Activin A Plays a Critical Role in Proliferation and Differentiation of Human Adipose Progenitors

Activin A (Preprotech) was prepared as a 100 μg/ml stock solution in culture medium supplemented with 0.5% FCS. SB431542 was purchased from Sigma-Aldrich. SB203580 was from CalbioChem. Rosiglitazone (BRL4953) was a gift from Dr. J.F. Dole (GlaxoSMithKline, King of Prussia, PA).

Cells were rinsed with PBS and solubilized in stop buffer containing 50 mmol/l Hepes, pH 7.2; 150 mmol/l NaCl; 10 mmol/l EDTA; 10 mmol/l Na₄P₂O₇; 2 mmol/l Na₃VO₄; and 1% Triton X-100 supplemented with Protease Inhibitor Cocktail (Roche). The primary antibodies and dilutions used were as follows: mouse anti–activin A (1/1,000; R&D Systems); C/EBPβ (C19, 1/200; Santa Cruz); Smad2/3 and phospho-Smad2 (1/1,000; Cell Signaling); extracellcular signal–related kinase (ERK)1/2 and phospho-ERK1/2 (1/2,000 and 1/1,000, respectively; Cell Signaling); enolase 1 (ENO1) (1/1,000; Santa Cruz); KLF4 (1/200; Abcam); β-tubulin-1 (1/5,000; Sigma Aldrich); and fatty acid–binding protein (FABP)4 (1/1,000), which was a gift of Dr. D.A. Bernlohr (Department of Biochemistry, Molecular Biology and Biophysics, University of Minnesota-Twin Cities, Minneapolis, Minnesota). Secondary horseradish peroxidase–conjugated antibodies were purchased from Promega.

In the first group, needle biopsies of subcutaneous white adipose tissue from 12 morbidly obese (age 41.5 ± 3.2 years; BMI 45.8 ± 1.8 kg/m²) and seven lean (age 43.71 ± 4.51 years; BMI 21.45 ± 0.45 kg/m²) women of the same ages were collected. In these groups, we measured gene expression of INHBA. In a second population, subcutaneous and visceral white adipose tissue samples from 11 morbidly obese women who were candidates for bariatric surgery (age 37.7 ± 2.7 years and BMI 49.9 ± 2.6 kg/m²) were also collected for expression analysis of the genes described above. Visceral adipose tissue refers to the omental depot. Informed consent was obtained from all subjects. All clinical investigations were performed according to the Declaration of Helsinki and approved by the ethics committees of the Hôtel-Dieu (Paris, France).

hMADS cells were transfected with small interfering RNA (siRNA) duplexes using HiPerfect reagent (Qiagen) either during the exponential growth cell phase (2.8 × 10⁴ cells/well) to investigate the effects of cell proliferation or a day before confluence (7 × 10⁴ cells/well) to investigate effects on differentiation. siRNAs for human Inhba were as follows: 5′ gaaacaguucuuugucagu 3′ and siSmad2, 5′ ucuuuugugcagagccccaa 3′; siRNA control duplex was from Eurogentec. Cells were transfected with 8 nmol/l siRNA in medium supplemented with the low concentration of serum described composed of 60% Dulbecco's modified Eagle's medium low glucose, 40% MCDB-201, 10 μg/ml insulin, 5 μg/ml transferrin, 50 ng/ml selenium, 10⁻⁹ mol/l dexamethasone, 50 μg/ml ascorbic sodium acid, and 2.5 ng/ml fibroblast growth factor (FGF)2, supplemented with 0.5% FCS.

293 FT cells (Invitrogen) were transfected with pLenti6/Ubc-Slc7a1 (Addgene plasmid 17224) along with packaging mix using lipofectamine 2000 according to the supplier's instructions (Invitrogen). Forty-eight hours after transfection, the supernatant was collected, filtered through a 0.45-mm pore-size filter, and transferred to dishes containing proliferating hMADS3 cells for 24 h in the presence of 5 μg/ml polybrene. Cells expressing the mouse receptor for retrovirus were then selected in the presence of 2.5 μg/ml blasticidin. hMADS3-EcoRec cells displayed a similar proliferation rate and the same capacities to undergo differentiation into adipocytes and osteoblasts (not shown).

Ecotropic retrovirus vector expressing C/EBPβ-LIP (Addgene plasmid 15714), C/EBPβ-LAP (Addgene plasmid 15712), Klf4 (Addgene plasmid 17219), or green fluorescent protein (GFP) (kindly provided by Dr. Kitamura, University of Tokyo, Tokyo, Japan) were produced in PLAT-E cells (Cell Biolabs). Twenty-four hours after transfection, medium was collected and transferred to hMADS3-EcoRec dishes for 24 h in the presence of 5 μg/ml polybrene. Twenty-four hours after cells reached confluence, they were induced to undergo differentiation. With GFP used as a reporter, 80% of hMADS3-EcoRec cells were transduced using this protocol.

All primer sequences are described in supplemental Table 1. Real-time PCR assays were run on an ABI Prism 7000 real-time PCR machine (PerkinElmer Life Sciences) or in a GeneAmp 7500 detection system (Applied Biosystems, Courtaboeuf, France). Normalization was performed using the geometric averages of the housekeeping genes G6PDH, POLR2A, and TBP. Quantification was performed using the comparative DCt method.

Cells were plated onto 12-well plates (2,900 cells/cm²), and after the appropriate time cells were trypsinized and counted with a Coulter counter. For each experiment, three wells per condition were counted.

Statistical significance was determined by t tests or ANOVA with post hoc comparisons by Duncan's multiple-range test. Probability values <0.05 were considered statistically significant and are indicated with a single asterisk; <0.01 is indicated by double asterisks. Nonparametric tests were used to compare adipose tissue gene expression between lean and obese subjects (i.e., Wilcoxon test) and to compare subcutaneous versus visceral adipose tissue in obese subjects (i.e., paired Wilcoxon ranked test).

According to the results of a transcriptome analysis that we previously reported (20), we noticed that the INHBA gene was downregulated during adipogenic differentiation of hMADS cells. To ascertain that our conclusions are not limited to one specific system, we used three types of human adipose progenitors, i.e., hMADS cells, primary culture of preadipocytes, and native CD34⁺/CD31⁻ cell populations (see supplementary Fig. S1 for the isolation procedure), to confirm the expression of INHBA/activin A during adipogenesis. As shown in Fig. 1,A, the INHBA gene was expressed by undifferentiated hMADS cells and its level of expression decreased dramatically early upon induction of adipocyte differentiation. Protein levels of activin A secreted by undifferentiated and differentiated cells were in agreement with RNA levels (supplementary Fig. S2). Primary preadipocytes that were isolated from subcutaneous, mesenteric, and omental fat depots of adult subjects expressed the INHBA gene and also displayed a downregulation of this gene expression when induced to undergo differentiation (Fig. 1 B). These data indicate that activin A is expressed in adipose progenitors from various human fat depots independently of the age of the donors and is downregulated as progenitors undergo adipogenesis.

To check that higher INHBA/activin A expression was indeed correlated with the undifferentiated state of adipose progenitors in vivo, we monitored the INHBA gene expression in fractioned human adipose tissue. Human stromal-vascular fraction (SVF) is known to be a heterogeneous cell population containing adipose progenitors and other cell types, which can be isolated based on expression of cell-surface markers (4). Isolation of different cell subpopulations from the SVF and quantitative RT-PCR revealed that INHBA was expressed at a higher level in native CD34⁺/CD31⁻ adipose progenitors than in the adipocyte fraction (Fig. 1 C). This result did not exclude the possibility that INHBA could also be expressed by other cell types resident in the SVF (see discussion). Altogether, these data show that activin A is expressed by undifferentiated adipose progenitors and that this expression is inversely correlated with adipogenic differentiation. To reveal the potential relevance of this expression profile, we then investigated the functional role of activin A in human adipogenesis.

As shown in Fig. 1 and supplementary Fig. S2, activin A was secreted by undifferentiated cells and its level decreased dramatically upon induction of adipocyte differentiation, suggesting its role in regulating the number of undifferentiated hMADS cells. Thus, we analyzed the effect of activin A addition in serum-reduced culture conditions that were previously set up to sustain hMADS cell proliferation. As displayed in Fig. 2,A, activin A supplementation enhanced cell number in proliferating hMADS cell cultures. Enhancement of adipocyte progenitor proliferation by activin A did not impact on their subsequent capacity to undergo adipocyte differentiation; undifferentiated activin A– treated cells displayed a potential to undergo adipocyte differentiation similar to that of untreated cells (not shown). Cell number was decreased when cells were treated with an activin A–neutralizing antibody, showing the existence of a basal activin A autocrine/paracrine signaling pathway involved in undifferentiated hMADS cell number. This conclusion was supported by a gene-silencing approach in which hMADS cells were transfected with Inhba siRNA (supplementary Fig. S3). Smad2 became phosphorylated in hMADS cells upon activin A treatment (supplementary Fig. S4A), strongly suggesting that, as in other cell types, Smad2 mediates the biological effects of activin A (21). Therefore, undifferentiated hMADS cells were transfected with siRNA duplex directed against Smad2 to investigate the role of the activin A/Smad2 pathway in regulating number of adipose progenitors (see supplementary Fig. S4B for siRNA effect on Smad2 protein levels). Twenty-four hours after transfection, cells were or were not stimulated with activin A, and cell number was quantified 5 days later. Fig. 2 B shows that Smad2 silencing decreased cell number and inhibited the effects of activin A. Collectively, these data show that activin A regulates the number of undifferentiated hMADS cells through autocrine/paracrine pathways and that this effect employed Smad2-dependent mechanisms.

As shown in Fig. 1, activin A is downregulated upon induction of differentiation. To identify whether activin A downregulation is necessary for commitment toward the adipocyte lineage, hMADS cells were induced to undergo adipocyte differentiation in the absence or presence of exogenous activin A. Chronic exposure of cells to activin A inhibited formation of lipid-containing cells and lowered the triglyceride-synthesizing glycerol-3-phosphate dehydrogenase (GPDH) activity in a dose-dependent manner (Fig. 3,A and B). hMADS cells were then treated with 100 ng/ml activin A for defined time intervals, and adipogenesis was scored at day 12. As shown in Fig. 3,C, exposure of hMADS cells to activin A for the first 3 days of the differentiation process led to a dramatic inhibition of differentiation. The highest level of GPDH inhibition was obtained when cells were exposed to activin A for the first 6 days. Expression of adipogenic genes, such as fabp4, pparγ2, and leptin, was inhibited when cells were exposed to activin A for the first 6 days (Fig. 3 D). These data indicate that activin A played a crucial role during the initial stages of the adipocyte differentiation program.

Expression of most of the adipogenic genes studied was downregulated in activin A–treated cells compared with untreated cells (supplementary Fig. S5B). KLF4 protein was expressed at low levels and was not altered by activin A treatment (supplementary Fig. S5C), suggesting that activin A acts either downstream, or independently, of KLF4. In contrast, C/EBPβ transcript (kinetics of expression during hMADS cell differentiation described in Fig. S6) and C/EBPβ-LAP/LIP isoform levels were downregulated in response to activin A (Fig. 3,E and supplementary Fig. S6). Therefore, we hypothesized that antiadipogenic effects of activin A occur through inhibition of C/EBPβ-LAP. If this hypothesis is correct, maintenance of C/EBPβ-LAP should abolish the inhibitory effect of activin A on adipogenesis. To test this hypothesis, we generated an hMADS-EcoRec cell line permissive to infection with murine retrovirus, and cells were transduced with a retroviral vector expressing LAP or LIP isoforms or GFP as control. We ascertained that overexpression of LAP promoted adipogenesis of hMADS cells and that overexpression of LIP inhibited this process (supplementary Fig. S7). Then, we investigated the consequences of this forced expression on activin A effects. Whereas inhibition of adipocyte differentiation by activin A was not affected in GFP- or LIP-hMADS cells, as assessed by GPDH activities and FABP4 protein levels, activin A–mediated inhibition of adipogenesis was abolished in LAP-hMADS cells (Fig. 4). These data are consistent with the hypothesis that the inhibitory effect of activin A on adipocyte differentiation is mediated by regulating C/EBPβ-LAP levels.

We showed that activin A is secreted by hMADS cells, while these cells expressed receptors for activin A, ACVR1A, ACVR1B, and ACVR2A (data not shown). Thus, we postulated the existence of a basal autocrine/paracrine signaling pathway that is involved in hMADS cell differentiation. We used complementary approaches to investigate the role of endogenous activin A. First, we used activin A antibodies to neutralize secreted activin A and SB431542, a potent and selective inhibitor of the ALK5 receptor, which binds activin A. hMADS cells were maintained in adipogenic medium in the absence or presence of 1 μg/ml anti–activin A antibodies or 5 μmol/l SB431542. As shown in Fig. 5,A, GPDH activity levels were increased in the presence of anti–activin A antibodies or SB431542. Enhanced expression of PPARγ and of adiponectin mRNAs, two markers of adipocytes, confirmed at the molecular level that inhibition of the activin receptor-like kinase (ALK) 5 induced adipogenesis (not shown). Because ALK5 can also bind transforming growth factor-β, which was expressed by hMADS cells (not shown), we specifically blocked the activin A signaling pathway using siRNA-mediated gene-silencing approaches. GPDH activity was increased in early differentiated hMADS cells transfected with Inhba-siRNA or with Smad2-siRNA compared with GPDH activity in cells transfected with a scrambled siRNA (Fig. 5 B). Inhibition of Smad2 dramatically promoted formation of lipid-containing cells and levels of the adipogenesis-induced FABP4 protein (supplementary Fig. S8). Together, these data demonstrate the existence of an autocrine/paracrine loop comprising the activin A/Smad2 pathway that plays a critical negative role in adipocyte differentiation of hMADS cells.

We investigated factors regulating activin A expression in human adipose progenitors. First, we analyzed the specific component(s) of the adipogenic cocktail that might be involved in downregulating INHBA gene expression. To address this, hMADS cells maintained in serum-free medium supplemented with transferrin and insulin were exposed to the various components of the adipogenic induction cocktail either alone or in combination. As shown in Fig. 6,A, T3, rosiglitazone, and 3-isobutyl-1-methylxanthine (IBMX) did not alter INHBA gene expression. In contrast, its expression was selectively downregulated by dexamethasone added alone or combined with the other components. Levels of activin A secreted by dexamethasone-treated cells were in agreement with the inhibition observed at RNA levels (Fig. 6 B). We also wanted to evaluate whether the expression of the INHBA gene could be regulated by the in vivo microenvironment of adipose progenitors. In this context, we compared the INHBA gene expression levels in subcutaneous fat depots of lean and obese individuals. Clinical and biological parameters measured in lean and obese women are shown in Tables S3 and S4.

As shown in Fig. 7 A, RNA levels of INHBA were significantly increased in subcutaneous adipose tissue of obese subjects compared with lean subjects. Activin A secretion was also investigated in paired subcutaneous and visceral adipose tissue explants from obese subjects. Activin A was similarly secreted by subcutaneous and visceral obese adipose tissue explants (not shown). Expression of activin receptors type 1 receptors ACVR1A and ACVR1B and type 2 receptors ACVR2A and ACVR2B has been investigated by RT-PCR. We observed that gene expression of type 1 (ACVR1B) and type 2 (ACVR2A and ACVR2B) activin receptors was not statistically different in adipose tissue of lean and obese subjects (not shown).

To identify the microenvironment components that might stimulate activin A expression in obese subjects, we tested the potential interaction between macrophages and adipose progenitors. For that purpose, preadipocytes isolated from lean adipose tissue were stimulated with factors secreted by macrophages selected from the stromal fraction of obese subcutaneous adipose tissue (adipose tissue microphages [ATMs]). As shown in Fig. 7,B, RNA levels of INHBA were significantly increased in human preadipocytes that were maintained for 10 days in media conditioned by ATMs compared with the control condition. As we reported previously, we observed an increase in preadipocyte number compared with the control condition (11). RNA levels of INHBA were also increased when native CD34⁺/CD31⁻ human adipose progenitors were treated for 24 h with ATM-secreted factors (Fig. 7,C). Finally, we investigated levels of activin A secreted by undifferentiated hMADS cells after a 24-h stimulation with ATM-conditioned medium. As shown in Fig. 7 D, factors secreted by ATMs selected from three different biopsies dramatically increased levels of activin A secreted by hMADS cells. No activin A was detected in ATM-conditioned media, excluding macrophages as the source of activin A. Levels of activin A secreted by ATM-conditioned media–treated cells were in agreement with the increase observed at the INHBA RNA levels (supplementary Fig. S9). Altogether, these data show that human macrophage–secreted factors induce an upregulation of activin A in human adipose progenitors.

The identification of factors regulating self-renewal of adipose progenitor cells could provide a means for better understanding the mechanisms that lead to hyperplasia and excessive development of adipose tissue. To our knowledge, only FGF1 (22) and FGF2, as we have shown previously (23), were shown to stimulate proliferation of human adipose progenitors and subsequently to increase their capacity to undergo differentiation. However, the involvement of FGFs in human white adipose tissue growth remains to be investigated. It has been proposed recently that the bone morphogenetic protein pathway, which shares signaling components with the activin pathway, regulates both adipose cell fate determination and differentiation of committed preadipocytes as well as function of mature adipocytes in mouse models (24). The present study specifies the function of activin A in this context and proposes that macrophages are niche components that are, at least in part, responsible for the activation of the activin A pathway in human adipocyte progenitors. INHBB subunits are expressed in mouse and human adipose tissues, suggesting that activin B and/or activin AB could also play a role in adipogenesis (25,26). In contrast to what we show for INHBA, INHBB is preferentially expressed in mature adipocytes, strongly suggesting that activin B/AB are not involved in early steps of adipocyte differentiation. Indeed, the authors propose that INHBB may play an autocrine/paracrine role in energy balance or the insulin insensitivity associated with obesity (25,–27). Altogether, these data support the hypothesis that activins are regulators of the different steps of adipogenesis.

Homozygous inhba mutant mice display a neonatal lethality phenotype (28). This precludes the study of the role of activin A during the development of white adipose tissue.

The Smad2 pathway mediates proliferative and antiadipogenic effects of activin A on hMADS cells. Even if other pathways in activin A–regulated self-renewal might be involved, the Smad2 pathway is likely to play a major role. Indeed, we observed a weak activation of ERK1/2 upon addition of activin A and that inhibition of p38 MAPK had no effect on activin A–induced cell proliferation (not shown).

Obese adipose tissue is characterized by accumulation of macrophages (6,7), and we show that human macrophage–secreted factors strongly induce an upregulation of activin A in human adipose progenitors (demonstrated here and previously [12]). Given that we report here that activin A promotes proliferation and inhibits adipocyte differentiation of adipose progenitors, it is tempting to propose that the pool of adipose progenitors increases in adipose tissue of obese patients compared with that of lean subjects, where macrophages are barely present. In agreement with this, we recently reported an increase of proliferative markers in the CD34^+/CD31⁻ cell population of obese adipose tissue (8). Moreover, adipose tissue macrophages have been shown to inhibit adipogenesis of the CD34^+/CD31⁻ cells (29). Altogether, these data support a model of self-renewal of adipose progenitors regulated by the obese microenvironment, where activin A plays a critical role. ATM-secreted factors involved in stimulation of activin A expression remain to be identified. Interleukin-1β and/or tumor necrosis factor-α are potent candidates; previous studies have shown in another cell model that activin A secretion is increased upon treatment with these two cytokines (30). The disappearance of macrophages (31) and reducing activin A levels in adipose tissue, for instance as a consequence of dieting, might be favorable to the formation of additional adipocytes from adipose progenitors upon ending dietary restriction—a situation reminiscent of the “yoyo” phenomenon. Activin A is also a critical mediator of inflammation (rev. in [19]). It is interesting to note that dexamethasone, an anti-inflammatory molecule, regulates activin A levels (Fig. 6). In addition to its stimulating effect on expression of proinflammatory cytokines in macrophages, activin A could promote fibrosis in adipose tissue (12). Therefore, activin A signaling could be a target for therapeutic interventions. At the extracellular level, follistatin can bind to activin A with high affinity and antagonizes binding to the activin A receptor. Recently, follistatin has been shown to reduce inflammatory disease severity in mouse animal models (32). More recently, it has been shown that recombinant follistatin treatment promotes adipogenic differentiation of progenitor cells in vitro (33). Activin A and follistatin can be synthesized by several cell types in the body. Based on expression of INHBA, preliminary experiments suggest that activin A is more expressed in adipose tissue than in muscle and less expressed in adipose tissue than in liver. We observed that Inhba is highly expressed in the CD34⁻/CD14⁺ cell population isolated from adipose tissue. Activin A might also be expressed by macrophages resident in adipose tissue (not shown). It is interesting to note that activin A was not detected in adipose tissue macrophage–conditioned media excluding macrophages as an abundant source of secreted activin A, whereas adipocyte progenitors secrete high levels of activin A in the presence of macrophage factors (see Fig. 7). Expression of activin A and of follistatin in different cell types has to be taken into account when studying activin A as a potential therapeutic target. Further studies to determine activin A circulating levels in lean subjects and in obese patients are required to better define the clinical role of activin A and to validate activin A as a secreted biomarker. Blocking activin A in adipose tissue of obese patients could represent a new therapeutic avenue to modulate the pool of adipose progenitors.

This work was supported by the Centre National de la Recherche Scientifique (CNRS), Fondation pour la Recherche Médicale (Equipe FRM and grant DVO20081013470) and by the European project Adipokines as Drug Targets to Combat Adverse Effects of Excess Adipose Tissue (ADAPT) (to A.B. and K.C.'s team and A.B.'s team). J.L.K. and T.T. were supported by grant AG13935 from the National Institutes of Health. Z.L.E. was supported by a fellowship from Association pour la Recherche sur le Cancer. D.C. is an INSERM established investigator.

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

L.-E.Z. research data and contributed to writing the manuscript. B.W., P.V., M.K., M.M., T.T., V.B., T.M.-K., A.L., C.E., M.S., and Y.K. researchd data. E.-Z.A., D.L., C.S., K.C., and A.B. collected data and contributed to discussion. G.A. contributed to discussion. J.L.K. collected data and reviewed the manuscript. C.D. researched data, collected data, provided administrative and financial support, and wrote the manuscript.

The authors thank the members of the Stem Cells and Differentiation team (UMR6543 CNRS/UNSA, Nice, France) and Professor Austin Smith (Centre for Stem Cell Research, Cambridge, U.K.) for helpful discussions.