SREBP-1c/MicroRNA 33b Genomic Loci Control Adipocyte Differentiation

Abstract

White adipose tissue (WAT) is essential for maintaining metabolic function, especially during obesity. The intronic microRNAs miR-33a and miR-33b, located within the genes encoding sterol regulatory element-binding protein 2 (SREBP-2) and SREBP-1, respectively, are transcribed in concert with their host genes and function alongside them to regulate cholesterol, fatty acid, and glucose metabolism. SREBP-1 is highly expressed in mature WAT and plays a critical role in promoting in vitro adipocyte differentiation. It is unknown whether miR-33b is induced during or involved in adipogenesis. This is in part due to loss of miR-33b in rodents, precluding in vivo assessment of the impact of miR-33b using standard mouse models. This work demonstrates that miR-33b is highly induced upon differentiation of human preadipocytes, along with SREBP-1. We further report that miR-33b is an important regulator of adipogenesis, as inhibition of miR-33b enhanced lipid droplet accumulation. Conversely, overexpression of miR-33b impaired preadipocyte proliferation and reduced lipid droplet formation and the induction of peroxisome proliferator-activated receptor γ (PPARγ) target genes during differentiation. These effects may be mediated by targeting of HMGA2, cyclin-dependent kinase 6 (CDK6), and other predicted miR-33b targets. Together, these findings demonstrate a novel role of miR-33b in the regulation of adipocyte differentiation, with important implications for the development of obesity and metabolic disease.

INTRODUCTION

Obesity is one of the largest burdens on the health care systems of developed countries and is a major risk factor in the development of insulin resistance, dyslipidemia, and hypertension, leading to a condition known as metabolic syndrome (1). However, obesity is not always associated with metabolic syndrome, as in the case of mice overexpressing adiponectin (2). On an ob/ob background, these mice have dramatically increased adipose tissue mass, but this does not promote insulin resistance, as the mice actually have improved insulin sensitivity. Alternatively, lipodystrophy, a condition caused by mutations that impair the expansion or differentiation of white adipose tissue (WAT), leads to severe forms of metabolic syndrome (3). Overall, the association between both impaired and excessive WAT accumulation and the development of metabolic syndrome emphasizes the critical role of WAT in maintaining metabolic homeostasis.

Recent work has demonstrated an important role for WAT in regulating whole-body metabolism through the release of signaling molecules, such as leptin and adiponectin, which can regulate insulin sensitivity and appetite regulation in other tissues. Moreover, it has been known for some time that an inability of WAT to properly remove and store circulating lipids results in accumulation of lipids in nonadipose tissues, promoting diseases such as type II diabetes and atherosclerosis (4, 5). De novo lipid biosynthesis is controlled by sterol regulatory element-binding proteins (SREBPs), which are activated in response to changes in intracellular and membrane levels of fatty acids and cholesterol (6–8). The SREBP family of transcription factors consists of the SREBP-1a, SREBP-1c, and SREBP-2 proteins, which are encoded by the genes SREBP-1 and SREBP-2. SREBP isoforms differ in their expression patterns and target gene selectivity (9). SREBP-2 functions primarily in the regulation of cholesterol-related genes, while SREBP-1c primarily regulates genes involved in fatty acid metabolism, and SREBP-1a can promote the expression of both sets of genes (10). In the liver, the ability of SREBPs to regulate gene expression in response to changes in lipid profiles is critical for maintenance of lipid homeostasis. SREBP-1 has also been shown to play an important regulatory role in promoting the expression of important adipogenic genes and is highly induced during adipocyte differentiation (11).

In addition to classical transcription regulators, a class of noncoding RNAs, termed microRNAs (miRNAs), have emerged as critical mediators of a variety of cellular functions, including lipid metabolism and adipogenesis (12). After being processed from primary transcripts by the sequential actions of Drosha and Dicer enzymes, these short, double-stranded regulatory noncoding RNAs are incorporated into the cytoplasmic RNA-induced silencing complex (RISC). This allows miRNAs to bind partially complementary target sites in the 3′ untranslated regions (3′ UTRs) of mRNA, resulting in translational repression and/or mRNA destabilization (13–15). A number of miRNAs have been shown to be differentially regulated upon differentiation of WAT. Among them, the miRNA cluster miR-17-92, miR-21, miR-30, miR-103, miR-143, miR-204/211, miR-210, miR-338/338*, miR-375, and miR-637 promote adipogenesis (16–25), while Let-7, miR-15a, miR-22, miR-27a/b, miR-130, miR-138, miR-155, miR-221/222, miR-205, miR-224, miR-369-5p, and miR-448 impair adipocyte differentiation (26–36). However, the mechanism by which many of these miRNAs regulate adipogenesis is unknown, and while some of the miRNAs are differentially expressed in WAT of patients with obesity and metabolic syndrome (37–39), their roles in the development of these conditions have not been established.

Work done by our laboratory and others has established miR-33a and miR-33b as important regulators of cholesterol, fatty acid, and glucose metabolism (40–45). In humans, two isoforms of miR-33 exist, miR-33b, which is within the SREBP-1 gene, and miR-33a, which is located in the SREBP-2 gene. In the liver, miR-33a and miR-33b are coexpressed with their host genes, working synergistically with SREBPs in their regulation of intracellular lipids. miR-33 regulates cholesterol trafficking and high-density lipoprotein (HDL) biogenesis by targeting the ATP-binding cassette (ABC) transporters ABCA1 and ABCG1 and the endolysosomal transport protein Niemann-pick protein C1 (NPC1), regulates fatty acid β-oxidation through targeting of carnitine O-octanyltransferase (CROT), carnitine palmitoyltransferase 1a (CPT1a), hydroxyacyl coenzyme A (hydroxyacyl-CoA) dehydrogenase/3-ketoacyl-CoA thiolase/enoyl-CoA hydratase (HADHβ), PGC1α, and AMPK, and regulates insulin signaling through targeting of insulin receptor substrate 2 (IRS2) (40–45).

SREBP-1 is an important regulator of the peroxisome proliferator-activated receptor γ (PPARγ) gene and other critical adipogenic genes, and loss of SREBP-1 results in impaired adipogenesis in vitro. While the roles of miR-33a and miR-33b in the liver have been well established (44, 45), it is not known whether miR-33b is induced along with SREBP-1 during adipogenesis or what role it may play in the regulation of adipocyte differentiation and function. Additionally, while the sequences of both miR-33 isoforms are fairly well conserved across species, mice and other small mammals express only miR-33a, due to variation in the region of the SREBP-1 gene harboring miR-33b. Therefore, investigation into the role of miR-33b in adipocyte differentiation and function may improve our understanding of how obesity and metabolic syndrome develop in humans and could help elucidate some of the differences in fat storage between humans and commonly utilized animal models.

In this report, we demonstrate that miR-33b is induced, along with SREBP-1, during differentiation of human preadipocytes. We further demonstrate that overexpression of miR-33b causes reduced preadipocyte proliferation and impaired differentiation, while inhibition of miR-33b enhanced lipid accumulation in differentiating adipocytes. The effects of miR-33 on proliferation are likely mediated by the regulation of the known miR-33 targets cyclin-dependent kinase 6 (CDK6) (46) and HMGA2 (47), a chromatin-remodeling factor critical for the clonal-expansion phase of adipogenesis (27). Consistent with prior work, we find that HMGA2 is overexpressed in tissue from patients with lipoma and, further, demonstrate that miR-33b expression is reduced in the same samples. Gene expression analysis of genes involved in adipogenesis has allowed us to identify many genes whose expression is altered in cells overexpressing miR-33b, including an overall impairment in the induction of PPARγ target genes. Knockdown of HMGA2 also impaired adipocyte differentiation and the induction of PPARγ target genes in a manner similar to that observed with miR-33b overexpression, although differences in the regulation of many adipogenic genes suggest that other targets of miR-33b are also likely to play an important role. Together, these findings implicate miR-33b in the regulation of adipocyte differentiation, which may have important effects on obesity and the development of lipid-derived tumors in humans.

MATERIALS AND METHODS

Cell culture.

The Simpson-Golabi-Behmel syndrome (SGBS) cell strain was kindly provided by Martin Wabitsch (University of Ulm, Ulm, Germany) and cultured under conditions developed in his laboratory (48). Briefly, cells were maintained in Dulbecco's modified Eagle's medium (DMEM)–F-12 supplemented with biotin (8 μg/ml), pantothenate (4 μg/ml), penicillin (50 U/ml), streptomycin (50 U/ml), and non-heat-inactivated fetal bovine serum (FBS) (10%). To induce differentiation into mature adipocytes, nearly (∼90%) confluent SGBS cells were washed thoroughly with phosphate-buffered saline (PBS), followed by 4 days of culture in DMEM–F-12 supplemented with biotin (8 μg/ml), pantothenate (4 μg/ml), transferrin (10 μg/ml), insulin (20 nM), cortisol (100 nM), triiodothyronine (0.2 nM), dexamethasone (25 nM), 1-methyl-3-isobutylxanthine (250 μM), rosiglitazone (2 μM), penicillin (50 U/ml), and streptomycin (50 U/ml). After 4 days, the cells were cultured in DMEM–F-12 medium containing biotin (8 μg/ml), pantothenate (4 μg/ml), transferrin (10 μg/ml), insulin (20 nM), cortisol (100 nM), triiodothyronine (0.2 nM), penicillin (50 U/ml), and streptomycin (50 U/ml), which was replaced every 4 days. Primary human preadipocytes and media were purchased from ZenBio Inc. (catalog number SP-F) and cultured/differentiated according to the provided protocols. Briefly, cryopreserved subcutaneous preadipocytes were thawed and cultured in preadipocyte medium (catalog number PM-1) until confluent. Differentiation was initiated by culturing for 7 days with adipocyte differentiation medium (catalog number DM-2), followed by partial replacement with adipocyte medium (catalog number AM-1). Cos7 and 293T cells were maintained in DMEM supplemented with 10% FBS, penicillin (50 U/ml), and streptomycin (50 U/ml). All the cells were cultured at 37°C and 5% CO₂.

RNA isolation and quantitative real-time PCR.

Total RNA form cells and tissue was isolated using TRIzol reagent (Invitrogen) according to the manufacturer's protocol. Lipoma samples were kindly provided by Laurence Bianchini and have been described previously (49). For mRNA expression analysis, cDNA was synthesized using iScript RT Supermix (Bio-Rad), following the manufacturer's protocol. Quantitative real-time (qRT)-PCR analysis was performed in duplicate using SsoFast EvaGreen Supermix (Bio-Rad) on an iCycler real-time detection system (Eppendorf). The mRNA levels were normalized to 18S. For miRNA quantification, total RNA was reverse transcribed using the miScript II RT kit (Qiagen). Primers specific for human miR-33a and miR-33b (Qiagen) were used, and the values were normalized to SNORD68 (Qiagen) as a housekeeping gene. miR-33 quantification and quantitative real-time PCR were performed in triplicate using SYBR green master mix (SA Biosciences) on an iCycler real-time detection system (Eppendorf). Adipogenesis PCR arrays were performed according to the manufacturer's instructions (Qiagen). Briefly, cDNA was synthesized using an RT² HT first strand kit, and quantitative PCR (qPCR) analysis was performed using SYBR green master mix (SA Biosciences) on an iCycler real-time detection system (Eppendorf).

Collection and separation of human stromal vascular and mature adipocyte fractions.

Subcutaneous adipose samples from subjects undergoing elective abdominoplasty were waste materials generated as a by-product of the surgery. The tissue samples were placed in sterile saline solution, and tissue explants were washed with a modified Krebs-Ringer phosphate (KRP) solution (NaCl, 0.127 M; KCl, 5 mM; sodium phosphate dibasic, 2 mM; sodium phosphate monobasic, 8 mM). The tissue was then minced, washed in modified KRP solution, and centrifuged at 300 × g for 3 min to separate out red blood cells. The minced adipose tissue was digested in modified KRP (supplemented with 0.8 mM ZnCl₂, 1 mM MgCl₂, and 1.2 mM CaCl₂) with 3% fetal bovine serum (Gibco, Life Technologies) with collagenase type 2 at 1 mg/ml (Worthington) for 75 min at 37°C with constant shaking (120 rpm). The digested sample was then filtered through a 250-μm nylon filter and rinsed with the modified KRP with 3% Chelex-fetal calf serum (FCS). Samples were centrifuged at 300 × g for 3 min, and the floating (mature adipocyte) fraction was removed. The stromal vascular fraction (SVF) was washed with KRP, filtered through a 70-μm filter, and spun at 300 × g for 3 min. Excess buffer was removed, and the SVF was washed with KRP, filtered through a 40-μm filter, and spun at 300 × g for 3 min.

Lentivirus production and infection.

Lentiviral constructs for overexpression and inhibition of miR-33b and control constructs were purchased from SBI System Biosciences. Production of viral medium was performed by transfecting a lentiviral construct, along with a packaging construct (psPAX2) and an envelope construct (pMD2.G), into 293T cells by CaCl₂ transfection. The medium was replaced 6 h posttransfection with DMEM–F-12 medium supplemented with biotin (8 μg/ml), pantothenate (4 μg/ml), penicillin (50 U/ml), streptomycin (50 U/ml), and non-heat-inactivated FBS (10%). Medium containing viral particles was collected, filtered (0.45 μm), and stored at 4°C. Infection of SGBS cells was performed by incubation with viral medium supplemented with Polybrene (4 μg/ml) for 8 h on two consecutive days. Differentiation was induced 1 to 2 days after the second day of viral infection. Viral infection efficiency, based on green fluorescent protein (GFP) expression, was determined by fluorescence microscopy. Phase-contrast and fluorescence images of differentiating SGBS cells were taken using a Zeiss Axiovert 2000 microscope.

Oil Red O staining and triglyceride quantification.

Oil Red O staining of differentiated SGBS cells was performed by washing the cells with PBS, followed by 20 min of incubation at room temperature in 10% formalin. The fixed cells were then washed twice with H₂O, followed by 5 min of incubation at room temperature in 60% isopropanol. Following removal of the isopropanol, the cells were dried and treated with Oil Red O (Sigma) working solution (∼2 mg/ml in 60% isopropanol) for 10 min. The Oil Red O-stained cells were then washed four times with H₂O, and images were acquired on an Evos XL core microscope. Following image acquisition, the cells were dried completely, and dye was extracted in 100% isopropanol and quantified at 500 nm. Quantification of cellular triglycerides was performed using a commercially available kit (Biovision) according to the manufacturers' instructions.

Transfection of siRNA and miRNA mimics/inhibitors.

Subconfluent SGBS cells were transfected with miR-33b or control miRNA mimics (Dharmacon) according to standard Lipofectamine 2000 protocols (Invitrogen). Following transfection, the cells were either stimulated to differentiate as described above or maintained in normal growth medium. For proliferation experiments with HMGA2 small interfering RNA (siRNA), cells were first transfected with HMGA2 or control siRNA (Dharmacon) for 48 h, after which the cells were replated and cotransfected with siRNA and miRNA mimics. Cell numbers were determined by trypsinization and cell counting using a hemocytometer. The cell number was determined based on 2 or 3 counts of four quadrants.

Western blot analysis.

Cells were lysed in ice-cold buffer containing 50 mM Tris-HCl, pH 7.5, 0.1% SDS, 0.1% deoxycholic acid, 0.1 mM EDTA, 0.1 mM EGTA, 1% NP-40, 5.3 mM NaF, 1.5 mM NaP, 1 mM orthovanadate, 1 mg/ml of protease inhibitor cocktail (Roche), and 0.25 mg/ml 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride (AEBSF) (Roche). The cell lysates were sonicated and rotated at 4°C for 1 h before the insoluble material was removed by centrifugation at 12,000 × g for 10 min. After normalizing for equal protein concentrations, cell lysates were resuspended in SDS sample buffer before separation by SDS-PAGE. Following transfer of the proteins onto nitrocellulose membranes, the membranes were probed with antibodies for HMGA2 (Biocheck; catalog number 59170AP) or HSP90 (BD Biosciences). Protein bands were visualized using the Odyssey infrared imaging system (Li-Cor Biotechnology), and densitometry was performed using ImageJ software.

3′ UTR luciferase reporter assays.

A plasmid containing the entire 3′ UTR of HMGA2 in a psiCheck-2 vector (Promega) was kindly provided by Marcus Ernst Peter. The vector also contains a constitutively expressed firefly luciferase gene, which is used to normalize transfections. Mutation of miR-33 binding sites was carried out using a QuikChange Multi site-directed mutagenesis kit (Agilent Technologies) according to the manufacturer's instructions. All constructs were confirmed by sequencing. Twelve-well plates were plated with COS7 cells, and the cells were cotransfected with 1 μg of the indicated 3′ UTR luciferase reporter vectors and miR-33 or control mimics by Lipofectamine 2000 transfection. Luciferase activity was measured using the Dual-Glo luciferase assay system (Promega). Renilla luciferase activity was normalized to the corresponding firefly luciferase activity and plotted as a percentage of the control.

Statistical analysis.

Data are presented as means and standard errors of the mean (SEM). Statistical differences between groups were evaluated using an unpaired two-tailed Student t test.

RESULTS

miR-33b expression is induced during differentiation of human preadipocytes.

To investigate the regulation of miR-33a and miR-33b during WAT differentiation, we performed in vitro differentiation experiments in early-passage (passage 4 [P4]) primary human preadipocytes. When stimulated with proadipogenic medium, primary human preadipocytes differentiate into mature adipocytes, resulting in the formation and accumulation of intracellular lipid droplets (Fig. 1A) and the induction of adipogenic genes (Fig. 1B to D), including PPARγ, GLUT4, and the adiponectin gene (AdipoQ). As expected, adipocyte differentiation also induces a dramatic increase in the expression of SREBP-1c. While SREBP-1a is the predominant SREBP-1 isoform in primary human preadipocytes, its expression is not altered during adipogenesis (Fig. 1E). Importantly, miR-33b expression is also strongly induced upon differentiation into mature adipocytes, while miR-33a expression is largely unaltered (Fig. 1F). Moreover, analysis of tissue samples from patients undergoing bariatric surgery revealed that levels of miR-33b are approximately 20-fold higher in mature adipocytes than in cells in the stromal vascular fraction, including preadipocytes (Fig. 1G).

FIG 1.

miR-33b expression is induced during differentiation of human primary preadipocytes. (A) Representative phase-contrast images of undifferentiated and differentiated human primary preadipocytes at different time points. (B to D) qRT-PCR analysis of PPARγ (B), GLUT4 (C), and adiponectin gene (AdipoQ) (D) expression in undifferentiated (Undiff) and differentiated (Diff) human primary preadipocytes at different time points (d, days). (E) qRT-PCR analysis of SREBP-1a, SREBP-1c, and SREBP-2 expression in undifferentiated and differentiated human primary preadipocytes at different time points. (F) qRT-PCR analysis of miR-33a and miR-33b expression in undifferentiated and differentiated human primary preadipocytes at different time points. (G) qRT-PCR analysis of miR-33b expression in stromal vascular and mature adipocyte fractions of WAT from human patients. qRT-PCR data are expressed as means and SEM.

Primary preadipocytes lose the ability to differentiate after only a few passages in culture. For this reason, further exploration of the mechanisms by which miR-33b regulates human preadipocyte differentiation utilized a human preadipocyte cell line derived from a patient with SGBS. SGBS preadipocytes retain their adipogenic capacity for many generations in culture (48, 50). Upon stimulation, these cells differentiate into mature adipocytes in a fashion similar to that of early-passage primary preadipocytes, including the development of intracellular lipid droplets (Fig. 2A) and induction of adipogenic genes (Fig. 2B to D). SREBP gene and miR-33 isoform inductions were also similar between SGBS cells and primary human preadipocytes, although in these cells, the induction of SREBP-1c during adipogenesis makes it the more abundant isoform in mature adipocytes (Fig. 2E and F and data not shown). Overall, these findings indicate that this model will be useful in elucidating the role of miR-33b in human adipogenesis. Moreover, the expression patterns of SREBP-1c and miR-33b are similar over the time course of adipogenic differentiation, showing an initial decrease in expression during the first 24 h followed by substantial induction after 3 to 5 days. The primary miR-33b transcript and precursor mRNA also show a similar expression pattern, although a more substantial increase in the mature form suggests that miR-33b may also be stabilized under these conditions (data not shown). Together, these findings demonstrate that miR-33b expression is markedly induced upon differentiation of human preadipocytes in a manner similar to that of SREBP-1c. These findings are consistent with work done in hepatic cell lines demonstrating that miR-33b is cotranscribed along with SREBP-1 rather than being regulated by an independent promoter, and further, they suggest that miR-33b may play an important role in the differentiation and/or function of mature adipocytes.

FIG 2.

miR-33b expression is induced during human SGBS preadipocyte differentiation. (A) Representative phase-contrast images of undifferentiated and differentiated SGBS cells at different time points. (B to D) qRT-PCR analysis of PPARγ (B), GLUT4 (C), and adiponectin gene (D) expression in undifferentiated and differentiated SGBS cells at different time points. (E) qRT-PCR analysis of SREBP-1a, SREBP-1c, and SREBP-2 expression in undifferentiated and differentiated SGBS cells at different time points. (F) qRT-PCR analysis of miR-33a and miR-33b expression in undifferentiated and differentiated SGBS cells at different time points. qRT-PCR data are expressed as means and SEM.

miR-33b expression influences adipocyte differentiation.

To determine whether alterations in miR-33b influence adipocyte differentiation capacity, SGBS cells were infected with lentiviral vectors containing control, pre-miR-33b, or anti-miR-33b expression constructs. Following 15 days of differentiation, SGBS cells infected with lentiviral constructs were nearly 100% GFP positive, indicating strong infection efficiency. Moreover, infection of SGBS preadipocytes with lentiviral vectors for overexpression of miR-33b resulted in reduced lipid droplet formation, which was evident under phase-contrast microscopy or following staining of neutral lipids with Oil Red O (Fig. 3A). Extraction of the Oil Red O dye revealed that overexpression of miR-33b resulted in a significant decrease in the amounts of neutral lipids accumulated after 8 and 15 days of differentiation (Fig. 3B). This finding was further confirmed by direct quantification of triglycerides, which were also reduced in differentiated adipocytes treated with pre-miR-33b lentivirus (Fig. 3C). Conversely, after 15 days of differentiation, cells treated with anti-miR-33b constructs showed an increase in lipid droplets, Oil Red O staining, and triglyceride accumulation following 15 days of differentiation (Fig. 3A to C). Furthermore, SGBS cells overexpressing miR-33b showed less induction of two of the key transcription factors regulating adipocyte differentiation, SREBP-1 and PPARγ (Fig. 3D and E), as well as the adipogenic genes GLUT4 and adiponectin (Fig. 3F and G), while inhibition of miR-33b resulted in increased expression of PPARγ and adiponectin (Fig. 3D to G). Importantly, treatment with anti-miR-33b lentivirus was not found to have any impact on miR-33a expression levels (Fig. 3H). Together, these findings indicate that miR-33b acts as a negative regulator of adipogenesis, despite being highly upregulated during the later stages of adipocyte differentiation.

FIG 3.

miR-33b overexpression impairs adipocyte differentiation, while miR-33b inhibition promotes lipid droplet accumulation. (A) Representative phase-contrast, GFP, and Oil Red O images of SGBS cells transduced with control, pre-miR-33b, or anti-miR-33b lentivirus and differentiated for 15 days. (B) Spectrophotometric quantification analysis of Oil Red O staining in SGBS cells transduced with control, pre-miR-33b, or anti-miR-33b lentivirus and differentiated for 8 or 15 days. RFU, relative fluorescence units. (C) Triglyceride quantification in SGBS cells transduced with control, pre-miR-33b, or anti-miR-33b lentivirus and differentiated for 0, 8, or 15 days. (D to G) qRT-PCR analysis of SREBP-1c (D), PPARγ (E), GLUT4 (F), and adiponectin gene (G) expression in SGBS cells transduced with control, pre-miR-33b, or anti-miR-33b virus and differentiated for 0, 8, or 15 days. (H) qRT-PCR analysis of miR-33a and miR-33b in SGBS cells transduced with control or anti-miR-33b lentivirus. The data are expressed as means and SEM. *, P ≤ 0.05 versus cells transduced with control virus.

miR-33b overexpression impairs preadipocyte proliferation and clonal expansion.

To begin to address the mechanism by which miR-33b impairs adipogenesis, we first sought to determine how overexpression of miR-33b affects preadipocyte proliferation and the clonal-expansion phase of differentiation, under which conditions miR-33b expression is very low (Fig. 1F). Indeed, transfection of subconfluent SGBS cells with mimics of miR-33b resulted in reduced proliferative capacity, as evidenced by a modest decrease in cell numbers after 2 days, which became statistically significant by day 5 (Fig. 4A). These findings are consistent with previous work showing that miR-33 can regulate proliferation in cancer cell lines and other cell types. These effects have been attributed to the ability of miR-33b to target cell cycle genes, including CDK6 and cyclin D1 genes (46). Consistent with this, our data indicate that the expression of CDK6 is reduced in proliferating preadipocytes overexpressing miR-33b, although the difference was not found to be significant. Other reported miR-33 targets, including cyclin D1 and SREBP-1, were not found to be altered (Fig. 4B). Furthermore, we found that overexpression of miR-33b causes a significant reduction of the expression of the chromatin-remodeling factor gene HMGA2 (Fig. 4B). HMGA2 has been demonstrated to play an important role in the clonal-expansion phase of adipogenesis (27), and genetic alteration of HMGA2 in vivo has dramatic effects on body weight and adiposity (51–53). Although the exact mechanism by which HMGA2 impacts adipocyte clonal expansion and differentiation is not entirely known, it likely involves its ability to control critical genes involved in cellular proliferation. Indeed, HMGA2 is known to be a critical regulator of cell growth in a number of different types of cancer, and regulation of HMGA2 by miR-33 has been shown to regulate proliferation and metastasis in cancer cell lines (47, 54, 55). We next assessed the proliferative capacity of SGBS cells following stimulation of adipocyte differentiation, and similar to our findings under normal growth conditions, SGBS cells transfected with miR-33b mimics showed reduced capacity for clonal expansion, resulting in significantly reduced cell numbers following 4 days of differentiation (Fig. 4C). Additionally, the expression of CDK6 and HMGA2 was significantly reduced under these conditions (Fig. 4D and E).

FIG 4.

miR-33b overexpression impairs SGBS preadipocyte proliferation and clonal expansion. (A) Cell number analysis of SGBS preadipocytes transfected with negative-control miRNA (CM) or miR-33b mimics (miR-33b) at different time points. (B) qRT-PCR analysis of CDK6, Cyclin D1, HMGA2, and SREBP-1 expression in undifferentiated and differentiated SGBS preadipocytes at different time points. (C) Cell numbers of SGBS preadipocytes transfected with CM or miR-33b and differentiated for 2 or 4 days. (D and E), qRT-PCR analysis of CDK6 and HMGA2 expression in differentiating SGBS preadipocytes treated with CM or miR-33b. The data are expressed as means and SEM. *, P ≤ 0.05 versus cells transfected with CM.

To further characterize the role of HMGA2 in mediating the effects of miR-33b on adipogenesis, we examined how alterations in miR-33b affect the levels of HMGA2 in differentiating preadipocytes. Consistent with the effect we observed in cells treated with miR-33b mimics, cells infected with lentiviral constructs for overexpression of miR-33b showed decreased protein levels of HMGA2 3 days after induction of differentiation, while cells treated with anti-miR-33b lentiviral constructs had increased levels of HMGA2 (Fig. 5A and B). Additionally, mRNA expression of HMGA2 was found to be decreased in cells overexpressing miR-33b even in the later stages of adipogenesis, when HMGA2 levels are already reduced (Fig. 5C). Interestingly, the expression levels of HMGA2 mRNA appear to follow an inverse pattern of regulation compared to that observed for miR-33b. As such, HMGA2 is induced during the clonal-expansion phase of adipogenesis but reduced following terminal differentiation (Fig. 5D).

FIG 5.

Regulation of HMGA2 in human preadipocytes and possibly patients with lipoma. (A and B) Western blot (A) and quantification (B) of HSP90 and HMGA2 in SGBS preadipocytes in untreated, undifferentiated cells or cells transduced with control, pre-miR-33b, or anti-miR-33b virus followed by 3 days of differentiation. *, P ≤ 0.05 versus cells transduced with control virus. (C) qRT-PCR analysis of HMGA2 expression in SGBS cells transduced with control or pre-miR-33b lentivirus at 0, 8, and 15 days of differentiation. *, P ≤ 0.05 versus cells transduced with control virus. (D) qRT-PCR analysis of HMGA2 expression in SGBS preadipocytes at different time points after differentiation. (E) Identification of miR-33 binding sites (boldface) in the HMGA2 3′ UTR. (F) Luciferase activity assays of HMGA2 3′ UTR with or without mutations in binding sites for miR-33. (G to I) qRT-PCR analysis of HMGA2, miR-33a, and miR-33b expression in subcutaneous adipose tissue (SAT) or lipoma samples from human patients. *, P ≤ 0.05 versus SAT. All data are expressed as means and SEM.

To further confirm the direct targeting of HMGA2 by miR-33b, we obtained a luciferase reporter construct for the entire 3′ UTR of HMGA2, which contains four predicted binding sites for miR-33b (Fig. 5E). Overexpression of miR-33b mimics resulted in a dramatic reduction in HMGA2 3′ UTR activity. Inserting point mutations at each of the four miR-33b binding sites in the HMGA2 3′ UTR individually to disrupt miR-33b binding had little effect on the ability of miR-33b to dampen HMGA2 3′ UTR luciferase activity. However, this effect was largely abrogated when all four binding sites for miR-33b were mutated together (Fig. 5F). HMGA2 is known to be dysregulated in a variety of cancers, including lipoma and liposarcoma. While most human cases of lipoma have rearrangements in the HMGA2 3′ UTR, which are believed to disrupt miRNA regulation, other tumors show elevated HMGA2 expression without alterations in the 3′ UTR. Gene expression analysis of lipoma samples with intact 3′ UTRs indicated that miR-33b expression was significantly downregulated compared to normal subcutaneous white adipose tissue (Fig. 5I), while miR-33a levels were not significantly altered (Fig. 5H). These findings suggest that repression of miR-33b may be involved in the elevation of HMGA2 expression in these tumors (Fig. 5G).

To further determine the role of HMGA2 in mediating the effects of miR-33b on adipocyte proliferation and differentiation, we performed siRNA knockdown of HMGA2. Transfection of SGBS cells with miR-33b mimics reduced HMGA2 mRNA levels similarly to cells transfected with HMGA2 siRNA (Fig. 6A). Assessment of proliferation in these cells indicated that knockdown of HMGA2 resulted in reduced proliferation, similar to overexpression of miR-33b. Consistent with the idea that miR-33b regulates preadipocyte proliferation by targeting HMGA2, cotransfection of HMGA2 siRNA and miR-33b mimics did not have an additive effect on SGBS cell proliferation (Fig. 6B). Previous studies in 3T3-L1 cells have shown that loss of HMGA2 impairs adipocyte differentiation. We demonstrate similar effects, as knockdown of HMGA2 in SGBS cells was found to decrease lipid droplet formation and Oil Red O staining, resembling the effects observed in cells overexpressing miR-33b (Fig. 6D and E). Additionally, cells treated with HMGA2 siRNA had impairment in the induction of adipogenic genes similar to that observed with overexpression of miR-33b (Fig. 6F to I). Although the repression of HMGA2 appeared similar in cells treated with miR-33b mimics or HMGA2 siRNA (Fig. 6C), the ability to impair the induction of adipogenic genes appeared less pronounced.

FIG 6.

Knockdown of HMGA2 impairs SGBS cell proliferation and differentiation. (A) qRT-PCR analysis of HMGA2 expression in SGBS preadipocytes transfected with nonsilencing siRNA, miR-33b mimics, or HMGA2 siRNA. (B) Cell number analysis of SGBS preadipocytes transfected with nonsilencing siRNA, miR-33b mimics, or HMGA2 siRNA at different time points. (C) qRT-PCR analysis of HMGA2 expression in undifferentiated and differentiated SGBS preadipocytes transfected with nonsilencing siRNA, miR-33b mimics, or HMGA2 siRNA. (D) Representative phase-contrast images of unstained or Oil Red O-stained differentiated SGBS preadipocytes transfected with nonsilencing or HMGA2 siRNA. (E) Quantification of Oil Red O staining from panel D. (F to I) qRT-PCR analysis of SREBP-1c (F), PPARγ (G), GLUT4 (H), and adiponectin gene (I) expression in SGBS cells transfected with nonsilencing or HMGA2 siRNA and differentiated for 10 days. The data are expressed as means and SEM. *, P ≤ 0.05 versus cells transfected with nonsilencing siRNA.

To gain further understanding of how miR-33b regulates adipocyte differentiation, we performed gene expression arrays of genes involved in adipogenesis in SGBS cells infected with control or pre-miR-33b lentivirus. Among genes whose expression was found to be significantly altered by overexpression of miR-33b, those involved in promoting the differentiation and function of WAT and brown adipose tissue (BAT) were primarily downregulated both prior to and during differentiation (17:0 [down/up] at day 0 and 11:4 at day 8) (Fig. 7A). Effects on genes impairing adipogenesis were less clear, with more genes downregulated in undifferentiated cells (7:2), while antiadipogenic genes were both up- and downregulated during differentiation (3:4) (Fig. 7B). Importantly, there was a striking reduction in the expression of genes known to be directly regulated by PPARγ. These findings indicate an overall reduction in the expression and induction of adipogenic genes in response to overexpression of miR-33b. Among previously identified targets of miR-33, cyclin D1 (CCND1), IRS2, and SREBP-1 (SREBF1) were not found to be significantly altered in cells overexpressing miR-33b, while expression of PGC1α (PPARGC1A) was significantly reduced both prior to and after differentiation (terms in parentheses indicate the abbreviations used in the qPCR array analysis). Additionally, these expression arrays allowed us to identify numerous predicted miR-33b targets that are significantly downregulated in cells overexpressing miR-33b, including CEBPα (CEBPA), DLK1, WNT3A, and WNT10B.

FIG 7.

Overexpression of miR-33b alters expression of genes involved in adipogenesis. (A) Heat map of genes involved in promoting adipogenesis following transduction with control or pre-miR-33b virus after 0 or 8 days of differentiation. (B) Heat map of genes involved in inhibition of adipogenesis following treatment with control or pre-miR-33b virus after 0 or 8 days of differentiation. #, significantly downregulated (P ≤ 0.05) versus undifferentiated control; %, significantly upregulated versus undifferentiated control; *, significantly downregulated versus differentiated control; &, significantly upregulated versus differentiated control; +, predicted target of miR-33; ++, known target of miR-33.

To further determine how effectively knockdown of HMGA2 can mimic the effects of miR-33b overexpression on adipocyte differentiation, we also performed adipogenesis gene expression arrays in cells transfected with control or HMGA2 siRNA, followed by 10 days of differentiation. Similar to cells overexpressing miR-33b, knockdown of HMGA2 led to an overall reduction in most genes regulated by PPARγ, although these effects were less substantial and in many cases did not reach statistical significance. Of the genes that were found to be significantly altered in cells transfected with HMGA2 siRNA, proadipogenic genes were primarily downregulated (7:2 [down/up]) (Fig. 8A), but expression of significantly altered antiadipogenic genes was also primarily reduced (6:1) (Fig. 8B). While knockdown of HMGA2 was found to result in similar impairments in lipid droplet formation and induction of adipogenic genes, the more pronounced effects in miR-33-overexpressing cells and the differences between gene expression patterns indicate that other factors regulated by miR-33b are also likely involved in mediating its effects on adipocyte differentiation and function.

FIG 8.

Knockdown of HMGA2 alters expression of genes involved in adipogenesis. (A) Heat map of genes involved in promoting adipogenesis following transfection with nonsilencing or HMGA2 siRNA and 10 days of differentiation. (B) Heat map of genes involved in inhibition of adipogenesis following transfection with nonsilencing or HMGA2 siRNA and 10 days of differentiation. #, significantly downregulated (P ≤ 0.05) versus undifferentiated control; %, significantly upregulated versus undifferentiated control; *, significantly downregulated versus differentiated control; &, significantly upregulated versus differentiated control.

DISCUSSION

miR-33 is one of the best-studied miRNAs due to its important role in regulating cholesterol and fatty acid metabolism in arterial macrophages and the liver (40–45). Indeed, numerous studies have demonstrated the efficacy of anti-miR-33 therapy for raising plasma HDL cholesterol levels and reducing the atherosclerotic plaque burden in both rodents and nonhuman primates (41, 43, 56–59). However, the increased risk for development of hepatic steatosis observed in some of these studies emphasizes the potential for unintended effects of anti-miR-33 therapy (60, 61). For this reason, understanding the role of miR-33 in other metabolic tissues, such as WAT, is of critical importance.

In this study, we demonstrate that miR-33b is highly induced during the differentiation of human preadipocytes. This is consistent with previous work showing that miR-33b is cotranscribed along with SREBP-1 in other tissues, as SREBP-1c is also highly induced during adipogenesis (11). Assessment of absolute gene expression values indicated that while SREBP-1a is the predominant isoform of SREBP-1 in preadipocytes and appears to still be more abundant in differentiated primary human adipocytes, its expression is not induced during adipocyte differentiation. As such, the increased expression of miR-33b during adipogenesis is most likely due to its being cotranscribed along with SREBP-1c. Similarly, miR-33a appears to be more highly expressed in human preadipocytes; however, following its induction during adipocyte differentiation, miR-33b appears to be more abundant. In addition to being highly induced, SREBP-1 is required for in vitro adipogenesis of WAT and BAT. A recent study in porcine preadipocytes showed that transfection with miR-33 mimics decreased in vitro adipogenesis, although in that work, the authors did not observe an increase in the expression of either miR-33b or SREBP-1c during adipocyte differentiation (62). Here, we demonstrate that overexpression of miR-33b in human preadipocytes impairs proliferation and differentiation, while inhibition of miR-33b promotes lipid droplet accumulation. As miR-33a and miR-33b share a seed sequence and are therefore expected to target the same genes, we would expect that overexpression of miR-33a would be similarly effective at impairing adipogenesis. Importantly, anti-miR-33b lentiviral constructs were not found to impact the expression of miR-33a, and as miR-33a is not highly induced during adipocyte differentiation in the same manner as miR-33b, it is not anticipated that anti-miR-33a treatment would have as much of an impact on adipogenesis.

Gene expression analysis of genes involved in adipogenesis further demonstrates an overall blunting of the induction of adipogenic genes in cells overexpressing miR-33b, especially those genes directly regulated by PPARγ. Among the genes found to have significantly reduced expression in cells overexpressing miR-33b, PGC1α has previously been shown to be regulated by miR-33 (45, 63). Although PGC1α is an important regulator of mitochondrial function and browning of WAT, a reduced capacity for PGC1α to induce these changes would be expected to decrease energy expenditure, which does not offer a ready explanation for the impairment in adipogenesis observed here. Additionally, numerous predicted targets of miR-33b were downregulated in cells overexpressing miR-33b, including CEBPα, DLK1, WNT3A, and WNT10B. Among these, DLK1, WNT3A, and WNT10B are negative regulators of adipogenesis, so the downregulation of the genes prior to differentiation would be likely to have an effect on adipogenesis opposite to that observed. CEBPα, however, is an important regulator of adipogenesis involved in promoting the activation of PPARγ and its downstream targets (64). As expression of CEBPα was found to be downregulated both prior to and after adipocyte differentiation, targeting of CEBPα by miR-33b may play an important role in mediating the effects of miR-33b on adipogenesis and the induction of PPARγ targets.

These findings suggest that the induction of miR-33b during the later stages of adipogenesis may serve as a mechanism to prevent excessive adipocyte hypertrophy due to overaccumulation of lipids. Interestingly, SREBP-1c mRNA expression is known to be reduced in WAT under conditions of obesity (65). If miR-33b levels were also reduced, it would suggest that downregulation of miR-33b, along with SREBP-1c, could be involved in promoting the adipocyte hypertrophy observed in obese patients. Overall, these findings demonstrate the capacity of miR-33b to regulate adipocyte differentiation, which should be considered in studies involving anti-miR-33 therapy. Indeed, work done to characterize miR-33-deficient mice has demonstrated that loss of miR-33 causes an increased propensity for obesity and type II diabetes in mice fed a high-fat diet (60). Although these are knockout mice for miR-33a, and only miR-33b was found to be highly induced during differentiation of human adipocytes, miR-33a may play a more important role in adipogenesis in mice, as they lack miR-33b.

Additionally, the induction of miR-33b following the terminal differentiation of adipocytes may aid in maintaining these cells in a nonproliferative state. This theory is consistent with previous findings demonstrating that miR-33 can reduce expression of cell cycle proteins and impair proliferation in cancer cell lines (46). Our data also support this, as overexpression of miR-33b was found to impair preadipocyte proliferation and to reduce the expression of CDK6, including a further reduction in gene expression beyond that normally observed in cells following terminal differentiation.

Moreover, overexpression of miR-33b decreased both the protein and mRNA levels of HMGA2. This chromatin-remodeling factor is known to be upregulated during and involved in the clonal-expansion phase of adipogenesis but downregulated following terminal differentiation (27). The importance of HMGA2 for the development of adipose tissue in vivo is demonstrated by HMGA2 knockout mice, which exhibit a dramatic reduction in adipose mass (51), while transgenic HMGA2 mice develop obesity and have an increased incidence of lipomas (52, 53). Consistent with this, we found that knockdown of HMGA2 was sufficient to impair the proliferation and differentiation of SGBS preadipocytes in a manner similar to that observed with overexpression of miR-33b, although the more modest effects on adipogenic genes, as well as discrepancies in gene expression patterns, suggest other factors are also involved in mediating the effects of miR-33b.

Although the exact mechanisms by which HMGA2 impacts adipocyte clonal expansion and differentiation are not entirely known, they likely involve its ability to control critical genes involved in cellular proliferation. Indeed, HMGA2 is overexpressed in a number of different types of human cancer, including lipoma and liposarcoma, where levels of HMGA2 are almost always elevated (49, 66, 67). In many of these cases, the 3′ UTR of HMGA2 has been truncated, leading to dramatic overexpression. This is consistent with work showing that negative regulatory elements within the HMGA2 3′ UTR are responsible for controlling its expression (68). As the 3′ UTR is also the primary site of miRNA targeting, these findings suggest that miRNAs may be important for regulation of HMGA2. Indeed, miRNAs, including Let-7, have been shown to regulate HMGA2, and regulation of HMGA2 expression by Let-7 has been explored in the context of both adipogenesis and cancer (27, 69, 70). Similar to what we have observed with miR-33b, expression of Let-7 is inversely correlated with HMGA2 during adipogenesis, and overexpression of Let-7 impairs adipocyte differentiation (27). Additionally, the ability of Let-7 to reduce tumor growth in a number of cancer models has been attributed to its targeting of HMGA2 (70). However, when investigators examined gene expression in lipoma samples that did not have rearrangements in the 3′ UTR, they observed that HMGA2 levels were elevated in the lipoma samples but were not inversely correlated with the expression of Let-7 (49). Interestingly, we found that levels of miR-33b were reduced in these lipoma samples, suggesting that repression of miR-33b may help promote the overexpression of HMGA2 in some cases of lipoma. These findings are consistent with previous work demonstrating that targeting of HMGA2 by miR-33 regulates proliferation and metastasis of multiple cancer cell lines (47, 54, 55).

In this work, we demonstrate that miR-33b, like SREBP-1c, is highly induced during, and plays an important role in regulating differentiation of human preadipocytes in vitro. However, nothing is known about the role of miR-33b during adipogenesis in vivo, and studies on SREBP-1 knockout and transgenic mice do not demonstrate the same absolute requirement for SREBP-1 that is observed during in vitro adipocyte differentiation. Specifically, targeted disruption of the SREBP-1 gene in mice, although primarily resulting in embryonic lethality, does not appear to result in any overt differences in adipose mass or gene expression in animals that survive to adulthood (71). Additionally, transgenic mice overexpressing the transcriptionally active nuclear SREBP-1c, the predominant form of SREBP-1 in adipose tissue, under the aP2 promoter exhibit severe lipodystrophy (72). Alternatively, a similar strain of transgenic mice overexpressing nuclear SREBP-1a under the same promoter has increased lipid accumulation and greatly enlarged fat pads (73). Although these transgenic-mouse models demonstrate the capacity of SREBP-1 to dramatically alter adipogenesis in vivo, the exact role of SREBP-1 induction during in vivo adipogenesis is still unclear. Moreover, because miR-33b is lost in rodents, assessing the in vivo role of miR-33b is challenging, and body weight measurements have not been reported in currently published studies of anti-miR-33 therapy in nonhuman primates. The role of miR-33b under conditions of obesity and diabetes is especially interesting because of the responsiveness of SREBP-1c to insulin and its dysregulation under conditions of insulin resistance (65).

Recently, a mouse model was described in which the sequence for miR-33b was inserted into the mouse SREBP-1 gene (miR-33b KI) (74). Initial characterization of these mice indicated a reduction in some miR-33 target genes in liver and a decrease in circulating HDL levels. However, a more detailed characterization is needed to discern the specific role of miR-33b in the regulation of cholesterol metabolism and the development of atherosclerosis. Furthermore, this and similar models will provide a useful system in which to study how miR-33b is regulated under conditions of obesity and insulin resistance. Indeed, the Nobel laureates J. L. Goldstein and M. S. Brown proposed that the low levels of HDL in patients with metabolic syndrome could be attributed to the SREBP-1c-mediated increase of miR-33b in liver and subsequent downregulation of ABCA1 (75). Selective insulin resistance has been demonstrated to maintain the ability of SREBP-1 to induce lipogenic genes despite impaired insulin signaling (76). However, in adipose tissue, this has been shown to be primarily due to improved processing, while SREBP-1 mRNA expression is reduced both in mouse models of obesity/diabetes and in human patients with these conditions (65, 77). As miR-33b is cotranscribed along with SREBP-1 mRNA, these findings suggest that expression of miR-33b may also be lower under conditions of obesity and insulin resistance. Overall, the findings of this study indicate that miR-33b is highly induced upon the differentiation of human preadipocytes and demonstrate the capacity of miR-33b to regulate this process. Further work, using newly developed mouse models, will allow researchers to determine how miR-33b expression is regulated in different tissues under conditions of obesity and insulin resistance and what impact miR-33b has on the development of disease states associated with metabolic syndrome.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.