Unbiased gene expression analysis identifies two distinct adipose cell types derived from the subcutaneous fat depot

To investigate the molecular identity of fat cells from the subcutaneous fat depot, we isolated the SVF from subcutaneous fat in the inguinal region, and then subjected these undifferentiated cells to a classical 3T3 immortalization protocol (Todaro and Green, 1963). These cultures were then subjected to cloning of individual cells by limiting dilution; we thus derived 305 new clonal cell lines (Figure S1A). Among these clonal lines were more than 20 that were highly and reproducibly adipogenic, as evidenced by lipid accumulation after the induction of differentiation (Figure S1B). For purposes of subsequent comparison, multiple adipogenic clones were also generated by the same methods from the interscapular brown fat depot. Importantly, these immortalized clonal lines retain expression of previously reported interscapular depot and inguinal depot-enriched genes (Billon and Dani, 2011; Gesta et al., 2006; Seale et al., 2007) (Figure S1C).

To investigate possible heterogeneity in the clonal cells derived from the subcutaneous adipose tissue, multiple cell lines were differentiated, treated with forskolin and their RNA was subjected to microarray analysis. We used the established class finding algorithm ISIS (von Heydebreck et al., 2001) to search for the existence of subgroups among these inguinal cell lines. As shown in Figure 1D, these 23 cell lines separated into 2 distinct clusters, with 10 cell lines in one group and 13 in the other. Once these groupings were identified and the differentially expressed gene sets were determined, we included analysis of the cell lines derived from the classical brown fat depots into the dataset; all samples were then hierarchically clustered using complete linkage of differentially expressed genes. This clustering (Figure 1D) indicates that one of the subsets of cells derived from the inguinal depot has a gene expression pattern more similar to the bona fide brown fat cell lines than to the other cluster of inguinal cell lines. The separation of these three groups of cells was further established by principle component analysis (PCA; Figure S1D). These data strongly support the notion that a distinct pool of progenitors in the inguinal depot can give rise to cells (beige cells) that are similar but not identical to classical brown fat cells.

Beige cells have characteristics of both white and brown fat cells

We further characterized several cell lines from both subgroups (beige and white) derived from the subcutaneous depot in terms of their molecular and functional properties. All these lines are reproducibly highly adipogenic and presented similar levels of adipogenesis and expression of fat cell specific markers (e.g. aP2, adiponectin, adipsin and PPARγ) (Figure S2A). Both groups of the inguinal-derived cells showed a similar, low basal level of gene expression for Ucp1, Cox7a1 and Cidea; these levels of gene expression were much lower than those observed in all of the cell lines derived from the classical brown fat (Figure 2A). However, upon cAMP stimulation, the beige lines responded with a very large induction of Ucp1 gene expression, reaching similar absolute levels to that observed in the interscapular BAT lines (Figure 2B and 2C). In fact, although the absolute levels of Ucp-1 mRNA were similar between beige and brown cells, the fold induction in the beige cell lines (>150 fold) was actually greater than that observed in the BAT cell lines (about 40 fold) (Figure S2B). It has been observed that some of the precursor cells within heterogeneous cultures derived from epididymal adipose tissue can show an elevated expression of Ucp-1 in response to PPARγ agonist treatment (Petrovic et al., 2010). In our clonal cell lines, basal expression levels of Ucp-1 are increased upon long-term TZD treatments in both beige and white cell lines, although this induction is stronger in the beige cells (8–10-fold vs. 4–5-fold, respectively) (Figure S2C).

Open in a separate window

Figure 2

Beige fat cells show comparable thermogenic potential to brown fat cells upon stimulation

Representative lines from each group were further analyzed functionally. Bat1, bat6, bat8 for brown fat cells; X9, X3, D16, D12 for beige fat cells and N13, X7, J6, G18 for white fat cells. All these lines presented similar levels of adipogenesis and fat cell specific markers were measured via qPCR. See Figure S2A. Similar results were observed in more than three independent experiments. (A) Total RNA was isolated from the differentiated brown, beige and white clonal cell lines and was assayed for mRNA levels of brown fat-like genes (Ucp-1, Cidea and Cox7a1) by qPCR. (B) Total RNA was isolated from the differentiated brown, beige and white clonal cell lines treated with 10μM forskolin for 4 hours and was assayed for mRNA levels of Ucp-1. (C) Protein lysates were isolated from the differentiated brown, beige and white clonal lines either unstimulated or treated with 10 μM isoproterenol for 6 hours, and probed by immunoblot as indicated, with tubulin as a loading control. (D) Cell transplantations were done in the preadipose state as described in Experimental Procedures. After 6 weeks, fat pads were harvested and mRNA expression was analyzed by qPCR (n=5–8). For acute sympathetic stimulation, CL316,243, at 1mg kg⁻¹, was injected intraperitoneally 5 hours before harvesting the transplanted fat pads. (E) Oxygen consumption in cultured brown, beige and white fat cells was assayed as described in Experimental Procedures. Oligomycin (ATP synthase inhibitor) was added to cells to measure the uncoupled respiration rate. The cells were treated with dibutyryl-cAMP for 12 hours before harvesting. Values are mean±SD (n=4). *, p<0.05; **, p<0.01; ***, p<0.001.

We next investigated the potential of these white and beige cells to express UCP1 in vivo, via transplantation experiments. Multiple beige and white inguinal lines can differentiate in vivo and form ectopic fat pads 4–6 weeks after transplantation into immunodeficient mice. Cidea, Cox7a1 and basal Ucp1 levels were comparable in transplanted fat pads derived from beige fat cell lines (X3 and X9) and white inguinal lines (N13 and X7) in the basal, unstimulated state (Figure 2D). To mimic sympathetic nervous system input, we treated mice with CL316,243, a selective β3-adrenergic agonist and harvested the transplanted fat pads 5 hours later. The fat pads from the beige lines displayed a 10–30 fold increase in Ucp-1 mRNA compared to basal level, whereas fat pads from white lines showed a 5 fold increase. Thus, these molecular differences in Ucp1 gene expression are a stable property of these cells and can be readily observed in an in vivo setting.

Mitochondrial respiration is central to adaptive thermogenesis. To assay the functional characteristics of these cells, we investigated respiratory function in representative lines of the white, beige and brown groups. In brown cells (bat1) without cAMP stimulation, uncoupled respiration (respiration insensitive to oligomycin) contributed to 78% of their basal respiration rate, significantly higher than the relative contribution of uncoupled respiration in the beige cells (X9, 60%). In the presence of cAMP, uncoupled respiration went up almost 2 fold in beige fat cells compared to a 1.2 fold induction in the brown fat cells. Importantly, the uncoupled respiration in the beige cells equaled or even exceeded that in the classical brown fat cells with the cAMP stimulation. The white cells (J6) had 53% of uncoupled respiration without stimulation and showed little cAMP-dependent increase in uncoupled respiration. Both the basal respiration and uncoupled respiration rates in the white fat cells are significantly lower than both the brown and beige fat cells in the presence of cAMP. These experiments revealed that the beige cells show high respiratory capacity and an uncoupled respiration that is even more cAMP-responsive that the brown cells (Figure 2E).

Beige fat cells express a unique gene expression profile

Analysis of differentially expressed genes between the beige and brown fat cells demonstrates that they have related but distinct gene expression profiles (Supplemental Experimental Procedures). The expression of a subset of the beige-selective and brown-selective genes was verified in the basal (not cAMP stimulated) state via qPCR analysis (Figure 3A). The genes whose expression is enriched in brown cell lines included markers identified in previous studies, such epithelial V-like antigen 1 (EVA1, also known as myelin protein zero-like 2, MPZL2) (Seale et al., 2007). Importantly, expression of certain beige-selective genes can distinguish beige fat cells from both brown fat cells as well as from the white cells derived from the inguinal depot. Figure 3A shows that genes expressed in a beige-selective manner include a developmental transcription factor (Tbx1), a component of lipid metabolism pathways (Slc27a1), as well as molecules known to be important in immune and inflammatory response pathways (CD40, CD137).

Open in a separate window

Figure 3

Beige and brown fat cells exhibit distinct gene expression profiles

(A) Genes with a beige- or brown-selective expression pattern were first identified by microarray analysis as described in Supplemental Experimental Procedures. Relative mRNA expression for these genes was then measured by qPCR in differentiated brown, beige and white fat cells in an unstimulated state. (B) Analysis of beige- and brown-selective genes expression in adipose tissues isolated from inguinal WAT and interscapular BAT of 129 SVE mice. Values are mean±SD (n=5). *, p<0.05; **, p<0.01; ***, p<0.001. (C) Protein lysates of adipose tissues isolated from inguinal WAT and interscapular BAT of 129 SVE mice were probed with antibodies by immunoblot as indicated. Asterisk represents a nonspecific band as a loading control. (D) Microscopic and confocol images of mouse inguinal white and interscapular brown adipose tissue stained with antibodies recognizing CD137 (red) (microscopic images), TMEM26 (green) (confocal images) and UCP1 (green and red), with nuclei co-stained with DAPI as indicated. To induce the “browning” morphology and UCP1 expression in inguinal white adipose depot, 129SVE mice were injected intraperitoneally with CL316, 243, at 1mg kg⁻¹ daily for 4 days. Inguinal and interscapular adipose tissues were harvested on the fifth day.

Since these gene signatures were established from the clonal cell lines, we returned to mice kept at ambient temperature to examine expression levels of putative beige- and brown-selective genes in the inguinal and interscapular adipose depots. Despite the inevitable heterogeneity in the fat tissues, mRNAs encoding beige fat cell markers such as TMEM26, CD137 and TBX1 were all expressed at higher levels in the inguinal fat depot, compared to the interscapular brown fat. Conversely Eva1 and other brown fat genes were expressed at higher levels in the interscapular brown fat depot (Figure 3B).

Antibodies were commercially available for some of these potential markers, including CD137, TMEM26 and TBX1 for beige cells, and EVA1 for brown cells. We confirmed that the protein expression was highly concordant with mRNA expression, as shown by immunoblot and immunohistochemistry in vivo (Figure 3C, 3D and Figure S3). Importantly, immunohistochemical analysis showed that the UCP1+ cells contained CD137 and TMEM26 from the inguinal depot but not from the interscapular brown fat depot. (Figure 3D and Figure S3)

Beige cell surface proteins can be used to select primary beige fat cell precursors

Some of the identified beige-selective markers are cell surface proteins, allowing us to use fluorescence-activated cell sorting (FACS) to separate primary cells from the stromal-vascular fraction (SVF) expressing high levels of CD137 or TMEM26 versus those expressing low levels of these proteins (Figure S4). While these cells differentiated equally well, the CD137+ group clearly show greatly elevated expression of Ucp-1, compared to the low CD137 cells (Figure 4A), in this unstimulated state. A very similar phenomenon was observed when we used another beige-selective marker, TMEM26, to sort these cells by FACS (Figure 4B). Many other thermogenic genes shown statistically significant or a trend towards enriched expression in beige-marker positive cells, including mitochondrial genes Cox7a1, Cox8b, transcriptional coregulators Prdm16, Pgc-1β and the thermogenic hormone Fgf21 (Figure 4A and 4B).

Open in a separate window

Figure 4

Isolation of beige precursor cells from inguinal stroma with beige-selective cell surface markers

Total RNA was isolated from differentiated cultures of beige-selective marker CD137 (A) and TMEM26 (B) positive and negative cells purified from inguinal SV cells; these adipocytes were then assayed for mRNA levels of general adipogenic markers, Ucp-1 and other characteristic thermogenic/brown fat-like genes by qPCR. (C) Primary inguinal SV cells were purified as in (A) and differentiated into adipocytes for 6 days in the presence of FC fragment of human IgG (hFC) (100nM), fusion protein of irisin-FC (100nM) or recombinant FNDC5 (20nM). Total RNA was isolated and assayed for mRNA levels of general adipogenic marker aP2, thermogenic gene Ucp-1 and other brown-like genes by qPCR. Values are mean±SD (n=3). *, p<0.05, **, p<0.01, ***, p<0.001. Similar results were observed in more than three independent experiments.

We have recently shown that irisin, a polypeptide hormone secreted by muscle and increased with exercise, induced the “browning” of subcutaneous white adipose tissues; in contrast, this protein had little effect on the classical brown fat cells isolated from the interscapular depot (Bostrom et al., 2012). This differential regulation of a thermogenic program by irisin suggests that the response to this hormone could be a selective characteristic of beige cells. To test this hypothesis, we applied irisin, prepared either as a fusion protein with the FC fragment of human IgG, or FNDC5, the transmembrane precursor of irisin shown to be active in Bostrom et al, (2012) to the sorted primary inguinal precursor cells during differentiation. All cultures treated with irisin or vehicle differentiated well, with greater than 90% showing an adipocyte morphology and copious lipid droplets. As shown in Figure 4C, expression of Ucp-1 and other known brown-like genes such as Prdm16, Cox8b were robustly increased by both irisin-FC and FNDC5 treatments in the “CD137-high” expressing groups. Very little effect of these same polypeptides was observed on the expression of these genes in the “CD137-low” cellular population. These data strongly suggest that beige cell precursors have preferential sensitivity to the “browning” effects of irisin. It is also worth noting that expression of the fat cell gene and PPARγ target aP2 is slightly but significantly elevated in the irisin/FNDC5 treated primary beige adipocytes. Although both CD137 high and CD137 low cultures appeared to be very well differentiated, this data suggests that irisin might have subtle effects on the differentiation or PPARγ activity of the beige cell precursors.

Cell transplantations

Immortalized clonal inguinal cells (3×10⁷) were implanted subcutaneously into 7–9-week-old male NCR nude mice (Taconic, n=5–8 mice per group), accordingly to the methods described previously (Seale et al., 2007). Six weeks after injection, the fat pads were isolated for the analysis of gene expression. For acute sympathetic stimulation, CL316,243, at 1mg kg⁻¹, was injected intraperitoneally 5 hours before harvesting the transplanted fat pads.

Oxygen consumption assays

At day 6 of differentiation, oxygen consumption was measured at 37°C in cultured fat cells using a Mitocell S200 micro respirometry system. For cyclic-AMP-induced respiration assays, fully differentiated fat cells were incubated with 1mM dibutyryl cyclic AMP for 12 hr before measuring oxygen consumption. 1μM oligomycin (Sigma-Aldrich) was added to block ATP production, rendering uncoupled respiration. Basal and uncoupled (i.e. oligomycin-insensitive) respiration rates were normalized by total cell numbers. Similar results were observed when the results were normalized by total cell protein amount.

FACS analysis

FACS experiments were carried out on a FACSAria cell sorter at Beth Israel Deaconess Medical Center Flow Cytometry Core Facility. Briefly, primary SVF were isolated from inguinal depot of male 129SVE mice (4–6 weeks old) housed at ambient temperature. Subconfluent SV cultures passaged 2–3 times following the initial fractionation were trypsinized, centrifuged, resuspended in 1% FBS/PBS for labeling. Antibody incubations were performed on ice for 30 min for primary and 20min for secondary. Cells were washed and resuspended for sorting. Cells were initially selected by size, on the basis of forward scatter (FSC) and side scatter (SSC), followed by exclusion of immune cells (APC labeled F4/80 or CD19 positive cells). Cells were gated on both SSC and FSC singlets, ensuring that the staining of individual cells was analyzed. Next, the cells were separated on the basis of the cell-surface markers indicated. The following antibodies were used: rabbit anti-CD137 (abcam, ab64836); rabbit anti-TMEM26 (IMGENEX, IMG-6633A); Monoclonal antibody F4/80 Rat anti-Mouse (Invitrogen, MF48000); Monoclonal antibody CD19 Rat anti-Mouse (Invitrogen, RM7700); F(ab′)2 Donkey Anti-Rabbit IgG PE(ebioscience, 12-4739-81); F(ab′)2 Anti-Rat IgG APC (ebioscience, 17-4822-82).

Human subjects

Human brown fat biopsies from two independent cohorts of patients were analyzed for expression of beige- and brown-selective genes. Seven are from University of Turku, Finland and four from Maastricht University, the Netherlands. Details on subjects and procedures are as described previously (Orava et al., 2011; van Marken Lichtenbelt et al., 2009). Briefly, the site of biopsies of supraclavicular adipose tissue was decided by the cold exposure [¹⁸F]FDG-PET/CT image that showed activated BAT. A subcutaneous WAT sample was collected from the same incision. All subjects gave written informed consent before taking part in the study.

We thank Y. Wang (Center for Cancer Computational Biology at Dana-Farber Cancer Institute) for assistance with affymetrix gene chip analysis, S. White (Beth Israel Deaconess Medical Center Histology Core) for cryosectioning and A. Korde for technical help. We are grateful to Drs. P. Seale, R. Gupta, P. Cohen, S. Kleiner and J. Lo for useful discussions. B.M.S. is supported by NIH grants DK31405 and DK90861 and a grant from the JPB Foundation. P.S. is supported by a VICI research grant for innovative research from the Netherlands Organization for Scientific Research. J.W. was supported in part by a postdoctoral fellowship from the American Heart Association (Founders Affiliate #09POST2010078) and a Scientist Developmental Grant from the American Heart Association (National Center, #12SDG8070003).