• Open Access

MicroRNA-26 Family Is Required for Human Adipogenesis and Drives Characteristics of Brown Adipocytes

Authors

  • Michael Karbiener,

    Corresponding author
    1. RNA Biology Group, Institute for Genomics and Bioinformatics, Graz University of Technology, Austria
    • Correspondence: Marcel Scheideler, Ph.D., Graz University of Technology, Petersgasse 14/V, 8010 Graz, Austria. Telephone: 43-316-873-5334; Fax: 43-316-873105334; e-mail: marcel.scheideler@tugraz.at; or Michael Karbiener, Ph.D., Graz University of Technology, Petersgasse 14/V, 8010 Graz, Austria. Telephone: 43-316-873-5346; Fax: 43–316-873105346; e-mail: michael.karbiener@tugraz.at

    Search for more papers by this author
  • Didier F. Pisani,

    1. Université Nice Sophia Antipolis, iBV, Nice, France
    2. CNRS, iBV, Nice, France
    3. Inserm, iBV, Nice, France
    Search for more papers by this author
  • Andrea Frontini,

    1. Department of Experimental and Clinical Medicine, Obesity Center, United Hospitals-University of Ancona (Politecnica delle Marche), Ancona, Italy
    Search for more papers by this author
  • Lisa M. Oberreiter,

    1. RNA Biology Group, Institute for Genomics and Bioinformatics, Graz University of Technology, Austria
    Search for more papers by this author
  • Eleonore Lang,

    1. RNA Biology Group, Institute for Genomics and Bioinformatics, Graz University of Technology, Austria
    Search for more papers by this author
  • Alexandros Vegiopoulos,

    1. Joint Division Molecular Metabolic Control, DKFZ-ZMBH Alliance and Network Aging Research, German Cancer Research Center (DKFZ) Heidelberg, Center for Molecular Biology (ZMBH) and University Hospital, Heidelberg University, Heidelberg, Germany
    Search for more papers by this author
  • Karin Mössenböck,

    1. Joint Division Molecular Metabolic Control, DKFZ-ZMBH Alliance and Network Aging Research, German Cancer Research Center (DKFZ) Heidelberg, Center for Molecular Biology (ZMBH) and University Hospital, Heidelberg University, Heidelberg, Germany
    Search for more papers by this author
  • Gerwin A. Bernhardt,

    1. Department of Orthopedic Surgery, Medical University Graz, Graz, Austria
    Search for more papers by this author
  • Torsten Mayr,

    1. Institute of Analytical Chemistry and Food Chemistry, Graz University of Technology, Graz, Austria
    Search for more papers by this author
  • Florian Hildner,

    1. Red Cross Blood Transfusion Service of Upper Austria, Austrian Cluster for Tissue Regeneration, Linz, Austria
    Search for more papers by this author
  • Johannes Grillari,

    1. Department of Biotechnology, BOKU-VIBT University of Natural Resources and Life Sciences Vienna, Vienna, Austria
    Search for more papers by this author
  • Gérard Ailhaud,

    1. Université Nice Sophia Antipolis, iBV, Nice, France
    2. CNRS, iBV, Nice, France
    3. Inserm, iBV, Nice, France
    Search for more papers by this author
  • Stephan Herzig,

    1. Joint Division Molecular Metabolic Control, DKFZ-ZMBH Alliance and Network Aging Research, German Cancer Research Center (DKFZ) Heidelberg, Center for Molecular Biology (ZMBH) and University Hospital, Heidelberg University, Heidelberg, Germany
    Search for more papers by this author
  • Saverio Cinti,

    1. Department of Experimental and Clinical Medicine, Obesity Center, United Hospitals-University of Ancona (Politecnica delle Marche), Ancona, Italy
    Search for more papers by this author
  • Ez-Zoubir Amri,

    1. Université Nice Sophia Antipolis, iBV, Nice, France
    2. CNRS, iBV, Nice, France
    3. Inserm, iBV, Nice, France
    Search for more papers by this author
  • Marcel Scheideler

    Corresponding author
    1. RNA Biology Group, Institute for Genomics and Bioinformatics, Graz University of Technology, Austria
    • Correspondence: Marcel Scheideler, Ph.D., Graz University of Technology, Petersgasse 14/V, 8010 Graz, Austria. Telephone: 43-316-873-5334; Fax: 43-316-873105334; e-mail: marcel.scheideler@tugraz.at; or Michael Karbiener, Ph.D., Graz University of Technology, Petersgasse 14/V, 8010 Graz, Austria. Telephone: 43-316-873-5346; Fax: 43–316-873105346; e-mail: michael.karbiener@tugraz.at

    Search for more papers by this author

Abstract

Adipose tissue contains thermogenic adipocytes (i.e., brown and brite/beige) that oxidize nutrients at exceptionally high rates via nonshivering thermogenesis. Its recent discovery in adult humans has opened up new avenues to fight obesity and related disorders such as diabetes. Here, we identified miR-26a and -26b as key regulators of human white and brite adipocyte differentiation. Both microRNAs are upregulated in early adipogenesis, and their inhibition prevented lipid accumulation while their overexpression accelerated it. Intriguingly, miR-26a significantly induced pathways related to energy dissipation, shifted mitochondrial morphology toward that seen in brown adipocytes, and promoted uncoupled respiration by markedly increasing the hallmark protein of brown fat, uncoupling protein 1. By combining in silico target prediction, transcriptomics, and an RNA interference screen, we identified the sheddase ADAM metallopeptidase domain 17 (ADAM17) as a direct target of miR-26 that mediated the observed effects on white and brite adipogenesis. These results point to a novel, critical role for the miR-26 family and its downstream effector ADAM17 in human adipocyte differentiation by promoting characteristics of energy-dissipating thermogenic adipocytes. Stem Cells 2014;32:1578–1590

Introduction

It is today established that precisely the relative amount of non-protein-coding DNA (ncDNA), as opposed to protein-coding DNA, correlates well with organismic complexity across phyla [1]. As for metazoans, the majority of ncDNA is also transcribed [2]. This suggests that the corresponding ncRNAs are particularly important for the development of cell types and tissues unique to higher animals like mammals. Brown adipose tissue (BAT) is one such tissue found specifically in mammals. Active BAT burns lipids and carbohydrates to generate heat (thermogenesis), which protects against cold stress and obesity [3], whereas white adipose tissue (WAT), the other principle type of adipose tissue, stores surplus energy as lipids to be released during fasting. Understanding the development of these different types of fat is therefore relevant for understanding the pathophysiology of obesity and related disorders.

The developmental trajectories of BAT and WAT are still being debated. Transcriptomics revealed a myogenic signature for brown, but not white adipocytes [4], and genetic lineage tracing demonstrated the existence of a common myoblast/brown adipocyte precursor [5]. However, physiological (cold) and pharmacological (e.g., sympathetic or peroxisome proliferator-activated receptor γ [PPARγ] agonists) treatments have revealed a remarkable plasticity of essentially all adipose depots, as these stimuli evoke an explicit increase in brown(-like) adipocytes in WAT [6, 7]. These brown(-like) adipocytes most probably arise via de novo differentiation of WAT-residing precursors [8, 9] or by direct conversion of preexisting white adipocytes [10, 11], and have recently been termed “brite” (brown-in-white [9]) or “beige” [12] adipocytes, as opposed to the “classical” brown adipocytes found within “genuine” BAT depots. The exceptionally high catabolic capacity of brown and brite (i.e., thermogenic) adipocytes is due to the brown fat-specific uncoupling protein 1 (UCP1), which acts as a physiological barrier against weight gain [13]. Correspondingly, pharmacological treatments or genetic manipulations that increase the number and/or activity of brown and/or brite adipocytes result in resistance to diet-induced obesity and ameliorate associated complications like insulin resistance [14]. This is of particular interest because obesity and type II diabetes have reached epidemic proportions, and there is still no effective pharmacological anti-obesity treatment.

Human BAT was long thought to be present only during early infancy, after which the brown depots were considered to turn white. Although exceptions to this model were described decades ago [15-17], it has only recently become broadly recognized that many healthy adult humans also possess active BAT [18-21]. Since then, several protein regulators of the brown/myoblast and brite/white balances have been identified [22]. However, most of these studies were performed in mice, without validation in human models, despite the fact that, at least in some aspects, WAT and BAT differ across species. Furthermore, although a detailed protein network regulating the development of white and brown adipocytes has been revealed during the last decades [23, 24], the role of ncRNAs is still mostly unknown.

MicroRNAs (miRNAs) are a subclass of ncRNAs that play a central role in RNA interference (RNAi), a posttranscriptional gene silencing mechanism existing in many eukaryotes [25-28]. As small RNAs of approximately 23 nucleotides, miRNAs interact with partially complementary sites in the 3′UTR of mRNAs to diminish protein output, both via mRNA destabilization and inhibition of translation [29]. Despite their recent discovery, it is already known that miRNAs play pivotal roles in various biological processes. Importantly, fat-selective inactivation of Dicer, a necessary factor for miRNA biogenesis, resulted in mice that were almost devoid of WAT [30], implicating miRNAs also in adipocyte development. Correspondingly, we [31, 32] and others have identified several miRNAs influencing white adipocyte differentiation [33], whereas knowledge about miRNA involvement in brown/brite adipogenesis is still sparse. In this study, we have used human adipose-derived stem cells to uncover miR-26a and -26b as the first miRNAs that regulate both adipocyte development as well as the acquisition of brown adipocyte characteristics in humans. Furthermore, we have identified the sheddase ADAM17 as a previously unknown direct miR-26a/b target that also mediates the miRNA effects on human adipogenesis.

Materials and Methods

Cell Culture

Human multipotent adipose-derived stem (hMADS) cells were originally established from surgical scraps of subcutaneous adipose tissue of infants or children [34-36]. For experiments of this study, hMADS-2 cells (source: pubic fat of 5-year old male donor) and hMADS-3 cells (source: prepubic fat of 4-month old male donor) were used between passages 15 and 30. hMADS cells were proliferated in Dulbecco's modified Eagle's medium (DMEM, 1 g/l glucose, Lonza, Basel, Switzerland, www.lonza.com), 10% fetal bovine serum (FBS, Pan-Biotech, Aidenbach, Germany, www.pan-biotech.de), 10 mM HEPES, 2 mM l-glutamine (Life Technologies, Darmstadt, Germany, www.lifetech.com), 100 µg/ml normocin (Invivogen, Toulouse, France, www.invivogen.fr), and 2.5 ng/ml human fibroblast growth factor 2 (hFGF-2, Sigma-Aldrich, St. Louis, MO, www.sigmaaldrich.com). For adipocyte differentiation, cells were grown to confluence, when the medium was changed and hFGF-2 was omitted. After 2 days (=day 0), differentiation was induced by changing the medium to DMEM/Ham's F12 (Lonza) (50:50), 5 mM HEPES, 2 mM l-glutamine, 100 µg/ml Normocin, 860 nM insulin (Sigma-Aldrich), 10 µg/ml apo-transferrin (Sigma-Aldrich), 0.2 nM triiodothyronin (Sigma-Aldrich), and 100 nM rosiglitazone (Cayman Chemical, Tallinn, Estonia, http://www.caymaneurope.com), and for the first 3 days, 100 µM 3-isobutyl-1-methylxanthine (Sigma-Aldrich) and 1 µM dexamethasone (Sigma-Aldrich). Medium was replaced every 2-3 days. At day 9 (early white stage), rosiglitazone was withdrawn to enable white adipocyte differentiation until days 14-16 (late white stage), whereas it was still added to promote brite differentiation until days 14-16 (late brite stages). Human embryonic kidney 293 (HEK293) cells were grown in DMEM (4.5 g/l glucose, Life Technologies), 10% FBS, 4 mM l-glutamine, and 100 µg/ml normocin. For isolation and differentiation of primary human adipose-derived stromal cell (hASC) populations, see Supporting Information Methods.

Transfection of miRNA Mimics and siRNAs

Cells were seeded into 12-well or 6-well plates and transfection was performed shortly before or at confluence. Oligonucleotides were premixed with HiPerFect (QIAGEN, Hilden, Germany, www.qiagen.com) in DMEM at ratios of 0.5—1 pmol oligonucleotide/µl HiPerFect, and incubated for 10 minutes before addition to cells. Oligonucleotide sequences, manufacturers, and catalog numbers are listed in Supporting Information Table S1. The final oligonucleotide concentrations were 5 nM for miRIDIAN miRNA mimics and 25 nM for miRCURY LNA miRNA Power Inhibitors. The small interfering RNA (siRNA) screen was performed at a final concentration of 25 nM siRNA using ON-TARGETplus SMARTpool siRNAs, consisting of sets of four distinct siRNAs targeting the respective mRNA. As the ADAM17-targeting SMARTpool contained one siRNA that is capable of targeting a transcript variant that lacks the miR-26 binding site (ENST00000497134), a pool of the remaining three siRNAs (products J-003453-06, J-003453-07, and J-003453-08) was used for subsequent siRNA experiments. This enabled the selective targeting of the ADAM17 transcript variant bearing miR-26 binding sites (ENST00000310823).

Oil Red O Staining

Cells were washed with phosphate buffered saline (PBS, Life Technologies), fixed in 3.7% formaldehyde (in PBS) for 20 minutes, washed twice with PBS, stained by incubation with oil red O (0.5 g oil red O in 100 ml isopropanol diluted with double distilled water [40:60] and filtrated) for 1 hour, washed thrice with PBS, and then photographed. Images representative of at least three biological replicates are shown in the respective figures.

Triglyceride Assay

Quantification of intracellular triglycerides relative to total protein was performed using the Infinity Triglycerides Reagent (Thermo Scientific, Waltham, MA, www.thermofisher.com) and BCA Protein Assay kit (PIERCE/Thermo Scientific) as described previously [32].

Isolation and Analysis of RNA

Total RNA was obtained using TRIzol reagent (Life Technologies) according to the manufacturer's protocol. For analysis of mRNAs, 0.5–1 µg of RNA were reverse transcribed using the QuantiTect Reverse Transcription Kit (QIAGEN). For analysis of miRNAs, 25 ng total RNA were reverse transcribed using the Universal cDNA synthesis kit (Exiqon, Vedbaek, Denmark, www.exiqon.com). Quantitative real-time reverse transcription polymerase chain reaction (qPCR) and data evaluation were performed as described previously [32]. For primer sequences, see Supporting Information Table S2.

Western Blot Analysis

Detailed experimental procedures are provided in Supporting Information Methods. Primary antibodies were anti-UCP1 (calbiochem/Merck Millipore, Darmstadt, Germany, www.merckmillipore.de), anti-β-tubulin (Sigma-Aldrich), anti-ADAM17 (abm Richmond, Canada, www.abmgood.com), and anti-insulin receptor β (IRβ) (Santa Cruz, Heidelberg, Germany, www.scbio.de). Secondary antibodies were horseradish peroxidase (HRP)-conjugated anti-rabbit (Dako/Szabo-Scandic, Vienna, Austria, www.szabo-scandic.com, dilution 1:5,000) and anti-mouse immunoglobulins (Dako/Szabo-Scandic, dilution 1:5,000).

mRNA Microarray and Data Analysis

See Supporting Information Methods.

Measurement of Oxygen Consumption

Oxygen consumption was recorded using an oxygen microsensor (NTH-Pst1, PreSens, Regensburg, Germany, www.presens.de) connected to a transmitter device (Microx TX3, PreSens). For detailed experimental procedures see Supporting Information Methods.

Immunofluorescence and TEM

See Supporting Information Methods.

Luciferase Reporter Assay

A 1,347 bp fragment harboring 80% of the 3′UTR of ADAM17 (ENST00000310823) was PCR-amplified from hMADS cells cDNA using HighFidelity PCR Enzyme Mix (Fermentas/Thermo Scientific, Schwerte, Germany, http://www.thermoscientificbio.com, primer sequences in Supporting Information Table S3). The amplicon was cloned into the psiCHECK-2 vector (Promega, Mannheim, Germany, www.promega.de) between XhoI and NotI restriction sites immediately downstream of the Renilla luciferase coding sequence (CDS). The wild-type reporter served as template to generate two reporters with mutated miR-26 seed matches using the QuikChange Lightning Site-directed Mutagenesis Kit (Stratagene/Agilent, Vienna, Austria, www.genomics.agilent.com, primer sequences for mutation in Supporting Information Table S3). For luciferase reporter assays, HEK293 cells were seeded into 96-well plates (20,000 cells per well). After 20 hours, 100 ng of the respective reporter vector were cotransfected with either 50 nM miR-26a, miR-C, pASO-26 or pASO-C using DharmaFECT Duo transfection reagent (Dharmacon/Thermo Scientific, 0.2 µl/well). Cells were harvested 48 hours after transfection and assayed for Renilla and firefly luciferase activity using the Dual Luciferase Reporter Assay System (Promega) and the Orion II luminometer (Berthold, Bad Wildbad, Germany, www.berthold.com).

Animals and Tissues

See Supporting Information Methods.

Statistical Analysis

Data are presented as means ± SEM. Differences between groups were analyzed by applying Student's two-tailed t-test for independent samples, unless indicated otherwise.

Results

MiR-26a and -26b Are Dynamically Expressed and Functional in Human Adipocyte Differentiation

To study the involvement of miRNAs in human adipocyte differentiation, we exposed hMADS cells to a chemically defined medium permitting the development of terminally differentiated white adipocytes within 14–16 days. Microarray analysis at various time points during differentiation revealed several dynamically regulated miRNAs (unpublished observation), among which were the members of the miR-26 family, miR-26a and -26b. qPCR confirmed a significant upregulation of both miRNAs, along with the adipocyte markers PPARγ2 and perilipin (PLIN) (Fig. 1A). Similarly, primary hASC populations exhibited a significant increase in miR-26a/b levels during white adipocyte differentiation (Fig. 1B). We thus hypothesized that induction of miR-26a/b is a critical event for adipocyte development. To test this, we transfected hMADS cells with antisense oligonucleotides against miR-26a (pASO-26a). Although miR-26a was robustly detectable in cells transfected with a control (Ct-values ∼29), it could not be detected in pASO-26a-transfected cells confirming effective inhibition (Fig. 2A). A similar inhibitory effect was seen for miR-26b, which was expected as the high degree of sequence similarity between both family members should permit effective binding of pASO-26a also to miR-26b. Inhibition of miR-26a/b had a pronounced negative effect on adipocyte differentiation, as revealed by diminished oil red O staining and a significant reduction in intracellular triglycerides along with adipocyte markers (Fig. 2B–2D). Thus, miR-26 family members are indeed required for human adipocyte differentiation. Furthermore, transfection of miRNA mimics to elevate miR-26a or miR-26b levels (Fig. 2E; Supporting Information Fig. S1A) resulted in accelerated adipocyte differentiation (Fig. 2F, 2G; Supporting Information Fig. S1B, S1C): Although intracellular triglycerides were significantly higher compared with control transfected cells at the early stage of white adipocyte differentiation (day 9), this difference was no longer evident at the late stages (day 14–16). Correspondingly, the mRNA levels of several adipocyte markers were increased only at the early, but not late white stage (Fig. 2H). In sum, miR-26a and -26b are induced early during human adipogenesis and are required for its progression.

Figure 1.

miR-26a and -26b are upregulated during human adipocyte differentiation. (A): Human multipotent adipose-derived stem-2 cells (n = 3) were stimulated to undergo white adipocyte differentiation. At indicated time points, RNA was prepared and subjected to quantitative real-time reverse transcription polymerase chain reaction (qPCR) for adipocyte markers PPARγ2 and PLIN (normalized to TBP) as well as for miR-26a and miR-26b (normalized to 5S rRNA, relative to day 16). (B): Human primary adipose-derived stromal cells (hASC) populations (n = 3) were stimulated to undergo white adipocyte differentiation for qPCR analysis similar to (A). Data are presented relative to undifferentiated cells. (A, B) *p < .05; **p < .01; ***p < .001 versus undifferentiated cells. Abbreviations: PLIN, perilipin; PPARG, peroxisome proliferator-activated receptor γ; TBP, TATA box binding protein.

Figure 2.

miR-26 is required for human adipocyte differentiation. (A–H): Human multipotent adipose-derived stem (hMADS)-2/-3 cells were transfected with pASO-26a to inhibit miR-26a and -b, with miR-26a mimics, or with respective control oligonucleotides (pASO-C and miR-C). Subsequently, adipocyte differentiation was induced. (A): Forty-eight hours after transfection of miRNA antisense oligonucleotides, RNA was prepared and subjected to quantitative real-time reverse transcription polymerase chain reaction (qPCR) for miR-26a and -b (normalized to 5S rRNA, n = 3). n.d., not detectable. (B): Oil red O staining of hMADS-2 cells (image magnification: ×40), (C) quantification of triglyceride accumulation (n = 4), (D) qPCR analysis of adipocyte marker gene expression (normalized to TBP, relative to pASO-C, n = 4) at day 14 of adipocyte differentiation. (E): Forty-eight hours after transfection with miRNA mimics, RNA was prepared and analyzed as in (A). (F–H): At early (ew) and late white (lw) stages of adipocyte differentiation, cells were analyzed by (F) oil red O staining (image magnification: ×100), (G) measurement of triglyceride accumulation (n = 3), and (H) qPCR analysis of adipocyte marker gene expression (normalized to TBP, relative to miR-C/ew, n = 3–6). *p < .05; **p < .01; ***p < .001 versus respective control. Abbreviations: FABP4, fatty acid binding protein 4; GLUT4, glucose transporter 4; GPD1, glycerol-3-phosphate dehydrogenase 1; LPL, lipoprotein lipase; This is such a general abbreviation used from the beginning of this publication that it is not necessary to provide this in the abbreviation sections. pASO, antisense oligonucleotides; PLIN, perilipin; PPARG, peroxisome proliferator-activated receptor γ; TBP, TATA box binding protein; TG, triglyceride.

Transcriptomic Analysis Uncovers Pathways Responsive to miR-26a

To determine how miR-26a/b affect white adipocyte differentiation, we performed transcriptomics to compare overall changes in mRNA levels between miR-26a and control transfected cells at the early white stage. As previously reported for other miRNAs [37, 38], a large number of putative direct miR-26 targets bearing one or more seed match sequences in their 3′UTR was repressed by miR-26a (Supporting Information Fig. S2A). In line with the positive effect of miR-26a/b on white adipocyte differentiation, Gene Set Enrichment Analysis (GSEA) revealed related Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (e.g., “Fatty acid metabolism”) and Gene Ontology (GO) biological processes (e.g., “Cellular lipid metabolic process”) to be significantly enriched among upregulated genes (Supporting Information Table S4). Surprisingly, also brown/brite adipogenesis-related gene sets emerged: (i) the PPAR signaling pathway, the activation of which is necessary for both white and brown adipocyte development [39], yet it has additional specific “browning” effects on WAT [7, 9, 40, 41]; (ii) several gene sets related to mitochondria and submitochondrial components; and (iii) the GO biological process “Cation transport”, comprising the archetypal marker of thermogenic adipocytes, UCP1. Indeed, among all 4,515 transcripts considered for analysis, UCP1 was the mRNA most strongly induced by miR-26a (Supporting Information Fig. S2B). Thus, we decided to investigate brown adipocyte characteristics in more detail.

MiR-26a and -26b Promote Brown Adipocyte Characteristics During White and Brite Adipogenesis

hMADS cells constitute a valuable model for human adipogenesis as they can be selectively directed toward the white or brite fate, depending on whether they are exposed to the PPARγ agonist rosiglitazone only until early white stage, or chronically [36, 42]. In line with results observed for the late white stage, miR-26a/b did not affect lipid accumulation and expression of adipocyte marker genes at the late brite stage (Fig. 3A–3C; Supporting Information Fig. S3A, S3B). In contrast, qPCR revealed substantial miR-26a effects on brown adipocyte marker gene expression for all three analyzed stages (early white, late white, and late brite; Fig. 3D). UCP1 was markedly upregulated by miR-26a at the early white stage, and also late white and brite stages showed an increase in UCP1 mRNA levels compared with control transfected cells. Similar results were obtained for PPARγ coactivator 1α (PPARGC1A, PGC-1α), a brown fat-enriched protein promoting mitochondriogenesis [43], fatty acid binding protein 3 (FABP3), a critical determinant for fatty acid oxidation in mouse BAT [44], and β1-adrenoceptor, which is important for lipolysis [45], and BAT activation [46] in humans. Interestingly, the PPARγ- and PGC-1α-interacting protein PR domain containing 16, which has recently been described to drive browning of white adipocytes in mouse [47], was only weakly induced by miR-26a at the late brite stage. The promoting effect of miR-26a on UCP1 was even observed for primary hASC populations (Fig. 3E), and was also reflected at protein level. Specifically, while UCP1 was undetectable in control cells at early and late white stages, it was present in miR-26a-transfected hMADS cells, at levels comparable with control cells at the late brite stage (Fig. 3F). Also in the late brite stage, miR-26a strongly upregulated UCP1 protein levels, in excess of the levels reached by chronic rosiglitazone exposure alone. Similar results were obtained for miR-26b (Supporting Information Fig. S3C).

Figure 3.

miR-26a induces characteristics of brown adipocytes during human adipocyte differentiation. Cells were transfected with miR-26a or miR-C and adipocyte differentiation was subsequently induced. (A–C): At the late brite (lb) stage, human multipotent adipose-derived stem (hMADS)−2/3 cells were analyzed by (A) oil red O staining (image magnification: ×100), (B) quantification of triglyceride accumulation (n = 3), and (C) quantitative real-time reverse transcription polymerase chain reaction (qPCR) for general adipocyte marker gene expression (normalized to TBP, relative to miR-C, n = 3). (D): qPCR analysis of brown/brite adipocyte marker gene expression in hMADS-2/3 adipocytes at early white (ew), late white (lw), and lb stages (normalized to TBP, relative to miR-C/ew, n = 3). (E): qPCR analysis for UCP1 expression in hASC population-derived adipocytes at days 9 and 16 of differentiation (normalized to TBP, relative to miR-C/d9, n = 4). (F): Analysis of UCP1 and β-tubulin (TUBβ) protein levels in hMADS-3 adipocytes at ew, lw, and lb stages of adipocyte differentiation by Western blot. (G): Analysis of basal (left) and uncoupled (right) respiration in hMADS-2 adipocytes (lb stage, n = 4–5). *p < .05; **p < .01; ***p < .001 versus respective control. Abbreviations: ADRB1, β1-adrenoceptor; FABP3, fatty acid binding protein 3; FABP4, fatty acid binding protein 4; GLUT4, glucose transporter 4; GPD, glycerol-3-phosphate dehydrogenase 1; LPL, lipoprotein lipase; PLIN, perilipin; PPARG, peroxisome proliferator-activated receptor γ; PPARGC1A, peroxisome proliferator-activated receptor gamma, coactivator 1 alpha; PRDM16, protein PR domain containing 16; TBP, TATA box binding protein; TUBβ, β-tubulin; UCP1, uncoupling protein 1.

As both GSEA of the transcriptomic data and the further characterization of thermogenic markers were indicative of miR-26-mediated increases in energy expenditure, we measured oxygen consumption of adipocytes at the late brite stage. Basal respiration was significantly enhanced by miR-26a, as was residual uncoupled respiration after inhibition of ATP synthase (Fig. 3G).

We further analyzed the effects of miR-26 on UCP1 at the single-cell level using immunofluorescence microscopy. Although at the early white stage, no UCP1+ adipocytes could be detected in the control condition, miR-26a transfection resulted in ∼5% of adipocytes staining positive for UCP1 (Fig. 4A). This number increased to ∼30% as differentiation proceeded until the late white stage. As expected, ∼30% UCP1+ adipocytes were detectable at the late brite stage for control transfected cells. miR-26a transfection evoked an increase in UCP1+ adipocytes up to ∼50%. Ultrastructural analysis by transmission electron microscopy (TEM) further revealed a shift of mitochondrial morphology toward brown adipocyte characteristics (Fig. 4B). For all three analyzed stages, the density of cristae was slightly increased by miR-26a compared with respective controls. Additionally, mitochondria of adipocytes from miR-26a-transfected cells were bigger in size and more roundish (i.e., brown-like) in the late brite stage. Altogether, miR-26a/b promote characteristics of energy-dissipating thermogenic adipocytes during human adipocyte differentiation, in a synergistic manner with PPARγ agonism induced by rosiglitazone.

Figure 4.

miR-26a increases the number of UCP1+ adipocytes and reshapes mitochondrial morphology. (A): Immunofluorescence staining for UCP1 in human multipotent adipose-derived stem-2 adipocytes transfected and induced to differentiate as described in Figure 2. Representative pictures are shown for early white (ew), late white (lw), and late brite (lb) stages. Left side of each condition shows UCP1+ adipocytes (green). Right side shows same field of view under transmission light to visualize the presence of UCP1 negative cells, which do not appear under fluorescent light. Scale bars = 100 µm. (B): Transmission electron microsocopy was performed at the same stages as in (A). For each condition, lower magnification images are depicted on the left (scale bar = 6 µm). For each of these images, an image section with representative mitochondria is shown in higher magnification on the right (scale bar = 0.4 µm). The low magnification is representative of the morphology of the examined cells and allows the evaluation of the degree of differentiation and lipid accumulation present at each time point. Higher magnification reveals the morphology of mitochondria. Abbreviation: UCP1, uncoupling protein 1.

Complementary Experimental and In Silico Approaches Reveal Potential Direct miR-26 Targets with Function in Adipogenesis

To identify direct miR-26 targets with a function in human adipogenesis, we performed a transcriptomic analysis with specific constraints. First, cells were harvested early (48 hours) after transfection with miR-26a or control to capture predominantly the immediate effects of the miRNA, which should involve a large portion of direct miR-26a-mRNA interactions. Indeed, the cumulative distribution of potential direct miR-26 targets was significantly shifted toward lower expression (Supporting Information Fig. S4A). Second, we focused on the 1.5% of detected transcripts that were most downregulated by miR-26a (i.e., 80 mRNAs), as we hypothesized that the main component(s) of the miRNA effect would likely be among these “strong responders”. Third, we subselected mRNAs that were either predicted by at least one of 10 bioinformatic miRNA-mRNA interaction algorithms, or contained at least one miR-26 seed match (Supporting Information Fig. S4B) in their 3′UTR. Based on literature search to identify interesting cellular functions, we chose a subset of the remaining genes (Supporting Information Fig. S4C) to determine their role in adipogenesis using an RNAi screen, in which a meaningful candidate to be further investigated should ideally fulfill two criteria upon knockdown: (i) increased adipocyte differentiation, as well as (ii) promotion of UCP1 expression. Although the knockdown of four mRNAs (SMURF2, ATPAF1, ADAM17, and PLOD2) positively affected lipid accumulation (Fig. 5A), induction of UCP1 was only observed upon knockdown of one candidate (Fig. 5B): ADAM metallopeptidase domain 17 (ADAM17), which was among the three transcripts most downregulated by miR-26a (Supporting Information Fig. S4C).

Figure 5.

Screen for putative direct miR-26 targets mediating the adipogenic effects. (A, B): Human multipotent adipose-derived stem-3 cells were transfected at confluence with siRNAs against potential direct miR-26 targets or a control (siC) and subsequently adipocyte differentiation was induced. (A): Bright-field microscopy images acquired at day 9 of adipocyte differentiation are shown to assess triglyceride accumulation (image magnification: ×40). (B): Analysis of silencing efficiency and UCP1 expression. Upper panel: quantitative real-time reverse transcription polymerase chain reaction (qPCR) analysis of siRNA knockdown efficiency. Cells were harvested 2 days posttransfection to measure mRNA levels of the respective transcripts. Lower panel: qPCR analysis for UCP1 at day 9 of adipocyte differentiation. Transcript abundance was normalized to TBP and is presented relative to siC-transfected cells. Abbreviations: ADAM17, ADAM metallopeptidase domain 17; ATPAF1, ATP synthase mitochondrial F1 complex assembly factor 1; HMGA1, high mobility group AT-hook 1; LOXL2, lysyl oxidase-like 2; MAPK6, mitogen-activated protein kinase 6; PLOD2, procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2; PRKAG1, Protein kinase, AMP-activated, gamma 1 non-catalytic subunit; siC, control siRNA; SMURF2, SMAD specific E3 ubiquitin protein ligase 2; TOP1, topoisomerase (DNA) I; TXNDC17, thioredoxin domain containing 17; UCP1, uncoupling protein 1.

ADAM17 Is a Direct miR-26 Target with Anti-Adipogenic and Anti-Browning Function

Thus, we hypothesized that the miR-26 family has pro-adipogenic and pro-browning effects due to blunting of ADAM17 activity, possibly via physical interaction with its 3′UTR. Indeed, although a direct miR-26-ADAM17 interaction was not predicted by any bioinformatic algorithm, we identified two miR-26 seed matches within the human ADAM17 3′UTR, one of them being evolutionarily conserved (Supporting Information Fig. S5). Consistent with ADAM17 being a miR-26 family target, qPCR confirmed the downregulation of ADAM17 mRNA by miR-26a and -26b while inhibition of miR-26a/b caused a slight but significant upregulation (Fig. 6A). Similar effects were observed for the ADAM17 protein (Fig. 6B). Finally, luciferase reporter assays confirmed a direct interaction of miR-26a with the ADAM17 3′UTR, and site-directed mutagenesis revealed that this interaction is predominantly mediated via the second, evolutionarily conserved miR-26 binding site (Fig. 6C). Analyzing the ADAM17 knockdown effects throughout the entire course of adipocyte differentiation revealed a striking similarity to miR-26a/b overexpression: whereas lipid accumulation was significantly increased only at the early white stage (Fig. 6D, 6E), UCP1 expression was promoted also at late stages (Fig. 6F), resulting in the presence of UCP1 in cells differentiated in a medium that normally leads to the development of bona fide white adipocytes (Fig. 6G). Importantly, ADAM17 knockdown was also able to elevate UCP1 in brite adipocytes. Altogether, these results strongly suggest that the miR-26a/b effects in human adipogenesis are mediated at least partly by ADAM17.

Figure 6.

ADAM17 is a direct downstream mediator of the miR-26 effects in human adipocyte differentiation. (A): RNA from human multipotent adipose-derived stem (hMADS)−2/-3 cells (n = 3–4) was isolated 2 days posttransfection with miR-26a, miR-26b, pASO-26a, or respective controls (pASO-C and miR-C) and analyzed by quantitative real-time reverse transcription polymerase chain reaction (qPCR) for ADAM17. (B): hMADS-3 cells were transfected with siRNAs targeting ADAM17 (siADAM17), miR-26a, pASO-26a, or respective controls (miR-C, pASO-C) and analyzed by Western blot for ADAM17 and IRβ protein levels. (C): Luciferase reporter vectors containing either the wild-type 3′UTR of ADAM17 (wt) or 3′UTRs with mutated miR-26 seed matches (mut-1 and mut-2), were cotransfected with miR-26a, pASO-26a, or respective control oligonucleotides into HEK293 cells. After 48 hours, cells were harvested and Luciferase assays were performed. Data are presented as mean Renilla normalized to firefly luciferase (n = 3–4) (D–G) hMADS-3 cells were transfected at confluence with siADAM17 or siC and adipocyte differentiation was induced 2 days later. At early white (ew), late white (lw), and late brite (lb) stages of adipocyte differentiation, cells were analyzed by (D) oil red O staining (image magnification: ×100), (E) quantification of triglyceride accumulation (n = 3), and (F) qPCR for adipocyte marker gene expression (normalized to TBP, relative to siC/ew). (G): Western blot analysis of UCP1 and TUBβ protein levels at lw and lb stages of adipocyte differentiation. *p < .05; **p < .01; ***p < .001 versus respective control. Abbreviations: ADAM17, ADAM metallopeptidase domain 17; FABP4, fatty acid binding protein 4; IRβ, insulin receptor β; LPL, lipoprotein lipase; pASO, antisense oligonucleotides; siC, control siRNA; TUBβ, β-tubulin; UCP1, uncoupling protein 1.

MiR-26a Is Enriched in BAT and Induced in WAT upon Cold Exposure

As several genes governing the fate of white/brite/brown adipocytes have been described to be differentially expressed between the respective adipose tissues, we analyzed the levels of miR-26a in the adipocyte fraction (AF) and the stromal vascular fraction (SVF) of murine BAT and WAT. Interestingly, miR-26a abundance was two to fourfold higher in the SVF of BAT compared with WAT (Fig. 7A). Correspondingly, also the AF of BAT exhibited 1.5- to twofold higher levels of miR-26a than the WAT AF. We furthermore exposed mice to a cold environment, which is the classical stimulus for BAT expansion and browning of WAT. Strikingly, in addition to the well-described induction of Ucp1, we also found miR-26a to significantly increase upon cold stress, whereas its direct target Adam17 was downregulated (Fig. 7B). In sum, these results point to the involvement of the miR-26 family in the promotion of energy dissipation, that is, brown adipocyte characteristics.

Figure 7.

miR-26a is more abundant in BAT than WAT and is responsive to cold in WAT. (A): Stromal vascular fraction (SVF) and adipocyte fraction (AF) were prepared from epididymal WAT and interscapular BAT of wild-type C57BL/6 mice, and total RNA was isolated. Two biological replicates (BR1 and BR2), each a pool of 15 mice, were subsequently analyzed by quantitative real-time reverse transcription polymerase chain reaction (qPCR) for expression of miR-26a (normalized to RNU5G), Leptin (Lep), and Ucp1 (both normalized to Uxt). Data are presented relative to WAT-SVF of BR2. (B): Wild-type NMRI mice (seven per group) were housed at 23°C or 5°C for 10 days before total RNA was isolated from perigonadal WAT. miR-26a abundance (normalized to 5S rRNA) and Ucp1 and Adam17 mRNA levels (normalized to Uxt) were analyzed by qPCR. **p < .01, ***p < .001 versus 23°C. Abbreviations: Adam17, ADAM metallopeptidase domain 17; AF, adipocyte fraction; BAT, brown adipose tissue; BR, biological replicate; Lep, leptin; SVF, stromal vascular fraction; Ucp1, uncoupling protein 1; Uxt, ubiquitously-expressed, prefoldin-like chaperone; WAT, white adipose tissue.

Discussion

During the past decades, a wealth of studies performed in rodents, or in vitro models derived thereof, has elucidated key molecular events regulating the development of white and brown fat cells. However, there are several crucial differences between mouse and human, which limit their extrapolation to humans. For example, the frequently investigated mouse epididymal WAT has no clear counterpart in humans. Furthermore, comparative transcriptomics of BAT and WAT in both species have revealed a surprisingly small overlap of coordinately regulated genes [48]. Similarly, the impact of single genes (e.g., LMO3 [49]), pathways (e.g., Hedgehog [50, 51]), and paracrine factors (e.g., BMP7 [42, 52]) on adipogenesis per se and also on browning has been shown to diverge. Finally, BAT is obligatory in mouse, but facultative in humans (not all adults are BAT-positive in PET/CT examinations [20]), which also implies species-specific molecular routes leading to the development of brite/brown adipocytes. Thus, we suggest that the use of human in vitro models is a crucial component of research on adipocyte physiology and correspondingly have used hMADS cells and non-immortalized hASC populations as models. hMADS cells have a normal karyotype, a high self-renewal ability, and the plasticity to differentiate into osteoblasts, chondrocytes, and adipocytes at the clonal level for more than 160 population doublings [34, 35, 53, 54]. In contrast, non-immortalized hASC populations isolated from adults usually exhibit multipotency at the population level and lose this multipotency after a few population doublings. We have used hMADS cells from two very young male donors at passage 15–30 and hASC populations from three adult female donors at passage 1–2. Thus, our study design does not allow dissecting specific effects of donor age, donor gender, and in vitro population doublings on the results obtained on adipogenesis.

In this study, we have focused on ncRNAs, which almost entirely await their functional characterization in various biological contexts. Our work reveals, to the best of our knowledge, the first miRNAs that simultaneously regulate adipocyte development and brown adipocyte characteristics in humans. Reemphasizing the strong involvement of miRNAs in cell type specification, we demonstrate that the miR-26 family is a key player in adipocyte development. In view of reports describing that miR-26a is required for myogenesis [55, 56], and that miR-26a/b influence osteogenesis, albeit with conflicting results [57-59], our findings suggest the miR-26 family to be of general importance for mesodermal cell fate.

Several miRNAs are already known to govern adipogenesis, but the global effects of individual miRNAs on the transcriptome of developing adipocytes have rarely been investigated. These approaches are a valuable starting point for subsequent, more detailed analyses of the phenotype caused by a gene of interest. Using GSEA, we indeed identified several gene sets that were significantly turned on or off by miR-26a. Among those, the upregulated pathways “pyruvate metabolism,” “tricarboxylic acid (TCA) cycle,” and “fatty acid metabolism” suggested increased de novo synthesis of lipids, in line with the increased triglyceride accumulation due to miR-26 at the early white stage. A further finding that supported the validity of our transcriptomic analysis was the significant enrichment of transforming growth factor (TGF)-β and Notch signaling among miR-26a-suppressed mRNAs (Supporting Information Table S5). Both pathways have been described as anti-adipogenic [60-62]; their downregulation can thus explain the pro-adipogenic miRNA effect. Finally, our transcriptomic screen also revealed pathways related to energy dissipation, and subsequent detailed analyses indeed confirmed the miR-26 family to promote brown adipocyte characteristics during adipocyte differentiation. It should be noted that, in humans, there is a constant and substantial turnover of adipocytes throughout adult life which is similar for normal weight and obese individuals [63]. Hence, the ability of the miR-26 family to drive brite adipogenesis might be of clinical relevance with respect to weight loss. Apart from the effect on differentiation, a similar involvement of miR-26a/b in the direct conversion of white to brite adipocytes is also conceivable. Thus, studies investigating the effects of the miR-26 family on mature white adipocytes—especially those from obese individuals—appear as highly interesting future prospects.

Interestingly, hMADS-derived adipocytes appeared as heterogeneous population with respect to UCP1 abundance and thus probably the browning process: as expected, only a negligible number of white adipocytes stained weakly positive for the brown marker. In contrast, miR-26a evoked the appearance of some strongly UCP1+ adipocytes and also elevated the number of weakly UCP1+ adipocytes, yet did not promote UCP1 accumulation in all cells. Thus, subsets of hMADS cells might be more susceptible to brite differentiation than others. In this respect, it will be interesting to generate subpopulations of cells, for example, based on cell surface marker sorting, as recently shown for mouse preadipocytes [64]. It is notable that also the browning due to chronic rosiglitazone treatment (late brite stage) was substantially elevated by miR-26a, suggesting that the molecular actions downstream of miR-26a involve additional mechanisms besides PPARγ activation that finally increase UCP1 abundance, the number of UCP1+ adipocytes, as well as basal and uncoupled respiration. Interestingly, the size of lipid droplets in brite adipocytes tended to be inversely correlated with intensity of UCP1 staining, which further supports the notion that the miR-26a-induced elevation of UCP1 promotes energy expenditure. Although in vivo, UCP1 is inactive as a proton carrier unless free fatty acids are provided [3] (e.g., by catecholamine-induced lipolysis), it should be noted that the serum-free culture conditions used for hMADS adipocytes go along with a certain level of constitutive, basal lipolysis (unpublished observation), which could thereby explain the observed increase in basal respiration.

TEM analysis further corroborated the pro-browning effect of miR-26a in terms of mitochondrial size and submitochondrial morphology. However, these effects were only modest compared with the profound induction of UCP1. This might be due to a suboptimal experimental setting (in vitro cultivation of adipocytes without serum). Thus, an appropriate, three-dimensional environment (as existing in the in vivo situation) and/or extracellular factors (cytokines and extracellular matrix signals) could lead to a more pronounced impact of miR-26a/b on mitochondriogenesis. It should be noted that, although some miRNAs have recently been demonstrated to impact on brown/brite adipocyte differentiation and thermogenesis [65-68], effects of miRNAs on mitochondrial morphology/density of adipocytes have not been directly investigated until now.

We further identified the sheddase ADAM17 as a previously unknown direct target of the miR-26 family. Most importantly, silencing of ADAM17 phenocopied the effects of miR-26a/b. Remarkably, the miR-26—ADAM17 interaction was not predicted by any of 10 distinct algorithms and was only revealed by screening of available 3′UTR sequences for the presence of miR-26 seed matches. Furthermore, it was interesting that many of the >300 validated direct miR-26a targets (available from miRWalk [69]) that were detectable in our transcriptomic data were not downregulated by miR-26a in hMADS cells (unpublished results). This suggests that miRNA–mRNA interactions vary considerably across different cell types. Altogether, our findings underscore the power of global gene expression studies to identify biologically meaningful targets of miRNAs, but simultaneously highlight the necessity for comprehensive and complementary in silico analyses.

ADAM17 was originally identified as tumor necrosis factor α converting enzyme (TACE) [70, 71], but since then has been shown to promote ectodomain shedding of a large number of other proteins [72]. Importantly, both Tace+/− and Tace−/− mice were shown to be remarkably resistant to diet-induced weight gain, presumably due to the promotion of energy expenditure in both WAT and BAT [73, 74]. Although the positive effect of ADAM17 deficiency on brown marker gene expression has been proposed to be due to increased sympathetic outflow [74], we demonstrate for the first time that also cell-autonomous suppression of ADAM17 promotes UCP1 expression. Correspondingly, future efforts to identify ADAM17 substrates that are shedded from the surface of adipocyte precursors and subsequently promote/impair the development of brite adipocytes will be of high interest.

In mouse, the miR-193b—365 cluster has recently been identified as necessary for brown adipogenesis [65], while miR-196a was shown to promote brite adipocyte differentiation [66]. In line with these physiological functions, the miRNAs were enriched in BAT versus WAT, or induced in WAT by cold, respectively. Similarly, we found that miR-26a is more abundant in both AF and SVF of BAT compared with WAT and that prolonged cold exposure upregulates miR-26a in WAT. Therefore, it is conceivable that high levels of miR-26a promote thermogenic adipocyte characteristics also in vivo and that the miRNA could thereby act as physiological mediator of cold acclimation. Further studies characterizing the effects of the miR-26 family in vivo, as well as identifying upstream factors regulating miRNA expression, are thus in demand.

Conclusion

In conclusion, we have identified the miR-26 family and its novel direct target ADAM17 as key players in human adipogenesis and the development of energy-dissipating, that is, brown adipocyte characteristics. Notably, antagonizing or restoring miRNA function has already been identified as attractive intervention in other diseases: Although the first miRNA inhibitor against hepatitis C virus is currently in clinical trial phase II [75], miRNA mimics for cancer treatment have just entered the clinic [76, 77]. In light of these studies and the results presented herein, miR-26 might be able to open up new avenues for miRNA-based drugs that may pave the way toward novel treatments for obesity via promotion of energy expenditure in brite or brown adipocytes.

Acknowledgments

We thank Hubert Hackl and Christoph Fischer for input on bioinformatic analyses, Peter M. Krempl for help with ArrayExpress, and Claudia Gaug, Florian Stöger, Cristina Zingaretti, and Mansour Djedaini for excellent technical assistance. Human multipotent adipose-derived stem cells were provided by Christian Dani and Ez-Zoubir Amri (University of Nice Sophia Antipolis) under the conditions of a material transfer agreement (Centre National de la Recherche Scientifique [CNRS]). This work was supported by the GEN-AU project “non-coding RNAs” (no. 820982), the Austrian Science Fund (FWF, P25729-B19), the French Agence Nationale de la Recherche (ANR-10-BLAN-1105 miRBAT), the Deutsche Forschungsgemeinschaft (He3260/8-1), and by the EU FP7 project DIABAT (HEALTH-F2–2011-278373).

Author Contributions

M.K.: conception and design, collection of data, data analysis and interpretation, manuscript writing; D.F.P., T.M., G.A., and E.A.: conception and design, data analysis and interpretation; A.F., L.M.O., E.L., K.M, and S.C.: collection of data, data analysis and interpretation; A.V., G.A.B., F.H., J.G., and S.H.: provision of study material; M.S.: conception and design, data analysis and interpretation, financial support, final approval of manuscript.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.

Ancillary