Differentiating SGBS adipocytes respond to PPARγ stimulation, irisin and BMP7 by functional browning and beige characteristics

Brown and beige adipocytes are enriched in mitochondria with uncoupling protein-1 (UCP1) to generate heat instead of ATP contributing to healthy energy balance. There are few human cellular models to reveal regulatory networks in adipocyte browning and key targets for enhancing thermogenesis in obesity. The Simpson-Golabi-Behmel syndrome (SGBS) preadipocyte line has been a useful tool to study human adipocyte biology. Here we report that SGBS cells, which are comparable to subcutaneous adipose-derived stem cells, carry an FTO risk allele. Upon sustained PPARγ stimulation or irisin (a myokine released in response to exercise) treatment, SGBS cells differentiated into beige adipocytes exhibiting multilocular lipid droplets, high UCP1 content with induction of typical browning genes (Cidea, Elovl3) and the beige marker Tbx1. The autocrine mediator BMP7 led to moderate browning with the upregulation of the classical brown marker Zic1 instead of Tbx1. Thermogenesis potential resulted from PPARγ stimulation, irisin and BMP7 can be activated in UCP1-dependent and the beige specific, creatine phosphate cycle mediated way. The beige phenotype, maintained under long-term (28 days) conditions, was partially reversed by withdrawal of PPARγ ligand. Thus, SGBS cells can serve as a cellular model for both white and sustainable beige adipocyte differentiation and function.


Results
SGBS cells express surface markers similarly to primary preadipocytes and are heterozygous for the FTO risk allele rs1421085. Primarily, we examined undifferentiated SGBS cells by surface antigen expression analysis. We found that hematopoietic/monocyte markers (CD34, CD47), endothelial markers (CD54), fibroblast markers (CD73, CD90), integrins and CAMs (integrin ß1, CD44, CD325) were expressed on the surface of undifferentiated SGBS preadipocytes ( Supplementary Fig. S1a). Then, we compared the surface antigen expression pattern of SGBS preadipocytes to SVF cells isolated from human abdominal subcutaneous fat 41 . Most of the investigated markers were expressed similarly in SGBS and primary preadipocytes. However, CD34, CD44, CD146 and HLA-DR expression levels were higher in SGBS preadipocytes, while CD105, CD49a and CD31 antigens were expressed at a lower level compared to primary preadipocytes ( Supplementary Fig. S1a). Next, we tested the presence of the C risk-allele of the rs1421085 locus; DNA sequencing ( Supplementary  Fig. S1b) and qPCR-based genotyping analysis (data not shown) determined that SGBS cells are heterozygous for the C risk allele.
SGBS preadipocytes respond to sustained PPARγ ligand and irisin or BMP7 treatment by inducing either beige or classical brown marker genes. We applied previously described white (initiated by four days treatment with the PPARγ-ligand rosiglitazone) 36 and browning (with the continuous presence of rosiglitazone during differentiation) 29 protocols to differentiate SGBS preadipocytes and compared the expression of selected thermo-and adipogenic marker genes 27 in the two settings. The browning cocktail highly induced Ucp1 mRNA expression. Similarly, the presence of human recombinant irisin or BMP7 on the top of the white differentiation protocol resulted in enhanced Ucp1 mRNA expression; presence of irisin or BMP7 in the browning cocktail did not increase Ucp1 expression further (Fig. 1a). mRNA of brown-fat specific genes, like Cidea and www.nature.com/scientificreports www.nature.com/scientificreports/ Elovl3 were also enriched during the administration of the browning cocktail and when irisin was added to the white differentiation cocktail (Fig. 1b). In contrast, we observed decreased expression of Lep, a white adipocyte marker gene, in response to the browning differentiation (the effect of irisin and BMP7 was not significant in that respect). The expression of other adipogenic markers, Cebpb and Pparg, did not differ between white and browned SGBS adipocytes. Fabp4 was expressed at a significantly higher level in browned adipocytes compared to the white ones. Out of these markers, only the expression of Pparg was increased in response to irisin or BMP7 treatment ( Supplementary Fig. S2).
The expression levels of the mitochondrial enrichment marker, Cyc1 and the master regulator of mitochondrial biogenesis, Pgc1a were significantly higher in browned SGBS cells compared to white adipocytes and irisin Figure 1. Browning of SGBS cells is induced by PPARγ-driven differentiation cocktail, irisin or BMP7 treatment. SGBS preadipocytes were differentiated to white (W) or brown (B) for two weeks; human recombinant irisin treatment at 250 ng/ml concentration (green bars) or BMP7 treatment at 50 ng/ml concentration (red bars) were applied to induce browning of SGBS cells from day 1. Expression of Ucp1 (a), Cidea, Elovl3 (b), Cyc1 (c), Tbx1, Cited1, Zic1, Pdk4 (e) genes in SGBS adipocytes. Gene expression was determined by RT-qPCR and target genes were normalized to Gapdh. Relative mitochondrial DNA content of preadipocytes and differentiated SGBS cells determined by quantitative PCR (d) n = 5 *p < 0.05; **p < 0.01. www.nature.com/scientificreports www.nature.com/scientificreports/ treatment had the same effect (Fig. 1c). In the undifferentiated SGBS preadipocytes we could detect high mitochondrial DNA content. Differentiated white adipocytes have relatively lower mitochondrial DNA content and irisin treatment resulted in significantly elevated mitochondrial DNA amount in them while the effect of BMP7 was moderate. The mitochondrial DNA amount was the highest in the case of browned cells after the application of the PPARγ-driven browning differentiation cocktail (Fig. 1d).
Next, we asked the question whether the beige-selective marker genes, including Tbx1 42 and Cited1 12 or classical brown adipocyte markers, like Zic1 23 were induced in SGBS cells. Irisin or the browning protocol resulted in marked upregulation of Tbx1 and Cited1 but no Zic1 induction. There was no further increase of Tbx1 and Cited1 expression when irisin was added on top of the browning protocol. BMP7, on the other hand, only upregulated Zic1 markedly and even prevented Tbx1 induction when it was combined with the browning differentiation cocktail (Fig. 1e). The expression of Tmem26 and Shox2, proposed beige-markers 43 reported in mouse studies, was not elevated significantly in the browned SGBS adipocytes as compared to the white ones ( Supplementary Fig. S3). Furthermore, Tnfrsf9 (CD137) was expressed markedly only in undifferentiated preadipocytes (data not shown). Interestingly, Pdk4, originally described as a classical brown marker in mice 44 , showed a similar expression pattern as the beige-selective Tbx1 and Cited1, in SGBS adipocytes (Fig. 1e).
Then, we determined the concentration dependence of the irisin and BMP7 effects during white adipocyte differentiation in SGBS cells. Jedrychowski et al. 45 found that irisin was present at 3-4 ng/ml concentration in human blood plasma in the case of sedentary life but increased to 4-5 ng/ml if individuals took part in aerobic training. Itgav, the gene which encodes the specific subunit of the recently identified irisin receptor 15 is expressed both in undifferentiated SGBS preadipocytes and in differentiated adipocytes to the same extent ( Supplementary  Fig. S3). At 5 ng/ml concentration of irisin only a slight increase could be observed in Ucp1 expression compared to untreated cells. However, irisin was efficient above 50 ng/ml concentration. The same trend was observed when the expression of Tbx1, Cyc1, Elovl3 and Pgc1a was analyzed ( Supplementary Fig. S4). The supplementation of BMP7 to the white adipogenic differentiation media in increasing concentrations resulted in upregulated expression of Ucp1 and Cidea brown marker genes compared to the untreated white adipocytes. In contrast, we observed downregulation of Lep at 100 ng/ml BMP7 concentration. However, we could not detect significant changes in the expression of Tbx1, Cyc1, and Elovl3 as a result of BMP7 treatment ( Supplementary Fig. S5).
As a next step, we investigated the morphological characteristics of the white and browned SGBS cells by assessing the textural parameters and UCP1 protein content of the individual adipocytes using laser-scanning cytometry (Fig. 2a, Supplementary Fig. S6). We found higher UCP1 content in single SGBS adipocytes (Fig. 2b) and in total cell lysates (Fig. 2c) as a result of the 14 day-long PPARγ-driven browning differentiation, as compared to white. Moreover, we detected increased UCP1 protein content in response to irisin or BMP7 administration (Fig. 2b). In contrast to primary adipocytes 27,46 the texture sum variance did not change significantly during the induction of browning marked by higher UCP1 expression (Fig. 2d).
Differentiated SGBS adipocytes respond to activation as functional beige cells using the creatine phosphate cycle. To test the functional capacity of differentiated SGBS adipocytes, mitochondrial oxygen consumption rate (OCR) was measured (Fig. 3a,b). We found higher basal OCR after the browning was induced by either PPARγ ligand, irisin or BMP7. When we stimulated adipocytes with cAMP, the OCR of browning adipocytes was elevated more robustly compared to the white adipocytes (Fig. 3b). A similar trend was observed as a result of oligomycin treatment which is an inhibitor of H + -ATP synthase and provides the measurement of proton leak respiration (Fig. 3b). In parallel, we detected significantly elevated extracellular acidification rate (ECAR) both in untreated and cAMP-stimulated browned adipocytes as compared to white cells (Fig. 3c).
We found it important to determine what proportion of cAMP induced OCR is related to the UCP1 independent creatine phosphate futile cycle, a characteristic feature of beige adipocytes, utilizing ß-guanidinopropionic acid (β-GPA) which is an inhibitor of this cycle 21 . The creatine-cycle related OCR was markedly higher in case of the browned SGBS cells, as well as in irisin and BMP7 treated adipocytes, in contrast to untreated white adipocytes (Fig. 3d). Ckmt1a and Ckmt2 mitochondrial creatine kinases were expressed at a greater extent in the browned SGBS adipocytes as compared to those which were differentiated to white. The expression level of these genes was below the detection limit in undifferentiated preadipocytes (Fig. 3e). This suggests that the creatine kinase futile cycle is active in browned SGBS adipocytes induced either by PPARγ-driven browning differentiation or by the administration of irisin and BMP7 during white differentiation, demonstrating that the browned SGBS adipocytes functionally resemble beige cells.
Notably, Pm20d1 which encodes the secreted enzyme peptidase M20 domain containing 1 was induced in response to the administration of the browning cocktail and when irisin was added to the white differentiation cocktail ( Supplementary Fig. S3). This means that browned SGBS adipocytes have the potential to produce N-acyl amino acids which can function as endogenous uncouplers of UCP1-independent thermogenesis 47 .

The brown/beige adipocyte phenotype of differentiated SGBS cells is maintained in the absence of PPARγ-ligand.
In order to determine whether the PPARγ-driven differentiation could maintain a beige phenotype on a longer time frame, we performed a long-term experiment after 2 weeks for 21 or 28 days; and in parallel samples the browning cocktail was replaced by the white differentiation cocktail for the third (B14,W7) and the fourth week (B14,W14). The highest Ucp1 mRNA expression was detectable at the end of the long-term browning differentiation program, at day 28 (Fig. 4a). Ucp1 mRNA level was increased when we replaced the browning cocktail to white, at least until day 21, then sank to the level of day 14. UCP1 protein expression was detected by Western blot at day 14 of PPARγ-driven differentiation and it was further enhanced by after one or two weeks of further stimulation that is long-term rosiglitazone induction resulted in robust upregulation of UCP1 at the protein level. When the PPARγ stimulus was removed after 2 weeks of browning differentiation program, the expression level of UCP1 protein was slightly elevated by the end of the third week, (2019) 9:5823 | https://doi.org/10.1038/s41598-019-42256-0 www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ and it was still detectable one week later at a decreased level (Fig. 4a, Supplementary Fig. S7). This was verified with 3 independent anti-UCP1 antibodies ( Supplementary Fig. S8).
Parallely, we investigated the expression of Cidea and Elovl3 brown-fat specific genes; similarly to Ucp1, further upregulation was observed when we continued the administration of the browning protocol; when PPARγ-stimulation was withdrawn, their expression levels remained elevated as compared to B14 (Fig. 4b). The replacement of the cocktails from browning to white without rosiglitazone did not induce de novo white adipogenesis, marked by the continuous low expression of Lep (Fig. 4c).
After four-weeks of differentiation, we observed by immunohistochemistry a six-fold higher UCP1 protein content as a result of the browning cocktail compared to white. When the PPARγ-agonist was replaced by the www.nature.com/scientificreports www.nature.com/scientificreports/ white differentiation protocol for the last 2 weeks the UCP1 protein content per cell did not change significantly (Fig. 4d) as compared to adipocytes maintained continuously in the presence of the PPARγ-driven differentiation regimen. In parallel, similar texture sum variance was detected in both cell populations (Fig. 4d). Expression of mitochondrial complex subunits detected by immunoblotting (e). Relative optical density was determined by Image J software. ß-actin was used as endogenous control, n = 3. All gels were run under the same conditions. The blots were cropped. The original pictures of the full-length blots are displayed in Supplementary Fig. S7. *p < 0.05; **p < 0.01. www.nature.com/scientificreports www.nature.com/scientificreports/ Mitochondrial oxidative phosphorylation (OXPHOS) plays a fundamental role in energy production in brown/beige adipocytes 48 . We found that protein expression of mitochondrial respiratory components in complex II, III and IV followed the pattern of UCP1 expression in SGBS adipocytes (Fig. 4e, Supplementary Fig. S7) that is higher expression of respiratory chain proteins were found in browned adipocytes compared to control white adipocytes indicating stimulated mitochondrial biogenesis. The withdrawal of PPARγ agonist resulted in only a slight reduction of the level these proteins suggesting the clearance of a small fraction of the mitochondria. These results suggest that the signaling pathways driving the beige phenotype in SGBS cells can be maintained at least for two consecutive weeks after the PPARγ agonist is removed.
Recently, we showed that hADMSCs from subcutaneous fat could be differentiated into beige adipocytes and we observed upregulation of brown-specific marker proteins (UCP1, CIDEA) in single adipocytes by using laser-scanning cytometry. Furthermore, we demonstrated that irisin could induce beige type of browning when added during white adipocyte differentiation as indicated by Tbx1 induction while BMP7 addition resulted in Zic1 induction pointing to the possibility of the classical brown adipocyte differentiation pathway 27 . However, the major limitations of experiments using hADMSCs are the restricted availability of biopsy material and the potentially large variability between donors, which may greatly affect the reproducibility of the obtained data. Fat donors may vary in nutrition, metabolism, life-style and genetic background, e.g. the presence of the risk-allele of the FTO locus (rs1421085), which fundamentally affects adipocyte browning and represents robust genetic association with obesity 31 .
Recently, several attempts were reported to overcome the limitations described above. Artificially immortalized cell lines, like PAZ6 cells, which were obtained from SVF of human infant BAT, and transformed with SV40 T and t antigens were able to differentiate into brown adipocytes in vitro 49 . PAZ6 is considered to be a brown preadipocyte cell line, because high UCP1 expression was shown in differentiated PAZ6 cells which could be further induced by norepinephrine 50 . Shinoda et al. compared the differentiation of 65 clonal preadipocyte lines originated from supraclavicular fat biopsies with 35 lines which were generated from the SVF of subcutaneous WAT. RNA sequencing was performed from three clonal brown and white adipocyte cultures, differentiated from supraclavicular and subcutaneous progenitor clones, respectively. They found that human browned adipocytes from supraclavicular regions display molecular signatures which resemble thermogenic beige adipocytes 32 . In another study, clonally derived adipocyte progenitors from the deep neck area were capable of differentiating to functional thermogenic adipocytes and responded to BMP7 administration, in contrast to those which were isolated and generated from subcutaneous WAT 33 .
The SGBS human preadipocyte cell line is often used in experiments to study human adipocyte biology and previously it served as a representative model of white adipocyte differentiation 34-37 . The first results, showing the possibility that SGBS cells can differentiate into thermogenic adipocytes were described in 2013. 2-oxoglutarate dependent dioxygenase (encoded by the proposed Fto gene) deficient SGBS preadipocytes differentiated to white adipocytes had elevated UCP1 expression and uncoupled respiration, without any changes in mitochondrial mass or structure 39 . Later, Tews et al. 40 found that TENM2 (teneurin-2), which might inhibit the classical brown marker, Zic1 51 , is enriched in white adipocyte progenitor cells from subcutaneous neck WAT compared to deep neck ones. They demonstrated that TENM2 knockdown with siRNA in SGBS cells resulted in increased UCP1 at mRNA and protein level. Furthermore, basal and proton leak mitochondrial respiration was enhanced. Interestingly, TENM2 deficient SGBS adipocytes had the same amount of mitochondria as the wild type ones and gained larger lipid droplets than the control cells. The aforementioned results suggest that TENM2 knockdown in SGBS adipocytes results in a beige gene expression program (the expression of Zic1 did not change) without enhanced mitochondrial biogenesis and accumulation of small lipid droplets in a multilocular arrangement 40 . In these studies, SGBS cells differentiated to white adipocytes for two weeks expressed UCP1 at low, but detectable levels which we could confirm in the present study. In response to the knockdown of 2-oxoglutarate dependent dioxygenase or TENM2 or the administration of the atypical antipsychotic clozapine 28 , the browning potential of the SGBS cells could be significantly enhanced.
In contrast to these studies and to results presented here, Guennoun et al. found that the SGBS adipocytes switch their phenotype during a four week-long differentiation program. Within two weeks, they noticed high UCP1 expression and a thermogenic phenotype even as a result of the white differentiation protocol without adding any browning stimuli. Then, the expression of UCP1 greatly declined in response to the continuation of the white differentiation for additional two weeks 52 . Yeo et al. seemingly strengthened these results by comparing the gene expression pattern of the differentiated SGBS and primary hADMSC-derived subcutaneous adipocytes 53 . Of note, hADMSCs in this case were obtained from obese patients and there might be a difference between the differentiation capacity of these and healthy hADMSCs and SGBS preadipocytes. They also observed that UCP1 protein expression was high in SGBS cells differentiated for 12 days even when either rosiglitazone or T3 were omitted from the media. In these experiments they considered a band on immunoblots with an apparent molecular mass of approximately 25 kDa as UCP1.
In our experiments, however, we could demonstrate UCP1 protein expression in response to long-term rosiglitazone and T3 administration (browning protocol) at ~33 kDa, but not when the white cocktail was applied. In order to confirm that we take the right band for UCP1, three kinds of anti-UCP1 (both monoclonal and polyclonal) antibodies were used to investigate the UCP1 protein expression. The application of polyclonal antibodies (U6382, PA1-24894) resulted in three distinct bands at ~36 kDa, ~33 kDa and below 28 kDa ( Supplementary  Fig. S8). However, when a monoclonal antibody (MAB6158) was applied, we could clearly demonstrate which band corresponds to the UCP1 protein, supported by its appearance at the predicted molecular weight of 33 kDa.
www.nature.com/scientificreports www.nature.com/scientificreports/ Furthermore, UCP1 expressed in a mouse BAT lysate was detected at the very same molecular weight as the UCP1 band in the browning SGBS adipocyte samples (Supplementary Fig. S8).
Using laser-scanning cytometry 27 and applying functional assays we aimed to investigate whether classical brown or beige adipocyte differentiation (browning) can be induced in SGBS cells and if these adipocytes maintain their morphology during long-term culturing. The browning of SGBS cells was induced by multiple approaches: sustained PPARγ stimulation, irisin and BMP7 treatment. Irisin induces browning of subcutaneous white adipocytes ex vivo 17,25,27 , BMP7 drives brown fat cell differentiation of both mesenchymal progenitor cells and committed brown preadipocytes. In vivo, BMP7 is able to increase brown fat mass and thermogenic energy expenditure in mice 18 . The present study demonstrates that the PPARγ-driven browning protocol (including rosiglitazone) and irisin treatment could be successfully used to induce browning of SGBS adipocytes, which resulted in a beige phenotype. BMP7 administration had a more moderate effect and induced a distinct gene expression program without the upregulation of the beige-selective markers, Tbx1 and Cited1. Both in primary subcutaneous and in SGBS adipocytes, BMP7 treatment enhanced the expression of the Pparg and Zic1 genes suggesting that this mediator rather induce a classical brown-like differentiation 27 . This program, similarly to irisin administration, resulted in increased mitochondrial DNA and Ucp1 levels and elevated mitochondrial respiration. Of note, the majority of the browning marker genes were detectable in white SGBS adipocytes but their expression remained markedly lower compared to the browned cells.
The functional experiments detected UCP1-dependent proton leak (heat production) and high extracellular acidification in the case of beige adipocytes induced by either rosiglitazone or irisin. These parameters were significantly lower in the case of white adipocytes, however, could be stimulated in response to cAMP supporting the notion that some of the differentiated white adipocytes are masked beige cells. Besides UCP1-dependent thermogenesis, UCP1-independent heat-producing mechanisms were described as a beige specific feature. In our experiments, the involvement of the creatine-phosphate futile cycle 21 was investigated in the heat production of SGBS beige adipocytes. Using a specific compound (β-GPA) which depletes creatine and inhibits the cycle, we demonstrated the induction of this futile cycle in response to a β-adrenergic cue in beige SGBS adipocytes.
Several single nucleotide polymorphisms located in the fat mass and obesity-associated gene, Fto were described by genome-wide association studies as a genetic trait predisposing for obesity 54,55 . The T-to-C conversion at rs1421085 position affects a mesenchymal super-enhancer site and leads to increased IRX3 and IRX5 expression levels during early adipocyte differentiation, because of the lack of ARID5B binding and repression, which results in a developmental shift from energy-dissipating beige cells to energy-storing white adipocytes. Therefore, the potential of mitochondrial thermogenesis is reduced in those individuals who carry the C risk allele which occurs at a high frequency (approx. 44%) in the European population 31 . So far no information has been available about the status of this allele in SGBS cells. A CC to TT rescue by genome editing of primary adipocytes resulted in an elevated Ucp1 expression by two and an increased stimulated oxygen consumption by seven fold 31 . To our knowledge, no data is available how a heterozygous composition affects these parameters. In this study, we demonstrated that the SGBS cell line carries one copy of the C risk allele; still these preadipocytes were able to differentiate into functional beige adipocytes if browning stimuli (e.g. rosiglitazone or irisin treatment) were continuously present during their differentiation.
In summary, the induction of browning in SGBS cells followed the beige pathway, as a result of the PPARγ-driven differentiation (rosiglitazone) or irisin treatment. The beige phenomenon is supported by the elevated expression of brown and beige marker genes, multilocular morphology, enrichment of mitochondria, high oxygen consumption in response to cAMP and significant involvement of the creatine-phosphate futile cycle in heat production. In our experiments, when the browning-inducer, rosiglitazone was eliminated from the differentiation media, UCP1 expression and mitochondrial enrichment could be partially maintained. In summary, we propose that the human SGBS preadipocyte cells represent an uncommitted preadipocyte stage with one FTO risk allele, which are able to differentiate into both white and beige adipocytes and can be used as an easily applicable and valuable model to study human adipocyte browning in a reproducible as well as high-throughput manner contributing to the development of novel therapeutic approaches against obesity.
Human recombinant irisin (Cayman Chemicals) was used at 250 ng/ml concentration, and human recombinant BMP7 (R&D systems, Minneapolis, USA) was administered at 50 ng/ml concentration 26,27 . Where indicated, the browning-inducers were added to the differentiating cells in increasing concentrations.
Oxygen consumption and extracellular acidification measurements. Real-time oxygen consumption and extracellular acidification rates were measured by using an XF96 oximeter (Seahorse Biosciences, North Billerica, MA, USA). SGBS cells were seeded onto 96-well XF96 cell culture microplates. Cells were kept in growth medium at the longest for 24 hours and then the formerly-described differentiation process started. After recording the baseline oxygen consumption, cells received a single bolus of dibutyril-cAMP at 500 µM concentration to induce adrenergic stimulation. Then, stimulated oxygen consumption was measured in every 30 minutes. The final reading took place at 6 h post-treatment. Differentiated adipocytes were treated with 2 mM ß-guanidinopropionic acid (Sigma) to block the creatine-driven substrate cycle 21 . In addition, proton leak respiration was measured by oligomycin (Enzo, USA) treatment at 2 µM concentration, which blocks the ATP synthase. For baseline correction, cells received a single bolus of Antimycin A (Sigma) treatment at 10 µM concentration. After the measurements, oxygen consumption rate was normalized to protein content 27,28,56 . Laser-scanning cytometry. SGBS cells were plated on eight-well ibidi micro slides (Ibidi GmbH, Germany) and differentiated as previously described. Immunofluorescence staining was carried out as described previously 27 . Scanning was done by an iCys Research Imaging Cytometer (iCys, Thorlabs Imaging Systems, Sterling, VA, USA). The images were processed and analyzed (n = 3, 2000 cells per SGBS sample) by our high-throughput automatic cell-recognition protocol 27 with some modifications using the iCys companion software (iNovator Application Development Toolkit version 7.0, CompuCyte Corporation, Westwood, MA, USA). As previuosly, cells were identified according to their Hoeschst 33342 nuclear staining. Then, the fluorescence signal intensity of the UCP1 immunostaining and the texture sum variance of the light scatter signal of lipid droplets were quantified in each cell within the 30-pixel immediate outward vicinity of the nucleus contour by the Cell Profiler software (The Broad Institute of MIT, MA, USA). Afterward, based on these fluorescence and light scatter signal of single cells, a semi-automated classification and enumeration of the differentiated white and brown adipocytes and undifferentiated preadipocytes was carried out applying the trained classification "Fast Gentle Boosting" of the Cell Profiler Analyst software (The Broad Institute of MIT, MA, USA).

Flow cytometry.
To investigate the phenotype of the undifferentiated SGBS preadipocytes, a multiparametric analysis of surface antigen expression was performed by three-color flow cytometry using fluorochrome-conjugated antibodies with isotype matching controls. See reference ref. 41 for further details about the analysis 41 .