Thyroid Hormone Receptor Sumoylation Is Required for Preadipocyte Differentiation and Proliferation*

Background: Thyroid hormone receptor (TR) sumoylation is essential for thyroid hormone regulation of gene expression. Results: TR sumoylation mutants impair differentiation though down-regulation of C/EBPs, constitutive interaction with NCoR, interference with PPARγ signaling, and disruption of the Wnt canonical signaling pathway important for preadipocyte proliferation. Conclusion: TR sumoylation site mutations impair preadipocyte proliferation and differentiation. Significance: TR sumoylation is required for adipogenesis. Thyroid hormone and thyroid hormone receptor (TR) play an essential role in metabolic regulation. However, the role of TR in adipogenesis has not been established. We reported previously that TR sumoylation is essential for TR-mediated gene regulation and that mutation of either of the two sites in TRα or any of the three sites in TRβ reduces TR sumoylation. Here, we transfected TR sumoylation site mutants into human primary preadiocytes and the mouse 3T3L1 preadipocyte cell line to determine the role of TR sumoylation in adipogenesis. Reduced sumoylation of TRα or TRβ resulted in fewer and smaller lipid droplets and reduced proliferation of preadipocytes. TR sumoylation mutations, compared with wild-type TR, results in reduced C/EBP expression and reduced PPARγ2 mRNA and protein levels. TR sumoylation mutants recruited NCoR and disrupted PPARγ-mediated perilipin1 (Plin1) gene expression, associated with impaired lipid droplet formation. Expression of NCoRΔID, a mutant NCoR lacking the TR interaction domain, partially “rescued” the delayed adipogenesis and restored Plin1 gene expression and adipogenesis. TR sumoylation site mutants impaired Wnt/β-catenin signaling pathways and the proliferation of primary human preadipocytes. Expression of the TRβ K146Q sumoylation site mutant down-regulated the essential genes required for canonical Wnt signal-mediated proliferation, including Wnt ligands, Fzds, β-catenin, LEF1, and CCND1. Additionally, the TRβ K146Q mutant enhanced the canonical Wnt signaling inhibitor Dickkopf-related protein 1 (DKK1). Our data demonstrate that TR sumoylation is required for activation of the Wnt canonical signaling pathway during preadipocyte proliferation and enhances the PPARγ signaling that promotes differentiation.

Thyroid hormone and thyroid hormone receptor (TR) play an essential role in metabolic regulation. However, the role of TR in adipogenesis has not been established. We reported previously that TR sumoylation is essential for TR-mediated gene regulation and that mutation of either of the two sites in TR␣ or any of the three sites in TR␤ reduces TR sumoylation. Here, we transfected TR sumoylation site mutants into human primary preadiocytes and the mouse 3T3L1 preadipocyte cell line to determine the role of TR sumoylation in adipogenesis. Reduced sumoylation of TR␣ or TR␤ resulted in fewer and smaller lipid droplets and reduced proliferation of preadipocytes. TR sumoylation mutations, compared with wild-type TR, results in reduced C/EBP expression and reduced PPAR␥ 2 mRNA and protein levels. TR sumoylation mutants recruited NCoR and disrupted PPAR␥-mediated perilipin1 (Plin1) gene expression, associated with impaired lipid droplet formation. Expression of NCoR⌬ID, a mutant NCoR lacking the TR interaction domain, partially "rescued" the delayed adipogenesis and restored Plin1 gene expression and adipogenesis. TR sumoylation site mutants impaired Wnt/␤-catenin signaling pathways and the proliferation of primary human preadipocytes. Expression of the TR␤ K146Q sumoylation site mutant down-regulated the essential genes required for canonical Wnt signal-mediated proliferation, including Wnt ligands, Fzds, ␤-catenin, LEF1, and CCND1. Additionally, the TR␤ K146Q mutant enhanced the canonical Wnt signaling inhibitor Dickkopf-related protein 1 (DKK1). Our data demonstrate that TR sumoylation is required for activation of the Wnt canonical signaling pathway during preadipocyte proliferation and enhances the PPAR␥ signaling that promotes differentiation.
Thyroid hormone receptor (TR) 3 and triiodothyronine (T 3 ) have an essential role in metabolic homeostasis. Patients with reduced or increased thyroid hormone levels experience changes in body weight and body composition due to thyroid hormone-mediated action on metabolic targets. In white adipose tissue (WAT), thyroid hormone enhances catecholaminemediated lipolysis by stimulating expression of adrenergic receptors and enhancing sensitivity (1), which leads to activation of hormone-sensitive lipase (2). In brown adipose tissue (BAT), thyroid hormone and catecholamines stimulate UCP1 expression and promote thermogenesis and oxidative phosphorylation, which increases energy expenditure and reduces fat accumulation (3). Knock-out of the type 2 5Ј-deiodinase (D2) gene, which converts thyroxine to the T 3 active form, is associated with reduced BAT activity and glucose intolerance. D2 knock-out mice have a defect in diet-induced thermogenesis, resulting in glucose intolerance, obesity, and enhanced hepatic steatosis (4). TR mutations alter fat deposition in several TR␣ mutant mouse models. TR gene point mutations that disrupt ligand binding produce a dominant-negative TR that antagonizes the wild-type TR. TR␣1 pv/pv mice have markedly reduced WAT and fat mobilization due to TR␣1 pv/pv inhibition of PPAR␥-mediated transcription (5). TR␣ P398H mice have significantly increased visceral fat due to impaired catecholamine-stimulated lipolysis and ␤-oxidation in WAT (6). TR␣1 R384C mice have reduced fat depots, increased lipid mobilization in WAT, and BAT activation due to increased sympathetic outflow, which normalizes when the animals are kept at thermoneutral conditions (7). TR␤ pv/pv mice do not display abnormalities in WAT but have excessive hepatic lipid deposition (8).
A widely used in vitro model used to study adipogenesis is the mouse 3T3L1 cell line. 3T3L1 cells must be grown to confluence to reach growth arrest, a prerequisite for preadipocyte differentiation. However, another model, 3T3-F422 cells, is grown in suspension and differentiates without reaching confluence (16). During in vitro adipogenesis, 3T3L1 cells require induction by a hormonal mixture (insulin, dexamethasone, and 3-isobutyl-1-methylxanthine) for 48 -72 h. The cells undergo two rounds of clonal expansion prior to final growth arrest and differentiation. Human primary preadipocytes require a constant presence of the hormone mixture and differentiate without further cell division. Differentiated adipocytes are filled with intracellular lipid droplets before maturation. The lipid droplets are coated by perilipin1 (Plin1), a protein in which expression is largely restricted to adipose tissue and is highly induced during adipogenesis. Plin1 prevents unregulated lipolysis by hormone-sensitive lipase, facilitates lipid transfer into the lipid droplet, and enables adrenergic signal-stimulated lipolysis by allowing phosphorylated hormone-sensitive lipase to enter the lipid droplets (17)(18)(19). Unregulated lipolysis increases lipid deposition in tissues, activates inflammation, and enhances insulin resistance. Null mutations of the hormone-sensitive lipase (HSL) gene are associated with an increased risk of type 2 diabetes (20). Plin1-deficient mice have lipodystrophy and atherosclerosis, as lipid is deposited in tissues outside of the normal fat storage areas (21,22). In humans, Plin1 deficiency, due to a Plin1 frameshift mutation, induces partial lipodystrophy, severe dyslipidemia, and insulin-resistance (23).
TR isoform specificity in gene regulation has been linked to the relative level of TR isoform expression in a specific tissue, response element configuration in the regulated gene, and intrinsic properties of the receptor. A recent genome-wide study of genes regulated by TR␣ and TR␤ suggests that TR isoform selectivity is not due to the response element sequence but to intrinsic receptor properties that influence the interaction with coactivator or corepressor (25). SUMO modification, therefore, may have a role in TR isoform specificity. Mutation of any sumoylation site impairs TR-dependent gene repression or induction (22).
We have demonstrated previously TR-SUMO conjugation in white adipose tissue (24). Here, we mutated TR sumoylation sites to reduce the capacity for sumoylation and investigated whether TR sumoylation plays a role in adipocyte differentiation. We utilized the mouse 3T3L1 preadipocyte differentiation model for mechanistic studies and human primary preadipocytes to confirm that these mechanisms were relevant for human adipocyte differentiation. We demonstrated that TR-SUMO conjugation is necessary for preadipocyte proliferation, PPAR␥-mediated gene expression, including Plin1 expression, and differentiation. The total number of fat cells in humans is set in late adolescence, although adipocytes in adults are constantly formed from preexisting preadipocytes, and ϳ10% of subcutaneous adipocytes are replaced every year (26). Understanding the role of TR in adipocyte differentiation provides important insights into the regulation of adipose tissue as well as potential novel therapeutic targets for metabolic diseases.

EXPERIMENTAL PROCEDURES
Human Primary Preadipocyte Culture, Transfection, and Differentiation-Human subcutaneous primary preadipocytes purchased from Lonza Inc. (Basel, Switzerland) were grown in preadipocyte medium to 80% confluence and then collected for transfection. Transfection conditions were optimized using 4 g of plasmid, 0.75 million cells, Nucleofector Kit V, and program V-033. Transfections were performed using a Nucleofector 2D device (Lonza). After electroporation, cells were grown in preadipocyte medium until growth arrest. The plating density after transfection was ϳ11,000 cells/cm 2 . Cells were then differentiated in differentiation medium (Lonza) and maintained in the same medium for 9 -12 days. The lipid droplets appeared after day 4 of differentiation.
Plasmids, 3T3L1 Cell Culture, Lentiviral shRNA Knockdown, Stable Transfection, and Differentiation-Expression vectors for sumoylation site mutants (hTR␤ K50Q, K146Q, and K443Q and hTR␣ K283Q/K288R and K389Q) were cloned by site-directed mutagenesis into pCMVTnT and/or pCMV6 entry. 3T3L1 preadipocytes were maintained in DMEM containing 10% bovine calf serum. Cells were transduced with lentiviral shRNA-TR␤ to silence mouse TR␤ gene expression. Cells (1 million) were then transfected using Nucleofector Kit V, program T-028, with plasmids (4 g each) expressing empty vector, hTR␤, hTR␣, or sumoylation site mutations hTR␤ K50Q, hTR␤ K146Q, hTR␤ K443Q or hTR␣ K283Q/K288R. After G418 selection, for 7 days, cells were grown to confluence followed by hormone mixture-induced differentiation. Stable transfection (without knockdown of endogenous TR) of 3T3L1 cells with TR and TR sumoylation site mutant experiments were conducted using nucleofection as described above. After 7 days of G418 selection, cells were grown to confluence and differentiated using the standard protocol.
C/EBP Reporter Assay-A C/EBP luciferase reporter vector carrying six C/EBP-binding elements (CCAAT) was purchased from Qiagen. Stably transfected 3T3L1 cells, as described above, were plated to 96 wells and grown to confluence. After 48 h of confluence, cells were treated with the adipocyte hormonal mixture to induce differentiation. After 24 h of induc-tion, cells were transfected with the luciferase reporter using Attractene transfection reagent. Luciferase activity was recorded in a Multimode plate reader (Promega). The data represent the mean value of six replicate reporter assays.
Cell Proliferation Assay-Cells were transfected as described above. After selection, cells were plated in a 96-well dish and grown for 36 h (ϳ80% confluence for control cells). Proliferation of the cells was analyzed (Cell Titer 96 aqueous kit, Promega). The absorbance at 490 nm was recorded in a Multimode plate reader (Promega). Proliferation was calculated as the mean value Ϯ S.E. of 10 duplicates and presented as percent control (cells transfected with empty vector).
Oil Red O Staining-Cells were fixed with 4% paraformaldehyde for 30 min and washed twice with PBS followed by one wash with 60% isopropanol. Cells were then incubated with Oil Red O for 5 min, washed with H 2 O and photographed.
Free Fatty Acid Assay-Free fatty acids were determined using a kit (catalog No. ab65341, Abcam) that detects octanoate and longer chain free fatty acids by an enzyme-based method. Cells were grown in a 96-well dish. After differentiation for 9 days, free fatty acids in the media were quantitated following the instructions with the kit.
Co-immunoprecipitation and Western Blot-3T3L1 cells were transfected with NCoR-FLAG and pCMV-TNT TR␤ expressing vectors using Nucleofector Kit V, program T-028. The FLAG immunoprecipitation kit, FLAGPT1 (Sigma-Aldrich), was used following the manufacturer's instructions. In brief, cells were lysed in lysis buffer. After centrifugation to remove the debris, the protein concentration in each sample was adjusted for uniformity. Each sample (1.0 ml of lysate) was incubated with anti-FLAG resin (60 l) overnight. Immunoprecipitated proteins were boiled with 2ϫ SDS sample buffer (20 l) and run on a 10% SDS-PAGE. The antibodies used for the Western blot were anti-TR␤ (1:500, ThermoFisher Scientific Inc., MA1-216) and anti-NCoR (1:500, Abcam, ab3482). Cells were transfected with empty vector for the co-immunoprecipitation control. Cell lysate (1.0 ml) was incubated with the anti-FLAG antibody. For input, cell lysate was immunoprecipitated using anti-TR␤. One-half of the immunoprecipitated protein was used in the Western blot as TR␤ input.
PPAR␥ expression was measured in 3T3L1 cells after 9 days of differentiation and lysis with immunoprecipitation lysis buffer (Pierce). Protein from the crude lysate (30 g) was loaded in 10% SDS-PAGE. The antibodies used were anti-PPAR␥ 2 (1:500, rabbit IgG, Santa Cruz Biotechnology) and anti-GAPDH, which recognizes a ϳ35 kDa protein (1:1,000, mouse IgG, Santa Cruz Biotechnology). GAPDH was used as a loading control, and the Western blot was imaged using the BioSpectrum 310 system.
Real-time PCR Profiling Wnt Signaling Pathway and Quanti-Tech Primer Analysis of Gene Expression-Human primary subcutaneous preadipocytes were transfected using Nucleofector as described above. Cells were then plated in a 6-cm dish and grown for 3 days (control cells were grown to ϳ80% confluence). RNA was isolated and used for a Wnt signaling profiling array (Qiagen). The values were presented as the mean value from three separate experiments. The data were analyzed using Qiagen software. RNA was isolated from cells and reverse-transcribed using a Qiagen RT 2 first-strand synthesis kit to measure expression of PPAR␥ 2 , CEBP␤, CEBP␣, Plin1, adipoQ, and aP2. QuantiTech primers (Qiagen) were used. The PCR was done with triplicates, and the data presented are a mean value from two separate experiments.
ChIP Enrichment-In brief, TR␤-BioChIP cells were generated by stable transfection of HepG2 cells with TR␤ and BiRA (biotin ligase) expression vectors. Stably transfected TR-BioChIP cells were then transduced with lentiviral shRNA SUMO1 to knock down SUMO1 mRNA. Cells were grown in Eagle's minimum essential medium with 10% serum replacement (Invitrogen) for 24 h and then treated with or without T 3 (50 nM) for 4 h. Cells were cross-linked by formaldehyde at a final concentration of 1%. Chromatin was fragmented using a Bioruptor sonicator (Diagenode Inc.). Biotinylated TR␤-associated DNA fragments were precipitated using a streptavidinbased purification system. After ChIP-seq analysis, peak regions were identified, and ChIP enrichment at the peak region was analyzed using qPCR.
Statistical Analysis-Statistical analysis for two group comparisons was performed using a two-tailed Student's t test. A value of p Ͻ 0.05 was taken as statistically significant. The p value was analyzed using the Qiagen software for the array data.

TR Sumoylation Is Required for Differentiation of Pre-
adipocytes-TR sumoylation is essential for the regulation of thyroid hormone target genes in a range of transient transfection

Role of TR Sumoylation in Adipocyte Differentiation
models (24). The biological importance of TR sumoylation, however, has not been established. We reported previously that TR sumoylation was observed in mouse WAT (24). We therefore generated TR sumoylation site mutations in TR␣ and TR␤, TR␣ K283Q/K288R and TR␤ K50Q, K146Q, and K443Q, to test the effects of TR sumoylation on adipogenesis in adipocyte differentiation models. Because Lys-283 and Lys-288 are interchangeable sites for sumoylation in TR␣, a double mutation, K283Q/K288R, was generated.
To investigate the role of TR sumoylation in preadipocyte differentiation, we transfected vectors expressing TR␣ and TR␤ with sumoylation site mutations into human primary subcutaneous preadipocyte cells. We optimized transfection conditions using cell type-specific transfection solution in combination with electroporation programs and obtained a transfection efficiency up to 75% (Fig. 1A). Growth arrest, a prerequisite condition for preadipocyte differentiation, was confirmed by a stable cell number at 48 h. Control cells (transfected with empty vector or with a vector expressing TR␣ or TR␤) reached confluence and growth arrest after 4 days of transfection. Cells expressing TR sumoylation site mutations required 6 days to reach growth arrest. The cells were then differentiated, and intracellular lipid droplets were visualized by BODIPY 495/503, which stains neutral lipids. Control cells (transfected with empty vector or a vector expressing TR␣ or TR␤) were filled with cytoplasmic lipid-storing vesicles after 9 days of differentiation, indicating commitment to a terminal adipocyte lineage (Fig. 1B). Cells expressing TR␤ and TR␣ sumoylation site mutants, however, had significantly fewer cells committed to differentiation and fewer and smaller lipid droplets in differentiated cells (Fig. 1, C and D).
We next determined whether TR sumoylation mutants similarly disrupted preadipocyte differentiation in the well characterized mouse 3T3L1 cell line preadipocyte model. TR␤ mRNA knockdown was used to reduce endogenous TR␤, and the cells were then stably transfected with TR␤ sumoylation mutants. After differentiation for 9 days, cells expressing the TR␤ sumoylation mutants, TR␤ K146Q and K443Q, had only a few lipid droplets ( Fig. 2A). TR␤ K50Q-expressing cells had near C.

B.
Plasmid: TRβ-GFP 55~65% efficiency Plasmid: GFP 60~75% efficiency A. normal differentiation, although the size and number of fat droplets were much less than in control cells ( Fig. 2A). We quantitated the impact of TR sumoylation mutants on preadipocyte differentiation by measuring free fatty acids in the medium (Fig. 2B). TR␤ K146Q-expressing cells had a 46% reduction compared with control cells and TR␤ K4443Q-expressing cells a 62% reduction. TR␤ K50Q-expressing cells had reduced free fatty acid levels but were not significantly different from control, consistent with the differentiation phenotype. TR␣ and TR␤ mRNA levels were similar in 3T3L1 cells (Fig.  2C). These data demonstrate that TR sumoylation is important for adipocyte differentiation and fat droplet formation, although the extent of disruption of adipocyte differentiation varied among the sumoylation mutants. TR Sumoylation Site Mutations Repress PPAR␥ Gene Expression-PPAR␥ gene expression is essential for adipocyte differentiation. Reduced PPAR␥ mRNA expression in white adipose tissue was reported previously in TR␣ PV mice, a TR mutation associated with resistance to thyroid hormone (RTH) (5). We analyzed PPAR␥ mRNA and protein levels in human preadipocytes differentiated for 9 days using real-time PCR and Western blot. PPAR␥ 2 mRNA expression was reduced 32% in TR␤ K50Q-, 46% in TR␤ K146Q-, 40% in TR␤ K443Q-, and 26% in TR␣ K283Q/K288R-expressing cells compared with control cells (Fig. 3A). A similar pattern of reduced PPAR␥ 2 mRNA expression was seen in 3T3L1 cells transfected with TR␤ sumoylation site mutants (Fig. 3B). The protein level of PPAR␥ 2 was reduced 14% in cells transfected with TR␤ K50Q and 33% in cells transfected with TR␤ K146Q and TR␤ K443Q (Fig. 3C).
Down-regulation of C/EBPs in Mitotic Expansion Phase by TR Sumoylation Mutants-Initiation of 3T3L1 cells differentiation by hormonal mixture allows growth-arrested cells synchronously reentering the cell cycle to undergo mitotic clonal expansion. During this period, C/EBP␤ is strongly induced, which stimulates the genes required for adipogenesis, such as C/EBP␣, and later the induction of PPAR␥ and C/EBP␣ (31,32). C/EBP is known to be regulated by TR (33). We hypothesized that reduced PPAR␥ in TR sumoylation mutant-expressing cells may be due to reduced C/EBP expression in the mitotic expansion phase. To investigate, C/EBP␤ and C/EBP␣ mRNA    were analyzed after 24 and 48 h of hormonal induction. After 24 h of hormonal induction C/EBP␤ was reduced, but this was not statistically significant compared with control cells (Fig.  4A). After 48 h of induction, the C/EBP␤ mRNA level was dramatically increased in control cells, whereas C/EBP␤ was reduced to 58% (p Ͻ 0.05) of control in TR␤ K146Q-transfected cells and 60 and 65% of control (p Ͻ 0.05) in TR␤ K443Q-and TR␣ K283Q/K288R-transfected cells, respectively (Fig. 4B). C/EBP␣ was significantly down-regulated although expressed at a low level after 24 h. After 48 h C/EBP␣ mRNA was further reduced to 13, 20, and 21% of control level in cells expressing TR␤ K146Q, TR␤ K443Q, and TR␣ K283W/288R (Fig. 4, A and  B). In a complementary study, we measured functional C/EBP in a reporter assay and saw a significant reduction of C/EBP-RE reporter expression with co-transfection of TR sumoylation mutants compared with the control (Fig. 4C). Only a trace of PPAR␥ 2 protein was detectable after 48 h of induction; however, the levels were lower in TR sumoylation mutant-expressing cells compared with control (Fig. 4C). These data demonstrate that TR sumoylation mutants inhibit C/EBP␤ and C/EBP␣, which likely contributes to reduced PPAR␥ expression. TR Sumoylation Site Mutations Attenuate PPAR␥-mediated Plin1 Gene Expression-The marked decrease in cytoplasmic lipid droplets observed with the TR sumoylation site mutations suggested a deficit in lipid droplet formation (34). Plin1 is the primary protein that coats the fat droplet. The actions of Plin1 include facilitating cytoplasm fatty acid transfer into fat droplets and regulating fatty acid release from adipose tissue (18). We analyzed Plin1 expression in differentiated human primary adipocytes and 3T3L1 adipocytes. Plin1 mRNA was expressed at a high level in control cells, but expression was significantly reduced in cells expressing TR sumoylation site mutants. In human primary adipocytes, Plin1 was expressed at 37.7% of control levels in TR␤ K50Q-expressing cells, 14.6% in TR␤ K146Q-expressing cells, 11.1% in TR␤ K443Q-expressing cells, and 15.3% in TR␣ K283Q/K288R-expressing cells (Fig. 5A). In 3T3L1 adipocytes, Plin1 was expressed at 15.4 and 9.8% of control levels in cells expressing TR␤ K146Q and TR␤ K443Q (Fig.  5B). In cells transfected with TR␤ K50Q, Plin1 was expressed at 71% of control level (p Ͻ 0.058), which was not significantly different from control, consistent with the near normal cell morphology. These results indicate that TR sumoylation is important for Plin1 gene regulation, although mutation of the TR␤ K50Q sumoylation site did not significantly reduce Plin1 expression.
We next determined whether Plin1 was directly regulated by TR. We identified two potential TR binding sites in the 5Ј-distal region of the Plin1 gene by sequence inspection and analyzed them with luciferase reporter assays. We found that both TR and the TR sumoylation site mutant had only a modest effect on reporter gene expression, either in the presence or absence of T3.
It has been reported that PPAR␥ regulates Plin1 gene transcription via a PPRE in the Plin1 5Ј-flanking region (35,36). Similar patterns of reduced mRNA expression were also observed in the PPAR␥-regulated genes FABP4 and adipoQ (Fig. 5C). The reduced PPAR␥ protein levels we observed in the presence of TR sumoylation mutants (Fig. 3) likely contribute to reduced PPAR␥-regulated gene expression. The reduction in PPAR␥-regulated gene expression, such as Plin1, however, is much greater than the reduction in the PPAR␥ protein level, indicating that additional mechanisms leading to reduced PPAR␥ signaling are likely important.
Because TR-PPAR␥ cross-talk has been demonstrated for several genes (5,37,38), we considered the possibility that TR sumoylation site mutants interfere with PPAR␥ signaling by binding to the Plin1 gene PPRE. To test this hypothesis, we performed ChIP assays in 3T3L1 cells transfected with expression vector expressing TR␤ or TR␤ with sumoylation site muta-tions (Fig. 6A). As expected, PPAR␥ binding to the PPRE was detected and was enhanced by the addition of the PPAR␥ ligand TZD, and TR␤ did not bind (Fig. 6A, lanes 1-3). In cells expressing TR␤ K146Q or TR␤K443Q, PPAR␥ binding to the PPRE was reduced and TR binding was augmented (Fig. 6A, lanes 4 -9), except that in TR␤ K443Q-expressing cells TZD enhanced binding of PPAR␥ and TR to the PPRE (lane 9). Similar to what we observed with the TR sumoylation mutants, TR␣ PV mutants have been reported to heterodimerize with PPAR␥ on the acyl-CoA promoter PPRE and block PPAR␥mediated acyl-CoA transcription (5).
We further analyzed the occupancy of TR␤ sumoylation site mutants on the PPRE of aP2 and adipoQ promoters. In the presence of wild-type TR, PPAR␥, but not TR␤, bound the PPRE of both promoters (Fig. 6B, lanes 1-3). TR␤ sumoylation mutants, however, bound to the aP2 and adipoQ PPREs in some ligand treatments (Fig. 6B, lanes 4, 6, and 9), and PPAR␥ binding was not consistently reduced (Fig. 6B, lanes 4 -9). The mRNA level of PPAR␥ 2 in the cells used for the promoter occupancy assay showed reduced mRNA expression with transfection of TR sumoylation mutants, similar to that previously shown (Fig. 3C). We concluded that TR sumoylation site mutations enhance the binding of the mutant TR to the PPRE associated with reduced PPAR␥-mediated gene expression.

Role of NCoR in TR Sumoylation-mediated Inhibition of Plin1
Expression-Sumoylation influences the activity of transcription factors by conformational changes that modulate the inter- FIGURE 7. NCoR⌬ID partially rescued impaired Plin1 mRNA expression and adipocyte differentiation. A, 3T3L1 cells expressing NCoR⌬ID and/or TR␤ mutant were differentiated for 9 days. Plin1 mRNA level was determined by real-time PCR. B, medium-free fatty acid levels were analyzed as described previously (see Fig. 2 legend). C, the neutral triglyceride and lipid in differentiated cells were stained by Oil Red O. D, PPAR␥ mRNA was determined in 3T3L1 cells with co-transfected TR␤ sumoylation mutants as shown. *, p Ͻ 0.05 compared with control. ⌬ID, NCoR⌬ID; pGS5, empty vector action with ligand, substrate, or other proteins. Unliganded TR binds the corepressor NCoR, and the binding is disrupted by the addition of ligand. TR dominant-negative mutations associated with RTH bind to corepressor NCoR, but the addition of ligand does not disrupt the binding, leading to tonic gene repression (39). To investigate whether inhibition by TR sumoylation site mutations may also be linked to interaction with NCoR, we utilized an NCoR mutation, NCoR⌬ID, which lacks the interacting domains (N2 and N3) and does not bind to TR (40 -42). This mutation, when co-transfected or expressed in an animal model with an RTH-associated TR mutant, "rescues" the abnormal phenotype. We performed a co-immunoprecipitation assay in cells transfected with TR␤ K443Q and NCoR or NCoR⌬ID and measured the TR␤ K443Q-NCoR complex. Overexpression of NCoR⌬ID should limit TR sumoylation mutant interaction with NCoR and release the transcription repression. Co-immunoprecipitation analysis showed that TR␤ bound NCoR only in the absence of T 3 as expected (Fig.  6D, lanes 1 and 2). TR␤ K443Q mutant bound to NCoR both in the absence and presence of T 3 (Fig. 6D, lanes 3 and 4). Overexpression of NCoR⌬ID significantly diminished the TR␤ K443Q-NCoR interaction (Fig. 6D, lanes 5 and 6 compared  with 3 and 4), indicating that NCoR⌬ID reduced NCoR interaction with the sumoylation mutant TR. We analyzed the Plin1 mRNA level and free fatty acid in medium of cells co-transfected with TR␤ sumoylation muta-tions and NCoR⌬ID. In NCoR⌬ID co-transfected cells, Plin1 mRNA was increased to 53% of the control level in TR␤ K146Qexpressing cells and 75% in TR␤ K443Q-expressing cells (Fig.  7A). The medium-free fatty acid content was also increased to 52 and 60% of control level in these cells, respectively (Fig. 7B). Cell differentiation was significantly enhanced in the presence of NCoR⌬ID, with some increase in neutral triglyceride and lipid droplets as shown by Oil Red O staining (Fig. 7C), although it was not fully restored to the level in control cells. These data indicate that the ligand-resistant interaction of the TR sumoylation mutant TR with NCoR contributes to altered gene expression and reduced free fatty acid and that this phenotype is partially rescued by the expression of NCoR⌬ID.
Genome-wide Study of the Impact of TR Sumoylation on DNA Binding and Co-factor Recruitment-The importance of TR sumoylation for normal ligand-reversible interaction with NCoR is supported by our genome-wide study of TR␤ DNA binding after SUMO1 knockdown. Four genes, C/EBP␤, CREB, SOX7, and SREBF1, all important for adipocyte differentiation and metabolism, were evaluated (Fig. 8). SUMO1 knockdown had no significant effect on TR␤ DNA binding to any of these genes. The TR recruitment pattern of coactivators and corepressors to these genes, however, was altered. After SUMO1 knockdown, TR did not recruit coactivator SRC-1 and was not able to release NCoR in response to T 3 . These data demonstrate that TR-SUMO conjugation is a key factor for coactivator and corepressor interaction.

Role of TR Sumoylation in Proliferation of Human Primary
Preadipocytes-Human adipocytes are differentiated from preadipocytes generated from mesenchymal stem cells in adipose tissue. The number of adipocytes in WAT is dependent on the proliferation of preadiocytes (9,26). Newly formed adipocytes are more insulin-responsive and metabolically active. In in vitro models of adipogenesis, the proliferation of preadipocytes is also a critical step in downstream differentiation.
We explored the role of TR sumoylation in regulating the proliferation of preadipocytes using human primary preadipocytes. Control cells (transfected with empty vector, TR␤, or TR␣) were grown to confluence in 4 days. In the same period, cells expressing the TR sumoylation site mutants TR␤ K146Q and TR␣ K283Q/K288R were only ϳ40% confluent (Fig. 9A). The proliferation rate was similar in control cells transfected with empty vector, TR␤, or TR␣. The proliferation rate for TR␤ K146Q-expressing cells was 40.5% (p Ͻ 0.001) of that of control cells transfected with empty vector and for TR␣ K283Q/ K288R-expressing cells 37.3% (p Ͻ 0.03) (Fig. 9B). In a parallel study, we used 3T3L1 preadipocytes stably transfected with TR␤ sumoylation mutants (K50Q, K146Q, and K443Q). A similar reduction in the rate of proliferation was seen. The proliferation rate for TR␤ K50Q-expressing cells was 81.5% (p Ͻ 0.03) of control cells, for K146Q-expressing cells 43.6% (p Ͻ 0.045), and for K443Q-expressing cells 40.3% (p Ͻ 0.05) (Fig.  9C).
Differential Effects of TR␣ and TR␤ Sumoylation Site Mutants on the Wnt Signaling Pathway-Wnt signaling is inhibited during differentiation of mesenchymal cell to precommitted preadipocytes and during terminal differentiation of preadipocytes (43). In vitro differentiation requires proliferation of precommitted preadipocytes until growth is rested. The proliferation of precommitted preadipocytes is via cell cycle regulation, which is partly mediated though the canonical Wnt/␤-catenin signaling pathway. Activation of this pathway leads to increased transcription of the genes controlling G 1 /S transition, such as cyclin D1, c-Jun, and c-Myc (41). Disassembling of the ␤-catenin-LEF/TCF complexes leads to the G 2 /M phase in cell cycle progression (44). We noticed premature proliferation in human primary preadipocytes. To investigate the influence of TR sumoylation on Wnt/␤-catenin signaling pathway-mediated proliferation, we used human primary preadipocyte cells transfected with TR␣ K283Q/K288R or TR␤ K146Q, compared with those transfected with wild-type TR␣ and TR␤, and analyzed the cells with the Wnt signaling profiling array. Compared with control cells, the mRNA of six genes were significantly down-regulated in the presence of either TR␣ or TR␤ sumoylation site mutants. These down-regulated genes included, frizzled 9 (Fzd9), frizzled-related protein (Frzb or SFRP3), matrilysin (MMP7), C-terminal-binding protein 1 (CTBP1), and casein kinase-1␣1 and -2␣1 (CSNK1A1 and CSNK2A1) ( Table 1 and Fig. 8). FRZB is known to antagonize Fzd by directly binding to Wnt and inhibits both canonical and noncanonical Wnt signaling pathways (45). CSNK proteins interact with AXIN1, a component of the ␤-catenin disruption complex. CTBP1 regulates tumor suppressor and promotes cellular proliferation in cancer cells (46) MMP7 is the ␤-catenin/ TCF/LEF target (47). The reduction in these genes suggested altered canonical Wnt activity. Besides these effects seen with either TR␣ or TR␤, sumoylation site mutants had effects on Wnt ligands, Wnt receptors, and Wnt signaling inhibitors that differed between TR␣ and TR␤ ( Table 2 and Fig. 10). TR␤ K146Q expression was associated with significant down-regulation of the genes involved in the canonical Wnt signaling pathway, including CTNNB1 (3.4-fold) coding for ␤-catenin, LEF (2.9-fold), CCND1 (3.3-fold), which codes for cyclin D1, Wnt3 (3.17-fold), and Wnt8A (4.58-fold). FZD8, involved in both canonical and noncanonical Wnt signaling (48,49), was reduced 2.85-fold. TR␤ K146Q expression was associated with up-regulation of two genes (Wnt10A and DKK1) that are inhibitors of adipogenesis. Wnt10A stimulates osteoblastogenesis and suppresses adipogenesis (50). Dickkopf-related protein 1 (DKK1) binds to LPR5/6 and prevent formation of the Wnt-FZD-LPR5/6 complex, which lead to cytoplasmic ␤-catenin degradation. In contrast to TR␤ K146Q mutant, TR␣ K283Q/ K288R showed greater influence on the noncanonical Wnt signaling pathway. DKK3, Wnt5A, Princkle1, and Vangl2 gene

DISCUSSION
We have demonstrated that TR sumoylation plays an important role in preadipocyte proliferation and differentiation. Mutation of sumoylation sites in either TR␣ or TR␤ showed a similar effect on adipocyte differentiation, including fewer and smaller lipid droplets compared with control cells and reduced PPAR␥-mediated gene expression. There was, however, some    MARCH (Fig. 11). Expression of C/EBPs during mitotic clonal expansion is important for stimulation of pro-adipogenic genes. TR sumoylation mutants down-regulated C/EBP␤ and C/EBP␣, resulting in reduced PPAR␥2 expression. In addition to the C/EBPs, other factors have been reported to regulate PPAR␥ expression and influence adipogenesis, including GATA (56), KLFs (11,12), FOXO1 (57), SREBP (58,59), and TNF␣/NFB (60). Whether these factors are additionally influenced by TR sumoylation mutants is not known.

Role of TR Sumoylation in Adipocyte Differentiation
TR isoform-specific actions have been reported in heart, brain, pituitary, liver, bone, and testes (3). These isoform-specific actions are thought to be related to the distribution of TR isoforms in tissue as well as to their intrinsic properties. We showed some TR isoform preference for the disruption of some adipogenic pathways, but the sumoylation mutants in the context of TR␣ or TR␤ had similar effects.
It has been speculated that there is a TR␤ and TR␣ selectivity for the T 3 response element (TRE) contained in the gene regu-latory region. A recent genome-wide ChIP-seq that analyzed TR isoform binding sites concluded that receptor isoform-selective DNA binding is not observed in neural cells, although receptor-selective response is observed (25). This finding indicates that other properties, such as posttranslational modification, may confer TR isoform specificity. TR sumoylation may be one of the mechanisms for TR isoform-selective regulation. TR␣ and TR␤ have a different number of sumoylation motifs. SUMO conjugation modifies the surface of TR, resulting in general conformational change and may result in alterations of protein-protein interactions such as TR interactions with co-repressors. These changes are dynamic, as sumoylation conjugation is rapid and reversible. A further level of regulation may be related to the multiple TR sumoylation motifs and may be influenced by whether TR is fully or selectively conjugated. Although both mono-and polysumoylation have been observed in TR (24), there is no evidence so far for selective sumoylation, because it is very difficult to isolate the TR with site-specific SUMO conjugation.
There are similarities between TR sumoylation mutants and RTH-associated TR mutants with respect to their action on PPAR␥ signaling, as both reduce PPAR␥ protein and disrupt PPAR␥ signaling. TR-SUMO conjugation plays an important role in Plin1 gene expression. Mutations of the TR␤ sumoylation site markedly diminished Plin1 expression due to crosstalk with the PPAR␥ signaling pathway. As shown by ChIP assay, TR␤ K443Q bound strongly to the Plin1 PPRE (Fig. 6A). The DNA-binding complex is either a TR/TR homodimer or a TR/RXR heterodimer. The addition of TZD stimulated the FIGURE 11. Summary of the influence of TR sumoylation on preadipocyte proliferation and differentiation. A, under normal conditions, TR-SUMO conjugation is dynamic. TR-SUMO promotes recruitment of either the coactivator or corepressor depending on the promoter configuration, ligand availability, and cellular signals. Predipocytes undergo normal proliferation and growth arrest. When induced, cells are differentiated into adipocytes filled with lipid droplets. The differentiation is driven by PPAR␥-mediated gene expression. B, TR sumoylation site mutants alter protein-protein interface and/or TR conformation, resulting in increased recruitment of NCoR resistant to disruption by T 3 and a diminished ability to recruit the coactivator SRC1. Additionally, TR mutants may lose or reduce the capacity to interact with factors carrying SUMO interaction motifs. Our data show that TR sumoylation site mutants inhibit Plin1expression by interfering with PPAR␥ expression and PPAR␥ signaling, leading to impaired lipid droplet formation and adipogenesis. In the proliferation phase, TR sumoylation site mutants reduce preadipocyte proliferation by inhibiting Wnt canonical signaling. Green fluorescence (BODIPY) depicts neutral triglycerides and lipids in adipocytes. Blue fluorescence (DAPI) depicts nuclei.
binding of both PPAR␥ and TR␤ K443Q to the PPRE, suggesting that TR␤ K443Q interacted directly with PPAR␥. Mutants associated with RTH, TR␣ PV and TR␣ P398H, have been shown to influence PPAR signaling by binding to a PPRE. Wildtype TR is also capable of cross-talk with PPAR, although it does not bind the PPRE but competes for the common heterodimer partner, RXR.
The interference of TR␤ sumoylation mutant K443Q with PPAR␥-mediated Plin gene expression was due, in part, to constitutive interaction with NCoR. Utilizing NCoR⌬ID to disrupt TR-NCoR interaction partially reversed the inhibition by the mutant TR, consistent with the mutation at Lys-443 influencing the interface with NCoR. A recent study reported that the altered adipogenesis seen in TR␣PV mice was completely rescued by expression of NCoR⌬ID (61). Although the model system and the TR mutation differ from the TR sumoylation mutants that we studied, the mechanism is similar. These similarities emphasize the central role of TR co-regulator interactions and the various ways that it can be disrupted by mutations that influence SUMO conjugation or ligand binding.
TR-SUMO conjugation, however, likely has regulatory roles in addition to co-regulator interaction. The addition of NCoR⌬ID only partially reversed the inhibition associated with expression of TR sumoylation site mutants, with partial recovery of Plin1 gene expression and adipocyte differentiation. TR sumoylation site mutations likely have effects beyond cross-talk with PPAR and constitutive interaction with NCoR. We have identified two sumoylation motifs in TR␣ and three in TR␤. It is possible that monosumoylation in TR promotes interaction with factors in addition to NCoR. Moreover, polysumo chains have been found in many SUMO substrates (62,63) including TR in mouse liver, heart, and adipose tissue (24). Polysumoylation may create additional binding sites for factors containing SUMO interaction motif (Fig. 8), which have been demonstrated in transcription regulation of HIF-1␣ by SUMO polymers. The biological relevance of the polymeric SUMO signal has been studied predominantly in yeast. It is related to DNA repair and sporulation efficiency (64,65). Polysumoylation may function, for example, as a signal for ubiquitination, which would lead to selective removal of specific proteins (66,67). Therefore, mutations at the SUMO conjugation site of a SUMO acceptor (TR) may lead to protein accumulation and alter signaling pathways. Further characterization of the TR protein interactions influenced by mono-and polysumoylation in adipogenesis will provide potential targets for metabolic regulation.
TR sumoylation is required for normal adipocyte differentiation and proliferation. We have identified key pathways influenced by TR sumoylation including PPAR␥ protein levels and signaling, C/EBPs expression, NCoR interactions with TR, and Wnt signaling (Fig. 11). The relative roles of these various actions may vary depending on the tissue studied, as well as the integration of other nutritional signals. The use of primary cells and cell lines showed largely complementary results and supports the relevance of TR sumoylation in the preadipocyte to adipocyte model. The relative roles of the various mechanisms studied, however, and identification of therapeutic targets rel-evant to metabolic diseases will need to be studied in whole animal models. Fig. 1 was incomplete in the version of the article that was published as a Paper in Press on January 8, 2015. The correct image is now shown.