Tbx15 Defines a Glycolytic Subpopulation and White Adipocyte Heterogeneity

Over the past decade, it has become clear that adipose tissue and its contribution to metabolic syndrome are more complex than originally believed. Accumulation of visceral white adipose tissue (WAT) is associated with insulin resistance and increased risk of metabolic disease, whereas subcutaneous WAT may be protective against the development of metabolic disorders (1). We and others (2,3) have shown that differences in WAT depots are also marked by differential expression of developmental genes.

Tbx15 is a member of the T-box gene family of developmental genes and has an established role in skeletal formation (4). We have recently shown that in skeletal muscle, Tbx15 is highly and specifically expressed in glycolytic myofibers and regulates the metabolism of these myofibers (5). Furthermore, Tbx15 is differentially expressed among different WAT depots in both rodents and humans and is markedly downregulated in obesity (2). Genome-wide association studies have shown that a single nucleotide polymorphism near the TBX15 gene correlates with adiposity in patients of both European and African ancestry (6).

In the current study, we show that even within a single WAT depot, the adipocytes and preadipocytes are heterogeneous with high Tbx15 expression in only a subset of cells. Metabolic profiling demonstrates that these Tbx15^Hi preadipocytes and adipocytes are highly glycolytic, whereas Tbx15^Low preadipocytes and adipocytes are more oxidative. Expression of TBX15 also correlates to markers of glycolytic metabolism in human fat. Finally, we demonstrate that overexpression of Tbx15 is sufficient to drive an increase in glycolytic metabolism in cultured adipocytes. Thus, Tbx15 differentially regulates metabolism in white adipocytes and leads to a functional heterogeneity in white fat cells similar to that found among muscle fibers.

C57BL/6 mice (The Jackson Laboratory), Tbx15-LacZ mice on a mixed genetic background (4), and Immortomouse (7) were housed with a 12-h light/dark cycle in a temperature-controlled room and allowed ad libitum access to water and food (Mouse Diet 9F 5020; PharmaServ). Animal care and study protocols were approved by the Animal Care Committee of Joslin Diabetes Center in accordance with National Institutes of Health guidelines.

Immunofluorescence using an anti-Tbx15 antibody (8) and fluorescence in situ hybridization (9) were performed as described previously. Regions corresponding to 1,815–2,600 bp of Tbx15 mRNA and 39–796 bp of Adipoq mRNA were cloned into PCRII-TOPO as probe templates. The Tbx15 probe was labeled by incorporation of uridine-5’-triphosphate–digoxigenin, and the Adipoq probe was labeled by incorporation of uridine-5’-triphosphate–biotin. Probes were detected with Alexa Fluor 488 and 594 Tyramine signal amplification kits (Thermo Fisher).

Preadipocytes were isolated, grown, and differentiated as previously described (10). Preadipocytes were plated, expanded, stained with fluorescein di(β-d-galactopyranoside), and sorted, as previously described (11). Immortomouse clonal lines were created from the subcutaneous fat of a single 8-week-old male mouse. The stromal vascular fraction was isolated, plated, and grown at 33°C in media supplemented with 10 units/mL interferon-γ. A total of 190 individual cells was FACS sorted and grown to clonal cell lines. Twenty of the reproducibly adipogenic cell lines were chosen for further analyses.

Oxygen consumption rates (OCRs) and extracellular acidification rates (ECARs) were measured using the XF^e24 Extracellular Flux Analyzer as previously described (8) using glucose (10 mmol/L), Glutamax (2 mmol/L), and sodium pyruvate (1 mmol/L) as the substrate.

Tbx15 was stably overexpressed in 3T3-L1 cells using a pBABE-puro retrovirus (8) and knocked down in 3T3-L1 cells by lentiviral short hairpin RNA as previously described (5).

Oil Red O staining and triglyceride quantification were performed as previously described (8).

Gene expression was examined in fat samples obtained from 65 previously described human patients (12). All study protocols have been approved by the Ethics Committee of the University of Leipzig (363–10–13122010 and 017–12–230112).

All differences were analyzed by ANOVA or Student t test if normally distributed. A Wilcoxon t test was used for data that were not normally distributed. Results were considered significant if P was <0.05.

We have previously observed that Tbx15 is more highly expressed in subcutaneous than visceral WAT (2) and also much higher in subcutaneous preadipocytes compared with perigonadal preadipocytes (10). Indeed, Tbx15 mRNA levels were ∼60-fold higher in subcutaneous WAT compared with perigonadal WAT of 8-week-old C57BL/6 mice (Fig. 1A). To determine if this difference in expression of Tbx15 among different fat depots was because of an increase of Tbx15 in all subcutaneous fat cells or to the presence of a subpopulation of cells with high Tbx15 expression, we FACS sorted preadipocytes isolated from fat pads of Tbx15-LacZ reporter mice. Preadipocytes (negative for Ter¹¹⁹, CD45, and CD31 and positive for SCA1 and CD34) (13) were collected and further sorted for Tbx15 by staining cells with fluorescein di(β-d-galactopyranoside), which is cleaved by β-galactosidase (encoded by the LacZ gene) to release fluorescein isothiocyanate (Fig. 1B).

Consistent with the low levels of Tbx15 in perigonadal fat (2,10), Tbx15^Hi preadipocytes were virtually absent in perigonadal fat. By contrast, ∼5% of preadipocytes from subcutaneous fat were Tbx15^Hi (Fig. 1C and D and Supplementary Fig. 1A). Quantitative real-time PCR (qPCR) analysis of mRNA from these cells confirmed a six-fold higher level in the expression of Tbx15 compared with Tbx15^Low preadipocytes. The Tbx15^Hi preadipocytes demonstrated 50–80% lower levels of expression of markers of oxidative metabolism, including Pgc1a and cytochrome c oxidase subunits 8b and 5b (Cox8b and Cox5b). Conversely, Tbx15^Hi adipocytes showed 60–120% increases in the markers of glycolytic metabolism hexokinase 2 (HK2) and Gapdh (Fig. 1E). The expression of other markers of oxidative phosphorylation, glycolytic metabolism, and the pentose phosphate pathway was not changed (Supplementary Fig. 1B).

Oil Red O staining and quantitation of triglyceride accumulation revealed no morphological differences between Tbx15^Hi and Tbx15^Low adipocytes after differentiation (Fig. 1F). qPCR analysis of differentiated adipocytes confirmed 4.5-fold higher levels of Tbx15 mRNA in Tbx15^Hi versus Tbx15^Low adipocytes but similar levels of markers of adipogenic differentiation (Fig. 1G). However, like the preadipocytes, Tbx15^Hi adipocytes had ∼30–70% lower levels of oxidative metabolism markers Pgc1a, Cox5b, and Cox8b and ∼20–140% higher levels of the glycolytic metabolism markers HK2 and Gapdh (Fig. 1H). Furthermore, Tbx15^Hi adipocytes had a trend toward decreasing markers of brown/brite fat compared with Tbx15^Low adipocytes (Supplementary Fig. 1C). Thus, Tbx15 expression marks a distinct subpopulation of preadipocytes and white adipocytes with high expression of markers of glycolytic metabolism and low expression of markers of oxidative metabolism.

To further investigate how heterogeneity in Tbx15 expression affects the adipocyte biology, we isolated single preadipocytes by FACS and created conditionally immortalized clonal adipogenic cell lines from the WAT of an Immortomouse (7). qPCR analysis demonstrated that 17 of these clonal cell lines had relatively low expression of Tbx15 mRNA (Tbx15^Low), whereas 3 of these cell lines had expression levels approximately five-fold higher (Tbx15^Hi) (Fig. 2A). This higher level of Tbx15 persisted after adipogenic differentiation, whereas lipid accumulation and the expression of markers of differentiation were unchanged (Fig. 2B). However, like the primary adipocytes, these Tbx15^Hi adipocytes had ∼70% lower expression of Pgc1a and ∼30% reduction in Cox8b compared with Tbx15^Low adipocytes. In contrast, in these immortalized cells, there were no differences in the expression of glycolytic enzymes or expression for markers of brown, brite, or white fat (14) (Fig. 2B and C).

During culture of these clonal cell lines, we observed differences in media acidification (ranging from pH 8.4 to 6.2) visualized by the color of the media (Fig. 2D). Lactic acid production, an end product of glycolysis, contributes to media acidification (15). The highly glycolytic Tbx15^Hi cell lines (circled wells) all demonstrated high media acidification rates with an average pH of 6.4, whereas the Tbx15^Low cells exhibited a mixture of high and low acidification rates and had an average pH of 7.2. Seahorse flux analysis confirmed these metabolic differences with a 35–40% reduction in basal and maximal OCRs and 50–65% higher basal and maximal ECARs in Tbx15^Hi versus Tbx15^Low preadipocytes (Fig. 2E and F). This resulted in an approximately three-fold increase in the ratio of ECAR/OCR in Tbx15^Hi cells, indicating a shift from oxidative to glycolytic metabolism (Fig. 2G). Treatment with known inducers of thermogenic markers in brite adipocytes, including treatment with 3.3 nmol/L BMP7 for 3 days prior to differentiation (16) and acute treatment with 10 μmol forskolin for 4 h (17), resulted in no differences in Ucp1 expression between Tbx15^Hi and Tbx15^Low adipocytes. Although forskolin treatment was able to markedly upregulate Ucp1 expression in both Tbx15^Hi and Tbx15^Low adipocytes, expression levels of Ucp1 in these cells remained 50-fold lower than those observed in brown adipocytes (Supplementary Fig. 2).

Heterogeneity of Tbx15 expression in white adipocytes was also observed in vivo. Compared with adiponectin expression, a marker of all mature white adipocytes, fluorescence in situ hybridization demonstrated that Tbx15 could be detected in only ∼25% of mature adipocytes in subcutaneous WAT and was virtually absent in adipocytes in perigonadal WAT (Fig. 3A and B). Similarly, immunofluorescence demonstrated that Tbx15 protein was not uniformly expressed, but was expressed in ∼30% of subcutaneous adipocytes and in only ∼7% perigonadal adipocytes (Fig. 3C and D). Although Tbx15 is expressed in preadipocytes, >95% of Tbx15-positive cells were also positive for adiponectin, presumably reflecting higher numbers of adipocytes to preadipocytes in the mature fat pad. These results correlate with the differential Tbx15 expression between these depots and demonstrate that Tbx15 expression defines adipocyte subpopulations within a single fat depot.

In human adipose tissue, TBX15 mRNA expression in adipose tissue was negatively correlated with BMI in both males and females. This was true of mRNA expression from both subcutaneous fat and visceral fat. Furthermore, TBX15 expression was significantly correlated with GAPDH and HK2 expression in both the visceral and subcutaneous WAT of males, with similar trends found in the subcutaneous WAT of females (Fig. 4A and B). To more directly investigate if Tbx15 directly regulates oxidative and glycolytic metabolism, we used a pBABE retrovirus to stably overexpress Tbx15 in a 3T3-L1 cell line, pBABE-Tbx15. These cells had a 25-fold overexpression of Tbx15 at the mRNA level (Fig. 4C). As we have previously reported (18), Seahorse flux analysis confirmed pBABE-Tbx15 cells had a 40–45% reduction in basal and maximal OCRs (Fig. 4D). pBABE-Tbx15 cells also have a 30% higher maximal ECAR that was fully suppressed by addition of 2-deoxyglucose (Fig. 4E and Supplementary Fig. 3A and B). These changes led to an approximately two-fold increase in the ratio of ECAR/OCR in pBABE-Tbx15 cells, indicating a shift from oxidative to glycolytic metabolism (Fig. 4F), as well as the change in the expression of enzymes that regulate glycolysis and oxidative phosphorylation (Fig. 4G and Supplementary Fig. 3C). In contrast, stable lentiviral reduction of Tbx15 in 3T3-L1 cells (shTbx15 cells) did not lead to observable changes in triglyceride accumulation, adipocyte differentiation, ECAR, OCR, or the expression of regulators of oxidative phosphorylation and glycolysis (Supplementary Fig. 3D–I).

In the current study, we have used three independent methodologies—FACS analysis, clonal cell lines, and histology—to demonstrate that Tbx15 is a marker of white adipocytes and preadipocyte heterogeneity that marks a highly glycolytic subpopulation of cells. These data closely parallel our previous finding that Tbx15 is expressed specifically in glycolytic skeletal muscle fibers. We also find that TBX15 levels in WAT of human subjects are negatively correlated with BMI and positively correlated with markers of glycolytic metabolism (Fig. 4A and B) (2,19). Taken together, these data strongly suggest that reduced TBX15 expression in obesity may be, at least in part, because of reduced numbers of highly glycolytic Tbx15^Hi cells. Thus, obesity leads not only to adipocyte hypertrophy and hyperplasia, but also to a shift in the cellular composition of the WAT.

We have previously shown in 3T3-L1 cells that overexpression of Tbx15 negatively regulates oxidative metabolism by reducing mitochondrial mass and activity (Fig. 4D) (18). In this article, we extend these studies to demonstrate that the reduced oxidative capacity of these cells is accompanied by an increase in glycolytic capacity. Thus, the increased glycolytic metabolism associated with high Tbx15 expression observed in primary adipocytes may be an adaptive response to compensate for reduced oxidative capacity. Indeed, obesity is correlated with both mitochondrial dysfunction (20) and increased basal lactate release (21), although the contribution of Tbx15 to these effects is not yet clear.

Although much attention has been paid to identifying differences between white, brown, and beige adipocytes, recent studies begun to show further molecular intradepot heterogeneity within these classifications (22–24). These studies serve to complement to earlier studies that demonstrate the functional heterogeneity of adipocytes in processes including glucose uptake (25), lipogenesis (26), lipolysis (27), and free fatty acid uptake (28). Importantly, although some cell-culture models have suggested that Tbx15 might be a marker of brite fat (29), we find that even after induction with BMP7 and forskolin that Tbx15^Hi white adipocytes tend to have reduced expression of brown/brite fat markers compared with the Tbx15^Low adipocytes.

In summary, our results show that the differential expression of Tbx15 leads to metabolic heterogeneity of white preadipocyte and adipocytes, even within a single depot. In both mice and, to some extent, humans, these Tbx15^Hi cells display lower oxidative and higher glycolytic metabolism than Tbx15^Low cells. Thus, just as we consider muscle fiber type when investigating muscle metabolism, it will be necessary to consider fat cell subtype when investigating adipocyte metabolism. Further investigation of these adipocyte subpopulations and the specific role of glycolytic and oxidative adipocytes in adipose tissue will potentially provide new targets in the treatment of obesity and its related disorders.

Acknowledgments. The authors thank Christie Penniman (Joslin Diabetes Center) for animal care, Jonathan Dreyfuss (Joslin Diabetes Center) for statistical analysis, Girijesh Buruzula and Joyce LaVecchio (Diabetes Research Center Flow Cytometry Core, Joslin Diabetes Center), Michele Pate (Ohio University Flow Cytometry Core), Misako Hata (Edison Biotechnical Institute, Ohio University), and Chris Cahill (Diabetes Research Center Microscopy Core, Joslin Diabetes Center) for technical assistance.

Funding. This work was supported by National Institutes of Health Grant R01-DK-082655 (to C.R.K.), a Research Center Grant P30-DK-036836 (to Joslin Diabetes Center), and a Training Grant T32-DK-007260 (to Joslin Diabetes Center); an American Diabetes Association mentor-based award (to C.R.K.); the Mary K. Iacocca Professorship (to C.R.K.); start-up funds from Ohio University College of Osteopathic Medicine (to K.Y.L.); American Diabetes Association Junior Faculty Development Award 1-17-JDF-055 (to K.Y.L.); and German Research Council Grant DFG KI728/3 (to A.K.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. K.Y.L. designed and performed experiments, analyzed data, and wrote the paper. R.S., G.G., S.U., and M.B. designed and performed experiments. Y.L. and L.W. analyzed data. D.E.B. and A.K. designed experiments and analyzed data. C.R.K. designed experiments, analyzed data, and wrote the paper. All authors discussed the results and implications and commented on the manuscript at all stages. C.R.K. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.