B) vimentin labeling of human cells (brown) within the center of the 3D scaffold. Fibrous encapsulation (arrows) comprising mouse cells in scaffold periphery are shown to be negative for vimentin, indicating human cell-specific labeling. (C, D): Scanning electron micrographs with secondary and backscattered electrons overlaid to show variation in composition (orange represents lipids). Representative images are taken from n = 8 mice with scaffold implants.


Previously, we have discovered a putative population of stem cells derived from mediastinal brown adipose depots [23]. Here, we further characterized this population as brown fat stem cells with unique characteristics compared to white adipose stem cells and examined several potential applications. Given the well-described role of brown fat in the dissipation of stored energy [24], the identification of BADSCs may hold added therapeutic potential for the treatment of obesity and its related complications compared to white adipose-derived stem cells [25, 26].

Our results suggest that human mediastinal brown adipose depots possess a pool of resident beige fat marker, TMEM26 and CD137, positive cells [18] with stem cell properties including the capacity for self-renewal, multipotency, and the ability to differentiate into the functional mature phenotype of the differentiated cell type. Multilineage differentiation demonstrated that BADSCs were capable of undergoing osteogenesis, chondrogenesis, and white and brown adipogenesis (Fig. 2). These results support the identification of a brown fat stem cell.

Previous reports have demonstrated that FNDC5 can induce brown fat differentiation in white adipose derived stem cells [15]. We used this peptide to induce differentiation in BADSCs over the course of 21 days and saw gene expression profiles characteristic of brown fat found in vivo [27], unlike BADSCs not exposed to FNDC5 (Fig. 2B). Western blots demonstrated UCP1 expression in brown fat differentiated cells (Fig. 2H). Real-time quantitative polymerase chain reaction (PCR) of the FNDC5-induced BADSCs demonstrated expression of UCP1, ELOVL3, and PGC1A (a major regulator of mitochondrial biogenesis), unlike the non-FNDC5-induced cells (Fig. 2B). Conversely, leptin, a gene associated with white fat development [28], was downregulated in brown adipose differentiated cells. Mitochondrial respiration and fatty acid uptake was used to determine the functional potential of BADSCs induced to undergo brown fat adipogensis. BADSCs differentiated with FNDC5 exhibited elevated levels of fatty acid uptake and higher OCR compared to the noninduced cells (Fig. 3). These results indicate a strong conversion to a brown adipose phenotype.

Mediastinal adipose-derived stem cells are similar to MSCs derived from subcutaneous adipose depots and bone marrow, based on their growth kinetics, cell surface marker profile, and multilineage potential [21, 29, 30]. They express typical MSC markers, such as CD44, CD73, CD90, CD105 and are negative for the hematopoietic markers CD14, CD34, CD45. However, these cells also express brown adipose specific genes such as PRDM 16, UCP-1, IRS2, and NRF1 among several others (Fig. 2A), giving them a unique phenotype compared to white fat stem cells. Some of the BADSC population were dimly positive for CD146, a known pericyte marker, but previously shown to be positive in MSCs [31]. CD31 was negative, confirming no contaminating epithelial cells.

Although white adipose derived stem cells have been shown to convert into brown adipocytes [15], this conversion may be less efficient than BADSCs already expressing several required genes such as UCP-1 and PRDM 16—known brown fat markers [15]. When comparing metabolism, BADSCs differentiated into brown adipocytes had higher OCR than WADSCs differentiated into brown adipocytes (data not shown). Furthermore, BADSCs may be able to spontaneously convert to brown fat. This idea is suggested in Figure 2B where a single group of unidifferentiated brown adipose stem cells (group 1) depicts several spontaneously upregulated brown adipose genes. Future work will test differentiation efficiency of WADSCs compared to BADSCs into brown adipocytes as well as BADSCs' potential for spontaneous brown fat adipogenesis both in vitro and in vivo.

The potential use of brown adipose cells as therapeutic agents has been raised by other groups [26, 32, 33]. However, to our knowledge, no report identifies multipotent BADSCs as a potential therapeutic agent. To analyze the functionality of these brown adipose-derived stem cells and their therapeutic potential, these cells were injected into immune-compromised NOD-SCID mice tolerant to human cell xenogeneic implantation. These mice were fed a high fat, high carbohydrate diet and, following injection of the unsupported and undifferentiated stem cells, had a slight decrease in blood glucose levels, but no decrease in weight (Supporting Information Fig. S4). This was a small cohort that did not experience substantial weight gain prior to starting this study (data not shown), which may explain why no weight loss was seen.

We sought to determine if similar or improved results could be seen with differentiated BADSCs. As brown fat cells have many intracellular lipids, we tested them in vitro to determine if they could be passaged from a surface, centrifuged, and prepared for in vivo injection. After 21 days of differentiation (Fig. 4A), cells were passaged, counted, and centrifuged. Substantial fractions of cells were dead after passaging (27%, data not shown). After centrifugation, only cells with few or no lipids adhered to the surface of a plate (Fig. 4B). The remaining cells with large or many lipids remained floating in the supernatant (Fig. 4C), precluding use as a parenteral cell therapy. Due to the difficulty in passaging brown fat-like cells, we applied a 3D cell culture scaffold to possibly better serve as a delivery system for differentiated brown fat cells. Cells were seeded onto 3D scaffolds made of processed human adipose-derived extracellular matrix to provide a contextual environment for these cells. Cells were capable of being evenly distributed and remained viable throughout 1 mm thick scaffolds (Fig. 4). Seeded cells were also shown to adhere well to the scaffolds even with multiocular lipid formation (Fig. 4). Nondifferentiated BADSCs were also seeded onto the scaffolds and showed good attachment (Fig. 4).

To determine if seeded scaffolds could serve as a viable cell delivery system, they were implanted subcutaneously in the backs of NOD-SCID mice. To our knowledge, no other group has implanted a tissue engineered brown fat construct. The cells in these scaffolds remained viable in vivo for 5 weeks, as shown by human-specific vimentin staining of transplanted cells within the scaffold (Fig. 5A). We also observed a cellular and fibrous encapsulation by host murine cells surrounding the outer edges of the scaffolds (arrows in Fig. 5B) that stained negative for vimentin, confirming the specificity of the vimentin staining for only human cells within the scaffold. Explanted scaffolds revealed high densities of small lipid vesicles within the size range of brown fat cells of 5–10 µm [7, 34], indicating that these cells possessed phenotypic aspects of brown fat cells even 5 weeks postimplantation (Fig. 5C, 5D). Preliminary PCR results also confirmed the presence of several brown fat genes (Supporting Information Fig. S6). Preliminary data obtained in high-fat fed NOD-SCID mice transplanted with differentiated BADSCs supported by a scaffold showed a significant reductions in weight and blood glucose levels compared to saline only injected controls (Supporting Information Fig. S7). Future work will implant empty and white adipose scaffolds as more robust controls. These animal implant studies sought to determine if: (a) undifferentiated or differentiated BADSCs could remain viable and functional in vivo; and (b) 3D scaffolds could serve as an effective cellular delivery vehicle. Future work will develop this preliminary study with a larger cohort that is weight matched, glucose fasted, and has additional tests and controls that will better elucidate changes due to the presence of brown fat. It is important to note, however, that even with a small pilot study, the BADSCs show promise in correcting the weight gain and hyperglycemia associated with high fat feeding and that 3D scaffolds are capable of retaining viable transplanted cells for at least 5 weeks postimplantation.

Many groups are studying BAT as a metabolic model to identify target molecules or other pharmaceuticals that can be used to treat or induce endogenous brown fat to treat obesity and its associated metabolic alterations [10, 15]. However, these treatments would likely be systemic and may have limited efficacy. Our cell therapy approach takes advantage of verifiably active brown fat that intrinsically regulates metabolic homeostasis and can be applied to specific sites. Furthermore, the regulatory hurdles for cell therapy are far less extensive than for pharmaceuticals, thus shortening the bench-to-bedside timeline. By using stem cells isolated from a mediastinal adipose depot and delivered in a biological scaffold, we can generate a therapeutic dose of cells to potentially treat obesity and related metabolic disorders in the near-term.


Our results demonstrate the existence of a resident multipotent stem cell population within depots of BAT in the adult human mediastinum. Cells from this tissue exhibit multilineage potential with the capacity to undergo osteogenesis, chondrogenesis, and both brown and white adipogenesis. Directionally differentiated brown adipocytes exhibit a distinct morphology and gene expression profile, with functional properties characteristic of BAT. These results also illustrate the potential of these BADSCs to serve as therapeutic agents in vivo with particular relevance to metabolic disease. Our study also highlights a novel 3D scaffold delivery system for these unique cells. BADSCs may offer a new means to activate and restore energy homeostasis for the treatment of obesity and related metabolic disorders.


We acknowledge the partial support of the BioRestorative Therapies Corp. We also thank Jeff Theisen for contributions in creating scaffolds.

Author Contributions

F.J.S.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; D.J.H.: design, collection and/or assembly of data, data analysis and interpretation, and manuscript writing; V.V.: collection and/or assembly of data; J.Y., W.S., and D.A.: collection and/or assembly of data and data analysis and interpretation; S.B.: collection and/or assembly of data, data analysis and interpretation, and manuscript writing; D.W.G.: manuscript writing; M.P.R.: collection and/or assembly of data and data analysis and interpretation; D.A.B.: manuscript writing and final approval of manuscript; A.N.P.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript. F.J.S., D.J.H., and A.N.P. contributed equally to this article.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.